Académique Documents
Professionnel Documents
Culture Documents
=
+
e
s
C e
t
C
h t
C (2)
This is comparable to the equation of straight line of the form:
Y= m.X +A (3)
Thus, if a plot is obtained in between
) ( h t
C
+
and
t
C , it would yield a straight line and slope of this line (m) represent the
value of
.a.h K
L
e from which .a
L
K can be calculated. Similarly, the intercept on Y-axis (A) represents the term
)
.a.h
L
K
1 (
e
s
C from which the value of
s
C can be known. Thus it clearly indicates that in order to determine the value of
.a
L
K , it must be carried out at a uniform interval of time.
In the present study the curves have been drawn between
) ( h t
C
+
against
t
C and time interval h is taken as 15 minutes. The
slope of this line is known and the value of overall oxygen transfer coefficient .a
L
K is calculated. This method is used for
calculation of performance of different aerators. The plot between C
(t+h)
and C
t
is obtained for different configuration of rotor
aerator at different speeds of 36, 42, 48 and 60 rpm. The data generated from different rotors followed different curves and
indicated different values of K
L
a and aeration efficiency.
4.2 Variation of DO with Time for various depth of immersion
We refer to figure 4, which illustrates that the increase in DO concentration is very high in first 15 minutes, and then it gradually
attains a saturation value pertaining to the performance of the respective aerator. It is observed that the maximum increase in DO
concentration, that is, from 0.0 mg/L to 8.2 mg/L is attained by CBR aerator when the blade tip angle is 47 degrees and the depth
of immersion is 5.5 cm. All the curves were plotted at 48 rpm aerator speed. The literature cited by the author reveals that the
optimum performance is generally obtained at blade angle of 45 degrees. Therefore the above blade angle is chosen.
4.3 Variation of K
L
a and power for various blade angles and speed
From Figure 5, it can be observed that the variation of K
L
a and power for various blade angles and speed, with respect to CBR
aerator and considering the depth of immersion 5.5 cm is an optimum value. It is very clear from figure that as the power and
speed increases the corresponding values of K
L
a also increases. It is also evident from the figure that at a speed of 48 rpm the
power requirement is observed to be in the medium range that is 73.8 W, whereas there is a considerable increase in value of K
L
a
in the tune of 10.33 h
-1
. As the power consumption decreases below 73.8 W, the value of K
L
a also decreases accordingly, which
further decreases the aeration efficiency, which is not acceptable as per the values quoted by the researchers.
When the power consumption increases beyond 73.8 W, there is a marginal increase in K
L
a value, but aeration efficiency drops
sharply from 2.269 kgO
2
/kWh to1.37 kgO
2
/kWh. As the speed of CBR aerator is increased from 48 rpm to 60 rpm, it is observed
that splashing of wastewater outside the ditch takes place, which deteriorates the surrounding atmosphere and thereby causing
problems for working people present at the plant. This may also create problems in driving mechanism as the motor is mounted
below the ditch.
Bhuyar et al./ International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 1-15
6
0
1
2
3
4
5
6
7
8
9
0 15 30 45 60 75 90 105 120
Time Interval min
D
O
C
o
n
c
e
n
t
r
a
t
i
o
n
m
g
/
l
DOI 4.8CM DOI 5.5CM DOI 6.3CM DOI 7.2CM
Figure 4.Variation of DO with Time for different depth of immersion and blade tip angle of 47 deg.
0
10
20
30
40
50
60
70
80
90
100
110
120
130
140
150
KLa Power KLa Power KLa Power
< 27 Degree < 47 Degree < 60 Degree
Speed 42 RPM
Speed 48 RPM
Speed
60RPM
Figure 5. Variation of K
L
a and Power for CBR for various Blade angles and Speeds, depth of immersion 5.5 cm
4.4 Variation in DO for various blade tip angle with Time
The variation in DO for various blade tip angle with time is shown in Figure 6. The figure 6 represents the curves plotted at 42
rpm and 5.5 cm depth of immersion. Three types of blade tip angles were tested, which suggest that maximum DO concentration is
achieved when the blade tip angle is at 47 degrees. With this configuration, DO level of 8.2 mg/L is attained in 45 minutes and
thereafter the DO concentration curve remains constant which indicates that the saturation deficit of oxygen is met.
Bhuyar et al./ International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 1-15
7
0
1
2
3
4
5
6
7
8
9
0 15 30 45 60 75 90 105 120
Time Interval min
D
O
C
o
n
c
e
n
t
r
a
t
i
o
n
m
g
/
l
DO 27 deg DO 47 deg DO 60 deg
Figure 6. Variation of DO Vs Time for various angles of CBR, speed 42 rpm and DOI- 5.5 cm
Figure 7 and Figure 8 represent the same plot as above follow the similar trend as that of Figure 6. Combined analysis of all the
three figures reveals that the optimum performance is recorded at blade tip angle of 47 degrees.
0
1
2
3
4
5
6
7
8
9
0 15 30 45 60 75 90 105 120
Time Interval min
D
O
C
o
n
c
e
n
t
r
a
t
i
o
n
m
g
/
l
DO 27 deg DO 47 deg DO 60 deg
Figure 7. Variation of DO Vs Time for various angles of CBR, speed 48 rpm and DOI- 5.5 cm
4.5 Plot between C
(t+h)
and C
t
for determination of K
L
a
As stated earlier, the plot between C
(t + h)
and C
t
will yield a straight line. The slope of the line is known and the value of K
L
a is
calculated. The time interval h is maintained as 15 minutes (Figure 9). The data obtained for different depth of immersion
produced different curves. It is quite evident from the figure that the curve obtained for 5.5 cm depth of immersion, 48 rpm speed,
and 47 degree blade tip angle yielded maximum value of K
L
a.
Bhuyar et al./ International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 1-15
8
0
1
2
3
4
5
6
7
8
9
0 15 30 45 60 75 90 105 120
Time Interval min
D
O
C
o
n
c
e
n
t
r
a
t
i
o
n
m
g
/
l
DO 27 deg DO 47 deg DO 60 deg
Figure 8. Variation of DO Vs Time for various angles of CBR, speed 60 rpm and DOI-5.5 cm
y = 0.0754x + 7.4954
R
2
= 0.9944
0
1
2
3
4
5
6
7
8
9
0 2 4 6 8 10
DO Level (mg/l)at any time (t),Ct
D
O
L
e
v
e
l
(
m
g
/
l
)
a
t
t
i
m
e
(
t
+
h
)
,
C
(
t
+
h
)
DOI=4.8 cm DOI=5.5cm DOI=6.3cm DOI=7.2cm
Figure 9. DO Level at time (t+h) Vs DO Level at time t for Determination of K
L
a
The variation in K
L
a and power required are presented in Table1. It also suggests that depth of immersion is directly proportional
to the power consumption. On the other hand the increase in K
L
a is not that significant. As a golden mean, the value of K
L
a is
finally taken as 10.33 h
-1
when the power required is 73.8 W and it is observed that the value of AE is the highest for these values,
which is discussed later on.
Table 1.Variation of K
L
a and Power with Depth of Immersion for 48 rpm Speed
S.No. Depth of Immersion (cm) K
L
a (h
-1
) Power (W)
1 4.8 8.414 69.9
2 5.5 10.33 73.8
3 6.2 10.58 93.5
4 7.2 10.93 136
Bhuyar et al./ International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 1-15
9
4.6 Variation in Aeration efficiency and K
L
a with respect to power
Figure 10 describes the variations of K
L
a and AE for various blades tip angles and speeds.
6.84
9.72
6.71
10.33
11.31
7.43
8.2
1.468
1.683
2.269
1.434
1.239
1.17
11.52
11.4
1.365
1.549
1.37
0
1
2
3
4
5
6
7
8
9
10
11
12
Kla AE Kla AE Kla AE
K
L
a
(
h
-
1
)
&
A
e
r
a
t
i
o
n
e
f
f
i
c
i
e
n
c
y
(
k
g
O
2
/
k
W
h
)
KLa,<27 KLa, <47 KLa, <60 AE,< 27 AE, <47 AE, <60
42 RPM
48 RPM
60 RPM
Figure 10. Variation in k
L
a and AE with different blade angles and speeds
Out of the three configurations tested, it can be seen that at 48 rpm and 47 degree blade tip angle, the optimum value of K
L
a is
observed to be 10.33 h
-1
and recorded the highest value of AE that is 2.269 kgO
2
/kWh. This is because, the AE is the function of
power and as stated earlier the power required increases with increase in speed and depth of immersion. Even though the values of
K
L
a at 60 rpm and 27, 47, 60 degree blade tip angles are more than that of 48 rpm, the power required to attend the same is quite
higher. Therefore, the AE values for above configuration and speed decreases drastically to 1.365, 1.37, 1.17 kgO
2
/kWh as against
the AE values of 1.683, 2.269 and 1.239 kg O
2
/kWh at speed of 48 rpm.
The performance of aerator is better judged by a parameter known as aeration efficiency (AE), which is defined as amount of
oxygen transferred per unit power under field conditions. Standard Aeration Efficiency of the aerator is defined as the amount of
oxygen transferred per unit of power, at standard conditions of 20
C to 17.5
C during the experimentation. Secondly, after addition of sodium sulphite to the water,
the initial concentration of water was sometimes observed to be on negative side, even though the prescribed quantity of sodium
sulphite was added to the water as mentioned in the literature. (Kumar,1991). Therefore optimum AE expressed in terms of SAE
i.e. at standard conditions. The SAE of aerator designed for present study comes out to be 2.95 KgO
2
/kWh. Table 2 shows the
comparison of different aerators in terms of aeration efficiency.
Also in the present study, the inner surface of the ditch was coated with anticorrosive paint thereby minimizing the frictional
losses. Therefore, it was observed that the flow of water in the ditch remains in turbulent stage from rotor1 to rotor 2 and
subsequently loss of kinetic energy was also found to be less because of negligible frictional losses. Hence the efficiency is
Bhuyar et al./ International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 1-15
10
observed to be on the higher side as compared to Boyd (1998). More-over the limitation of the present study is that, it was a
laboratory setup and therefore only few influencing parameters which are responsible for aeration efficiency were concentrated
upon of out of many, otherwise, the standard aeration efficiency in the present case may have increased at least 1.3 times than the
efficiency obtained presently.
Table 2. Comparison of different aerators in terms of standard aeration efficiency (SAE)
Sr.No Aerator type
Standard aeration
efficiency
kg oxygen/kWh
References
1. Taiwanese aerator 1.17 Boyd and Watten 1989; Boyd 1998
2. Japanese aerator 1.03 Boyd and Watten 1989; Boyd 1998
3. Auburn University 2.25 Boyd and Watten 1989; Boyd 1998
4. Present Study 2.95
4.7 Model
In this paper attempts are made to correlate the data generated in the laboratory and the data calculated by using mathematical
model. In this phenomenon, the author used two models. From the study of the data generated in laboratory, the attempts are made
to identify the dependent and independent variables and by using Buckingham theorem concept, these variables are correlated by
developing dimensionally homogeneous equations.
( ) ( )
=
031 . 0
038 . 1
768 . 1
000746 . 0
D
h
N a
L
K
(4)
( ) ( ) ( )
=
1729 . 0 1942 . 0
1412 . 1
062 . 2
0694 . 0 a
L
K
D
h
N P (5)
where,
K
L
a overall oxygen transfer coefficient min
-1
,
N Speed of aerator in rpm,
h/D Ratio of depth of immersion to the diameter of aerator,
- Blade tip angle in degrees,
P Power required in watts (W).
The calculated values using the above models yielded good coefficient of determination (R
2
) with the experimentally observed
values which are in good agreement with standard error of estimation. A program in CPP is developed to interrelate the various
influencing parameters mentioned above and are solved by multiple linear regression analysis method. Figure 11 and Figure 12 are
the plots between the experimentally determined values and the calculated values from the model formulated for K
L
a and power
respectively.
It has been estimated that the simulation equations (4) and (5) predict the values of K
L
a and power with an average standard error
of estimation of 0.0164 and 7.66 respectively and with R
2
values of 0.9797 and 0.9892 respectively, when compared with
experimentally determined values. Thus equations (4) and (5) are justifiable by taking into consideration all of the experimental
errors themselves. Therefore, the above equations may be used with confidence for predicting the K
L
a and power (P) and can be
considered as design equations for CBR aerator. The standard error of estimation for power happens to be large (7.66) because of
the fluctuation in the power supply from the source. It was observed during the experimentation that the voltage fluctuation was
from 210 V to 235 V. To minimize the error for power measurement, the average readings were taken.
Bhuyar et al./ International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 1-15
11
y = 0.9706x - 0.0035
R
2
= 0.9797
0
0.05
0.1
0.15
0.2
0.25
0.3
0 0.05 0.1 0.15 0.2 0.25 0.3
Calculated KLa(h-1)
O
b
s
e
r
v
e
d
K
L
a
(
h
-
1
)
Figure 11. Relationship between observed and calculated values of k
L
a
y = 1.0246x - 1.46
R
2
= 0.9892
0
20
40
60
80
100
120
140
160
0 20 40 60 80 100 120 140 160
Calculated Power (W)
O
b
s
e
r
v
e
d
P
o
w
e
r
(
W
)
Linear (Cal.Power(W) Vs Obs Power(W))
Figure 12. Relationship between observed and calculated values of Power (P)
5. CFD Model
The computational fluid dynamic model of an aerator used for treatment of waste water in oxidation ditch is a crucial part of this
study. The results of this phase have a huge impact on the quality of the results obtained by experimental setup and the
mathematical model developed. The main problem was to model the geometry of the aerator and the oxidation ditch. Actually the
CFD part of this study is only to find out the correlation between the experimentally observed data and data generated by the CFD
model and if the CFD model thus developed establishes good correlation with the observed data, it could be used to analyze the
various parameters of the said system without fabricating the physical model. It is relatively trivial to draw the aerators in Gambit,
but the problem is the overall number of elements. A single aerator has a number of groups with 12 blades each. Going further
with the computation, it can be seen that in the oxidation ditch there are 2 aerators with 2 groups of 12 blades each. Overall, it is
about 48 elements. The source of momentum in the ditch is the rotating aerators. To model the aerators the moving mesh option in
fluent is used and the part of aerator that is submerged in the water is created.
5.1 Modeling the Aerator
The aerator can be modeled in different ways. The one that is finally used is the one phase, 3D model. To model the moving wall
in fluent a volume with the shape of aerator was created and to define the constant velocity of it a user defined function is used.
This way the cells that were source of momentum and were working like moving wall were created. There are low and high
pressure regions before and after the rotor blades respectively, while a strong turbulence is created by the blades. Because it was
decided to use a one phase model, instead of creating the whole aerator, only the part of the aerator under water level is used.
5.2 Computational domain
The model dimensions are 32 x12 x 4.3 cm (length x width x height). Only the part of the aerator that is submerged in the water
is modeled. The sides of the model are symmetrical boundary conditions. That way, we represent a number of groups of rotating
blades. After meshing 2, 40,000 cells were created.
Bhuyar et al./ International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 1-15
12
5.3 Setting up fluent
Model:
Solver: 3 D, Pressure based, steady state
Viscous: Standard k (k-epsilon) fluid model
Material:
Water: Density 1000 kg/m
3
Viscosity 0.001003 kg/m-s
Operating conditions: Gravity OFF
5.4 Boundary conditions
Inlet pressure inlet Pressure inlet with relative pressure equals 0, instead of creating any flow it is providing as much water to
the model as it is needed. This way aerator is the only source of momentum in the ditch.
Outlet Pressure outlet Similar to inlet pressure equals 0. As inlet is not creating any flow, it is allowing water to go out of the
model.
Side walls symmetry - This way a very wide aerator which contains a number of blades groups is represented.
Bottom wall Roughness and other parameters are not important because they do not have influence on the flow very close to
the aerator.
Free surface symmetry It is a standard method to represent a free surface.
Mesh: The tetrahedral 3-D mesh is used with mesh inflation applied at inner wall.
5.5 Solving
The case was computed after approximately 2000 iterations. In this research paper, an attempt is made to correlate the data
generated in the laboratory with the model developed using Computational Fluid Dynamics. A series of simulation experiments
were carried out on the CFD model developed.(Software ANSYS - FLUENT).
Flow governing equations used were, the Reynolds-averaged, Navier-Stokes equations, governing the three dimensional, steady,
incompressible flow in the oxidation ditch.
The turbulence kinetic energy k, and its rate of dissipation , are computed from the transport equations for the Standard k
turbulence model.
Figure 13: Streamlines of Oxygen Transfer in the water present in Oxidation Ditch
Figure 13 shows the variation in the oxygen transfer in different zones. It is clearly visible from the figure that oxygen transfer in
the vicinity of aerators is quite predominant than that of other parts of the ditch.
The velocity profile was copied very close to the brush. It was done in order to minimize the influence of the flow in the ditch on
the velocity profile. The copied values were measured 10 cm from the end of the rotors blade. As the turbulence is created, the
dissolved oxygen (DO) transport was modeled using finite rate chemistry. The reaction was set to change DO in the water.
Bhuyar et al./ International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 1-15
13
Figure 14. Streamlines of velocity of water flow in Oxidation Ditch
5.6 Determination of K
L
a from CFD model
A model with large amount of moving parts would be too complicated to be created and because of that it is sometimes
impossible to solve to obtain accurate results. To overcome this situation simple logic was used. The velocity of flow in the
oxidation ditch should be in the range of 0.25 m/s to 0.6 m/s, (Abusam et al. 2002). During laboratory experimentation, the speed
of the aerator was maintained at 36, 42, 48 and 60 rpm, so that the above-mentioned velocity range could be satisfied. It was
observed that the lower limit of velocity was satisfied by maintaining the speed at 36 rpm and 0.5 m/s velocity was observed at 60
rpm. As the speed was increased beyond 60 rpm, there was rise in velocity but splashing of water from the ditch was observed,
which is not tolerable, because oxidation ditch is used to treat the sewage.
It was a difficult task to model the brush rotation. It was, therefore decided to maintain the velocity of flow from 0.25 m/s to 0.5
m/s so that the speed of rotation of aerator could be determined from the experimental observations. Hereafter, the rise in DO
concentration was measured for equal interval of time and a plot is obtained between C
t
and C
(t+h).
As stated earlier the plot
between C
(t + h)
and C
t
will yield a straight line. The slope of the line is known and the value of K
L
a is calculated.
Figure 15 illustrates the relationship between the observed values of K
L
a and simulated values of K
L
a obtained from CFD model
simulation. The R
2
value of 0.9609 is obtained by plotting the above values against each other, which suggests that the CFD model
and the simulation values thus generated are in good agreement with the experimental model and the observation recorded therein.
y = 1.0523x - 1.146
R
2
= 0.9609
0
1
2
3
4
5
6
7
8
9
10
11
12
0 1 2 3 4 5 6 7 8 9 10 11 12 13
Experimental values of KLa
C
F
D
m
o
d
e
l
V
a
l
u
e
s
Figure 15. Relationship between Experimental values and CFD Model values
Bhuyar et al./ International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 1-15
14
6. Conclusions
The following conclusions are drawn from the study conducted:
1. The aerator configured in the present study was explicitly used for oxidation ditch process only, and the speed of
rotation was in the range of 36 rpm to 60 rpm. At 48 rpm speed, 47 degree blade tip angle and 5.5 cm depth of
immersion, the optimum value of aeration efficiency (AE) was obtained and was observed to be 2.269 kgO
2
/kWh.
Secondly, with such a low speed, the back splash condition did not occur, as was observed in the case of Peterson E.L
and Walkar M.B. (2002).Whereas the speed of paddle wheel aerator is of the tune of 135 rpm (Moulick et al, 2005) and
therefore, the occurrence of backsplash is very significant at higher speed using paddle wheel as an aerator.
2. The standard aeration efficiency of curved blade aerator was observed to be 2.95 kgO
2
/kWh.
3. It has been estimated that the simulation equations (4) and (5) predict the values of K
L
a and power with R
2
values of
0.9797 and 0.9892 respectively, when compared with experimentally determined values. Thus equations (4) and (5) are
justifiable by taking into consideration all of the experimental errors themselves. Therefore above equations may be
used with confidence for predicting the K
L
a and power (P) and can be considered as design equations for CBR aerator.
4. The model of the brush was created to understand the influence of an aerator on the flow velocity.
5. The oxygen transfer analysis gave all the results as were expected, but tracing the particles is not enough to model all
the processes that takes place in the ditch, because the chemical, biological and physical processes have large influence
on settling of particles.
6. A CFD model thus developed shows a good coefficient of correlation (R
2
= 0.9609) with the experimentally observed
values and proves to be best-fit model with respect to oxidation ditch.
Finally, it can be concluded that aeration can be made more effective by optimizing various factors affecting it viz, speed of
aerator, depth of immersion, blades angles etc. The CFD model developed leads to the improved quality of the waste-water
treatment while saving money as required for running and maintaining it.
Nomenclature
dc/dt Rate of oxygen transfer in water.
C
s
Saturation concentration of dissolved oxygen in water (mg/L).
C
t
Concentration of dissolved oxygen in water at any time t.
h Time interval for taking observations (min).
C
(t+h)
Concentration of dissolved oxygen in water after time (t+h).
DO Dissolved oxygen (mg/L)
AE Aeration efficiency (kgO
2
/kWh).
SAE Standard aeration efficiency (kgO
2
/kWh).
DOI Depth of immersion (m).
CFD Computational fluid dynamics.
K
L
a Overall oxygen transfer coefficient min
-1
,
N Speed of aerator in rpm,
h/D Ratio of depth of immersion to the diameter of aerator,
Blade tip angle in degrees,
P Power required in watts (W).
References
Abusam.A., Keesman K.J., Spanjers H., Straten G. and Meinema. K. 2002. Effect of oxidation ditch horizontal velocity on the
nitrogen removal process. European Water Management Online, EWA 2002.
APHA., 1985. Standard methods for the examination of water and wastewater, 15
th
Ed., American Public Health Association,
America Water Works Association, and Water Pollution Control Federation, Washington D.C.
Boyd C.E., Watten, B.J. 1989. Aeration systems in aquaculture. CRC Critical Reviews in Aquatic Science, Vol.1, No. 3, CRC
press, pp. 425-472.
Boyd C.E., 1998. Pond water aeration systems. Aquacultural Engineering, Vol.18, pp.9-40.
Cancino, B., Roth, P., and Reu, M., 2004. Design of high efficiency surface aerators part 1. Development of new rotors for
surface aerators. Aquacultural Engineering, Vol. 31, pp. 83-98.
Cancino B., 2004. Design of high efficiency surface aerators part 2. Rating of surface aerator rotors. Aquacultural Engineering,
Elsevier, Vol.31, pp. 99-115.
Eckenfelder W.W., 1956. Process design of aeration system for biological waste treatment. Chemical. Engineering Progress,
Vol.52, No 7, pp. 286-292.
Bhuyar et al./ International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 1-15
15
Metcalf and Eddy Inc., 2001. Waste water Engineering; Treatment disposal and reuse. 3
rd
ed. Tata McGraw Hill, New Delhi
India, 1200p.
Moulick S. Mal, B.C. Bandyopadhyay, S., 2005. Design characteristics of single hub paddle wheel aerator. Journal of
Environmental. Engineering, Vol.131. No. 8, pp. 1147-1154.
Horvath I., 1984. Modeling in the technology of waste water treatment, Pergamon press, Tarrytown, N.Y.
Kumar R., 1991. Oxygen transfer by mechanical aeration. M.E thesis, Dept. of Civil. Eng. University of Roorkee. Roorkee, India.
Ognean T., 1993. Aspects concerning scale-up criteria for surface aerators. Water Research, Vol.27. No.3, pp.485-488.
Peterson E.L and Walkar M.B., 2002. Effect of speed on Taiwanese paddle wheel aeration. Aquacultural Engineering, Vol.26, pp.
129-147.
Rao A. R., 1999. Predication of reaeration rate in square, stirrers tanks. Journal of Environmental. Engineering, Vol.133, No 4, pp.
411 418.
Rao A. R., Kumar, B., 2007. Neural modeling of square surface aerators. Journal of Environmental. Engineering, Vol.125, No. 3,
pp. 215-223.
Schmidtke N.W., Horwath, T., 1977. Scale-up methodology for surface aerated reaerators. Progress in Water Technology, Vol.9,
pp. 477-493.
Thakre S.B., Bhuyar, L.B. and Deshmukh, S.J., 2008-a. Effect of different configurations of mechanical aerators on oxygen
transfer and aeration efficiency with respect to power consumption. International Journal of Mechanical System Science and
Engineering, WASET. Vol 2., No.2., pp. 100-108.
Thakre S.B., Bhuyar L.B. and Deshmukh S.J., 2008-b. Mathematical models for overall gas transfer coefficient using different
theories and evaluating their measurement accuracy. International Journal of Chemical and Biomolecular Engineering.
(WASET), Vol. 1, No. 4., pp. 203-207.
Thakre S.B., Bhuyar L.B. and Deshmukh S.J., 2009. Oxidation ditch process using curved blade as aerator. International Journal
of Environmental Engineering, Science and Technology, Vol.6. No.1, pp.113-122.
Wesner G.M., Ewing, J. J., Lineck, T.S., Jr., and Hinrichs, D.J.1977. Energy conservation in municipal wastewater treatment.
EPA-130/9-77-011, NTIS No PB81-165391, U.S. EPA Res., Washington, D.C.
Zlokarnik M. 1979. Scale up of surface aerators for waste water treatment. Advance Biochemical Engineering, Vol.11, pp.157-179.
Biographical notes
Dr. Lalit B. Bhuyar is Professor & Head, Department of Mechanical Engineering, PRM Institute of Technology and Research, Badnera,
Amravati (India). He has engaged in teaching and research activities since the last 20 years. His field of specialization is Thermal and Fluid
Engineering. Dr. L.B. Bhuyar has published several papers in various national, international conferences and journals. He has guided 8 students
for their Ph.D. work.
Prof. Shashank B. Thakre is working as a Professor in Mechanical Engineering Department, PRMIT & R, Badnera, Maharashtra, India. He has
done his research in the field of reutilization of domestic waste water by designing the aerators to be used in oxidation ditch. He has published 30
papers on various research areas such as renewable energy, environmental pollution and design, in national and international journals. Presently,
he is focusing on experimentation and analysis of solar chimney power plant.
Dr. Nitin W. Ingole is Principal, IBSS College of Engineering , Ghatkheda, Amravati, Maharashtra, India. He has done his M.Tech. in
Environmental Engineering. Dr. Ingole is engaged in Biomass energy development since the last 6 years. He has done extensive work on biogas
development from aquatic plants and biomass. He has twenty and more international publication in national and international journals.
Received February 2009
Accepted February 2009
Final acceptance in revised form May 2009
MultiCraft
International Journal of Engineering, Science and Technology
Vol. 1, No. 1, 2009, pp. 16-32
INTERNATIONAL
JOURNAL OF
ENGINEERING,
SCIENCE AND
TECHNOLOGY
www.ijest-ng.com
2009 MultiCraft Limited. All rights reserved
A variational analysis for large deflection of skew plates under uniformly
distributed load through domain mapping technique
Debabrata Das, Prasanta Sahoo
*
and Kashinath Saha
Department of Mechanical Engineering, Jadavpur University, Kolkata 700032, India.
*
Corresponding author (e-mail: psahoo@vsnl.net, psjume@gmail.com, Tele-fax: +91 33 2414 6890)
Abstract
In the present paper, the static behaviour of thin isotropic skew plates under uniformly distributed load is analyzed with the
geometric nonlinearity of the model properly handled. A variational method based on total potential energy has been
implemented through assumed displacement field. The computational work has been carried out on a square normalized domain
based on an appropriate domain mapping technique. Validation study for the present work has been carried out quite extensively
to establish its accuracy and stability. The developed method is quite general to be applied readily for any classical boundary
condition, but to maintain brevity results have been furnished for clamped and simply supported boundaries only.
Keywords: Skew plate; Variational method; Domain mapping; Large deflection; Jacobian
1. Introduction
Skew plates have quite a good number of applications in modern structures. Skew plate structures can be found frequently in
modern construction in the form of reinforced slabs or stiffened plates. Such structures are widely used as floors in bridges, ship
hulls, buildings, etc. Skew plates are also used in the construction of wings, tails and fins of aircrafts and missiles. Simulation of
static behaviour of skew plates is an interesting area of work for the researchers.
Alwar and Rao (1973) presented the non-linear analysis of orthotropic skew plates of constant thickness subjected to uniform
transverse load. They developed a numerical technique of dynamic relaxation to carry out the analysis. Dynamic relaxation is
essentially a step by step integration of critically damped vibration using viscous damping to ensure the attainment of a steady-
state solution. Alwar and Rao (1974) used the same technique of dynamic relaxation. They have investigated the non-linear
behaviour of clamped isotropic skew plate of constant thickness under uniformly distributed load. Srinivasan and Ramachandran
(1976) presented large deflection elastic behaviour of clamped, uniformly loaded orthotropic skew plates. They obtained the
solution by using Newton-Raphson method to solve the non-linear algebraic equations transformed from governing non-linear
partial differential equations. Buragohain and Patodi (1978) investigated the large deflection skew plate problem and formulated it
by energy principle to derive a set of nonlinear algebraic equations which are solved by using Newton-Raphson iterative method
with incremental loading. They claimed their formulation to be independent of boundary conditions and furnished results for
clamped and simply supported isotropic skew plates of constant thickness with immovable edges and subjected to uniformly
distributed transverse load. Chia (1980) considered the moderately large deflection elastic behaviour of homogenous isotropic and
laminated anisotropic rectangular as well as skew plates by analytical method. Xiang et al. (1995) studied the elastic buckling
behaviour of skew Mindlin plate under shear load. Based on the Rayleigh-Ritz method, new critical shear load factors and
buckling mode shapes for two kinds of shear loads - rectangular shear (R-shear) and skew shear (S-shear) - have been investigated.
Liew and Han (1997) presented the bending analysis of a simply supported thick
skew plate based on the first-order shear
deformation using Reissner/Mindlin theory.
They introduced the geometric transformation of the governing differential equations
and boundary
conditions of the plate from the physical
domain into a unit square computational domain. Subsequently they derived
a set of
linear algebraic equations from the transformed differential
equations via differential quadrature method (DQM) and the
approximate solutions of the problem are obtained by
solving the set of algebraic equations. Using Boundary Element Method
Das et al./ International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 16-32
17
(BEM), a fundamental solution in oblique coordinates for the analysis of isotropic skew plates under transverse loading has been
obtained by Rajamohan and Raamachandran (1997) based on the analogy with the anisotropic plates bending governing equation
in Cartesian coordinates. They have used Charge Simulation Method (CSM), a version of BEM, to obtain the solution and the
simplicity of the method was demonstrated by avoiding numerical integration due to the fact that the domain integrals, which
appear in any boundary element method is avoided by assuming a suitable polynomial particular integral function.
Saadatpour et al. (1998) developed a numerical method for the stability analysis of arbitrary quadrilateral-shaped elastic plates
with internal supports and used the Galerkin method in conjunction with the natural coordinates of the plate as a basis for the
analysis of general plates. Wang et al. (2003) proposed a new version of differential quadrature method (DQM) to obtain buckling
loads of thin anisotropic rectangular and isotropic skew plates for clamped and simply supported boundary conditions. In this new
technique, two degrees of freedom in boundary points for determination of the weighting coefficients have been introduced in the
existing differential quadrature element method (DQEM). Duan and Mahendran (2003) had developed a new non-linear
quadrilateral hybrid/mixed shell element with five degrees of freedom at each node, using oblique coordinate systems to analyze
the large deflection behavior of skew plate under uniformly distributed and concentrated loads. The authors have claimed to
overcome the shortcomings for carrying out the large deflection analyses of the skew plates that have a 60 skew angle under a
concentrated load. Singh and Elaghabash (2003) presented a numerical method for the linear and geometrically nonlinear static
analysis of thin plates having quadrangular boundary with four straight edges. They have derived the matrix equation of
equilibrium using the work-energy principle with the assumed displacement fields expressed by algebraic polynomials.
Liew et al. (2004) employed a mesh-free radial basis function (RBF) method for the buckling analysis of non-uniformly loaded
thick plates. They obtained initial (i.e., pre-buckling) stresses by discretizing the variational form of the static system of equations
and calculated the static buckling loads of the plates by solving the resultant Eigen value equation. Muhammad and Singh (2004)
presented an energy method for the linear static analysis of first order shear deformable plates of various shapes. In this method,
the displacement fields are defined in terms of the shape functions, which correspond to a set of predefined points and are
composed of significantly higher order polynomials. Malekzadeh and Fiouz (2007) have presented two different differential
quadrature (DQ) approaches based on the thin plate theory (TPT-DQ) and the first order shear deformation plate theory (FSDT-
DQ) to investigate the large deformation analyses of thin and moderately thick orthotropic skew plates with rotationally restrained
edges. In both approaches, they have modeled the geometrical nonlinearity of the plate by using Greens strain and Von Karman
assumption. Daripa and Singha (2009) studied the influence of corner stresses on the stability behaviour of composite skew plates.
Prakash et al. (2008) analyzed thermal post-buckling of skew plates having functionally graded materials while Wu and Shih
(2006) considered the dynamic instability of arbitrarily laminated skew plates.
Several researchers have addressed the linear and non-linear dynamic problems of skew plates incorporating various
complicating effects. Singh et al. (2006) considered the vibration behaviour of thermally stressed composite plates whereas Ashour
(2009) analyzed free vibration behaviour of symmetrically laminated clamped skew plates. Das et al. (2008, 2009) used a
variational method to study large amplitude dynamic problem of skew plates under simply supported and clamped boundary
conditions. Differential quadrature method has also been used for non-linear vibration problems of laminated composite skew
plates (Malekzadeh, 2007, 2008; Malekzadeh and Karami, 2006).
In the present work, a simulation model for large deflection static analysis of a thin isotropic skew plate under uniform transverse
pressure has been presented. The mathematical formulation is based on variational form of energy principle. The displacement
fields are assumed as orthogonal polynomials in the computational domain that satisfy the plate boundary conditions and the
resulting nonlinear set of equations are solved by direct substitution method using relaxation parameter. Results have been
validated successfully with the available ones and a good amount of new results has been presented which will be of great help to
designers. It is found that convergence in results for large skew angle plates poses numerical difficulty and to this end a new
extrapolation technique has been prescribed.
2. Mathematical formulation
A skew plate ( a x b xt ) with skew angle is shown in Figure 1. For large displacement formulation, it is assumed that the
material of the plate is isotropic and homogeneous. The thickness of the plate is considered to be sufficiently small so as to avoid
the effect of shear deformation. Also the stress and strain measures are based on the original dimensions of the plate.
2.1 Mapping of physical domain into computational domain
The physical domain of interest is mapped to the computational domain in coordinate system as shown in Figure 2. In
this domain, the intersections of constant and constant lines passing through the gauss points are selected as the reference
points for computation. The coordinates of all the gauss points in the computational domain are calculated numerically in the x-y
plane by suitable grid generation techniques. This technique of numerical grid generation is not used extensively by any other
researcher, and it can be used for similar other problems involving irregular geometry such as rectangular plate with a central
circular cut-out.
Das et al./ International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 16-32
18
Figure 1. Skew plate in x-y plane.
(a) Physical domain
(b) Computational domain
Figure 2. Gauss point locations for skew plate.
Das et al./ International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 16-32
19
2.2 Variational principle
Variational principle states that
( ) 0 U V + = (1)
where,
b m
U U U = + is total strain energy,
b
U is strain energy due to pure bending and
m
U is strain energy due to stretching of
its middle surface, V is the potential energy of the external forces and is the variational operator in configuration space. It is to
be noted here that geometric nonlinearity, i.e., the nonlinear strain-displacement relationship is taken into consideration through
inplane stretching.
b
U is given by (Timoshenko and Woinowsky, 1959),
2 2
tan cos 2 2 2 2
2 2 2 2
0 tan
2
2
y a b
b
y
D w w w w
U
x y x y
+
| | | |
= + +
| | `
| |
\ \
)
( )
2
2 2 2
2 2
2 1
w w w
dxdy
x y x y
(
| |
(
| `
|
(
\
)
(2)
where,
( )
3
2
12 1
Et
D
, flexural rigidity of plate and u , v and w are components of displacement along x, y and z directions
respectively.
m
U is given by (Timoshenko and Woinowsky, 1959),
( )
2
2 2 2
2 2 2 tan cos
2
0 tan
1
4
2 1
y a b
m
y
Et u v u w v w w w
U
x y x x y y x y
+
| | | | | | | | | | | |
= + + + + +
` ` | | | | | |
\ \ \ \ \ \
) )
2 2
2 2
1 1 1
2 2
2 2 2
u v v w u w u v u v
x y y x x y y x y x
| | | | | | | |
+ + + + + +
` | | | |
\ \ \ \
)
2
u v w w
dxdy
y x x y
( | |
+ +
( ` |
( \ )
(3)
For a plate loaded with uniform transverse pressure p , potential energy V is given by,
( )
tan cos
0 tan
y a b
y
V pw dxdy
+
=
(4)
2.3 Approximate displacement field
The displacement fields w , u and v are expressed by linear combinations of unknown parameters
i
d as follows:
( ) ( )
( ) ( )
( ) ( )
, , ,
, , ,
, ,
nw
i i
i 1
nw nu
i i nw
i nw 1
nw nu nv
i i nw nu
i nw nu 1
w d
u d
v d
=
+
= +
+ +
= + +
=
=
=
(5)
where, ( ) , , ( ) , and ( ) , are sets of orthogonal functions and nw, nu and nv are number of functions for w , u and
v respectively.
The start functions of these orthogonal sets are selected to satisfy the flexural and membrane boundary conditions of the plate.
As the entire computation is carried out in normalized computational domain, all the start functions are defined in computational
domain. The basic functions for the definition of plate deflection w come from the flexural boundary conditions, either clamped
or simply supported. The start functions for u and v are selected to satisfy the zero displacement conditions at the boundary, being
considered as the only form of in-plane boundary condition. The higher order functions are generated following a two dimensional
implementation of GramSchmidt scheme and this is done for the first time in this type of study.
In conventional method of generating higher order functions, first the higher order one dimensional functions are generated from
the corresponding start functions and then the higher order two dimensional functions are generated by ordered multiplication of
the one dimensional functions corresponding to two orthogonal directions. In the present method, higher order functions in a
particular orthogonal direction are generated corresponding to different coordinate values in the other orthogonal direction through
Gram-Schmidt scheme. Then these values at a particular coordinate are collected to generate the complete set of higher order two-
dimensional functions through the implementation of Gram-Schmidt scheme. The advantage of using this technique is that unlike
the earlier case, the start functions can be of two variables that need not be separable in spatial coordinates.
Das et al./ International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 16-32
20
2.4 Derivation of governing system of equations
The governing equations are obtained in the following form using Eqs. (2), (3), (4) and (5) in Eq. (1),
[ ]{ } { } K d f = (6)
where, [ ] K =
b
k
(
+
m
k
(
is the stiffness matrix of dimension ( ) nw nu nv + + ,
b
k
(
is the stiffness matrix due to bending action,
m
k
(
is the stiffness matrix due to stretching action, { } d is the unknown coefficient vector of dimension ( ) nw nu nv + + and { } f
is the load vector of dimension ( ) nw nu nv + + .
b
k
(
is of the form given below:
[ ]
(
(
(
=
b b b
b b b
b b b
b
k k k
k k k
k k k
k
33 32 31
23 22 21
13 12 11
(7)
where,
2 2 2
1 1 2 2 2
11
2 2 2 2 2 2
1 1
0 0
nw nw
j j j b i i i
j i
k D
x x y y y x
= =
| | | | | |
| | | | | |
( | | | = + + | | |
| | |
| | |
\ \ \
\ \ \
( )
2 2 2
2 2 2
2 2 2 2 2 2
1
j j j
i i i
x y y x x y
| | | | | |
| | | | | |
| | | + + | | | `
| | |
| | |
\ \ \
\ \ \ )
2
2
2 det
j
i
J d d
x y x y
( | |
| |
( | | `
|
|
(
\
\ )
12
b
k
(
=
13
b
k
(
=
21
b
k
(
=
22
b
k
(
=
23
b
k
(
=
31
b
k
(
=
32
b
k
(
=
33
b
k
(
=
0
m
k
(
is of the form given below:
[ ]
(
(
(
=
m m m
m m m
m m m
m
k k k
k k k
k k k
k
33 32 31
23 22 21
13 12 11
(8)
where,
( )
2 2
1 1
11
2
1 1 1 1
0 0
2 1
nw nw nw nw
j j
m i i i i
i i
j i i i
Et
k d d
x x x y y y
= = = =
| | | |
( = +
| |
\ \
( )
1 1 1
1
nw nw nw nu
j j j
i i i i i nw i
i i i
i i i nw
d d d
x y x y y x y x y
= = = +
| || | | |
+ + +
` | | |
\ \ \ )
( )
1
1 det
nw nu nv
j
i nw nu i
i
i nw nu
d J d d
x x y
+ +
= + +
( | |
+
( |
\
( )
1 1
12
2
1 1 1 1
0 0
2 2
2 1
nw nw nu nw nw
j j
m i i nw i i nw
i i
j i nw i i
Et
k d d
x x x y x y
+
= = + = =
| | | |
(
= +
| |
\ \
( )
1
1 det
nw
j
i i nw
i
i
d J d d
y y x
=
( | |
+
( |
\
( )
1 1
13
2
1 1 1 1
0 0
2 2
2 1
nw nw nu nv nw nw
j j
m i i nw nu i i nw nu
i i
j i nw nu i i
Et
k d d
y y y x y x
+ +
= = + + = =
| | | |
(
= +
| |
\ \
Das et al./ International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 16-32
21
( )
1
1 det
nw
j
i i nw nu
i
i
d J d d
y x x
=
( | |
+
( |
\
( )
1 1
21
2
1 1 1 1
0 0
2
2 1
nw nu nw nw nw
j nw j nw
m i i i i
i i
j nw i i i
Et
k d d
x x x y y x
+
= + = = =
| | | |
(
= +
| |
\ \
( )
1
1 det
nw
j nw
i i
i
i
d J d d
x y y
=
( | |
+
( |
\
( )
( )
1 1
22
2
1 1
0 0
2 1 det
2 1
nw nu nw nu
j nw j nw
m i nw i nw
j nw i nw
Et
k J d d
x x y y
+ +
= + = +
(
(
= +
(
( )
1 1
23
2
1 1
0 0
2
2 1
nw nu nw nu nv
j nw
m i nw nu
j nw i nw nu
Et
k
y x
+ + +
= + = + +
(
=
( ) 1 det
j nw
i nw nu
J d d
x y
(
+
(
( )
1 1
31
2
1 1 1 1
0 0
2
2 1
nw nu nv nw nw nw
j nw nu j nw nu
m i i i i
i i
j nw nu i i i
Et
k d d
y y y x x y
+ +
= + + = = =
| | | |
(
= +
| |
\ \
( )
1
1 det
nw
j nw nu
i i
i
i
d J d d
x y x
=
( | |
+
( |
\
( )
1 1
32
2
1 1
0 0
2
2 1
nw nu nv nw nu
j nw nu
m i nw
j nw nu i nw
Et
k
x y
+ + +
= + + = +
(
=
( ) 1 det
j nw nu
i nw
J d d
y x
(
+
(
( )
1 1
33
2
1 1
0 0
2
2 1
nw nu nv nw nu nv
j nw nu j nw nu
m
j nw nu i nw nu
Et
k
y y
+ + + +
= + + = + +
(
=
( ) 1 det
j nw nu j nw nu
J d d
x x
(
+
(
The load vector { } f is of the form[ ]
T
f f f
13 12 11
,
where,
{ }
1 1
11
1
0 0
det
nw
j
j
f p J d d
=
=
and { }
12
f ={ }
13
f =0, as there is no in-plane loading.
Here, J is the Jacobian of the transformation from y x plane to plane given by,
x y
J
x y
( | | | |
( | |
\ \
(
=
(
(
| | | |
( | |
\ \
3. Solution Methodology
The set of governing equations (Eq. (6)) is nonlinear in nature and solved by direct substitution technique using successive
relaxation scheme (Cook et al., 1989). For each load-step, the values of { }
i
d are assumed to evaluate the stiffness matrix. Using
the stiffness matrix and based on assumed values, new values of { }
i
d are calculated by matrix inversion technique from the
expression { } [ ] { }
1
d K f
= . Calculated values of { }
i
d are compared with their values in the previous iteration. If the difference
comes below a predefined value of error limit, , the convergence of deflection field is assumed, otherwise the values of { }
i
d are
modified with a relaxation parameter and it is taken as next approximation for the values of { }
i
d . Once the solution is obtained
for a load, stress and strain fields are computed following routine procedure. Subsequently an increment is given on load and
iteration starts with the present solution or with another assumed deflection field, which may be extrapolated from the previous
Das et al./ International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 16-32
22
solution set. In the present extrapolation technique, the ratios of the unknown coefficients { }
i
d for the preceding two load steps
are used to calculate the guess values of { }
i
d at the present load step. Although this technique has not been tried extensively, this
can be utilized as a useful numerical tool for the solution of nonlinear set of equations, when convergence gets difficult. The entire
program for the present study is developed in Fortran.
4. Results and Discussions
All the results in this paper have been presented for uniform transverse pressure. Although, the present method is quite general to
be applied for any classical boundary condition, normalized load-deflection plots are presented for skew plates for two types of
flexural boundary conditions: all edges clamped (CCCC) and all edges simply supported (SSSS). In both the cases, the edges of
the plate are modeled as immovable by imposing zero in-plane displacement boundary conditions. Here C and S denote
clamped and simply supported boundaries, respectively, for a single edge of the plate. Results for different possible combinations
of clamped and simply supported boundary conditions are omitted here to maintain brevity, but it will be taken up in the
subsequent study. The normalized load (
*
p ) is defined as
* 4
( ) /(16 ) p pa Dt = and the maximum normalized deflection is defined
as
*
max
/ w w t = . The results are generated using the following material and geometric parameter values, E = 210 GPa, = 1/3,
a=1.0 m and t = 0.01 m. The skew angle and the aspect ratio (a/b) of the plate have been separately indicated while presenting the
results.
The convergence of the present method with the number of functions used to approximate the displacement fields is carried out
for CCCC skew plate with / a b =1. The convergence study is carried out for deflections at two different locations namely, (0.5,
0.5) and (0.25, 0.25). It is to be noted that the coordinates of the points are given in normalized form and (0.5, 0.5) represents the
point at which maximum deflection occurs. The results of the convergence study in terms of normalized displacement ( / w t ) are
shown in Table 1 for two different skew angles, 30
0
and 45
0
at
*
p =7.94. The numbers in the parentheses provide a break up for
the number of the functions for each plate displacements corresponding to two orthogonal directions. Table 1 clearly shows the
convergence of displacements with increase in number of functions. All the results presented in this paper are generated using (7 X
7) number of functions for each plate displacements.
Table 1: Convergence of displacements in normalized form ( / w t ) at two different locations of rhombic CCCC skew plate for two
different skew angles at
*
p =7.94
Number of functions in two orthogonal directions Normalized
coordinates
( , )
4 X 4 5 X 5 6 X 6 7 X 7 8 X 8
(0.5, 0.5) 0.0976 0.0974 0.0974 0.0972 0.0972 30
0
(0.25, 0.25) 0.0233 0.0232 0.0232 0.0232 0.0231
(0.5, 0.5) 0.0478 0.0478 0.0479 0.0478 0.0478 45
0
(0.25, 0.25) 0.00824 0.00827 0.00816 0.00814 0.00807
4.1 Validation of the present approach
In order to validate the present formulation, the results from the present analysis for CCCC boundary condition are contrasted
with the results available in the existing literature and also with the results generated by the commercial finite element package
ANSYS. The results for
*
w against
*
p from present study for aspect ratio 2/3 with two different skew angles have been compared
with analytical solution (Chia, 1980) and with the results generated by ANSYS (version 8.0) and shown in Figure 3(a). The results
in ANSYS are generated using SHELL 63 element with 80 120 uniform meshes. The plot shows agreement is better with the
results of ANSYS. It also indicates that the amount of deviation in the results of Chia (1980) is more pronounced for 60
0
skew
angle. The validation of the present approach is further shown in Figure 3(b) with the results of Singh and Elaghabash (2003) for
rhombic ( / a b =1) skew plates having 30
0
and 45
0
skew angles. The normalized load-deflection behaviour of simply supported
rhombic skew plate is also compared with the results of Singh and Elaghabash (2003) for 45
0
skew angle and it is shown in Figure
3(c). In view of Figures 3(a) 3(c), it may be concluded that the present formulation has the accuracy and stability to analyze large
deflection behaviour of skew plates.
Das et al./ International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 16-32
23
(a) CCCC, a/b=2/3
(b) CCCC, a/b=1
Das et al./ International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 16-32
24
(c) SSSS, a/b=1
Figure 3. Validation plots for load-deflection behaviour: (a) CCCC, a/b=2/3, (b) CCCC, a/b=1 and (c) SSSS, a/b=1.
(a) a/b=1
Das et al./ International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 16-32
25
(b) a/b=2/3
(c) a/b=1/2
Figure 4. Normalized load-deflection plots of CCCC plates with varying aspect ratios and skew angle.
Das et al./ International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 16-32
26
Figure 5. Normalized load-deflection plots of CCCC plates with 75
0
skew angle for varying aspect ratios.
Figure 6. Normalized load-deflection plots of CCCC plates with 30
0
skew angle for varying aspect ratios.
4.2 Results for plate with all edges clamped
The variation of normalized loaddeflection behaviour of skew plates with the variation of skew angles with all edges clamped
has been obtained for different values of aspect ratios ( / a b ). The skew angles selected in each case are 15
0
, 30
0
, 45
0
and 60
0
.
Figures 4(a) - 4(c) show
*
p vs.
*
w plots for / a b =1, 2/3 and 1/2 respectively. It is to be mentioned that the results for skew
angles 15
0
and 75
0
for all three aspect ratios and for skew angles 30
0
and 45
0
for aspect ratio 1/2 are new in this paper. It is seen
that deflection (
*
w ) decreases as the skew angle increases for all values of load parameter. This is due to the fact that with increase
in skew angles, the system stiffness increases due to increase in stretching force in the middle plane of the plate. Further, for any
particular value of load parameter, the slope of the load-deflection plots in the normalized plane decreases with increase in skew
angles. This nature of the plots shows a consistent trend for all values of aspect ratio.
The normalized load deflection behaviour for =75
0
has been separately shown in Figure 5 corresponding to same values of
aspect ratios. These plots have not been accommodated in Figure 4 to maintain greater clarity and to avoid significant difference in
normalized load scale. However, this plot becomes indicative to show the effect of different aspect ratio for a particular value of
Das et al./ International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 16-32
27
skew angle in normalized load-deflection behaviour. To get further insight to this phenomenon, another plot is shown in Figure 6
for 30
0
skew angle, which also agrees well with the previous trends in results. From the above results, it is quite obvious that the
nature of the normalized load-deflection plot tends to a straight line with increase in skew angle and this is true for three different
cases of aspect ratio values. Although general trend for normalized load-deflection behaviour for a fixed value of skew angle is
shown in Figure 6, it may change significantly as can be seen from the plots for aspect ratio 2/3 and 1/2 in Figures 5 and 6. A close
examination of Figures 5 and 6 reveal that the normalized load-deflection behaviour for aspect ratio 2/3 and 1/2 have interchanged
their positions with respect to the same for aspect ratio 1. This appears to be intriguing but similar trend had also been reported in
Duan and Mahendran (2003) for =60
0
of clamped skew plate. This finding points towards the existence of a strong interaction
among aspect ratio, skew angle and boundary condition on the load-deflection behaviour of the plate in normalized plane, which
need further investigation through generation of additional results. It should be mentioned that the normalized loads in Figures 4(a-
c), Figure 5 and Figure 6 are different because the system stiffness changes with change in skew angle and aspect ratio. Generally,
the stiffness of the skew plate increases with increase in skew angle and/ or increase in aspect ratio.
4.3 Results for plate with all edges simply supported
The variation of normalized loaddeflection behaviour of skew plates with the variation of skew angles with all edges simply
supported has been obtained for different fixed values of aspect ratios ( / a b ) and they are shown in Figs. 7(a) 7(c) for / a b =1,
2/3 and 1/2 respectively. The trends for these plots remain the same as discussed for CCCC plates. It is to be mentioned that the
results for skew angles 15
0
and 75
0
for all three aspect ratios, for skew angles 30
0
and 45
0
for aspect ratio 2/3 and for skew angles
45
0
for aspect ratio 1/2 are new in this paper.
The normalized load deflection behaviour with the variation of aspect ratio for =75
0
has also been shown separately in Figure
8 due to the same reason mentioned in section 4.2. A similar plot for =45
0
has been shown in Figure 9. Both these plots show
that for a particular value of skew angle, the slope of the normalized load-deflection plot increases with decrease in aspect ratio.
Unlike the case of clamped plates for =75
0
(Figure 5), the normalized load-deflection behaviour for simply supported plates with
similar skew angle show a consistent trend in conformity with the general behaviour as mentioned in section 4.2
In order to visualize the static deflection of the plate, an enlarged plot for deflected position of a 60
0
skew plate with clamped
boundary for aspect ratio 2/3 has been shown in Figure 10. It is to be noted that the deflected shape of the plate is represented by
considering only the vertical displacement. Singh and Elaghabash (2003) indicated that there are many numerical results in the
literature but only a limited quantity of them is truly available. Moreover it is found during computation that the convergence of
the solution is really a critical issue. In general, it is found that the convergence of the solution has serious difficulties for skew
angles 60
0
and above, and in such situations an assumed displacement field extrapolated from previous solutions, yields better
convergence. It should be noted that the normalized loads in Figures 7(a-c), Figure 8 and Figure 9 are different due to the same
reason stated in section 4.2 for clamped plate.
(a) a/b=1
Das et al./ International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 16-32
28
(b) a/b=2/3
(c) a/b=1/2
Figure 7. Normalized load-deflection plots of SSSS plates with varying aspect ratios and skew angle.
Das et al./ International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 16-32
29
Figure 8. Normalized load-deflection plots of SSSS plates with 75
0
skew angle for varying aspect ratios.
Figure 9. Normalized load-deflection plots of SSSS plates with 45
0
skew angle for varying aspect ratios.
Figure 10. Enlarged view of deflected shape of a CCCC plate with 60
0
skew angle.
Das et al./ International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 16-32
30
5. Conclusions
In this paper, a simulation model for large deflection static analysis of a thin isotropic skew plate under uniform transverse
pressure has been presented based on variational form of energy principle. The displacement fields are assumed as admissible
orthogonal polynomials and the resulting nonlinear set of equations is solved by direct substitution method using relaxation
parameter. The validation of the present method is carried out successfully with the available results, thus establishing the accuracy
of the present method. The results indicate a strong interaction between aspect ratio and skew angle on the load-deflection
behaviour of skew plates. Also, a new extrapolation technique for the solution of the set of nonlinear algebraic equations has been
introduced. A good amount of new results are presented in this paper. This method of analysis establishes the basic kernel, which
can be extended to various other types of static and dynamic skew plate problems.
Nomenclature
a Length of the plate
b Width of plate along skew direction
{d} Vector of unknown coefficients
D Flexural rigidity of the plate
E Youngs modulus of the plate
{f} Load vector
J Jacobian of the coordinate transformation
[K] Stiffness matrix
[k
b
] Stiffness matrix due to bending
[k
m
] Stiffness matrix due to stretching
nw Number of functions for w
nu Number of functions for u
nv Number of functions for v
p Uniform transverse pressure
p
*
Non-dimensional load parameter
t Thickness of the plate
u Displacement along x direction
U Strain energy
U
b
Strain energy due to flexural action
U
m
Strain energy due to membrane action
v Displacement along y direction
V Work potential
w Transverse displacement
w
*
Maximum transverse displacement
Poissons ratio
i
Set of functions defining approximate displacement field w
i
Set of functions defining approximate displacement field u
i
Set of functions defining approximate displacement field v
Skew angle
References
Alwar R.S. and Rao N.R. 1974. Large elastic deformations of clamped skewed plates by dynamic relaxation. Computers &
Structures, vol. 4, pp.381-398.
Alwar R.S. and Rao N.R. 1973. Nonlinear analysis of orthotropic skew plates. AIAA Journal, vol.11, No.4, pp. 495-498.
Ashour A.S. 2009. The free vibration of symmetrically angle-ply laminated fully clamped skew plates, Journal of Sound and
Vibration, vol. 323, No. 1-2, pp. 444-450.
Buragohain D.N. and Patodi S.C. 1978. Large deflection analysis of skew plates by lumped triangular element formulation.
Computers & Structures, vol.9, No.2, pp.183-189.
Chia C. (1980): Nonlinear analysis of plates. USA:McGraw-Hill Inc.
Cook R.D., Malkus D.S. and Plesha M.E. 1989. Concepts and applications of finite element analysis. USA: John Wiley and
Sons.
Daripa R., Singha, M.K. 2009. Influence of corner stresses on the stability characteristics of composite skew plates, International
Journal of Non-Linear Mechanics, vol. 44, No. 2, pp.138-146.
Das et al./ International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 16-32
31
Das D., Sahoo P., Saha K. 2008. Large-amplitude dynamic analysis of simply supported skew plates by a variational method,
Journal of Sound and Vibration, vol. 313, No. 1-2, pp.246-267.
Das D., Sahoo P., Saha K. 2009. Nonlinear vibration analysis of clamped skew plates by a variational method, Journal of
Vibration Control, vol. 15, pp.985-1017.
Duan M. and Mahendran M. 2003. Large deflection analyses of skew plates using hybrid/mixed finite element method.
Computers & Structures, vol.81, pp.1415-1424.
Liew K.M. and Han J.B. 1997. Bending analysis of simply supported shear deformable skew plates. Journal of Engineering
Mechanics, vol.123, No.3, pp.214-221.
Liew K.M., Chen X.L. and Reddy J.N. 2004. Mesh-free radial basis function method for buckling analysis of non-uniformly
loaded arbitrarily shaped shear deformable plates. Computers Methods in Applied Mechanics and Engineering, vol.193, pp.205-
224.
Malekzadeh P. 2007. A differential quadrature nonlinear free vibration analysis of laminated composite skew thin plates, Thin-
Walled Structures, vol. 45, No. 2, pp.237-250.
Malekzadeh P. 2008. Differential quadrature large amplitude free vibration analysis of laminated skew plates based on FSDT,
Composite Structures, vol. 83, No. 2, pp.189-200.
Malekzadeh P. and Foiuz A.R. 2007. Large deformation analysis of orthotropic skew plates with nonlinear rotationally restrained
edges using DQM. Composite Structures, vol.80, No.2, pp.196-206.
Malekzadeh P., Karami G. 2006. Differential quadrature nonlinear analysis of skew composite plates based on FSDT,
Engineering Structures, vol. 28, No. 9, pp.1307-1318.
Muhammad T., Singh A. V. 2004. A p-type solution for the bending of rectangular, circular, elliptic and skew plates,
International Journal of Solids and Structures, vol. 41, No. 15, pp. 3977-3997.
Prakash T., Singha M.K., Ganapathi M. 2008. Thermal postbuckling analysis of FGM skew plates, Engineering Structures, vol.
30, No. 1, pp.22-32.
Rajamohan C. and Raamachandran J. 1997. Boundary element analysis of skew plates using a new fundamental solution.
Mechanics Research Communications, vol.24, No.4, pp.407-414.
Saadatpour M.M., Azhari M. and Bradford M.A. 1998. Buckling of arbitrary quadrilateral plates with intermediate supports
using the Galerkin method. Computer Methods in Applied Mechanics and Engineering, vol.164, pp.297-306.
Singh A.V. and Elaghabash Y. 2003. On finite displacement analysis of quadrangular plates. International Journal of Non
Linear Mechanics, vol.38, pp.1149-1162.
Singha M. K., Ramachandra L.S., Bandyopadhyay J.N. 2006. Vibration behavior of thermally stressed composite skew plate,
Journal of Sound and Vibration, vol. 296, No. 4-5, pp.1093-1102.
Srinivasan R.S. and Ramachandran S.V. 1976. Large deflection of clamped skew plates. Computer Methods in Applied
Mechanics and Engineering, vol.7, No.2, pp.219-233.
Timoshenko S.P. and Woinowsky K.S. 1959. Theory of plates and shells. Singapore: McGraw Hill International Editions.
Wang X., Tan M. and Zhou Y. 2003. Buckling analyses of anisotropic plates and isotropic skew plates by the new version
differential quadrature method.Thin-Walled Structures, vol.41, pp.15-19.
Wu G.Y., Shih Y.S. 2006. Analysis of dynamic instability for arbitrarily laminated skew plates, Journal of Sound and Vibration,
vol. 292, No. 1-2, pp.315-340.
Xiang Y., Wang C.M. and Kitipornchai S. 1995. Buckling of skew Mindlin plates subjected to in-plane shear loadings.
International Journal of Mechanical Sciences, vol.37, No.10, pp.1089-1101.
Biographical notes
Debabrata Das graduated from the Department of Mechanical Engineering, R. E. C., Durgapur and got his Master in Mechanical Engineering
degree from Jadavpur University. He has served the national Defence Research and Development Organization for five years as a scientist.
Thereafter he started pursuing research in the area of large amplitude free and forced vibration analysis of various structural elements in elastic
and post-elastic regime. He has completed his Ph. D. thesis from Jadavpur University. He has 11 publications in international journals and
conference proceedings. At present he serves as Senior Lecturer in M. S. Institute of Technology, Kolkata, India.
Prasanta Sahoo is a Professor at the Department of Mechanical Engineering, Jadavpur University, Kolkata, India. He received Bachelor of
Mechanical Engineering from Jadavpur University and MTech and PhD in Mechanical Engineering from Indian Institute of Technology,
Kharagpur, India. His research area is tribology in manufacturing, contact mechanics and structural mechanics. He is the recipient of First Track
Project for Young Scientists from Department of Science and Technology, India. He is the supervisor of 5 PhD theses and 15 Masters theses. He
has authored a textbook on Engineering Tribology (Prentice Hall of India, 2005) and a book-chapter in Tribology Research Advances (Nova
Science, 2009). He has co-authored more than 120 technical papers in international journals and conference proceedings. He is on the editorial
board of Journal of Tribology Research (Serials), Journal of Tribology and Surface Engineering (Nova Science), Materials and Design (Elsevier).
Kashi Nath Saha graduated from the Department of Mechanical Engineering, Jadavpur University and then served Indian industry for five years
as design engineer. Joined academia as lecturer in the Department of Mechanical Engineering, Jadavpur University in 1987 and pursuing
teaching and research in the area of Machine Elements Design and Computer aided drafting, Stress, deformation analysis and design of Structural
Das et al./ International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 16-32
32
Elements and Structural Stability Analysis. He has also carried out several experimental studies in the field of structural dynamics and supervised
a number of post graduate thesis and research projects and has over 50 publications in international journals and conference proceedings. At
present he serves as Associate Professor in the Department of Mechanical Engineering, Jadavpur University.
Received May 2009
Accepted July 2009
Final acceptance in revised form August 2009
MultiCraft
International Journal of Engineering, Science and Technology
Vol. 1, No. 1, 2009, pp. 33-42
INTERNATIONAL
JOURNAL OF
ENGINEERING,
SCIENCE AND
TECHNOLOGY
www.ijest-ng.com
2009 MultiCraft Limited. All rights reserved
Evaluation of the information servicing in a distributed learning
environment by using monitoring and stochastic modeling
R. P. Romansky
1
, E. I. Parvanova
2
1
Computer Systems Department, Technical University of Sofia, Bulgaria
2
College of Energetic and Electronics at Technical University of Sofia, Bulgaria
Abstract
The distributed learning is an instructional model that allows instructor, students, and content to be located in different nodes
in the global network. The authors main idea is to organize a distributed learning environment (DLE) based on information and
communication resources of global network in combination with the technologies for virtual reality and 3D simulation. In this
reason a conceptual model of the DLE architecture and learning processes is defined, and preliminary investigation is carried
out. The purpose of this paper is to present an evaluation of information servicing in DLE by using stochastic model designed on
the base of the results from program monitoring. A formalization of the DLE processes is made and a stochastic model based on
the apparatus of Markovs chains and queuing theory is designed. The results obtained from program monitoring are used for
determination of the values of the basic model parameters to secure high model adequacy. The program monitoring is realized
by using two program applications and a preliminary statistical analysis is made. Finally, an evaluation of the information
servicing in DLE is realized and the obtained assessment are presented in suitable manner.
Keywords: Distributed learning, information servicing evaluation, Markovs modeling, monitoring, statistical analysis
1. Introduction
The distributed learning (DL) is an instructional model that allows instructor, students, and content to be located in different,
noncentralized locations so that instruction and learning occur independent of time and place (Oblinger, 1996; Bowman, 1999;
Dede, 2004). This requires building a distributed learning environment (DLE) based on information and communication resources
of global network and web-space, including wireless networks (Predd et al., 2006). The development of the technologies for
virtual reality and 3D simulation (Funkhouser, 2005; Bierbaum, et al., 2008) create a possibility to organize a virtual environment
for e-learning based on computer generated 3D models.
The main idea of the authors of this article is to combine the technological principles of the web-environment with the
possibilities for e-learning based on the virtual reality and 3D learning objects. The main goal is to organize DLE and a conceptual
model (Romansky and Parvanova, 2008) and for this purpose principles of virtual environment (Romansky and Parvanova, 2009a)
are defined. A preliminary investigation of the distributed learning processes is needed to secure effective organization of the DLE
architecture. In this reason we carry out different experiments by using modeling based on deterministic (Romansky and
Parvanova, 2009b) and stochastic tools.
The purpose of this paper is to present an evaluation of information servicing in DLE by using stochastic model designed on the
base of the results from program monitoring. The apparatus of Markovs chains (Deshpande and Karypis, 2004) is selected to
design the model with the combination of the main principles of queuing theory. The results obtained from program monitoring are
used for determination of the values of the basic model parameters in order to secure high model adequacy.
The goal of the paper is to present basic steps of the evaluation organization and summary of the obtained analytical assessments.
In this reason a formalization of the environment is given and a Markovs model is designed. The program monitoring is realized
by using two program applications and a preliminary statistical analysis is made. Finally, an evaluation of the information
servicing in DLE is realized and obtained assessment are presented in suitable manner.
Romansky and Parvanova/ International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 33-42 34
2. Formalization and Abstract Description
The conceptual model for organization of DLE is proposed in (Romansky and Parvanova, 2008) and consists of some basic
components students, teachers, clients, information and education resources, communication tools, etc. The DLE is an interactive
environment for knowledge presentation and servicing based on requests sent by users. The access to the learning resources could
be realized from different remote nodes by communication resources of Internet. The main components defined in this conceptual
model are users, information learning resources and communication medium. All these components could be described as discrete
independent objects with their own internal structure and functionality. This formulation permits to build an abstract model of
DLE as an ordered discrete structure DLE = {U, R, T, D} (Figure 1) based on the following formalization:
a) abstract object:
9 Set of independent users U = {U
i
/ i=1N}, U;
9 Set of distributed learning resources in different nodes of the DLE R = {R
j
/ j=1M}, R;
9 Set of network communication tools (transmitters) T = {T
q
/ q=1K}, T;
9 Distributor (D) that routs all information objects in the communication medium by using information in the packets.
a) interactions between defined abstract objects:
9 requesting a learning resource initialized by user req:U
i
R
j
(for U
i
U & R
j
R);
9 responding by sending the requested learning object (information block) Inf:R
j
U
i
(for U
i
U & R
j
R).
U
i
U
R
j
R
Users
} 1 / { N i U U
i
= =
U
Resources
} 1 / { M j R R
j
= =
R
req
Distributed
Education System
D
Distributor
Transmitters
} 1 / { K q T T
q
= =
T
Inf
T
q
Distributed
Users
Communication
Medium
Figure 1. Abstract model of education process in the DLE
The information servicing in the DLE has a stochastic nature. This permits to use a stochastic apparatus for modeling of the DLE
components and interaction. In this reason, the apparatus of Markovs chains is suitable for the description of system behavior as a
sequence of discrete states <S(0), S(1), ..., S(k),...> for each sequence of times t
1
<t
2
<...<t
k
<t
k+1
<... . For model designing is
necessary to define a finite set of discrete states {s
1
, s
2
, ..., s
n
} presenting the main events or situations in the educational process
and determining the matrix of transitive probabilities ={p
ii
:[S(t
k
)=s
i
][S(t
k+1
)=s
j
]} and vector of initial probabilities P
0
={p
1
(0),
p
2
(0), ..., p
n
(0)}. The stochastic analytical model of the servicing in the DLE could be defined by using the following formulas:
9 formula for composite probability:
=
=
n
i
ij i j
n j p k p k p
1
,..., 2 , 1 ; ). 1 ( = ) (
;
9 condition for normalization:
=
= =
n
j
ij
n i k p
1
,..., 2 , 1 ; 1 ) (
.
3. Stochastic modeling by using Markovs chain
3.1. Model defining
Two basic situations could be realized in the distributed learning environment (Figure 2): (a) remote access of multiple users to
selected information-learning resource situated in node of the global network; (b) remote access of only one user to multiple
information-learning resources situated in different nodes of the global network.
Romansky and Parvanova/ International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 33-42 35
U1
U2
UN
D&T
Rj
a) multi-user remote access to selected distributed resource
Ui
D&T
Rj
RM
R1
b) one-user remote access to multiple distributed resources
Figure 2. Abstract presentation of the remote access to the learning resources
To describe the stochastic nature of the information servicing two basic intensities are defined the intensity of the input flow of
requests () and intensity of the flow of servicing in the node of the requested resource (). It is possible each learning resource R
j
in the DLE to be requested by different number of users and this number could be from 0 (minimum) to N (maximum). Each new
user will generate request with the intensity . After the full servicing of each user request by the resource (by the server in the
node) the Markovs process will be returned to the previous state by the intensity . In this reason the stochastic nature of the
information servicing in the DLE could be described by Markovs model shown in the Figure 3.
S
0
S
1
S
2
S
N
Figure 3. Markovs chain model of the multi-user access in DLE
The stochastic investigation is connected to the determination of assessments for the parameters for the steady-state regime
assessments of the final probabilities for all defined states in the Markovs model. It is known that to ensure the steady-state
regime it is necessary to by realized the condition = / < 1, i.e. < . The general Markovs model is presented below:
+
=
=
+ = +
=
N
i
i
N N
i i i
p
p p
N i for p p p
p p
0
1
1 1
1 0
1
. .
1 1 ; . . ). (
. .
(1)
After some transformations of the (1) we define the next form of the Markovs model:
.
1 ) ... 1 (
1 . ... . 2 . 1
. . . .
1 1 ; . . ). 1 (
. . . . . ). (
. .
2
0
0 0 0
1
0
0
0 1 1
0 1 1
1 1 1 1
0 1 0 1
= + + + +
= + + + + =
= = =
= + = +
+ =
+
+ = +
= =
=
+
+ +
N
N
N
i
i
N
N N N N
i
i i i i
i i i i i i
p
p p p p p
p p p p p
N i for p p p p p
p p p p p p
p p p p
(2)
The theory of probabilities gives some assessments for the main stochastic parameters if N that could be interpreted in our
case:
9 Expectation of the number of the active users E[N] = N
av
= /(1-) = /(-);
9 Variance of the number of the active users V[N] = N
av
+ N
av
2
;
Romansky and Parvanova/ International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 33-42 36
9 Expectation of the number of the users waiting servicing E[W] = W
av
=
2
/(1-);
9 Expectation of the waiting time for servicing E[T
W
] = N/ = /[ .(1-)].
The stochastic analysis of the information servicing in the learning environment is carried out in three basic directions:
9 Analysis of the influence of the number of active users;
9 Analysis of the users access to the resources;
9 Analysis of the influence of the learning contents size.
3.2. Preliminary model solution
The preliminary solution of the Markovs model is made for its verification and to define the model sensitivity. In this reason
concrete values for the basic model parameters are accepted: N = 3; = 0.3 (the average time between users requests is 3s); =
0.6 (the average time for a request servicing is 1.5s) = 1/2 < 1. The Markovs model is the following:
=
=
+ = +
+ = +
=
3
0
3 1 2
2 0 1
1 0
1
. 6 , 0 . 3 , 0 ). 6 , 0 3 , 0 (
. 6 , 0 . 3 , 0 ). 6 , 0 3 , 0 (
. 6 , 0 . 3 , 0
i
i
p
p p p
p p p
p p
1
. 5 , 0
. 5 , 0
. 5 , 0
3
0
0
3
3
0
2
2
0 1
=
=
=
=
= i
i
p
p p
p p
p p
(3)
The analytical solution of the model gives the assessments for the final probabilities presented in the Figure 4. Each value for the
p
i
(i=0,1,2,3) presents the probability of the number of active users in the steady-state regime. The measures of the stochastic
parameters values are: number for E[N] and V[N]; second for E[W] and E[Tw].
0,53
0,27
0,13
0,07
0
0,1
0,2
0,3
0,4
0,5
0,6
p0 p1 p2 p3
Probabilities for the active users
V
a
l
u
e
o
f
t
h
e
p
r
o
b
a
b
i
l
i
t
y
a) assessments of the final probabilities
1
2
0,5
1,67
0
0,5
1
1,5
2
2,5
E[N] V[N] E[W] E[Tw]
stochastic assessments
V
a
l
u
e
b) assessments of the stochastic parameters
Figure 4. Results obtained by the model solution
3.3. Organization of the stochastic evaluation
To organize the adequate analysis of the information servicing in the distributed learning environment based on the defined
stochastic model is necessary to determine the adequate values for the main model parameters intensities of the input requests
flow and of the servicing in the resources nodes. One of the suitable manners for determining of these parameters is by using
program monitoring in the real distributed medium. In this reason, an organization of the monitoring experiments and preliminary
statistical analysis of the measured data are presented in the next section.
4. Monitoring and preliminary statistical analysis
4.1. Monitoring organization
A program monitoring of the network traffic parameters is organized to collect real data for information processes and
distributed servicing (Romansky, 2006). The measurement is realized in two directions to collect details for the frames passed
via network medium and to determine statistical data about communication and servicing parameters. An adequate experimental
plan should be defined to obtain correct assessments for the probability distribution of the investigated parameters (in particular,
for the intensities of the input request flow and of the servicing processes).
Romansky and Parvanova/ International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 33-42 37
The monitoring of network traffic parameters and characteristics of the distributed information servicing could be made by using
different program tools for measurement, registration and analysis, for example: Webserver Stress Tool (HTTP client-server tool);
Iris (measurement and control of network traffic); Distinct Network Monitor (packets catching and analysis of the network
protocols); Microsoft Performance Monitor (standard instrument of the Windows NT system); Microsoft Network Monitor (typical
network analyzer in the basic version of Windows NT); LANAlyzer (instrument of Novell for long time observation of the network
traffic), etc.
4.2. Program application for statistical analysis
Two independent program applications are designed for interpretation of monitored data and to determine basic statistical
assessments.
9 NetMonitor is an application for data interpretation of real network traffic registered by using the standard monitoring
program Distinct Network Monitor. The application is developed by using Visual C++ 6.0 and MFC library of classes. It gives
information about packet length, time, source and destination of the packets, protocol type, etc. and presents graphical
visualization of the main statistical assessments (Figure 5).
Figure 5. Interpretation of monitored data by NetMonitor
9 Network Analyzer is an application realized by using of C# in the Microsoft Visual Studio.NET environment. It analyses
empirical data from monitoring and calculates basic statistical assessments (average value, correlation, variation, regression). The
application permits to determine the functional dependency between two parameters (variables) as a regression line based on the
registered empirical data
{ }
) ( q
i
x
for q=1, , N (Figure 6).
Romansky and Parvanova/ International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 33-42 38
Figure 6. Initial forms of the Network Analyzer
4.3. Preliminary statistical analysis
Some of the basic statistical parameters are presented below:
number of active users (users);
bandwidth for server (server bandwidth) and user access (user bandwidth);
average value of the time for information object selection (average click time);
waiting time for the selection activation (click time);
time for URL interpretation by using DNS server of the client machine (time for DNS);
time to connect with the server of the selected object (time to connect);
average time for request servicing (average request time);
time for the first byte receiving of the response from the server (time to first byte TFB);
average packet length;
The statistical analysis that is carried out using Network Analyzer is based on the formulas presented in table 1.
Table 1. Basic statistical assessments
Parameter Analytical formula
Expectation for x
i
(average value)
i
N
q
av i
q
i
x x x
N
EX = = =
=1
_
) (
1
Variance for x
i
[ ]
2
1
2
) (
1
i
N
q
i
q
i
x x
N
DX = =
=
Mean square deviation
DX + =
Coefficient of variance
av
V
x
C
=
Covariation between x
i
and x
j
[ ][ ]
=
=
N
q
j
q
j i
q
i ij
x x x x
N
1
) ( ) (
.
1
Coefficient of correlation
j i
ij
ij
r
.
=
Model of simple line regression
(b
0
displacement; b
1
coefficient of regression)
; .
1 0 i j
x b b x + =
i j
i
ij
x b x b b . ;
1 0
2
1
= =
The experiments simulate access to www-object and they are carried out for values of the number of active users from 1 to 10.
This permits to analyze the users number influence on the other statistical parameters. Some of the statistical assessments for the
functional dependency between parameters, calculated by using Network Analyzer, are presented in Figure 7.
Romansky and Parvanova/ International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 33-42 39
server bandwidth = f(users) click time = f(users) time to connect = f(users)
Figure 7. Assessment obtained by regression analysis
Statistical analysis of the measured data that was carried out using Network Analyzer permits us to determine basic stochastic
parameters for investigation of the designed Markovs model. The intensity of the input flow of requests () could be determined
on the base of the calculated average values for the monitored parameters average click time (4,83 s), click time (4,88 s), time to
DNS (0,118 s) and time to connect (0,118 s). These values permit to determine = 0,1 s
-1
. In the same way, the intensity of the
flow of servicing in the node of the requested resource () could be connected to the parameter average request time (4,83 s)
which determines value 0,2 s
-1
. These values for the intensities define = / = 0,5 s
-1
. All assessments are summarized in
table 2.
Table 2. Assessments obtained during the preliminary statistical analysis [s]
users average click time click time time to DNS time to connect average request time
2 1 0,57 0,024 0,028 1.0
3 3.5 3,6 0,076 0,047 3.5
5 4.0 4 0,090 0,100 4.0
6 3.5 3,8 0,120 0,132 3.5
8 6.0 6,3 0,190 0,190 6.0
10 11.0 11 0,210 0,210 11.0
Average: 4,83 s 4,88 s 0,118 s 0,118 s
Total time T
R
= 9,95 s = 0,1 s
-1
T
S
= 4.83 s
= 0,207 s
-1
5. Evaluation of the stochastic parameters of the information servicing in DLE
5.1. Analysis of the influence of the number of active users ( = const; N = var)
The investigation is carried out for constant value = 0,5 and different number of the active users from 1 to 10. The goal is to
evaluate the influence of the parameter N on the stochastic assessments of the final probabilities for steady-state regime. The
calculations are realized on the base of the main formula
0
. p p
i
i
=
(for i = 1, 2,, N) see (2). Examples for some of the
calculations are presented below and all obtained assessments are summarized in table 3.
N = 1: p
1
= .p
0
p
0
(1 + ) = 1 p
0
= 1/(1+) = 0,67 p
1
= 0,33.
N = 2: p
1
= .p
0
and p
2
=
2
.p
0
p
0
(1 + +
2
) = 1 p
0
= 1/(1,75) = 0,57 p
1
= 0,29 and p
2
= 0,14.
N = 3: p
0
(1 + +
2
+
3
) = 1 p
0
= 1/(1,875) = 0,53 p
1
= 0,27, p
2
= 0,13 and p
2
= 0,07.
. . .
N = 10: p
0
(1 + +
2
+
3
++
10
) = 1 p
0
= 1/(1,999225) = 0,50 p
1
= 0,25, p
2
= 0,125, etc.
Romansky and Parvanova/ International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 33-42 40
Table 3. Evaluation of the final probabilities for steady-state regime
N p
0
p
1
p
2
p
3
p
4
p
5
p
6
p
7
p
8
p
9
p
10
1 0,67 0,33
2 0,57 0,29 0,14
3 0,53 0,27 0,13 0,07
4 0,52 0,26 0,13 0,06 0,03
5 0,51 0,25 0,13 0,06 0,03 0,02
6 0,50 0,25 0,13 0,06 0,03 0,02 0,01
7 0,50 0,25 0,126 0,063 0,031 0,016 0,008 0,004
8 0,50 0,25 0,125 0,0626 0,031 0,0157 0,0078 0,0039 0,00196
9 0,50 0,25 0,125 0,0626 0,031 0,0157 0,0078 0,0039 0,00195 0,00098
10 0,50 0,25 0,125 0,0625 0,031 0,0156 0,0078 0,0039 0,00195 0,00098 0,00049
average 0,53 0,265 0,128 0,0626 0,03 0,0172 0,0083 0,0039 0,00195 0,00098 0,00049
The average values of the determined final probabilities for steady-state regime and graphical interpretation of the functions
p
i
(N) for i = 1,2,3,4 are shown in Figure 8. The results show that for N<4 the number of active users has a little influence on the
final probabilities and for N5 the probability for multi-user access to the informational resources is practically equal to zero.
0
0,1
0,2
0,3
0,4
0,5
0,6
P0 P1 P3 P4 P5 P6 P7 P8 P9 P10
Average values of the final probabilities
0
0,1
0,2
0,3
0,4
0,5
0,6
0,7
0,8
1 2 3 4 5 6 7 8 9 10
number of active users
f
i
n
a
l
p
r
o
b
a
b
i
l
i
t
i
e
s
p0 p1 p2 p3
functions p
i
(N) for i = 1,2,3,4
Figure 8. Interpretation of the obtained assessments
5.2. Analysis of the users access to the resources ( = var; N =const; = const)
The experiments are carried out for different values of the intensity (average time between two requests from 5,55 s to 100 s)
and the intensity of servicing is fixed as average value 0,2 s
-1
. All assessments obtained by the analytical analysis are
summarized in table 4 and some of them are graphically interpreted in Figure 9.
Table 4. Assessments from the analytical evaluation for = var
p
0
p
1
p
2
p
3
E[N] V[N] E[W] E[Tw]
0,18 0,9 0,29 0,26 0,24 0,21 9 90 8,1 45
0,16 0,8 0,34 0,27 0,22 0,17 4 20 3,2 20
0,14 0,7 0,39 0,28 0,19 0,14 2,33 5,76 1,67 11,67
0,12 0,6 0,46 0,27 0,17 0,10 1,5 3,75 0,9 2,5
0,1 0,5 0,53 0,27 0,13 0,07 1 2 0,5 5
0,08 0,4 0,61 0,25 0,10 0,04 0,67 1,12 0,27 3,33
0,06 0,3 0,71 0,21 0,06 0,02 0,43 0,61 0,13 2,14
0,04 0,2 0,8 0,16 0,03 0,01 0,25 0,31 0,05 1,25
0,01
0,2
0,1 0,9 0,09 0,009 0,0009 0,11 0,12 0,01 0,56
Romansky and Parvanova/ International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 33-42 41
0,00
0,10
0,20
0,30
0,40
0,50
0,60
0,70
0,80
0,90
1,00
0,01 0,04 0,06 0,08 0,10 0,12 0,14 0,16 0,18
Intensity of the request flow
P
r
o
b
a
b
i
l
i
t
y
p0 p1 p2 p3
Figure 9. Graphical interpretation of the final probabilities and expectations for users
0
1
2
3
4
5
6
7
8
9
10
1 2 3 4 5 6 7 8 9
Intensity x 0,1
E[N] E[W]
5.3. Analysis of the influence of the learning contents size ( = var; N =const; = const)
The experiments are carried out for different values of the intensity of servicing (the average time of requests servicing
depends on the size of the learning content in the separate nodes of DLE and range from 2 s to 8 s) and the intensity of input flow
of requests is fixed as average value 0,1 s
-1
. The main assessments obtained by the analytical analysis are summarized in table 5
and some of them are graphically interpreted in Figure 10.
Table 5. Assessments from the analytical evaluation for = var
p
0
p
1
p
2
p
3
E[N] V[N] E[W] E[Tw]
0,125 0,8 0,34 0,27 0,22 0,17 4 20 3,2 32
0,15 0,67 0,41 0,28 0,19 0,12 2 6 1,36 13,54
0,175 0,57 0,48 0,27 0,16 0,09 1,32 4,08 0,76 7,57
0,2 0,5 0,53 0,27 0,13 0,07 1 2 0,5 5
0,25 0,4 0,61 0,25 0,10 0,04 0,67 1,12 0,27 2,67
0,30 0,33 0,68 0,22 0,07 0,03 0,49 0,73 0,16 1,64
0,35 0,286 0,72 0,20 0,06 0,02 0,4 0,56 0,12 1,14
0,40 0,25 0,75 0,19 0,05 0,01 0,33 0,44 0,08 0,67
0,1
0,5 0,2 0,8 0,16 0,03 0,01 0,25 0,31 0,05 0,5
0
0,1
0,2
0,3
0,4
0,5
0,6
0,7
0,8
0,9
0,125 0,15 0,175 0,2 0,25 0,3 0,35 0,4 0,5
Intensity of servicing
P
r
o
b
a
b
i
l
i
t
y
p0 p1 p2 p3
0
0,5
1
1,5
2
2,5
3
3,5
4
4,5
0,125 0,15 0,175 0,2 0,25 0,3 0,35 0,4 0,5
Intensity of servicing
P
r
o
b
a
b
i
l
i
t
y
E[N] E[W]
Figure 10. Graphical interpretation of the final probabilities and expectations E[N] and E[W]
6. Conclusions
The following conclusions could be made from the study conducted:
1. The processes in the DLE have a stochastic nature that determines the usefulness of a stochastic apparatus like the Markovs
chain. In this case the theory of Markovs stochastic processes is combined with the principles of queuing theory for servicing of
an input flow of requests. This combination permits to use the model from Figure 3 defined by (1) and (2).
Romansky and Parvanova/ International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 33-42 42
2. The evaluation in this paper is different from the investigations presented in (Romansky and Parvanova, 2009b), and permits
to compare the main results of servicing and to define the specific parameters for DLE organization in a real medium. To define
the frame of the current investigation a preliminary model solution is made and program monitoring is realized.
3. The statistical analysis of the monitored data in part 4.3 permits to determine the functional dependency of the main
parameters by the number of active users and to define the tendencies for regression and variation needed for the next stochastic
evaluation.
4. The stochastic evaluation shows that the number of active users has a little influence on the variation of the other parameters
of the education servicing in DLE. This fact permits to use a fixed value for N during the evaluation of the influence of input
requests and servicing intensities.
5. The assessments in parts 5.2 and 5.3 permit to determine the range of variation for some important stochastic parameters like
the situation of servicing in steady-state regime, expectation of active users E[N], expectation of waiting users E[W], etc. These
analytical assessments define the optimal frame of servicing in DLE determined by ranges = 0,1s
-1
0,02s
-1
and = 0,2s
-1
0,05s
-1
.
References
Bierbaum, A., Just C., Harting P., Meinert K., Baker A., Crus-Neira C., 2008. VR Juggler: a virtual platform for virtual reality
application development. Intl Conf. on Computer Graphics and Interactive Techniques (ACM SIGGRAPH ASIA 2008),
Singapore, Article 42.
Bowman, M., 1999. What is Distributed Learning. News & Advice from the Technology Collaborative, Vol. II, No. 1, pp.1-5.
Dede, C., 2004. Distributed-learning communications as a model for education teachers. Proc. of Society for Information
Technology and Teacher Education Intl Conf., Chesapeake, VA, pp.3-12.
Deshpande, M., Karypis G., 2004. Selective Markov models for predicting Web page access. ACM Trans. on Internet Technology,
Vol. 4, No 2, pp. 163-184.
Funkhouser, T., 2005. Shape-based retrieval and analysis of 3D models. Communication of the ACM, Vol.48, No. 6, pp.58-64.
Oblinger, D.G., Maruyama, M.U., 1996. Distributed learning. CAUSE Professional Paper Series, No 14, pp. 1-27.
Predd, J. B., Kulkarni S.B., Poor H.V. 2006. Distributed Learning in Wireless Sensor Networks. IEEE Signal Processing, Vol.3, 4,
pp.56-69.
Romansky, R., 2006. An organization of program monitoring and distributed servicing evaluation. Proceedings of the 7
th
International Symposium on Intelligent Systems (INTELS2006), Russia, 2630 June 2006, pp. 35-40.
Romansky, R., Parvanova, E., 2008. Conceptual model of distributed architecture for 3D simulation learning. Proceedings of the
International Scientific Conference UNITECH08, Gabrovo, Bulgaria, Vol. III, pp. 287-292 (in Bularian).
Romansky, R., Parvanova, E., 2009a. Generation of virtual learning environment on the internet using VRML. Proceedings of the
23
rd
Intl Conf. SAER-2009 (in the frame of Intl Conf. on Information Technologies), St. Konstantine and Elena, Bulgaria, pp.
224-229.
Romansky, R., Parvanova, E., 2009b. Deterministic investigation of distributed learning environment by using Petri nets model. 5
th
International Conference E-Learning and the Knowledge Society. Berlin, Germany, 31 August 1 September 2009
(accepted for publishing).
Biographical notes
Assoc. Prof. Dr. Radi P. Romansky is working in the Computer Systems Department of the Technical University Sofia, (Bulgaria). He has
engaged in teaching and research activities since the last 28 years. His fields of scientific interest are Information and Network Technologies,
Computer Modeling, Computer Architectures, Data Protection, Applied Informatics, etc. He has over 15 published books and over 150
publications in various national, international conferences and journals.
Assist. Prof. Elena I. Parvanova is working in the College of the Energetic and Electronics at the Technical University of Sofia (Bulgaria). She
is PhD student in the field of the distributed learning and 3D Simulation. The scientific fields are 3D Simulation, Virtual Reality, Multimedia,
Information Technologies, etc.
Received: August 2009
Accepted: September 2009
Final acceptance in revised form: September 2009
INTERNATIONAL
JOURNAL OF
ENGINEERING,
SCIENCE AND
TECHNOLOGY
International Journal of Engineering, Science and Technology
MultiCraft
Vol. 1, No. 1, 2009, pp. 43-60
www.ijest-ng.com
2009 MultiCraft Limited. All rights reserved
A computational approach to the design of a cryogenic turbine
blade profile
Subrata Kr. Ghosh
1
*, R. K. Sahoo
2
, Sunil K. Sarangi
2
1
* Indian School of Mines University, Dhanbad Jharkhand, India,
2
National Institute of Technology, Rourkela, Orissa, India
Corresponding author (e-mail: subratarec@yahoo.co.in)
*
Abstract
The present day cryogenic gas turbines are in high use as they meet the growing needs for low pressure cycles. This calls for
improved methods of blade profile design. The present study is aimed at the design of the blade profile of mixed flow impellers
with radial entry and axial discharge. In this paper, details of a computer code have been developed for designing such blade
profile. The computational procedure developed describes the three-dimensional contours of the blades for the turbine wheel.
The flow of fluid in a turbine blade passage depends on the length of the flow path and the curvature of such path. This has been
given due weightage. Effect of some of the turbine operating and design parameters on the flow path and its curvature have been
analyzed and presented. Optimum solution for some of the free parameters and angles between velocity components was found.
On that basis, co-ordinates of this blade profile were computed. Although manufacturers design the blade profiles for their
production, till date no academic literature is available for this purpose. The work may help researchers for further
developmental work on this topic as well as the manufacturers of cryogenic expanders.
Keywords: Radial turbine, Blade profile parameter, Meridional streamlength, Axial coordinate, Radial coordinate, Angular
coordinate
1. Introduction
A medium or large cryogenic system needs many components - compressors, heat exchanger, expansion turbine,
instrumentation, vacuum vessel, etc. At present, in many developing countries, most of the cryogenic system components are
sourced from industrialized nations in view of their high demand locally. This has enormous foreign exchange implications; these
spending could be preserved for local industrial growth. Thus, there is a need for countries to be self reliant in this vital
technology. In India, all components of cryogenic systems are available locally; however the whole plant is being imported due to
lack of the single component, expansion turbine. Successful development of a turboexpander strongly depends on the performance
of the turbine wheel. The complete design of turbomachinery rotors requires aerodynamic analysis of the flow path and structural
analysis of the rotor including the blades and hub (Bruce, 1998). In order to develop such a design system, the rotor geometry
generation procedure should be clearly understood. A computerized method for specifying the shape of impellers and guide vanes
has been described by different literature (Wallace et al., 1975; Whitfield, 1978; Casey, 1983; Strinning, 1995). It is necessary that
all machine components operate efficiently. This can be achieved by selecting the most suitable combination of design parameters
with the following conditions ensured the highest possible value of efficiency, uniform and steady operating conditions and an
easy manufacture of the blades. For the turbine blade design varies from 0.75 to 1 and varies from 1 to 20 (Balje, 1970).
Turbine blade surface deterioration is strongly dependent on the turbine geometry and blade surface material. A numerical model
has been developed to show the effects of aerodynamic forces in the three dimensional turbine flow field (Hamed et al., 2005).
e
K
h
K
The paper is devoted to the design of the blade profile of mixed flow impellers with radial entry and axial discharge. The
computational process aims at defining a blade profile that maximises the performance with highest possible value of efficiency,
uniform and steady operating conditions and easy manufacture of the blades. The detailed procedure describes computation of the
Ghosh et al./ International Journal of Engineering, Science and Technology, Vol. 1, No.1, 2009, pp. 43-60 44
three-dimensional contours of the blades and simultaneously determines the velocity, pressure and temperature profiles in the
turbine wheel. The magnitude of the velocity and change in its direction determine the optimum blade profile of the turbine. In a
word, in this paper a parametric study has been carried out to determine the optimum blade profile for given specifications.
2. Basic design of turbine wheel
The design of turbine wheel has been done following the method outlined by some authors (Balje, 1981; Kun and Sentz, 1985),
which are based on the well known similarity principles. The similarity laws state that for any given Reynolds number, Mach
number and Specific heat ratio of the working fluid to achieve optimized geometry for maximum efficiency, two dimensionless
parameters: specific speed and specific diameter uniquely determine the major dimensions of the wheel and its inlet and exit
velocity triangles. Specific speed ( ) and specific diameter ( ) are defined as:
s
n
s
d
( )
=
3
s
3
4
in 3s
Q
n
h
(1) Specific speed
( )
=
1
4
2 in 3s
s
3
D h
d
Q
(2) Specific diameter
The basic input parameters for the cryogenic expansion turbine system are:
Working fluid: Air
Discharge pressure: 1.5 bar
Turbine inlet temperature: 122 K
Throughput: 67.5 nm
3
/hr
Turbine inlet pressure: 6.0 bar
Expected efficiency: 75%
From Balje (1981), the peak efficiency of a radial inflow turbine corresponds to the values of:
54 . 0 =
s
n and 4 . 3 =
s
d
Substituting these values in equations (1) and (2) respectively, yields
Rotational speed rad/s 22910 =
Wheel diameter mm D 0 . 16
2
=
Power produced KW 9 . 0 P =
The ratio of exit tip diameter to inlet diameter should be limited to a maximum value of 0.70 (Dixon, 1978; Rohlik and Harold,
1968) to avoid excessive shroud curvature. Corresponding to the peak efficiency point (Balje, 1981):
676 . 0 D D
2
tip
= = mm 8 . 10 =
tip
D
In order to determine the optimum turbine geometry, various studies have been conducted and published in relation to the rotor
blade profile (Thakker and Abdulhadi, 2007). According to Rohlik and Harold (1968), the exit hub to tip diameter ratio should
maintained above a value of 0.4 to avoid excessive hub blade blockage and energy loss. Kun and Sentz (1985) have taken a hub
ratio of 0.35 citing mechanical considerations.
425 . / = =
tip hub
D D 6 . 4 =
hub
D mm
Recommendation has given on the choice of number of blades by ensuring that the flow does not stagnate the pressure surface
(Denton, 1996). For small turbines, the hub circumference at exit and diameter of milling cutters available determine the number
of blades. In this design the number of blades (Z ) are chosen to be 10, and the thickness of the blades to be 0.6 mm throughout.
tr
3. Determination of meridional geometry
In selecting the most suitable combination of parameters, the rotor contour (tip and hub streamline) and the change of the flow
angle with flow path length are determined when a pressure balanced flow path is desired. This also means that an arbitrary
Ghosh et al./ International Journal of Engineering, Science and Technology, Vol. 1, No.1, 2009, pp. 43-60 45
selection of the rotor contour and angular change with flow path length is likely to yield a design with potentially high transverse
pressure gradients.
Thus complete pressure balance in practical flow path designs will be nearly impossible. The three dimensional effects can,
however, be minimized by keeping the relative velocity gradient low, i.e., by providing a high blade number in that portion of the
flow path where the suction side and pressure side streamlines begin to diverge, up to the point where the flow path inclination
angle approaches 90.
W2
U2
C2
r2
r3
W3
C3
C
U
U3
W
CM
Detail A
W
C
CM=WM
U
2
3
Inlet of
Turbine Wheel
(State 2)
Outlet of
Turbine Wheel
(State 3)
W3
C3
U3
3
Detail B
Detail B
Detail A
t
s
Figure 1. Illustration of flows in radial axial impeller
/2-
r
/2-
(u)
r
s(Cm)
z
n
b
t(W)
Figure 2. Coordinate system
The concave and convex profiles between blades are designed to reduce the pressure difference between the pressure and suction
surfaces near the endwall, as this causes the secondary flow and the end wall boundary layer to conform as much as possible to the
mainstream flow direction. At the same time, the trailing edge profile is designed such that as much of the primary flow as possible
leaves the blade smoothly in the mainstream flow direction (Eisaku et al., 2009).
The design of blade profile is usually carried out by Hasselgrubers approach (Hasselgruber, 1958). The equations are derived in
a body fitted orthogonal coordinate system (t, b, n). The coordinate t is the direction along the central streamline, the coordinate b
is the lateral coordinate between the suction and the pressure surfaces and the coordinate n refers to depth of the flow path in the
turbine passage. The author also defines the meridional coordinate s to correlate the body fitted coordinate system with the
Ghosh et al./ International Journal of Engineering, Science and Technology, Vol. 1, No.1, 2009, pp. 43-60 46
cylindrical coordinate system (r, , z) where the s coordinate lies on the r-z plane. Figure 1 and Figure 2 show the flow velocity
and coordinate transformation respectively.
3.1 Assumptions
The blade profiles have been worked out using the technique of Hasselgruber (1958), which was also employed by others (Kun
and Sentz, 1958; Balje, 1993, 1981). Hasselgrubers approach of computation for the central streamline is based on some key
assumptions such as:
(i) Constant acceleration of the relative velocity
The blades of pure radial impellers are so shaped that it always gives a constant retardation to the relative velocity. The
acceleration of relative velocity from wheel inlet to exit follows a power law relation. The substantial derivative under steady state
condition results,
1
1
=
=
e
K
t C
t
W
W
D
DW
(3)
where refers to the time coordinate and t stands for the distance along the central streamline.
Integrating equation (1) and substituting the following boundary conditions,
W = W
3
, at t = 0, and
W = W
2
at t = t
2
,
The solution can be written as,
( )
e
K
t
t
W W W W
+ =
2
2
3
2
2
2
3
2
(4)
(ii) Pressure is constant over the blade channel in the direction normal to the mean relative streamline, indicating the hydrostatic
pressure has negligible effect. Thus,
0 =
n
p
(iii) Relative flow angle at the wheel inlet = 90.
Both the earlier conditions relate to the width of the channel and shape of the curve in meridian section. Another equation has to
be formed which determine the shape of the curve in the circumferential direction. The variation of the relative velocity angle
along the flow path follows the relation:
h
K
s
s
C
s
t
+ =
=
2
2 2
1 cosec cosec (5)
2 3 2
sec sec Co Co C = where,
(iv) Equal meridional velocity at the wheel inlet and at exit
1
2
3
= =
m
m
I
C
C
k The meridional velocity ratio:
3.2 Input and output variables
The following tables give the list of input and output variables considered in the analysis. Consistent SI units have been used in
all cases.
Ghosh et al./ International Journal of Engineering, Science and Technology, Vol. 1, No.1, 2009, pp. 43-60 47
Table 1. Input data for blade profile analysis of expansion turbine
A. Variable
Variables Notation Units
Free parameter None
e
k
Free parameter None
h
k
Characteristic angle radian
3
s Meridional streamlength M
No of points for calculation n_points None
B. Constant thermodynamic properties
Constants Notation Units
Outlet temperature K
3
T
Outlet pressure Pa
3
p
Outlet density kg/m
3
3
(6)
where,
( ) ( ) ( ) ( ) ( )
( ) ( )
e
h
k
k
+ +
=
+
mean 2 h
1
2
2 mean 2 h
2
cosec cosec k
s
s
1 1 cosec cosec cosec 1 k
s
s
A
(7)
( ) ( ) ( ) { }
h
k
+
=
2
2 mean 2
2
2
s
s
- 1 cosec cosec cosec
1
s
s
f
(8)
2
2
2
2
1
2
3
s
s
f 1
s
s
f
s
s
f (9)
The function f
1
depicts the variation of the relative acceleration of the fluid from the wheel inlet to exit. The function f
2
gives the
relative flow angle along the flow path while function f
3
is a combination of f
1
& f
2
.
The radius of curvature of meridional streamline path is expressed in terms of the three characteristic functions
( )
( )
cos
r
s
s
f
tan r
r
s
s
f
s
s
f
R
2
2
3
mean
2
2
2
1
m
=
mean
(10)
( )
hub tip mean
r r r + = 5 . 0 where,
The angle between meridional velocity component and axial coordinate is derived to be:
Ghosh et al./ International Journal of Engineering, Science and Technology, Vol. 1, No.1, 2009, pp. 43-60 49
=
s
m
ds
R
0
1
(11)
The coordinate (r, , z) of the central streamline are calculated by using the following equations:
=
s
ds r
0
) (sin (12)
=
s
0
2
2
2
2
2
ds
s
s
f r
s
s
f - 1
(13)
=
s
ds z
0
) (cos (14)
Middle stream-surface is created by joining the points on the hub streamline to the corresponding points on the tip streamline.
The coordinates of hub and tip streamlines are calculated by using channel depth and the angle between meridional velocity
component and axial coordinate.
To represent the ratio of meridian to circumferential velocity, a characteristics factor is defined as:
3
3
3
tan = =
U
C
m
m
(15)
The velocities at different points of meridional streamline are
=
2
1 3
s
s
f C W
m m
(16)
=
2
2
1
sin
s
s
f
m
(17)
mean
m
r
r U
U
=
3
(18)
= cos W U C
u
(19)
=
2
2
2
1 3
s
s
f
s
s
f C C
m m
(20)
) (
2 2
u m
C C C + = (21)
The density along the fluid flow path is
1
1
2
3
2
3
2
3
2 2
2
3 3
2
) 1 (
1
+ =
m
m C
W U W U
m
m
M
(22)
The channel width and channel depth at each point is determined by using these equations respectively as,
tr
tr tr
tr
Z
t Z r
w
=
sin 2
(23)
Ghosh et al./ International Journal of Engineering, Science and Technology, Vol. 1, No.1, 2009, pp. 43-60 50
m tr tr
tr
C w Z
m
b
sin
.
= (24)
The r, , z coordinate of the hub and tip streamlines are calculated by using the following equations:
( )
( )
cos
2
b
r r
cos
2
b
r r
mean tip
mean hub
+ =
=
(25)
( )
( ) sin
2
b
z z
sin
2
b
z z
mean tip
mean hub
=
+ =
(26)
(27)
mean hub tip
= =
The surface so generated is considered as the mean surface within a blade. The suction and pressure surfaces of two adjacent
channels are computed by translating the mean surface in the +ve and -ve directions through half the blade thickness. The
suction side and pressure side surfaces (r, and z coordinates of streamlines) are obtained through the following equations:
(28)
mean suction pressure
r r r = =
(29)
mean suction pressure
z z z = =
cos *
* 2
mean
tr
mean pressure
r
t
+ = (30)
cos *
* 2
mean
tr
mean suction
r
t
= (31)
The average velocity can be split into two parts: one, due to the curvature of the blades and other, due to the rotation of the
blades the so-called channel-vortex. For turbine wheel as the curvature is backward ( is positive), the effects are opposite.
b
R
The blade angle along the hub and tip streamlines are calculated by using the following equations:
=
tan tan
1
hub
hub
r
r
(32)
=
tan tan
1
tip
tip
r
r
(33)
The velocities of fluid at the suction and pressure side are determined from the following expression.
sin . .
. 2
1 .
,
L
R
L
w w
b
ss ps
m
= (34)
cot
1
cos
1
2
2
2
=
h
K
h
b
s
s
C
ec
K
s
R where,
1
1
2
1
1
+
= C
3 1
tan =
Ghosh et al./ International Journal of Engineering, Science and Technology, Vol. 1, No.1, 2009, pp. 43-60 51
3.4 Outline of computer programme
Inputs: major dimensions and the flow conditions at the inlet and exit of the wheel
Determination of the middle streamline
Determination of the hub and tip streamlines
Determination of the suction and pressure side geometry of the blade
Start
Input
n_points , 001 .
, , , ,
, 0 , 0 , 0 , 0
2 3 3
3 3 3 3
=
= = = =
e h mean mean
k k r
z s
=
mean
r
s
ratio s
Assume
, 3
2
_
Iteration =1
s
2
= s_ratio * r
3,mean
points _
2
n
s
ds =
3
, 3
3
3
3
0 _
=
=
=
=
=
=
mean
r r
z z
s s
count r
A
B
Ghosh et al./ International Journal of Engineering, Science and Technology, Vol. 1, No.1, 2009, pp. 43-60 52
Compute f1, f2, f3, & R
m
Using equations (6) (10)
m
R ds d
ds dz
ds dr
/
cos
sin
=
=
=
1 _ _ + =
+ =
+ =
+ =
+ =
count r count r
ds s s
d
dz z z
dr r r
Print s, r, z,
Is
r_count
n_points
Is
2
2
r
r r
Iteration = Iteration+1
s_ratio = s_ratio * r
2
/r
A
C
No
Yes
B
No
Yes
Ghosh et al./ International Journal of Engineering, Science and Technology, Vol. 1, No.1, 2009, pp. 43-60 53
Calculate at each point by numerical
Integration.
Compute channel depth at each point by
using equations (24)
Compute z ; ; r at each point for hub and
tip by using equations (25 27)
C
Compute
suction press suction press
z ; z ; r ; r ;
suction press
; for the
mean streamline, using equations (28) (31)
Print results
Stop
Compute -the flow angle for hub and tip
streamline, using equations (32) (33)
Compute
suction press
w ; w for the mean
streamline, using equations (34)
Figure 3. Flow chart of the computer program for calculation of blade profile using Hasselgrubers method
4. Results and discussion
There are two free parameters for calculating the flow path contour and . The parameter controls the change in flow
path angle whereas the term is the acceleration exponent which controls the relative acceleration. Effect of these two
parameters on blade profile discussed below.
h
k
e
k
h
k
e
k
Ghosh et al./ International Journal of Engineering, Science and Technology, Vol. 1, No.1, 2009, pp. 43-60 54
Variation of
h
k
As may vary from 1 to 20, the greater the , greater is the curvature of the blade at the wheel exit, shown in Figure 4. More
compact is the construction, the shorter is the path of the stream. So, the distribution of the velocity at the wheel exit is non-
uniform and this means a strong friction. Again the probability of dumb-bell shaped blade is more in case of large value. The
net effect is the worsening of the efficiency. The axial co-ordinate does not change too much with changing of as shown in
Figure 7. But the angular co-ordinate of blade will be more in case of small value as shown in Figure 10, which is not
permitted. The characteristic angle (
h
k
h
k
h
k
h
k
h
k
3
, Figure 13) is more for large value i.e. to achieve the value
h
k
2
as 90, it is required to
choose more . To calculate the flow angle from exit to inlet it has been seen from Figure 16 that, for small , it is increasing at
a particular slope initially but at close to the inlet, it changes abruptly. Again for the large value of the flow angle increasing to
the particular value and then it remains unchage.
h
k
h
k
h
k
Variation of
e
k
The radial co-ordinate, axial co-ordinate and the angular co-ordinate do not change significantly with changing of as shown
in Figure 5, Figure 8 and Figure 11. As shown in Figure 14 the characteristic angle
e
k
2
is more in case of small and it
decreases for large . The flow angle is totally independent of is shown in Figure 17. For >1 the relative acceleration i.e.
the rate of change of relative velocity along the meridian streamlength from inlet to exit is more and effect is less for <1.
e
k
e
k
e
k
e
k
e
k
3
Variation of
3
By allowing a smaller at the exit of the wheel, the curvature of the blade is smaller at the wheel exit as shown in Figure 6 and
the chances of dumb-bell shaped can be eliminated. The axial co-ordinate and the angular co-ordinate are also less in case of some
value of
3
as shown in Figure 9 and Figure 12. But the characteristics angle at the inlet of the turbine wheel will be more for
more
3
as shown in Figure 15. The flow angle is almost independent on the initial value of characteristic angle as shown in
Figure 18.
ke=.75 and 3=9.0 deg
0
2
4
6
8
10
0 5 10 15
Meridional Streamlength (mm)
R
a
d
i
a
l
C
o
-
o
r
d
i
n
a
t
e
(
m
m
)
kh=1.0
kh=3.0
kh=5.0
kh=8.0
kh=10.0
kh=12.0
Figure 4. Variation of radial co-ordinate of turbine wheel with the variation of
h
k
kh=8.0 and 3=9.0 deg
0
2
4
6
8
10
0 2 4 6 8 10
Meridional Streamlength (mm)
R
a
d
i
a
l
C
o
-
o
r
d
i
n
a
t
e
(
m
m
)
ke=.25
ke=.50
ke=.75
ke=1.0
ke=1.25
Figure 5. Variation of radial co-ordinate of turbine wheel with the variation of
e
k
Ghosh et al./ International Journal of Engineering, Science and Technology, Vol. 1, No.1, 2009, pp. 43-60 55
kh=8.0 and ke=.75
0
2
4
6
8
10
0 5 10 15
Meridional Streamlength (mm)
R
a
d
i
a
l
C
o
-
o
r
d
i
n
a
t
e
(
m
m
)
delta3=0.0
delta3=5.0
delta3=9.0
delta3=12.0
3
Figure 6. Variation of radial co-ordinate of turbine wheel with the variation of
ke=.75 and 3=9.0 deg
0
2
4
6
8
10
12
0 5 10 15
Meridional Streamlength (mm)
A
x
i
a
l
C
o
-
o
r
d
i
n
a
t
e
(
m
m
)
kh=1.0
kh=3.0
kh=5.0
kh=8.0
kh=10.0
kh=12.0
Figure 7. Variation of axial co-ordinate of turbine wheel with the variation of
h
k
kh=8.0 and 3=9.0 deg
0
1
2
3
4
5
6
7
8
0 2 4 6 8 10
Meridional Streamlength (mm)
A
x
i
a
l
C
o
-
o
r
d
i
n
a
t
e
(
m
m
)
ke=.25
ke=.50
ke=.75
ke=1.00
ke=1.25
Figure 8. Variation of axial co-ordinate of turbine wheel with the variation of
e
k
k
h
=8.0 and k
e
=.75
0
2
4
6
8
10
0 5 10 15
Meridional Streamlength (mm)
A
x
i
a
l
C
o
-
o
r
d
i
n
a
t
e
(
m
m
)
delta3=0.0
delta3=5.0
delta3=9.0
delta3=12.0
3
Figure 9. Variation of axial co-ordinate of turbine wheel with the variation of
Ghosh et al./ International Journal of Engineering, Science and Technology, Vol. 1, No.1, 2009, pp. 43-60 56
ke=.75 and 3=9.0 deg
0
20
40
60
80
100
0 5 10 15
Meridional Streamlength (mm)
A
n
g
u
l
a
r
C
o
-
o
r
d
i
n
a
t
e
(
d
e
g
)
kh=1.0
kh=3.0
kh=5.0
kh=8.0
kh=10.0
kh=12.0
Figure 10. Variation of angular co-ordinate of turbine wheel with the variation of
h
k
kh=8.0 and 3=9.0 deg
0
5
10
15
20
25
30
0 2 4 6 8 10
Meridional Streamlength (mm)
A
n
g
u
l
a
r
C
o
-
o
r
d
i
n
a
t
e
(
d
e
g
)
ke=.25
ke=.50
ke=.75
ke=1.00
ke=1.25
Figure 11. Variation of angular co-ordinate of turbine wheel with the variation of
e
k
kh=8.0 and ke=.75
0
5
10
15
20
25
30
35
0 5 10 15
Meridional Streamlength (mm)
A
n
g
u
l
a
r
C
o
-
o
r
d
i
n
a
t
e
(
d
e
g
)
delta3=0.0
delta3=5.0
delta3=9.0
delta3=12.0
3
Figure 12. Variation of angular co-ordinate of turbine wheel with the variation of
ke=.75 and 3=9.0 deg
0
20
40
60
80
0 5 10 15
Meridional Streamlength (mm)
C
h
a
r
a
c
t
e
r
i
s
t
i
c
A
n
g
l
e
(
d
e
g
)
kh=1.0
kh=3.0
kh=5.0
kh=8.0
kh=10.0
kh=12.0
Figure 13. Variation of characteristic angle in the turbine wheel with the variation of
h
k
Ghosh et al./ International Journal of Engineering, Science and Technology, Vol. 1, No.1, 2009, pp. 43-60 57
kh=8.0 and 3=9.0 deg
0
20
40
60
80
100
0 2 4 6 8 10
Meridional Streamlength (mm)
C
h
a
r
a
c
t
e
r
i
s
t
i
c
A
n
g
l
e
(
d
e
g
)
ke=.25
ke=.50
ke=.75
ke=1.00
ke=1.25
Figure 14. Variation of characteristic angle in the turbine wheel with the variation of
e
k
kh=8.0 and ke=.75
-20
0
20
40
60
80
100
0 5 10 15
Meridional Streamlength (mm)
C
h
a
r
a
c
t
e
r
i
s
t
i
c
A
n
g
l
e
(
d
e
g
)
delta3=0.0
delta3=5.0
delta3=9.0
delta3=12.0
3
Figure 15. Variation of characteristic angle in the turbine wheel with the variation of
ke=.75 and 3=9.0 deg
0
20
40
60
80
100
0 5 10 15
Meridional Streamlength (mm)
F
l
o
w
A
n
g
l
e
(
d
e
g
)
kh=1.0
kh=3.0
kh=5.0
kh=8.0
kh=10.0
kh=12.0
Figure 16. Variation of flow angle in the turbine wheel with the variation of
h
k
kh=8.0 and 3=9.0 deg
0
20
40
60
80
100
0 2 4 6 8 10
Meridional Streamlength (mm)
F
l
o
w
A
n
g
l
e
(
d
e
g
)
ke=.25
ke=.50
ke=.75
ke=1.00
ke=1.25
Figure 17. Variation of flow angle in the turbine wheel with the variation of
e
k
Ghosh et al./ International Journal of Engineering, Science and Technology, Vol. 1, No.1, 2009, pp. 43-60 58
kh=8.0 and ke=.75
0
20
40
60
80
100
0 5 10 15
Meridional Streamlength (mm)
F
l
o
w
A
n
g
l
e
(
d
e
g
)
delta3=0.0
delta3=5.0
delta3=9.0
delta3=12.0
3
Figure 18. Variation of flow angle in the turbine wheel with the variation of
5. Conclusions
A numerical model of the cryogenic turbine wheel blade geometry has been developed covering meridional, suction, pressure,
hub and tip streamline to study the influence of design and operating parameters on the shape of the blade. In addition to its use in
the definition of impeller geometry, the method can be used to define the complete flow passages in a turbine. By using this
method the designer is able to get a better feel for the geometry during the design process, he spends less time on mundane data
preparation, and he has more freedom to make full use of this creativity to design a better machine. Analysis of result reveals that
the optimum value exists to avoid the dumb-bell shaped blade and to have an optimum path length. One of the optimum value of
= 5.0, =0.75 leads to exit radial component
h
k
e
k
3
as 9.
Nomenclature
b channel width
C absolute velocity
turbine wheel diameter D
2
eye tip diameter D
tip
eye hub diameter D
hub
specific diameter d
s
h enthalpy
h adiabatic enthalpy drop across turbine wheel
in-3s
free parameter K
e
free parameter K
h
.
m mass flow rate
M mach number
N number of revolution
specific speed n
s
P power produced
pressure p
Q volumetric flow rate
R
m
radius of curvature of meridional streamline
r radius
S direction and arc length of a meridional streamline
s entropy
t
b
blade thickness
temperature T
U circumferential velocity
W relative velocity
Z number of blades
r, , z cylindrical coordinate fixed to rotor
Greek Symbols
efficiency
rotational speed
Ghosh et al./ International Journal of Engineering, Science and Technology, Vol. 1, No.1, 2009, pp. 43-60 59
ratio of tip diameter to turbine wheel diameter
ratio of hub diameter to tip diameter
relative velocity angle
time coordinate
angle between meridional velocity component and axial coordinate
density of fluid
Subscripts
in inlet to nozzle
1 exit to nozzle
2 inlet to turbine wheel
3 exit to turbine wheel
ex exit from diffuser
hub hub of turbine wheel at exit
tip tip of turbine wheel at exit
m meridional
u circumferential
tr turbine
s isentropic condition
0 stagnation condition
Acknowledgements
The authors are grateful to the Board of Research in Nuclear Sciences (Department Atomic Energy) for sponsoring this work. The
technical inputs from Mr. Trilok Singh, Head Cryo Technology Division, BARC and his group are also gratefully acknowledged.
The authors are thankful to the Mechanical Engineering Department, NIT, Rourkela, for providing all testing facilities and
guidance.
References
Awatef A. Hamed, Widen Tabakoff, Richard B. Rivir, Kaushik Das, Puneet Arora, 2005, Turbine Blade Surface Deterioration by
Erosion, ASME, Journal of Turbomachinery, Vol. 27, pp. 445-452
Balje, O. E. 1970, Loss and flow path studies on centrifugal compressors, Part-II, Trans ASME J Eng Power, pp. 275-299.
Balje, O.E. 1981. Turbomachines: A guide to design, selection and theory. John Wiley and Sons, New York.
Bruce, G. J. 1998, Computer-aided turbomachinery design system, I Mech E Conference on Turbochargers I Mech E Publications
(Paper No. C554/026/98. pp. 3-22).
Casey, M. V. A 1983. Computational geometry for the blades and internal flow channels of centrifugal compressors, Trans ASME
J Eng Power, V105, pp. 288-295
Denton, J. D. 1996, The turboexpander-a design, make and test student project ASME-96-GT- 191.
Dixon, S. L. 1978, Fluid mechanics and thermodynamics of turbomachinery (3
rd
ed) Elsevier, USA.
Hasselgruber,H. 1958, Stromungsgerechte gestaltung der laufrader von radialkompressoren mit axialem laufradeintrict
Konstruction , Vol. 10, No. 1, pp. 22 (in German).
Ito E., Sakamoto Y., Tsukagoshi K., Muyama A., Masada J., 2009, Advanced three-dimensional aerodynamic design for next
generation gas turbines, Mitsubishi Heavy Industries Technical Review, Vol. 46 No. 1, pp. 9-11.
Kun, L .C. and Sentz, R. N. 1985, High efficiency expansion turbines in air separation and liquefaction plants, International
Conference on Production and Purification of Coal Gas & Separation of Air, Beijing, China, pp.1-21
RohliK, H.E. 1968, Analytical determination of radial inflow turbine geometry for maximum efficiency NASA TN D-4384.
Strinning, P. 1995, A computerised trial and error technique for impeller design using computers in the design and selection of
fluid machinery, I Mech E, No. 3, pp. 19-27.
Thakker A. and Abdulhadi R., 2007, Effect of blade profile on the performance of wells turbine under unidirectional sinusoidal
and real sea flow conditions, International Journal of Rotating Machinery, pp. 1-8.
Wallace, F.J., Whitfield, A. and Atkey, R. 1975, A computer-aided design for radial and mixed flow compressor, Computer Aided
Design, pp. 163-170.
Whitfield, A. 1978, Computer aided design and testing of radial and centrifugal impellers with straight and backward swept blades,
Proc Inst Mech Eng, pp. 89-100.
Ghosh et al./ International Journal of Engineering, Science and Technology, Vol. 1, No.1, 2009, pp. 43-60 60
Biographical notes
Dr. Subrata Kumar Ghosh is Senior Lecturer in the Department of Mechanical Engineering & Mining Machinery Engineering at ISMU,
University, Dhanbad, India. Before joining at ISMU He was Lecturer at BIT, Mesra. His research has focused on Thermal Engineering. He is the
author and coauthor of over 8 articles, which have appeared in journals such as Applied Thermal Engineering, Indian Journal of Cryogenics, and
several seminars.
Prof. Ranjit Kr. Sahoo is Professor and Head of Mechanical Engineering Department in the National Institute of Technology, Rourkela, India.
His research has focused on Heat Transfer, Cryogenic Engineering in Space Applications etc. He is the author of over 20 articles, which have
appeared in reputed International Journals.
Prof. Sunil Kr. Sarangi is Director of the National Institute of Technology, Rourkela, India. He is also Professor at the Cryogenic Centre of
Indian Institute of Technology, Kharagpur, India. His research has focused on Cryogenic Engineering in Space Applications, as well as Heat
Exchanger, Radiation etc. He is the author of over 50 articles, which have appeared in reputed International Journals.
Received July 2009
Accepted September 2009
Final acceptance in revised form September 2009
MultiCraft
International Journal of Engineering, Science and Technology
Vol. 1, No. 1, 2009, pp. 61-73
INTERNATIONAL
JOURNAL OF
ENGINEERING,
SCIENCE AND
TECHNOLOGY
www.ijest-ng.com
2009 MultiCraft Limited. All rights reserved
Single electron based binary multipliers with overflow detection
Souvik Sarkar
1
, Anup Kumar Biswas
2
, Ankush Ghosh
1
, Subir Kumar Sarkar
1
*
1*
Department of Electronics and Telecommunication Engineering, Jadavpur University,Kolkata, India.
2
Department of Commuter Science and Engineering, Jalpaiguri Govt. Engineering College, India
*
Corresponding Author: e-mail:sksarkar@jdvu.ac.in
Abstract
Low power consumption, high operating speed and high integration density equipment(s) are financially indispensable in
modern Electronics. Single Electron Device (SED) is one such equipment. Single Electron Devices are capable of controlling
the transport of only an electron. A single electron is sufficient to store information in the SED. This paper presents the approach
for designing multipliers by using single-electron based device. Multipliers with overflow detection based on serial and parallel
prefix computation algorithm are elaborately discussed analytically and designed. The overflow detection circuits works in
parallel with a simplified multiplier to reduce the overall area and to increase the speed compared to the classical digital circuits.
Power consumption in the single electron circuit is low irrespective of Bipolar junction transistor (BJT) or Complimentary Metal
Oxide Semiconductor (CMOS) circuits. Power consumption can be drastically reduced by reducing the nodes. The processing
speed of SED will be nearly close to electronic speed. Noise during processing becomes ultra-low when the mode built with
SEDs is in operation.
Keywords: Single-electron, Binary multiplier, overflow detection, binary decision diagram.
1. Introduction
Single electron based logic gates have already been constructed with binary decision diagram (BDD) (Asahi et al, 1998; Biswas
et al, 2003a) with clock pulses of 1ns each. The technique of tunneling of an electron is utilized for those gates. This technique
may also be used for more complex logical circuit (Korotkov et al, 1998; Biswas et al, 2003b) like a case of a binary multipliers
and overflow detection. For easily understanding the operation of some of the single electronic gates having different number of
inputs have been depicted. Single electron tunneling devices exploit effects that arise due to the quantized nature of charge. These
effects have been observed in systems of small metal structures (Asahi et al, 1995; Asahi et al, 1997), in semiconductors structures
and in structures made from conducting polymers. Because these effects are omnipresent in small structures, they are likely to
have an impact on any future nano-scale electronic circuits. These devices are able to use in low power circuits as only a few
electron is needed for carrying information (Biswas et al, 2003a). The speed power product of single electronic device is predicted
to lie close to the quantum limit set by the Heisenbergs Uncertainty Principle. The processing speed of such device will be close
to the electronic speed (Biswas et al, 2003b).
When the number of bits that do not fit into space available then overflow occurs. For example, when an arithmetic logic
operation creates a result outside of the range of the representable number, overflow occurs (Schulte, et al 2000). If overflow
occurs, an error flag is generated to indicate out of range. Well-known that, multiplication for two n-bit integers produces a
n+n=2n product. But some electronic architecture only return n-least significant bit out of the 2n-bits and overflow sets in the
product cannot be represented correctly with only n bits. For example, IBMs Power Microprocessor family supports a 32-bit by
32-bit twos complements multiply instruction which returns the least significant 32-bits of the 64-bit product and an overflow flag
(IBM, 2000). Similarly Java Virtual Machine supports two integer multiplication instructions; a 32-bit by 32-bit imul (IBM, 2000)
instruction which returns the 32 least significant bits of the product and 64 bit by 64-bit imul instruction which returns the 64 least
significant bits of the product (Lindholm and Yelin, 1996). In the present work, signed and unsigned multipliers with overflow
detection circuit are implemented using single electron tunneling phenomena.
Sarkar et al./ International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 61-73 62
2. Coulomb blockade and single electron transistor
A tunnel junction shown in Figure 1 is considered to be a thin insulating barrier between the two conducting electrodes. The
electrodes may be a superconducting or semiconducting if they are supercononducting, Cooper pairs with a change of two
elementary charges i.e., neither superconducting nor semiconducting, electrons with one elementary charge (1.610
-19
C) carry the
current.
In classical electrodynamics, no current can flow through an insulating barrier. But in quantum mechanics, there is a
nonvanishing (i.e., greater than zero) probability for an electron one side of the barrier to reach the others side. If we apply bias
voltage, there will be a current flow. Avoiding additional effects, according to first-orderapproximation-tunneling current is
proportional to the applied bias voltage. In electrical terms, a tunnel junction behaves like a resistor of a constant value depending
experimentally upon the barrier thickness. If two conductors connected with an insulating layer in between has not only a
resistance but also a capacitance. In this context the tunnel junction acts as a capacitor and the insulator is said to be dielectric. For
the discrete nature of electric charge, current following through a tunnel junction is a series of events in which only one electron
goes through the tunnel junction. As the electron tunnels the junction, the tunnel capacitance is charged with an elementary charge
building up a voltage V=e/C; C=junction capacitance. If the capacitance of the tunnel junction is very small, the voltage developed
in the tunnel junction may be adequate to prevent another electron to tunnel. The electrical current is suppressed then for the bias
voltage lower than the voltage developed in the tunnel junction and the resistance of the device no longer remains constant. The
increment of the differential resistance of the tuunel junction around zero bias is considered as the Coloumb blockade. So, we can
define the Coloumb blockade as increased resistance at very low bias voltages of an electronic device which is having at least one
low capacitance tunnel junction.
Figure 1. Tunnel Junction Figure 2. Single Electron Transistor (SET)
The fundamental principle of single-electronics is based on the Coulomb blockade. Single electron tunneling circuits seem to be
a promising candidate for future VLSI for its ultra-low power consumption, ultra small size and rich functionality.
Single electron Transistor (SET) is shown in Figure 2. A SET has two tunnel junctions having capacitances and conductances
C
1
, C2 and G
1
, G2 respectively, and shares one common electrode with a low capacitance known as island. The electric potential
of the island can be tuned by a third electrode, called gate, which is capacitively coupled (gate capacitance C
g)
to the island. The
drain, source and gate voltages are V
d
, V
s
and V
g
respectively. For proper operations of SET both of the conductances G1 and G2,
of course, are to be smaller than 1/Rq; where Rq = h/e
2
25.8 K and charging energy E
C
=e
2
/(2C) [where C = C
1
+C
2
+C
g
]
has to be greater than thermal fluctuations kT i.e., E
C
>kT.
3. About root node of single-electron Binary Decision Diagram (BDD)
The path selector depicted in Figure 3 is driven by a signal X
i
and a clock pulse
i
,
i =1, 2, 3, 4. This path selector consists of
two SETs and one capacitance C
1
. When an electron comes at point A and pulse
i
>5mV
is applied then the electron can cross
tunnel junctions (J
1
and J
2
) to B (or C) depending on whether the Coulomb energy [E
c
=e
2
/(2C)] + applied energy is greater than the
potential height of the barrier energy of junction(s) J
1
(or J
2
). Following this principle, the electron follows the path ABD (or ACE)
provided the signal X
i
>5mV(or X
i
) and the corresponding total energy Coulomb energy + applied energy is greater than static
potential junction energy of J
3
(or J
4
).This path ABD(or ACE) is said to be 1-arm (or 0-arm). Figure 4 is the symbol of this path
selector.
Sarkar et al./ International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 61-73 63
Figure 3. Path selector of an electron Figure 4. Symbol of path selector
4. Configuration of Electron to Voltage Converter
Figure 5 acts as a buffer. The configuration of the input electron to output voltage has been depicted in Figure 6 and its operation
has been explained below.
Figure 5. A buffer Figure 6. Electron to voltage converter Figure 7. Input output waveforms
It contains six tunnel junction capacitances and six capacitances (their values are depicted in the Figure 7). It has two input
terminals: input(for 1-arm) and input (for 0-arm )and one output terminal V
out
. The 1-arm and 0-arm of a single electron circuit
are connected, respectively, to the input(for 1-arm) and input (for 0-arm) of the Figure 6. If an electron reaches at the 1-arm, the
output voltage V
out
becomes approximately 5mV. If the electron reaches at 0-arm then V
out
will be approximately 0 (zero). The
input and output waveforms of Figure 6 are shown in Figure 7.
5. Some Single-Electronic Gates
Figure 8. Two-input AND gate Figure 9. Two-input OR gate Figure 10. Two-input XOR gate
Sarkar et al./ International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 61-73 64
6. Theorem related to overflow
6.1 Theorem-1:
For two unsigned binary codewords having l and m significant bits respectively, no overflow for the product (multiplication) of
these two codewords will happen if total number of significant bit of the product is nl+m.
Suppose two unsigned binary word A and B. Significant bits of A include the leftmost 1 and all the remaining bits right to that.
If A=00100010, then A has six significant bits. Similarly if B=10010001, then it has eight significant bits. Now M=AB, then
there will be no overflow if the significant bit of the product M is l+m >n i.e., we have to prove for non-overflow l+m>n.
Limit of the value of A is
2
l-1
A2
l
-1 (1)
Similarly limit of B is
2
m-1
A2
m
-1 (2)
Limit of the value of M=AB is
2
l+m-2
A2
l+m
-(2
l
+2
m
)+1 (3)
from the left hand side of the equation if 2
n
2
l+m-2
then overflow occurs.
i.e., if l+mn+2 (4)
overflow occurs, overflow does not occur if l+m<n+2 (5)
From the right hand side of the equation (iii) there will be no overflow if
2
l+m
-(2
l
+2
m
)+1 2
n
-1 (6)
but any positive integer (2
l
+2
m
) -2 or (2
m
+2
m
)+1-1 (7)
combining equation(6) and (7) we get 2
l+m
2
n
or l+m n (8)
from equations (v) and (viii) we can conclude that no overflow occurs if l+m n or if nl+m
6.2 Detection of Overflow:
From the equation (4) it is obvious that if l+mn+2 then overflow must happen. This condition is satisfied with the equation (24-
26) given below.
O
v1
=a
n-1
.b
1
+(a
n-1
+a
n-2
).b
2
+( a
n-1
+a
n-2
+a
n-3
).b
3
++( a
n-1
+a
n-2
+ +a
1
).b
n-1
(9)
For verification we take first term a
n-1
.b
1,
if its value is 1 i.e., a
n-1
.b
1
=1 then the total number of significant bits is at least n+2 as A
has at least n significant bits and B has at least 2 significant bits. In the same way, if the second term (a
n-1
+a
n-2
).b
2
is 1 then A has at
least (n-1) significant bits and B has at least 3 significant bits. Similarly, if the last term (a
n-1
+a
n-2
+ +a
1
).b
n-1
is 1, then A has at
least 2 significant bits and B has at least n significant bits so the total number of significant bits will be (n+2). The equation (9) is
written as
n-1 n-1
i=1 j=1
O
v1
= a
n-j
.b
i
,
where bit-dot-product and bit-summation indicate logical
AND and OR respectively. Comparing equation (4) and (8), we get that when l+m=n+1 the multiplied value may overflow but the
value is less than 2
n+1
. In this situation the product bits p
0
through p
n
are sufficient to indicate whether the product value will
overflow or not. If m
n
is equal to 1 then only overflow occurs. So we conclude that if l+mn+2 or m
n
=1 then overflow must occur.
But from equation (9), the essential number of AND and OR gates are respectively
1+2+3++(n-1)=(n-1).n /2 (10)
and (n-2)+(1+2+3+4++n-2) = {(n-1).n /2}-1 (11)
If the recursive (iterative) technique is applied then the number of AND or OR gates can be reduced.
Sarkar et al./ International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 61-73 65
6.3. Iterative technique:
The hardware reduction can be used here by using the iterative process. We take
r
k+1
=r
k
+o
k+1
.b
k
and o
k+1
= o
k
+a
n-k
where 2kn-1.
Initially we take, o
k=2
=a
n-1
and r
k=2
=a
n-1
.b
1
Then o
3
= o
2
+a
n-2
=a
n-1
+
a
n-2
r
3
=r
2
+o
3
.b
2
= a
n-1
.b
1
+( a
n-1
+
a
n-2
).b
2
o
4
= a
n-1
+
a
n-2
+a
n-3
r
4
=
r
3
+o
4
.b
3
= a
n-1
.b
1
+( a
n-1
+
a
n-2
).b
2
+(a
n-1
+
a
n-2
+a
n-3
).b
3
r
n
=O
v1
=a
n-1
.b
1
+(a
n-1
+a
n-2
).b
2
+( a
n-1
+a
n-2
+a
n-3
).b
3
++( a
n-1
+a
n-2
++a
1
).b
n-1
(12)
After (n-1) iterations we obtain the result of the equation (12) which indicates the overflow. if the significant bits of the product M
is n+1 then also overflow occurs As overflow when P>2
n
. So we can draw the block diagram for overflow using multiplier and
iterative circuit like as
Figure 11. Block diagram for overflow
6.4. Highest order of input AND and OR gates for r
n
(Big oh of r
n
)
:
From the equation (12), number of AND gates and OR gate are (n-1).n /2 and {(n-1).n /2}-1 respectively. Now, (n
2
/2-n/2)
n
2
/2 when n2
n
2
when n 2
So big oh of the number of AND and OR gates are O(n
2
) and O(n
2
) respectively.
6.5. Implementation of multiplier:
From Theorem-1 and from the iteration equation (12), We can implement an unsigned multiplier and a overflow circuit. First
eight bit unsigned digital multiplier has been implemented in Figure 12.
6.6. Reduction of delay:
The unsigned digital overflow can be reduced by using the prefix algorithm (Gok et al, 2000). The equation (12) are
rewritten as
O
v1
=A
n-1
.b1+A
n-2
.b
2
+A
n-3
.b
3
++A
1
.b
n-1
(13)
n-1
where a
k
(14)
k=1
6.7. Serial Prefix:
For a size n, assume that the inputs are d
1
, d
2
, d
3
, d
4
, , dn, and the related operation used is . After applying each output y
i
is written as
y
i
= d
1
d
2
d
3
d
4
d
i-1
d
i
(15)
where 1in .
Equation (15) can be utilized for sequentially computing y
i
by using (n-1) times of operations. Then this process is called serial
prefix computation.
6.8. Parallel prefix:
Sarkar et al./ International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 61-73 66
From equation (15), it is clear that we can apply the operation in any order. For example, we can determine (d
1
d
2
), (d
3
d
4
),or ,(d
i-1
d
i
) in parallel. This technique is called parallel-prefix (Cha et al, 2000; Ladver et al, 1980) technique. For
reduction of time delay parallel prefix technique can be applied.
k
A
k
= a
j
, 1kn-1, can be implemented using parallel prefix technique for an n = 8bit
j=1
using the operator OR. Parallel prefix technique is given in Figure 12 and Figure 13.
Figure 12. Parallel prefix circuit for A
k
= a
j
Figure 13. Parallel Prefix technique for overflow
(
O
v1
)determination.
6.9. Delay Optimization:
If we analyze the parallel prefix based implemented circuit, it is possible to reduce the node-number and the processing
delay time. After analysis, the Figure 12 and Figure 13 are implemented another way so that the processing delays are
minimized/optimized. The optimized figures are given in Figure 14 and Figure 15 respectively. The processing delay based on
SED and analyzed SED for parallel prefix circuit is shown in Table 1.
Table 1. Processing delay based on SED and analyzed SED for parallel prefix circuit
Figure No. No. of Delay (ns) Node
Combination
Total delay
time(ns) Nodes required
Figure-12 32 12 28
84 Figure-13 52 16
Figure-14 39 8 55 16
Figure-15 16 8
Sarkar et al./ International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 61-73 67
Figure 14. Analyzed Parallel prefix circuit of Figure 12 Figure 15. Analyzed Parallel prefix of Figure 13
Figure 16. Modified Half Adder Figure 17. Modified Full Adder
Sarkar et al./ International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 61-73 68
Figure 18. Multiplication and Overflow for unsigned numbers
7. Multiplier of sign numbers
To multiply the sign numbers we have to find the twos complements of the sign numbers. Next step is to determine the number
of significant bits of the twos complement number(s) as according to the Theorem 2 significant numbers are essential for
indicating the overflow condition. The significant bit numbers of a twos complement number is determined as:
The number of bits for first different rightmost bit with the sign bit of A and the bits right to that different rightmost bit is called
significant bit numbers. Suppose A=00010010 and B=11111000, the number of significant bit of A and B are in Table 2.
Table 2. Significant bit numbers computation
1
st
unlike bit position number Sign bit Significant bit
number from LSB
A 0 5
th
[10010] 5
B 1 3
rd
[000] 3
7.1. Theorem-2: If the significant bits of twos complement numbers of A and B are l and m then the overflow of the product
(multiplication) of these numbers will happen if l+m<n-1, (n=number of significant bits)
Say, A and B have significant numbers l and m respectively then A and B are bounded by
2
l-1
A<2
l
(16)
2
m-1
B<2
m
(17)
and the product is also bounded by
2
l+m-2
M= AB <2
l+m
(18)
We know that the overflow occurs if M2
n-1
(19)
when M is positive
or, M< -2
n-1
(20)
when M is negative
(As overflow happens when M2
n-1
or M< -2
n-1
)
Sarkar et al./ International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 61-73 69
Comparing equations (16) and (17), the left hand limit implies that overflow occurs if
2
l+m-2
2
n-1
or l+m-2 n-1
or l+m n +1 (21)
So, overflow will not happen if l+m n (22)
For the right hand limit from equation (16), overflow will not hold if
2
l+m
< 2
n-1
or l+m< n-1 (23)
Comparing equation (16) and (22) overflow will not occur if l+m< n-1.
7.2. Condition-a: when M is positive
From the Theorem 2 we observe that (i) overflow happen if l+m n +1 and
(ii) overflow will not happen if l+m< n-1
But what happens when l+m=n or n-1?
We investigate these two situations step by step if l+m=n-1, we get from equation (17)
2
l+m-2
M<2
l+m
or 2
n-3
M<2
n-1
(24)
It is known that for overflow condition M2
n-1
(25)
So overflow may happen but it does not exceed the limit 2
n-1
. When l+m=n, then
2
n-2
M<2
n
so overflow may occur but it does not exceed the limit 2
n
.
7.3. Condition -b: when M is negative
When l+m=n or n-1 and the sign of M is negative or positive
i) If both A and B are positive and M< 2
n
then it implies that m
n
=0 always (as M2
n-1
for overflow) and overflow holds only
if m
n-1
=1
ii) If A and B both are negatives then M2
n
indicates that p
n
=0 all times and p
n-1
=1only when overflow occurs.
iii) When A and B has different signs, then -2
n-1
<M means p
n
=1 always
and p
n-1
=0 only when overflow occurs.
Clearly from the above three situations it is clear that
or, m
n
m
n-1
=1 (26)
7.4. Verification of Output (O/P) of the Figure 19
We are to show that the output value of the circuit and the logic value of the logic function F= ab C D are the same. The result
will be 0 or 1 if the messenger electron reaches at the or valued node. For example, when [a b C D]=0110 then the electron
follows the 0 branch between the nodes (i) and (ii), 0 or 1 branch between the nodes (ii) and (iii), 1-branch between then nodes
(iii) and (iv) and 0-branch between the node (iv) and terminal and the messenger reaches the node i.e., the O/P will be 1.
From the Table- 3 the other binary words can be verified.
Sarkar et al./ International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 61-73 70
Table 3. Binary word verification
a b C D O/P ab ab F remark
Figure 19. Overflow detection circuit
8. Comparison of time delay
The parallel prefix circuit given in Figure 14 represents the maximum processing time delay is 44=16ns whereas if the circuit is
analyzed we shall be able to implement the circuit whose processing delay time must be lesser than the previous one i.e., the
processing time can be optimized The same circuit has been implemented another way (Figure 16) which is a optimized form no
doubt. In Figure-16, the processing delay time is 8ns, Similarly, the parallel circuit (Figure 15) consumes 16ns and its optimized
form, Figure 17, requires 8ns. If these two parallel circuits are implemented by using CMOS/TTL gates then the minimum delay /
processing time would have 48ns and 48 ns respectively. If the parallel prefix technique is not used then for ANDing of 7 bits will
be for
(i) CMOS/TTL logic 126=72ns (Milman, 2000)
(ii) SED based 46=24ns
(iii) Analyzed based = 8ns.
In Table 4, time delays and fastness for CMOS/TTL gates, SED based circuits and analyzed based circuits of serial (sequential)
and parallel prefix computations are given.
Table 4. Comparison of time delays and fastness for CMOS/TTL gates, SED based circuits
Sl.
No.
Circuit
Name
Serial prefix
Delay time
(ns)
Faster (times)
with respect to
CMOS/TTL
Parallel prefix
Delay time (ns)
Faster (times)
with respect to
CMOS/TTL
1 CMOS/TTL gates
(Milman 2000)
72 1 48 1
2 SED gates
(Asahi 1995; Asahi 1998)
24 3 16 3
3 Analyzed SED gates 8 9 8 6
9. Overflow completion
Combining the two conditions (i) n significant bits of the A and B, and (ii) the value of (m
n
m
n-1
), the resultant overflow is
obtained for unsigned multiplication. The logical equation of the resultant overflow is:
Overflow=O
v2
+ (m
n
m
n-1
) (27)
0
0
1
0
1
0
0
0
0
0
0
0
1
1
0
1
1
0
0
0
0
0
0
1
1
0
Value of
O/P and F
are same
1 0 0 0
1 1 1 0 1 1 1 1
1 1 0 0
1 1 1 1
Sarkar et al./ International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 61-73 71
where O
v2
is given in equation (28).For signed multiplication overflow must occur if l+m n +1 [from equation (21)]. To satisfy
this condition we can write below a logic equation [20-22], the value of which will be 1when overflow happens,
O
v2
= {b
n-2
1
+ (b
n-2
+ b
n-3
)
2
+(b
n-2
+ b
n-3
+b
n-4
)
3
+
+(b
n-2
+ b
n-3
+
+b
1
)
1
} (28)
that is O
v2
= 1 for overflow, where
j
= a
j
a
n-1
and b
j
= bj b
n-1
.
Figure 20. Overflow detection of sign-multiplier
10. Combination of signed and unsigned multiplication and overflow
We are able to combine the signed and unsigned multiplication with their overflows in a single multiplier for the similarities of
their designed circuits without loss of the generality. To do this, an external control line is connected with the circuit. The input
value of the control lone data (d) will be 0 or 1. We choose the general data input equation of a
j
and b
j
as
j
= a
j
(a
n-1
d) (29)
b
j
= b
j
(b
n-1
d) (30)
where 1j n-1.
When d=0 the equations (29) and (30) give a
j
and b
j
respectively and when d=1, give
j
and b
j
avoiding
j
=a
n-1
a
n-1
and b
j
= b
n-
1
b
n-1
in the last case. Clearly, for d=0, the circuit will contribute the unsigned product and for d=1, the circuit will contributes the
signed product.
Sarkar et al./ International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 61-73 72
Figure 21. Final overflow circuit
In accordance with the Theorem 1 and Theorem-2 and with the derived equations of (9) and (27), we can create the multiplied
value(s) and general cased overflow. Finally, the overflow is calculated from the overflow equation given as:
O
flow
= (O
v1
+m
n
) + (O
v2
+m
n
m
n-1
) d (31)
where +, and indicates OR, X-OR and AND operations. The related Final overflow detection circuit is given in Figure
23. Where m
n
and m
n-1
are taken from the multiplication circuit in Figure 20 and O
v1
and O
v2
are taken from the overflow parts of
the circuits of Figure 18 and Figure 20 respectively.
11. Results and discussions
If both the operands are of same sign (i.e., either positive or negative) the result of the multiplication will be of positive sign. In
such situation the Figure 19 is used for overflow detection. On the other hand, if the sign bit of the two operands are opposite signs
i.e., the XOR of these two sign bit is 1 then Figure 20 is used for multiplication and overflow detection. These two signed and
unsigned multiplication circuits are brought to a same platform for unifying them after a small modification of the input data. The
Figure 21 delivers the final overflow indication. The overflow detection circuit does not depend upon the internal carries which is
generated by the multiplier and does not require the partial product of the multiplier to be modified. So the circuit can be applied to
any type of simplified multiplier which produces (n+1) product bits. The parallel and analyzed circuits have less area and fewer
nodes than the serial prefix circuits. The designed multiplier circuit can be used efficiently for both signed and unsigned operands.
The table 3 manifests the speed as well as the delay time is given for comparing the CMOS/TTL logic gates based circuit(s) and
single electron based circuit. Obviously, Single electron device based circuit is at least 3 times faster than the classical logic based
circuits.
12. Conclusion
A general technique for multiplication along with the overflow based on single electron tunneling phenomena has been
implemented for positive (unsigned) and negative (signed) integer multiplication. This technique includes separately serial and
parallel prefix algorithm. Parallel prefix algorithm is used for time saving and decreasing the gates or nodes. Analyzed circuit also
optimizes the processing delay in comparison to the classical implemented techniques. Thermal agitation and stochastic errors
caused by probabilistic fluctuations are excluded in our design.
Acknowledgment
Subir Kumar Sarkar thankfully acknowledges the financial support obtained from UGC vide F. No.36-100/2008(SR)
References
Asahi N., Akazawa, M. and Amemiya, Y. 1995. Binary decision diagram device. IEEE Trans. ED. 42 11, pp. 1999-2003.
Asahi N., Akazawa, M. and Amemiya, Y.. 1997. Single electron logic device based on the binary decision diagram, IEEE Trans
on ED. 44, 7. pp. 1109-1116.
Sarkar et al./ International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 61-73 73
Asahi N. Akazawa, M. and Amemiya, Y. 1998. Single electron logic systems based on binary decision diagram. IEICE Trans.
Electron. E-81-C. 1. pp. 49-56.
Biswas A.K. Sarkar, S.K. 2002. Single electron logic circuits for the realization of a logic unit of a computer, IJICS, Vol.5, No.1,
pp.53-58.
Biswas A.K., Sarkar S.K. 2003. Error detection and Debugging on Information in Communication system using single electron
circuit based binary decision diagram. Semiconductor Physics Quantum Electronics and opt electronics. Vol. 6, pp. 1-8.
Biswas A.K. and Sarkar S.K. 2003. An Arithmetic logic unit of a computer based on single electron transport system.
Semiconductor Physics Quantum Electronics and opt electronics. Vol. 6, No.1 pp.91-96.
Cha Y.H., Cho G.Y., Choi H.H. Song H.B., 2001. Bit result integer multiplier with overflow detector, IEE Electronic Letters. Vol.
37. pp. 940-942.
Gok M., Schulte M.J., Balzola P.I., and Brocato R.W., 2000. Combined unsigned and twos complement saturating multipliers,
Proc. SPIE: Advanced Signal Processing Algorithms, Architectures and Implementations. pp.185-196.
IBM, 2000. Power PC Microprocessor Family. The Programming Environments for 32-bit Microprocessors, no. G522-0290-01.
Korotkov A.N., Likharev K.K. 1998. Single electron-parameter-based logic devices. J. Appl. Phys. Vol. 84, pp. 6114-6126.
Lindholm T. Yelin, F. 1996. Java Virtual Machine Specification. Addison-Wesley, Reading, MA.
Ladver R.E and Fiscer M.J. 1980. Parallel prefix computation. J. ACM, Vol. 27, pp. 831-938.
Milman J., Halkias C.C. 2000. Integrated Electronics, Edn. McGraw Hill.
Schulte M.J., Balzola P.I., Akkas A., Brocato R.W., 2000. Integer multiplication with overflow detection or saturation, IEEE
Trans. Computers, Vol.49, No. 7, pp. 681-691.
Biographical notes
Mr. Souvik Sarkar is a Research Scholar in the Department Electronics and telecommunication Engineering, Jadavpur University. His areas of
interest include nano, single electron and spintronic device based circuit modeling.
Dr. Anup Kumar Biswas is Lecturer in Commuter Science and Engineering in Jalpaiguri Govt. Engineering College (India). He has engaged in
teaching and research activities since the last 10 years. His field of specialization is Single Electron Device. Dr. Biswas has published several
papers in various national, international conferences and journals.
Mr. Ankush Ghosh is a Research Scholar in the Department Electronics and telecommunication Engineering, Jadavpur University under Prof.
Subir Kumar Sarkar. His areas of interest include nano, single electron and spintronic device based circuit modeling.
Prof. Subir Kumar Sarkar is a Professor of Department of Electronics and Telecommunication Engineering, Jadavpur University, Kolkata -700
032. He has published three Engineering text books (from CRC, USA) and more than 260 technical research papers in archival journals and peer
reviewed conferences. Thirteen students have been awarded PhD (Engg) degree under his guidance and twelve more are presently pursuing PhD
under his guidance. As Principal Investigator he has successfully completed four R&D projects and two more are running. He has visited several
countries like France, UK, Switzerland and Japan for Technical reasons like training, presenting papers or visiting sophisticated laboratories. His
most recent research focus is in the areas of simulations of nanodevice models, transport phenomenon, single electron & spintronics devices and
their applications in VLSI circuits, low power VLSI design, ad hoc wireless networks, wireless mobile communication and watermarking. He is
a life Fellow of the Institution of Engineers and Institution of Electronics and Telecommunication Engineers, life member of Indian Association
for the cultivation of Science.
Received August 2009
Accepted September 2009
Final acceptance in revised form September 2009
MultiCraft
International Journal of Engineering, Science and Technology
Vol. 1, No. 1, 2009, pp. 74-89
INTERNATIONAL
JOURNAL OF
ENGINEERING,
SCIENCE AND
TECHNOLOGY
www.ijest-ng.com
2009 MultiCraft Limited. All rights reserved
Transverse vibration of spinning disk with attached distributed patch and
discrete point masses using finite element analysis
Vinayak Ranjan
1*
and M.K.Ghosh
2
1
Department of Mechanical Engineering and Mining Machinery Engineering, Indian School of Mines, Dhanbad, Jharkhand, India
2
Department of Mechanical Engineering, Institute of Technology, Banaras Hindu University,Varanasi, 221005, India
*
Corresponding author (e-mail: vinayakranjan@yahoo.com)
Abstract
Free and forced transverse vibration characteristics of a thin spinning disc attached to a rigid core have been investigated by
finite element analysis using ANSYS software. The effect of discrete point masses and patches of distributed masses attached at
the periphery of the plate on free and forced vibration behavior of a spinning disc has been investigated. Discrete point masses
and distributed patch masses have been placed at higher strain regions to look into its influence on the natural frequencies, mode
shapes and response of the plate to external excitation. Results for eight and sixteen point masses and patches on the vibration
characteristics have been compared. It has been observed that discrete patches and point masses have significant influence on the
modal frequencies and these can also act as dynamic vibration absorbers in reducing vibration of a spinning disc. It has also been
shown that discrete patches of piezoelectric patches can also be used to actively control vibration of the spinning disc.
Keywords: Free and forced vibration, discrete mass and patch, vibration control.
1. Introduction
Spinning discs are important components in various mechanical applications such as circular saws, disk brakes, hard disks,
optical discs and gas turbines. Vibration control of spinning discs has been studied by a number of researchers. Lamb and
Southwell (1922) and Southwell (1922) have reported a study on the free vibration of flexible disks. Eversman and Dodson (1969)
studied the free vibration of a centrally clamped spinning disk. Barasch and Chen (1972) investigated on the natural frequencies of
rotating disk using modified Adams method. Iwan and Moeller (1976) explored the vibration behavior and stability of spinning
disks subject to stationary transverse loads. They employed eigenfunction expansion method to calculate the natural frequencies.
Chen and Bogy (1992) examined the natural frequencies of rotating discs with stationary load system. They presented a theoretical
study of a general gyroscope system and applied it to a spinning disc. Ramaiah (1981) studied the natural frequencies of spinning
annular plate for eight different boundary conditions using Rayleigh-Ritz method. Lakshminarayana (1986) analyzed the rotating
laminated composite annular disc using finite element method. Sinha (1998a, 1998b) investigated the free vibration problem of a
spinning annular disc with uniformly distributed masses or a ring attached to its outer edge by using Rayleigh Ritz technique.
Huang and Yu (1990) examined the vibration behavior of a rotating annular plate with elastic or rigid restraints at discrete
locations. They used energy principle and Galerkin method to obtain an approximate solution. Shahbab (1993) analyzed the
transverse vibrations of a spinning flexible variable thickness disc. He employed both Ritz method and finite element technique
and also carried out experimental investigation. Lee and Ng (1995) determined the natural frequencies and critical speeds of a
spinning annular plate of varying thickness based on Lagrangian approach and assumed mode method. Liang et al. (2002) studied
the free and forced vibration of a rotating polar orthotropic annual plate with a stationary concentrated transverse load. They used
Galerkin approximation to evaluate the eigenvalues of the system and stressed that discs with higher modulus ratios or poisons
ratios have higher natural frequencies. Jen-San Chan (2003) proposed an active control scheme based on the transfer function
model to suppress the vibration of a spinning disc. Ferretti et al. (2002) presented a methodology, based on dynamic modeling and
experimental measurements to study the vibrations in hard disc drives. Curadelli et al. (2004) dealt with the vibration control of a
plate excited by rotary machines by placing masses on the plate system and compared its performance with the conventional
Ranjan and Ghosh / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 74-89
75
vibration isolation using springs. Wang (2005) determined the natural frequencies of a circular plate with an attached core for
different boundary conditions. Koo (2006) calculated the natural frequency and critical speed for rotating composite disks by the
RayleighRitz method. He showed that the circumferentially-reinforced disk is more effective in increasing critical speed than the
radially-reinforced disk.
Bauer and Eidel (2007) determined the lower approximate natural frequencies of a spinning circular plate for various boundary
conditions. They determined the approximate lower natural frequencies as functions of the speed of spin. Duan et al. (2008) has
proposed that fundamental vibration modes of circular plates with free edges can be modified by increasing the bending rigidity of
the outer rim of the circular plate by using a larger thickness or by using a material with a larger Youngs modulus or both.
Bashmal et al. (2009) studied the in-plane modal characteristics of circular annular disks under combinations of all possible
classical boundary conditions. The in-plane free vibration of an elastic and isotropic disk is studied on the basis of the two-
dimensional linear plane stress theory of elasticity. The boundary characteristic orthogonal polynomials are employed in the
RayleighRitz method to obtain the natural frequencies and associated mode shapes. In the present work, free and forced
transverse vibration behavior of a spinning disc with a rigid core having discrete patches and discrete masses attached at its
periphery have been analyzed using finite element method. The actuation effect of piezoelectric patches on the transverse vibration
of rotating disc has also been looked into.
2. Theory and formulation
The equation of the motion for the free transverse vibration of a rotating thin circular plate in space fixed coordinates is given by
Wang (2005):
r
w
r
w
r
r r
h
w D
r 4
+
2
2
w
h
t
=
0
(1)
Notations used to write the above equation are given below:
) , , ( t r w denotes the transverse displacement with respect to ) , ( r frame,
is the density of plate,
is the Poissons ratio,
is the angular speed of the plate,
h is the thickness of the plate,
( )
2
3
1 12
=
Eh
D is the bending stiffness of the plate
2
2
2
2 2
2
4
1
r r r r
( , r
r
) ) and are stresses due to centrifugal effect and are given by ( , r
( ) , r
r
= ( )
2 2 2
8
3
r a
+
+
( ) ( ) ( ) [ ]
( ) { }
+ +
+ +
1
1 ( 1 8
1 3 1
2
2
2 2
2 2 2 2 2
r
a
a b
b a b
( ) , r
= ( ) ( ) [ ]
2 2
2
3 1 3
8
r R
+ +
( ) ( ) ( ) [ ]
( ) { }
+ +
+ +
1
1 ( 1 8
1 3 1
2
2
2 2
2 2 2 2 2
r
a
a b
b a b
where,
b
is the inner radius of the disk, where core of thickness 0.008 m is attached.
a is the outer radius of the disk
The boundary conditions are
0 = w ; for (2) b r =
0 =
r
w
; for (3) b r =
2
2
2 2
2
1 1
u
r r
u
r r
w
D = 0 ; for a r =
(4)
Ranjan and Ghosh / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 74-89
76
( )
( )
w
r r r
w
r
D
1 1
2
2
=0 ; for a r =
(5)
The fourth order partial differential equation given by (i) has historically been difficult to solve exactly. Finite Element Method
has been used as an efficient tool for modeling the vibration response of the plate, wherein a system of equations is constructed
which is solved as an eigenvalue/eigenvector problem. ANSYS has been used to determine the mode shapes also called
eigenvector and the natural frequencies of the plate system. The mode extraction method employed is Block Lanczos [ANSYS ]. A
block shifted Lanczos algorithm as found in Grimes et al. (1994) is the theoretical basis of the eigenvalue solver. The method
employs an automated shift strategy to extract the number of eigen value requested. The Block Lanczos method is the variation of
the classical Lanczos algorithm where the Lanczos recursions are performed using a block of vectors as opposed to a single vector
as explained by Rajakumar and Rogers (1991). The Lagrange Multiplier approach is implemented to treat constraint equations in
the Block Lanczos eigenvalue solver as discussed by Cook (1981). A circular plate of radius 0.1m having a thickness 0.002 m and
with core radius of 0.01m and core thickness of 0.008 m is used for the free vibration analysis. The four node shell element (shell
63 in Ansys 5.4) having six degrees of freedom at each node, three for translational displacements in x, y, z directions and three
for rotational displacements about x, y, z directions, has been used to model the plate. The plate is divided into 1404 elements as
shown in Figure 1 and is rigidly clamped at the core.
Ranjan and Ghosh / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 74-89
77
Location of piezo-electric patches and discrete masses are important consideration in structural vibration control. Crawley and
Javier (1987) suggested the location of high strain regions for placement of actuators for vibration control. However, the dynamic
response of the structure is due to contribution of several modes. High strain energy locations may be different for different modes.
This particularly becomes important for rotating structures which generate centrifugal stresses. Consequently higher strain regions
change. The locations of the patches are shown in the Figure 1. The physical properties of the plate and the patch are given in the
Table 1. The discrete patches are considered bonded perfectly to the plate. The dimension of each patch is shown in Figure 2.
Table 1 : The physical properties of the plate and patch
Plate Patch
Density [Kg/m
3
] 2790 7800
Youngs modulus [Pa] 0.72E11 0.79365E11
Poisson ratio 0.29 0.30
The sector patches are inserted into 0.001m depth groove of the plate. At the first instance the effect of rotation on the natural
frequencies of the plate with and without piezo-ceramic patches is investigated for different modes. Then the influence of point
masses on the free transverse vibration of rotating disk is examined. The weights of discrete patches and masses placed are equal
i.e. 1.9399E-3. Finally the case of forced vibration response and vibration control of the rotating plate with distributed patches is
considered when some of the patches are activated. In the case of forced vibration response the forcing function is a harmonic
excitation of 1.0 N in the transverse direction at the location (0.07 m, 45
0
). The transverse deflection of the plate is measured at
(0.1 m, 45
0
). Since the piezo-electric patches generate distributed load over the area, the total force is the resultant force on the area
at its centre of gravity. In order to determine forced vibration response the magnitude of the force has been chosen arbitrarily so
that a measurable transverse deflection is obtained.
3. Results and discussion
Results are now presented below for various cases which have been considered through figures and tables.
3.1. Natural frequencies and mode shapes with patches
Modal frequencies and mode shapes of the spinning plate rigidly clamped to a core are given in Table 2 for the first ten modes at
different angular speeds up to 1000rad/sec. From Table 2, it is clear that modal frequencies increase with speed for all the modes.
In fact, the rotation generates a centrifugal force field whose net effect is an apparent increase in disc stiffness causing in turn an
increase of the natural frequencies with speed. This is a well known fact and has been studied earlier. Further, it is also evident that
percentage increase in natural frequencies of plate is higher at higher speed. This gives an indication that disk stiffening is more
significant at higher speed. It can also be noted from Table 2 that for a given rotational speed, frequencies of modes 1 and 2 are
same. Modal frequencies occur in pairs with same value e.g. frequencies for modal pairs (4, 5), (6, 7) and (8, 9) also remain same
for each pair but their mode shapes change as can be seen in Figure 3. The diametric lines of symmetry change with the mode i.e.
mode 4 and 5 have same frequency but the diametric line of symmetry are different. Vibration control of rotating disc has always
been a challenge because of complexity and density of vibration modes. Modal frequencies and their corresponding mode shapes
provide a better insight into the location of piezo-ceramic patches for vibration control. To enhance vibration control it is better to
78 Ranjan and Ghosh / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 74-89
place the patches at higher strain/curvature regions. The first ten mode shapes of the stationary plate rigidly clamped to a core are
given in Figure 3. The first or the fundamental mode shape and the second mode of the stationary plate shown in Figures 3(a) and
(b) are characterized by a nodal diameter through which the relative phase of displacement changes by about 180
0
. These two
mode shapes have two maximum curvature regions each where the patches are to be located. The third mode shape in Figure 3(c)
has nodal point (no strain) at the center, which is rigidly fixed. The maximum deflection point is on the periphery of the disk where
the patches may be located. The fourth and fifth mode shapes in Figure 3 have four points of maximum curvature with 90
0
phase
displacement. At these four points patches may be located. In the case of sixth mode in Figure 3(f) and seventh mode in Figure 3
(g), there are three nodal diameters with six points of maximum curvature where the patches may be located.
Again the 8
th
and 9
th
mode shapes in Figure 3(h) and 3(i) have four nodal diameters with eight points of maximum curvature on
the circumference of the disk. The tenth mode shape is characterized by nodal ring in Figure 3(j). The maximum curvature region
consists of two circular rings along which patches may be placed. It is interesting to note that with increasing speed the points of
maximum curvature changes. Table 3 presents the effect of eight patches on the free vibration of the rotating plate. The eight
patches are located in positions 1, 3,5,7,9,11,13,15 as shown in Figure 1. The 1
st
, 2
nd
and 4
th
modes shapes of the plate with eight
patches change and that out of eight patches only four patches which lie in the regions of high strains contribute significantly to
drop the modal frequency while the contribution of the rest four patches are only partial. Only small drop in the corresponding
frequencies are noted in Table 3. For the third mode, additional patches do contribute to vibration control as they lie in the
maximum strain region. However, it is noticed that 5
th
modal frequencies drop for all the rotational speeds considered for
evaluation in Table 3. The 5
th
mode shape remains similar to Figure 3(e) and therefore only four out of eight patches lie in the
higher strain region and therefore only four patches contribute effectively to lower the modal frequencies. The 6th and 7
th
mode are
having six antinodes and are having six small regions of higher stain. Some of the patches fall directly in higher curvature region
while some of the patches cover partially. So decrease in frequency is noted in Table 3. In case of 8
th
mode shape, all the eight
patches fall directly in the region of higher curvature so drop in modal frequencies are significant. For the 9
th
mode, all the patches
cover the nodal diameters but due to the geometry of the patch and also due to eight small areas of higher strains owing to eight
antinodes, the patches cover partially the higher strain areas and therefore drop in frequency for 9
th
mode shape is also reported but
drop in frequency is not significant as it is for 8
th
mode. The tenth mode is characterized by a circular nodal ring. As the patch
locations are on higher curvature region of circumference, so fall in frequency is noticed. This indicates that the proposed eight
patches may be used to control the vibration of all the ten modes.
Figure 3 (a) Mode 1 of stationary plate with one nodal diameter, Freq. = 172.06 Hz
Ranjan and Ghosh / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 74-89
79
Table 2: Modal frequencies of the plate without any patch / point mass for different rotational velocities, ( rad / sec)
Figure 3(b). Mod 2 of stationary plate with one nodal diameter, Freq. = 172.06 Hz
Mode 0 = 100 = 200 = 300 = 400 = 500 = 600 = 700 = 800 = 900 = 1000 =
1 172.06 172.87 175.31 179.29 184.71 191.46 199.40 208.38 218.27 228.96 240.34
2 172.06 172.87 175.31 179.29 184.71 191.46 199.40 208.38 218.27 228.96 240.34
3 208.08 208.50 209.76 211.83 214.69 218.31 222.63 227.62 233.22 239.38 246.05
4 278.97 280.03 283.19 288.37 295.48 304.37 314.90 326.90 340.22 354.72 370.24
5 278.97 280.03 283.19 288.37 295.48 304.37 314.90 326.90 340.22 354.72 370.24
6 617.64 618.47 620.97 625.10 630.85 638.15 646.97 652.25 668.90 681.86 696.06
7 617.64 618.47 620.97 625.10 630.85 638.15 646.97 657.25 668.90 681.86 696.06
1156.7
1156.7
1297.0
8 1086.8 1087.5 1089.7 1093.3 1098.3 1104.7 111205 1121.6 1132.0 1143.8
9 1086.8 1087.5 1089.7 1093.3 1098.3 1104.7
255.6
1112.5
1261.7
1121.6
1269.0
1132.0
1277.3
1143.8
1286.6 10 1241.5 1242.0 1243.7 1246.6 1250.5 1
Ranjan and Ghosh / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 74-89
80
Figure 3(c). Mode 3 of stationary plate with one nodal point at centre, Freq. = 208.08 Hz
Figure 3. (d) Mode 4 of stationary plate with two nodal diameters, Freq. = 278.97 Hz
Figure 3 (e) Mode 5 of stationary plate with two nodal diameters, Freq. = 278.97 Hz
Ranjan and Ghosh / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 74-89
81
Figure 3 (f) Mode 6 of stationary plate with three nodal diameters, Freq. = 617.64 Hz
Figure 3. (g) Mode 7 of stationary plate with three nodal diameters, Freq. = 617.64 Hz
Figure 3 (h) Mode 8 of stationary plate with four nodal diameters, Freq. = 1086.8 Hz
Ranjan and Ghosh / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 74-89
82
Figure 3 (i) Mode 9 of stationary plate with four nodal diameters, Freq. = 1086.8 Hz
Figure 3 (j) Mode 10 of stationary plate with one nodal circle, Freq. = 1241.5 Hz
Another configuration with 16 patches as shown in Figure 1 has been examined. The drop in modal frequencies of all the modes
is reported in Table 4 for all the rotational speeds. This is because greater number of patches covers the higher strain region more
effectively. This suggests that sixteen-patch configuration can be used to control the vibration of spinning disk. It is interesting to
note that for 1000 rad/sec rotational speed, the fundamental mode changes its shape and becomes like umbrella similar to Figure
3(c) with highest strain region along the circumference in positive direction. It has been shown that the fundamental mode shape
of stationary disc attached to a core with piezoelectric patches fixed in the groove at the outer periphery in higher strain regions is
altered at a particular rotation speed of 1000 rad/sec due to increase in stiffness owing to rotation of the disk. This has considerable
bearing on modification of fundamental vibration modes of spinning circular disks with central core. In certain applications like
sensing and actuating devices, it may be necessary to have an axisymmetric mode shape of spinning disk for the fundamental
vibration mode (i.e. mode shape with no nodal diameter,
z
n = 0).
3.2. Natural frequencies and mode shapes with point masses
Table 5 represents the modal frequencies of the plate with eight point masses placed at the eight points A, B, C, D, E, F, G, H in
Figure 1.The modal frequencies for 1
st
, 2
nd
, 3
rd
, 5
th
, 6
th
, 7
th
, 8
th
and 10
th
decrease as additional point masses contribute further to
cause the drop in frequencies. The frequencies for 4
th
and 9
th
modes do not fall further due to addition of the point masses.
Table 6 presents the modal frequencies of the plate with 16 point masses placed at 16 locations from A to P in Figure 1. It is
clear that all the modal frequencies drop in comparison to Table 5 for all the speeds. This suggests that the 16 point masses may be
able to contain the vibration of the plate for all the ten modes.
It is also clear that with 8 patches, all the modal frequencies drop whereas with eight point masses, 4
th
and 9
th
modal frequencies
remain unaffected. However with 16 patches or point loads, all the modal frequencies fall. It is also noticed that with increase in
number of patches or point masses, the reduction in the modal frequencies for the higher modes are greater in case of patches than
the point masses.
83 Ranjan and Ghosh / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 74-89
4. Forced vibration response and vibration control of rotating plate
Forced vibration response and its control through actuation of some of the piezo-electric patches are shown in Figures 4-6.
Figure 4 shows the comparison of transverse vibration response of stationary plate without and with eight discrete point masses
and eight distributed patches attached at its periphery. As can be seen that all the response peaks do not show reduction in
magnitude e.g. in case of patches fourth peak does not show reduction whereas in case of masses third peak does not reduction.
This may be due to arbitrary location of measurement chosen. However, first two resonance peaks are reduced significantly in case
of plate with patches than plate with masses. Three patches 9, 11 and 13 of the stationary plate with eight patches are then actuated
by 0.1 N while patch 2 is actuated by 0.05 N. The magnitude of actuating forces has been selected so that the controlling forces
are within ten percent of the disturbing force which in this case is 1.0 N. The location of the control forces are also selected
randomly. However, the locations have been chosen so that actuating forces are in a quadrant opposite to the disturbing force. It is
presumed that in general this is likely to give required vibration control. Since the system is linear this approach would be
applicable to other locations of disturbing forces also. The corresponding vibration response is shown in Figure 5. From Figure 5
it is quite evident that amplitudes at resonant peaks 1, 3 and 4 drop due to actuation while the amplitude at 2
nd
resonant peak
almost remains same. Further, all the four peaks of the plate after actuation in Figure 5 are also lower than the corresponding peaks
of the stationary plate without patch / mass in Figure 4. This implies that actuating these four patches improves the vibration
control of the stationary plate having eight patches in the frequency range of 0-800 Hz.
Ranjan and Ghosh / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 74-89
84
Table 3: Modal frequencies of the plate with eight patches for different rotational velocities, ( rad / sec)
Mode 0 = 100 = 200 = 300 = 400 = 500 = 600 = 700 = 800 = 900 = 1000 =
1 162.34 163.21 165.78 169.99 175.70 182.78 191.07 200.42 210.68 221.73 223.45
2 162.34 163.21 165.78 169.99 175.70 182.78 191.07 200.42 210.68 221.73 233.45
3 195.57 196.02 197.37 199.60 202.68 206.56 211.18 216.50 222.44 228.96 235.99
4 256.31 257.45 260.83 266.38 273.95 283.38 294.50 307.13 321.08 336.19 352.31
5 256.31 257.45 260.83 266.38 273.95 283.38 294.50 307.13 321.08 336.19 352.31
6 549.85 550.75 553.46 557.95 564.17 572.07 581.57 592.61 605.09 618.93 634.04
7 549.85 550.75 553.46 557.95 564.17 572.07 581.57 592.61 605.09 618.93 634.04
8 875.94 876.72 879.08 882.99 888.44 895.40 903.83 913.69 924.94 937.52 951.37
9 1035.5 1036.3 1038.7 1042.7 1048.2 1055.3 1064.0 1074.0 1085.6 1098.5 1112.7
10 1198.2 1198.3 1200.7 1203.8 1208.1 1213.6 1220.4 1228.3 1237.4 1247.6 1258.9
Table 4: Modal frequencies of the plate with sixteen patches for different rotational velocities, ( rad / sec)
Mode 0 = 100 = 200 = 300 = 400 = 500 = 600 = 700 = 800 = 900 = 1000 =
1 154.06 154.97 157.68 162.10 168.08 175.46 184.08 193.76 204.36 215.72 227.65
2 154.06 154.97 157.68 162.10 168.08 175.46 184.08 193.76 204.26 215.72 227.74
3 184.99 185.48 186.92 189.30 192.58 196.70 201.60 207.21 213.47 220.30 227.74
4 238.62 239.83 243.42 249.29 257.29 267.22 278.87 292.05 306.54 322.19 338.81
5 238.62 239.83 243.42 249.29 257.29 267.22 278.87 292.05 306.54 322.19 338.81
6 501.14 502.12 505.03 509.85 516.52 524.98 535.12 546.87 560.12 574.76 590.70
7 501.14 502.12 505.03 509.85 516.52 524.98 535.12 546.87 560.12 574.76 590.70
8 799.60 800.45 803.00 807.23 813.11 820.61 829.68 840.27 852.33 865.79 880.59
9 898.67 899.54 902.15 906.48 912.51 920.20 929.51 940.39 952.79 966.65 981.91
10 1164.6 1165.3 1167.3 1170.7 1175.4 1181.5 1188.8 1197.4 1207.3 1218.4 1230.7
Ranjan and Ghosh / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 74-89
85
Table 5. Modal frequencies of the plate with eight point masses for different rotational velocities, ( rad / sec)
Table 6: Modal frequencies of the plate with sixteen point masses for different rotational velocities, ( rad / sec)
Mode 0 = 100 = 200 = 300 = 400 = 500 = 600 = 700 = 800 = 900 = 1000 =
1 161.11 161.98 164.57 168.80 174.55 181.67 190.00 199.39 209.69 220.78 232.53
232.53 2 161.11 161.98 164.57 168.80 174.55 181.67 190.00 199.39 209.69 220.78
3 195.10 195.56 196.92 199.16 202.25 206.15 210.80 216.14 222.11 228.65 235.71
4 258.90 260.02 263.24 268.79 276.24 285.53 296.49 308.94 322.71 337.64 353.57
5 258.90 260.02 263.34 268.79 276.24 285.53 296.49 308.94 322.71 337.64 353.57
6 564.04 564.91 567.51 571.81 577.78 585.37 594.52 605.14 617.17 630.38 645.12
7 564.04 564.91 567.51 571.81 577.78 585.37 594.52 605.14 617.17 630.38 645.12
8 901.59 902.32 904.49 908.11 913.14 919.52 927.37 936.50 946.93 958.60 971.48
9 1086.8 1087.6 1089.9 1093.8 1099.2 1106.0 1114.4 1124.2 1135.4 1148.0 1161.9
10 1165.5 1166.1 1168.0 1171.1 1175.4 1181.0 1187.7 1195.6 1204.7 1214.9 1201.3
Mode 0 = 100 = 200 = 300 = 400 = 500 = 600 = 700 = 800 = 900 = 1000 =
1 152.00 152.92 155.66 160.12 166.16 173.62 182.31 192.06 202.72 214.15 226.23
226.23 2 152.00 152.92 155.66 160.12 166.16 173.62 182.31 192.06 202.72 214.15
3 184.24 184.72 186.18 188.59 191.89 196.05 200.98 206.64 212.94 219.82 227.22
4 242.61 243.78 247.26 252.97 260.74 270.40 281.77 294.63 308.80 324.12 340.42
5 242.61 243.78 247.26 252.97 260..74 270.40 281.77 294.63 308.80 324.12 340.42
6 523.43 524.34 527.05 531.53 537.75 545.63 555.12 566.11 578.54 592.31 607.32
7 523.43 524.34 527.05 531.53 537.75 545.63 555.12 566.11 578.54 592.31 607.32
8 838.33 839.09 841.35 845.11 850.35 857.04 865.14 874.61 885.41 897.49 910.79
1057.7 9 981.53 982.33 981.71 988.64 994.13 1001.1 1009.6 1019.6 1031.0 1043.7
10 1117.5 1118.1 1120.2 1123.5 1128.3 1134.3 1141.6 1150.2 1160.0 1171.1 1183.3
Ranjan and Ghosh / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 74-89
86
Ranjan and Ghosh / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 74-89
87
Further, consider the case of the plate with eight patches rotating at 500 rad / sec with four patches 2, 9, 11 and 13 actuated by a
force 0.1 N. The vibration response of the plate is shown in Figure 6. It is noticed from Figure 6 that except 2
nd
peak the vibration
amplitude at all other resonance peaks drop significantly due to actuation. The magnitude of the 2
nd
peak in Figure 6 marginally
increases after actuation. The effect of actuation is evident and clearly indicates that actuation of patches may be used as a means
of vibration control of the rotating plate in the entire frequency range of 0-800 Hz.
Further, the speed of rotation of the plate with eight patches is increased to 800 rad /sec. The patches 2, 9 and 13 are actuated by
0.1 N and patch 11 is actuated by a force of 0.05N. The vibration response is shown in Figure 7. It is noticed from Figure 7 that
all the peaks drop due to actuation. Further the peaks due to actuation in Figure 7 are lower than the corresponding peaks of the
plate without patch / mass and rotating at 800 rad / sec in Figure 7. This further shows the effectiveness and demonstrates that
vibration control can be successfully achieved through actuation of some of the patches in a wide range of the frequencies between
0-800 Hz covering several natural frequencies. It is important to note that other combination of actuating forces and patch
locations may also be possible for effective control of the vibration of the plate.
5. Conclusions
Natural frequencies of thin spinning plate attached to a rigid core increase significantly with increase in speed. Further discrete
patches of distributed masses and discrete point masses when attached to the plate have significant influence on the mode shapes
and modal frequencies of the spinning plate. By locating the patches / point masses at higher deflection points / regions, vibration
absorption can be achieved. The proposed eight or sixteen discrete patches configuration or the sixteen point masses configuration
can be used for controlling the first ten modes of the spinning disk. The reduction in frequencies of higher modes is more in case
of plate with a definite number of patches in comparison to the plate with same number of point masses. However, number of the
discrete patches or point masses required and their locations depend on the modal frequencies at which vibration control is desired.
It has also been shown that discrete patches of piezoelectric ceramics can also be used to actively control the vibration of the
spinning disc. It has also been shown that the fundamental mode shape of the stationary disc attached to a central core having
piezoelectric ceramic patches fixed in the groove at the outer periphery in higher strain regions is altered at a high rotation speed of
1000 rad/sec due to increase in stiffness owing to rotation of the disk. This has considerable bearings on modification of
fundamental vibration modes of spinning circular discs with a central core. In certain applications like sensing and actuating
devices, it may be necessary to have an ax symmetric mode shape of spinning disc for the fundamental vibration mode (i.e. mode
shape with no nodal diameter, n = 0).
References
Barasch, S. and Chen, Y., 1972. On the vibration of rotating disk, American Society of Mechanical Engineers, Journal of Applied
Ranjan and Ghosh / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 74-89
88
Mechanics, Vol. 39, pp. 1143-1144.
Bashmal, S., 2009, In-plane free vibration of circular annular disks, Journal of Sound and Vibration, Vol. 322, pp. 216226
Bauer H.F. and Eidel W. , 2007, Transverse vibration and stability of spinning circular plates of constant thickness and different
boundary conditions, Journal of Sound and Vibration, Vol. 300, pp. 877895.
Chan, J.-S, 2003. Vibration control of a spinning disk, International Journal of Mechanical Sciences, 45, pp. 1269-1282.
Chen S. C. and Bogy, D. B., 1992. Effects of load parameters on the natural frequencies and stability of flexible spinning disk
with a stationary load system, American Society of Mechanical Engineers, Journal of Applied Mechanics, Vol. 58, pp. 230- 235.
Crawley EF, Javier, D.L., 1987. Use of Piezoelectric actuators as elements of intelligent structures, AIAA Journal, pp. 1773-1785.
Cook, R.D., 1981. Concepts and Applications of Finite Element Analysis, Second Edition, John Wiley and Sons, New York,
Curadelli,, R. O., Ambrosini R.D., Danesi, R.F., 2004. Vibration control by attaching masses to a plate excited by rotary
machinery, Journal of Sound and Vibration, Vol. 273 , pp. 1087-1100.
Duan, W.H., Wang C.M., Wang C.Y., 2008. Modification of fundamental vibration modes of circular plates with free edges,
Journal of Sound and Vibration, Vol. 317, pp. 709715.
Eversman, W. and Dodson, R. O.,1969. Free vibration of a clamped spinning circular disk, American Institute of Aeronautics and
Astronautics Journal, Vol. 7, pp. 2010-2012.
Ewan, W. D. and Moeller, T. I., 1976. The stability of a spinning elastic discs with transverse load system, American Society of
Mechanical Engineers, Journal of Applied Mechanics, Vol. 43, pp. 485-490.
Ferretti, G, Magnani, G., Rocco, P., 2002. Modeling and experimental analysis of the vibrations in hard disk drives, IEEE/ASME
Transactions on Mechatronics, Vol. 7, No. 2, pp.152-159.
Grimes, R.G., Lewis J.G., Simon H.D., 1994. A Shifted Block Lanczos Algorithm for Solving Sparse Symmetric Generalized
Eigen problems, SIAM Journal Matrix Analysis Applications, Vol. 15, No. 1, pp. 228-27.
Huang, S. C. and Yu, S.C., 1990. On the free vibration of rotating annular plates elastically restrained at discrete locations, Journal
of Chinese Society of Mechanical Engineers, 11, pp.488-498.
Huang, S.C. and Wu, C.H., 1998. Frequency Analysis of rotating plate with external beam supports, Journal of Sound &
Vibration, Vol. 210, No. 4, pp.415-429.
Koo K.-N., 2006, Vibration analysis and critical speeds of polar orthotropic annular disks in rotation, Composite Structures, Vol.
76, pp 67-72.
Lakshminarayana, H. V., 1986. Finite element analysis of rotating laminated composite annular disks, Composites, 17, pp.42-48.
Lamb, H. and Southwell, R.V., 1922. The vibration of spinning disk, Proceedings of the Royal Society, U.K., 99, pp. 272-280.
Liang, D. -S., Wang H. J., Chen L. W, 2002. Vibration and stability of rotating polar orthrotropic annular disks subjected to a
stationary concentrated transverse load, Journal of Sound and Vibration, Vol. 250, No. 5, pp.795-811.
Lee, H.P. and Ng, T.Y., 1995. Vibration and critical speeds of a spinning annular disk of varying thickness, Journal of Sound and
Vibration, Vol. 187, No. 1, pp. 39-50.
Rajakumar, C. and Rogers, C.R., 1991. The Lanczos algorithm applied to unsymmetric generalized eigenvalue problems,
International Journal for Numerical Method in Engineering, Vol. 32, pp.1009-1026.
Ramaiah, G. K., 1981. Natural frequencies of spinning annular plates, Journal of Sound and Vibration, 74, pp.303-310.
Shahbab, A. A. S., 1993. Finite element analysis for the vibration of variable thickness disks, Journal of Sound and Vibration, 162,
pp. 67-68.
Sinha, S. K., 1998a. Free vibration of a thick spinning annular disk with distributed masses at the outer edge, Journal of Sound and
Vibration, Vol. 122 pp. 217-231.
Sinha, S. K., 1998b. On free vibration of a thin spinning disk stiffened with an outer reinforcing ring, American Society of
Mechanical Engineers, Journal of Vibration, Acoustics, Stress and Reliability in Design, 110, pp.507-514.
Southwell, R.V., 1922. On the free vibrations of a uniform circular disc clamped at its center and on the effects of rotation,
Proceedings of the Royal Society, U.K., 101, pp. 133-153.
Wang, C.Y., 2005. Vibration of a circular plate with an attached core, Journal of Sound and Vibration, 280 , pp.1075-1082.
Biographical notes
Dr. Vinayak Ranjan received his graduate degree in Mechanical Engineering in 1991 from Bihar University, India and Ph.D. in Mechanical
Engineering from Banaras Hindu University, Varanasi, India in 2006. Dr. Ranjan joined as a Scientist (Mechanical Engineering) in Indian
Council of Agricultural Research, India in 1994 and remained at various scientific positions till 2007. He has executed a number of projects
sponsored by ICAR and industries. Dr. Ranjan joined as Assistant Professor in the Department of Mechanical Engineering and Mining
machinery Engineering, Indian school of Mines, Dhanbad, India in 2008. Dr. Ranjans research interest areas are: Finite element analysis,
Vibration, and Dynamics of Mechanical Systems.
Prof. M.K. Ghosh received his B.S. (Mechanical Engineering) in 1966 from Banaras Hindu University, Varanasi, India, M.Tech in 1968 and
Ph.D. in 1979 in Mechanical Engineering respectively from Indian Institute of Technology, Kharagpur, India. Professor Ghosh joined as a faculty
member in 1969 in the Department of Mechanical Engineering, Institute of Technology, Banaras Hindu University, and Varanasi, India and is a
Professor since 1993. He has worked as a NRC Research Associate during 1982-1984 as a senior NRC Research Associate during 1989-90 at the
Ranjan and Ghosh / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 74-89
89
NASA Lewis Research Center, Cleveland, OH (USA). He was a Professor of Mechanical Engineering in the Department of Mechanical
Engineering at I.I.T. Kharagpur during 1987-88. Professor Ghosh is a member of the American Society of Mechanical Engineers since 1984, a
Life Member of the Tribology Society of India (TSI) and was a member of the Executive Committee of the TSI during 1995-97. He is a member
of the honorary editorial board of the journal "Advances in Vibration Engineering", the scientific journal of the Vibration Institute of India. He
is a recipient of the University Grants Commission of India "Career Award" in Engineering in 1984. Professor Ghosh's research interest areas
are: Tribology, Vibration, and Dynamics & Control of Mechanical Systems. He has published about 45 papers in peer reviewed international
journals e.g. Tribology Transactions of ASME, Proc. of I. Mech. E., U.K., Tribology International, Int. J. of Wear, Int. J. of Mechanical Sciences,
Int. J. of Machine Tools & Manufacturing etc. Professor Ghosh has supervised about 35 graduate thesis and 6 Ph.D. students.
Received July 2009
Accepted September 2009
Final acceptance in revised form September 2009
MultiCraft
International Journal of Engineering, Science and Technology
Vol. 1, No. 1, 2009, pp. 90-105
INTERNATIONAL
JOURNAL OF
ENGINEERING,
SCIENCE AND
TECHNOLOGY
www.ijest-ng.com
2009 MultiCraft Limited. All rights reserved
Comparative performance analysis of Thyristor and IGBT based induction
motor soft starters
Ahmed Riyaz
1
, Atif Iqbal
1
*, Shaikh Moinoddin
1
, SK. MoinAhmed
1
,
Haitham Abu-Rub
2
1
* Department of Electrical Engineering, Aligarh Muslim University, Aligarh (UP), India
2
Electrical & Computer Engineering programme, Texas A&M University at Qatar, Doha, Qatar
*
Corresponding author (e-mail: atif_iqbal1@rediffmail.com)
Abstract
In this paper the performance of a split-phase motor and a three-phase induction motor drives for soft starting is evaluated. The
paper systematically investigates and compares the characteristics of a variable voltage fed induction motor drive for two
different types of soft starters; one based on IGBT and another based on Thyristor. Experimental validation is done using
analogue circuit based system, developed in the laboratory. The novelty of the work lies in the development of simple and
flexible models for simulation purpose and their experimental validation.
Keywords: Variable speed electric drives, Soft starter, Thyristor, IGBT
1. Introduction
Direct on-line starting of large induction motors poses a great problem to the grid as well as the motor itself due to sudden
voltage dips and large fluctuating torque, Willams and Griffth (1978), Bruce et al. (1984) and Nevelsteen and Aragon (1989).
Some time due to tripping of under-voltage and over-load relays the motor may fail to start. Alternatively reduced voltage starting
is being used especially for starting of large induction machines using auto-transformers. However, the volume and the cost of
auto-transformers limit its use. Power Electronics converter based induction motor starters more commonly called soft starter are
becoming popular and are rapidly replacing the conventional reduced voltage starters using auto-transformers, Shephard (1976),
Mazda (1973), Mozder and Bose (1976), Rashid (1993), Rowan and Merrison (1991), Lipo (1971a) and Lipo (1971b). Two
topologies of soft starters are more commonly employed
Back-to-back connected thyristor based soft starters also called ac voltage regulators.
Insulated Gate bipolar transistor (IGBT) and gate turn-off thyristor (GTO) based dc link converter.
Thyristor based soft starters are economical, simple, and reliable. They can be economically employed in medium voltage large
induction motor starting where the starting torque requirement is low. By using power electronics converter based starting, the
initial inrush current of motor can be reduced significantly. Additionally they offer smooth acceleration, ease in implementation of
current control, and energy savings with a partial load can be available. The thyristors are initially triggered at a large firing angle
(reduced voltage) when soft started, followed by decrease in the firing angle and subsequently increase in the applied voltage to
the motor. The initial starting torque at standstill of an induction motor is proportional to the square of the applied voltage. Thus
this surely affects the starting of induction motor on load. Additionally the output voltage quality is poor with high distortion and
poor power factor. The problem of higher THD is more prominent at large firing angle. This method is highly efficient if the motor
is working under light load condition as demonstrated in Eltamaly et al. (2007a), Eltamaly et al. (2007b), Vazquez et al. (2008)
and Hamed and Chalmers (1990).
Numerous attempts have been made worldwide to improve the performance of the soft starters and a number of different
variants are evolved, Jang and Choe (1998), Deleroi et al. (1989), Zenginobuz et al. (2004), Gastli and Ahmed (2005), and Kashif
and Saqib (2007). A step-up/down ac voltage regulator is used by Hyun and Choe (1998), in which the transformer with tap
Riyaz et al./ International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 90-105 91
changer and PWM chopper, is reported. The proposed regulator topology can step up or down the output voltage by the tap
changer of transformer and offers accurate output voltage to load by adjusting the PWM chopper. The major improvement with
this scheme is the continuous ac side input current in contrast to the conventional ac regulator where the input current happens to
be discontinuous. In Deleroi et al. (1989) a switching function is defined based on triggering angle of the thyristor and is
implemented in the controller. The transient response is improved considerably with such technique. The performance of induction
motor starter using soft starter is improved considerably by Zenginobuz (2004) where the starting is optimized in such as way to
eliminate the supply frequency torque pulsation. Artificial intelligence based techniques have been employed to improve the
starting characteristics by Gastli and Ahmed (2005) and Kashif and Saqib (2007). More recently Zhong et al. (2009) have adopted
the pump control soft start scheme to improve the starting behaviour of induction machine. PWM signals are generated by the
control unit which triggers the thyristors to keep the current within the specified limit. The pump control is a torque control method
which uses PID control to adjust the motor torque. On the other hand Bernatt et al. (2009), have proposed special design of rotor
cage of large induction machine for better starting performance. Eltamaly et al. (2009) have proposed digital firing scheme of
triacs based on the current shape of the input. The sensor detects the operating mode and accordingly the triacs gate pulses are
generated. Charles and Bhuvameswari (2009) have utilised shunt active filter between the ac voltage controller and input grid
supply to mitigate the power quality problem arises due to the ac voltage controller acting alone.
This paper focuses on the classical topology of the soft starter based on thyristor and IGBT power switches for both single-phase
and three-phase induction motors. A complete simulation model is obtained using Matlab/Simulink software for the proposed
topologies. Then a prototype analog circuit based soft starter is developed in the laboratory and the simulation results are verified
using experimental implementation of the same. Thus the major objectives of the paper include the development of simple and
flexible simulation model for the soft starters of induction machine. The behaviour of the starters can be easily verified by the
proposed simulation models. The paper is organised in five sections; the second section deals with the single-phase system, the
second section details the implementation of three-phase drive system, the third section details the experimental investigation
followed by the conclusion and references.
2. Matlab
/Simulink
model development
In the modern era, almost all the processes and techniques are at first simulated before their actual real time implementation.
This reduces a significant portion of effort and cost of real time implementation. The performance of the proposed
system/process/technique can be evaluated accurately by using proper simulation models. Thus the models should be flexible and
accurate to take into account the real time implementation issues as well. With the rapid development in computer hardware and
software, new simulation packages which are much faster and more user friendly are now available. This paper discusses the use
of one such software, the SIMULINK software of MATLAB. The main advantage of the SIMULINK over other programming
software is that, instead of compilation the simulation model is built systematically by means of function blocks. A set of machine
differential equations can be modelled by interconnection of appropriate blocks, each of which performs a specific mathematical
operation. Programming efforts are drastically reduced and the debugging of errors is easy. Since the SIMULINK is model
operation programme, the simulation model can be easily developed by addition of new sub-models to cater for various control
functions. The induction motor could be incorporated in the complete electric motor drive system from Simulink Library. Simulink
modelling of single-phase and three-phase voltage regulators of two types i.e., thyristor and IGBT based are done and the results
are evaluated in the subsequent sections.
2.1 Matlab
/Simulink
+ =
(
(
(1)
Note that V
o
can be varied from V
s
to 0 by varying from 0 to . The rms value of load current
R
V
I
o
o
= (2)
The input power factor
( )
2 1
2
2
1
/
s
o o
sin
V
V
VA
P
(
+ = =
(3)
The average thyristor current
( ) ( )
t d t sin V
R
I
s Thy , A
2
2
1
(4)
As each thyristor carries half the line current, the rms current in each thyristor is
2
o
Thy , o
I
I =
(5)
2.2 Matlab
/Simulink
+ =
(
(
(
(7)
The input current of Fig. 9 can be expressed in Fourier series as
( ) ( )
=
+ =
... , , n
n s
t n sin t i
5 3 1
2 (8)
where
0
1
= =
n
n
n
b
a
tan (9)
The rms value of the nth harmonic input current is given as
( ) |
\
|
= + =
2
2 2
2
1
2 1
2 2
n
sin I
n
b a I
a
/
n n sn
(10)
The rms value of the fundamental current is
|
\
|
=
2
2 2
1
sin I I
a s
(11)
The rms input current is found as
a s
I I = (12)
Harmonic factor
( ) ( )
2 1
2 1
2
1
1
1 4
1
/
/
s
s
cos I
I
HF
(
=
(
(
|
|
\
|
=
(13)
Displacement factor
Riyaz et al./ International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 90-105 97
( ) 1
1
= = cos DF (14)
Power factor
|
\
|
=
|
|
\
|
=
2
2 2
1
sin DF
I
I
PF
s
s
(15)
The advantage of using IGBTs is that due to same extinction angle as that of firing angle, the fundamental current in this case is
in phase with the voltage, hence displacement factor becomes unity. A plot of the power factor vs firing angle is illustrated in Fig.
10. It reveals that the power factor varies significantly with change in the firing angle especially for IGBT based soft starter.
Nevertheless, it is higher for thyristor based system.
Figure 10. Power factor vs firing angle for Thyristor and IGBT based soft starters.
A point must be noted here that the power factor vs firing angle curve shown above is of the case when no filter is incorporated
in the system, if the filters are also incorporated in the system, THD of the system reduces significantly and as already discussed
above displacement factor is unity for this case, hence overall power factor improvement will be there. Further, since the
harmonics in starters using IGBTs are of higher order, L and C requirement will be of lower value, hence of smaller size, on the
other hand starters with thyristors have lower order harmonics requires filters with larger value of L and C which are of large size
and non-economical in terms of both cost and area requirement. Thus while comparing the two types of starter based on filter
requirements, it can be concluded that the IGBT based starter is a cheaper option due to smaller filter components requirement.
2.3 Matlab
/Simulink
Model of Three-phase soft Starter using Thyristor
The basic three-phase induction motor drive scheme is illustrated in Figure 11, where the power switches can either be thyristors
or IGBTs.
3-Phase
INDUCTION
MOTOR
Three-phase Soft Starter
Analog
Interface
Gate
Drive
Figure 11. Schematic diagram of soft-starter controlled induction motor drive.
The simulation model of three-phase soft starter for three-phase induction motor is developed as shown in Figure 12 for thyristor
based circuit. Three-phase sinusoidal voltages are generated using three independent voltage sources having phase difference of
Riyaz et al./ International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 90-105 98
120
0
and magnitude of 220Vrms. The output of soft starter is given to a three-phase induction motor. The outputs response of the
motor is recorded and presented in Figure 13-15.
The output voltage waveform of thyristor based three-phase voltage regulator is shown in Figure 9 for = /6. The controllable
range of voltage regulator is from firing angle, = 0 to = /2.The Total Harmonic Distortion (THD) for higher values of
increases, which indicates increased harmonic content in the line current and voltage.
stator_fl ux_qd
Vs_qd
Vr_qd
v
+
-
Vc1
Vc
v
+
-
Vb1
Vb
v
+
-
Va1
Va
g
a
k
Thyri stor6
g
a
k
Thyri stor3
Te
In1
Out1
Out2
Subsystem3
In1
Out1
Out2
Subsystem2
In1
Out1
Out2
Subsystem1
In1Out1
Subsystem
Step
Scope4
Scope3
Scope2
Scope
Rotor_speed(rpm)
Rotor_fl ux_qd
Rotor_angl e_theta_m
In1
Out1
Out2
Out3
Out4
Out5
Out6
Out7
Out8
Out9
Out10
Out11
Mahi nes
Measurment
Demux
Is_qd
Is_abc
Ir_qd
Ir_abc
pi /6
Fi ri ng Angl e
i
+
-
Current Measurement
Tm
m
A
B
C
3 ph IM
g
a
k
Thyri stor5
g
a
k
Thyri stor4
g
a
k
Thyri stor2
g
a
k
Thyri stor1
Figure 12. Simulink Model of Induction Motor fed from thyristor based soft starter
a.
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
0
200
400
600
800
1000
1200
1400
1600
Time (sec.)
S
p
e
e
d
( r
p
m
)
b.
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
0
1
2
3
4
5
6
7
8
9
Time (sec)
T
o
r
q
u
e
( N
m
)
c.
Figure 13. Induction motor response for = /6; a. output voltage, b. Speed Vs Time curve and c. Torque Vs Time curve for no-
load.
Riyaz et al./ International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 90-105 99
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
-2500
-2000
-1500
-1000
-500
0
500
Time (sec.)
S
p
e
e
d
(r
p
m
)
Figure 14. Speed vs. time curve for = /3 when load is applied at t=1.2sec.
0 200 400 600 800 1000 1200 1400 1600
-1
0
1
2
3
4
5
6
7
8
9
Speed (rpm)
T
o
r
q
u
e
(
N
m
)
for /6
for /4
for /3.1
Figure 15. Torque vs. speed curve for different value of firing angle at no-load.
The drive system shows a typical behaviour for different values of firing angle i.e., = /6, /4, /(3.1) . The motor is taking up
the load well with a reduction in the corresponding speed. There is a reduction in the peak torque for change in the firing angle as
expected. From Figure 13 it is observed that motor performs satisfactorily at = /6, but when the magnitude of is increased
beyond the value /3.1, the motor become unstable (as shown in Figure 14). Hence, the value of should be / (3), for a
thyristor based soft starter for stable operation of the drive system.
In Figure 15, the speed-torque characteristics of motor are shown under various loading conditions and the variations of torque
speed curves with respect to the applied voltage are shown. It can be noted that increasing the firing angle results in decrease in
starting torque. These curves show that the slip (hence speed) at maximum torque s remains same (in both transient as well as
steady state condition). The speed range for stable operation remains the same. The slope and shape of the characteristics are
dependent ultimately on the machine parameters. The curve shows both the transient as well as steady state characteristic. It is to
be noted here is that, if the applied voltage is reduced, the voltage across the magnetizing branch also comes down. This in turn
means that the magnetizing current and hence flux level are reduced. Reduction in the flux level in the machine impairs torque
production. If, however, the machine is running under lightly loaded conditions, then operating under rated flux levels is not
required. Under such conditions, reduction in magnetizing current improves the power factor of operation. Some amount of energy
saving may also be achieved.
It is also noted that the starting torque is also lower at lower voltages. Thus, even if a given voltage level is sufficient for
achieving the running torque, the machine may not start. This method of speed control is best suited for loads that require very
little starting torque. It is further noted that the torque response are identical under steady state conditions. Thus only the transients
are affected by the choice of the firing angle.
2.4 Matlab
/Simulink
=
=
n
i
w
LB
y
n
L
1
2
1
(1)
where
w
y represent response abrasive wear and n denotes the number of experiments.
The S/N ratio can be calculated as a logarithmic transformation of the loss function as shown below.
S/N ratio for wear = -10log
10
(
LB
L ) . (2)
In order to study the effect of various variables (factors) such as applied load, sliding velocity and fibre length and their
interactions on abrasive wear, a L
27
(3
13
) orthogonal array is chosen. Table 1 indicates the factors to be studied and the
corresponding levels of factors. Using linear graph shown in Figure 2, the factors are assigned to columns of the OA array. The
first column is assigned to load (A), the second column assigned to sliding velocity (B) and fifth column assigned to the fibre
length (D). The third and fourth column is assigned to the interaction of load and sliding velocity (AxB), the sixth and seventh
column to the interaction of sliding velocity and fibre length (BxD) and the eight and eleven column assigned to the interaction of
load and fibre length (AxD). The response (abrasive wear, weight loss in g) is transformed into signal to noise (S/N) ratio using
lower is better condition. An analysis of means (ANOM) was performed to establish statistically significant parameters (Glen,
1993).
Table 1. Levels of the variable used in the experiment
Level
Control factors I II III Units
A: Load 10 15 20 N
B: Sliding Velocity 15 30 45 cm/s
D: Fibre length 1 5 10 mm
Figure 2. Linear graph for L
27
orthogonal array
3. Particle swarm optimization
Particle Swarm Optimization (PSO) is one of the latest evolutionary optimization techniques based on the metaphor of the social
interaction and communication of bird flocking and fish schooling (Kennedy et al., 2001). In PSO, the members of the entire
population are maintained through the search procedure so that information is socially shared among individuals to direct the
search towards the best position in the search space. PSO has been recognized as an evolutionary computation technique and has
features of both genetic algorithms (GA) and evolution strategies (ES). It is similar to a GA in that the system is initialized with a
population of random solutions. However, unlike a GA each population individual is also assigned a randomized velocity, in
effect, flying them through the solution hyperspace. As is obvious, it is possible to simultaneously search for an optimum solution
in multiple dimensions. Compared to GA, the information sharing mechanism in PSO is significantly different. In GAs,
A(1)
B(2)
D(5)
(9)
(10) (12) (13)
(3,4)
(6,7)
(8,11)
Mahapatra and Chaturvedi / International Journal of Engineering, Science and Technology,
Vol. 1, No. 1, 2009, pp. 123-135
127
chromosomes share information with each other. So the whole population moves like a group toward an optimal area. In PSO,
only global best gives out the information to others. It is a one-way information sharing mechanism. The evolution only looks for
the best solution. Compared with GA, all the particles tend to converge to the best solution quickly even in the local version in
most cases. The advantages of PSO are that PSO is easy to implement and there are few parameters to adjust. However, PSO
initially converges fast and then becomes extremely slow. Moreover, there is a tendency of being trapped at local optimum point or
premature convergence. In this work, these issues have been resolved using a mutation strategy. Extensive experimental results
prove that PSO with mutation outperforms basic PSO and genetic algorithm (GA).
In PSO, each member is called particle, and each particle moves around in the multidimensional search space with a velocity
which is constantly updated by the particles own experience and the experience of the particles neighbours or the experience of
the whole swarm. Application of PSO algorithm requires that parameters be initialized and the initial population generated
randomly.
After evaluation, the PSO algorithm repeats the following steps iteratively:
Each particle with its position, velocity, and fitness value updates its personal best (best value of each individual so far) if
an improved fitness value is found.
The best particle in the whole swarm with its position and fitness value is, on the other hand, used to update the global
best (best particle in the whole swarm).
Then each particle updates its velocity based on the experiences of personal best and the global best in order to update the
position of each particle with the velocity currently updated.
After finding the personal best and global best values, velocities and positions of each particle are updated using Equations (3) and
(4) respectively.
v
ij
t
=
t-1
v
ij
t-1
+ c
1
r
1
(p
ij
t-1
- x
ij
t-1
) + c
2
r
2
(g
j
t-1
- x
ij
t-1
) (3)
x
ij
t
= x
ij
t-1
+ v
ij
t
(4)
where v
ij
t
represents velocity of particle i at iteration t with respect to j
th
dimension (j =1,2,n). p
ij
t
represents the position value
of the i
th
personal best with respect to the j
th
dimension. x
ij
t
is the position value of the i
th
particle with respect to j
th
dimension. c
1
and
c
2
are positive acceleration parameters, called cognitive and social parameter, respectively and r
1
and r
2
are uniform random
numbers between (0,1). w is known as intertia weight which is updated as w
t
= w
t-1
where is a decrement factor. The
parameter w controls the impact of the previous velocities on the current velocity. Termination criterion might be a maximum
number of iteration or maximum CPU time to terminate the search.
4. Results and discussions
The neat polyester has strength of 0.0753 GPa in tension. It is interesting to note that this value drops to 0.053 GPa when fibre
length is 1 mm. As the fibre length increases, the tensile strength of the composite improves to 0.0746 GPa for fibre length of 5
mm and again this value drops to 0.0645 GPa for fibre length of 10 mm (Figure 3). The decline in strength may be attributed to
weak chemical reaction at the interface between the fibre and the matrix unable to transfer the tensile stress. Chemical surface
treatment might result in strong interface bond and tensile strength may improve. It is well known that sugarcane fibre contains a
high percentage of lignin (21% on the average) which is considered to prevent a good surface wetabillity between polymer matrix
and natural fibres (Monteiro et al., 1998). More efficient chemical cleaning to the fibre would increase the bond strength and
consequently give different results. Another slightly possible reason could be due to experimental aspect related to the voids and
pores which strongly affect the mechanical properties. The hardness of polyester matrix is 85.57. It is evident from Figure 4 that
hardness values go on decreasing monotonically as the fibre length increases.
The data on responses viz., weight loss collected for 27 experimental runs as per combination of abrasive wear process
parameters. The experimental data on weight loss are converted into S/N ratios using Equation 2. The data are shown in Table 2.
Taguchi method is adopted to analyze the effect of various process parameters viz., A, B, and D on weight loss. The analyses are
made using the popular software specifically used for design of experiment applications known as MINITAB 14. Figure 5 shows
graphically the effect of the three control factors on weight loss.
Mahapatra and Chaturvedi / International Journal of Engineering, Science and Technology,
Vol. 1, No. 1, 2009, pp. 123-135
128
0
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
1 3 5 7 9 10
Fibre length in mm
T
e
n
s
i
l
e
s
t
r
e
n
g
t
h
i
n
G
P
a
Figure 3. Variation of tensile strength with fibre length
73.50
74.00
74.50
75.00
75.50
76.00
76.50
77.00
77.50
78.00
78.50
79.00
1 3 5 7 9 10
Fibre length in mm
H
a
r
d
n
e
s
s
s
h
o
r
e
D
Figure 4. Variation of hardness with fibre length
M
e
a
n
o
f
S
N
r
a
t
i
o
s
20 15 10
26
25
24
23
45 30 15 3 2 1
10 5 1
26
25
24
23
3 2 1 3 2 1
A B AXB
D BXD AXD
Mai n Ef f ect s Pl ot ( dat a means) f or SN r at i os
Signal-to-noise: Smaller is better
Figure 5. Effect of various factors on weight loss
Mahapatra and Chaturvedi / International Journal of Engineering, Science and Technology,
Vol. 1, No. 1, 2009, pp. 123-135
129
Table 2. Experimental data
Expt.
No.
Factor A
Load in N
(Column 1)
Factor B
Sliding
Velocity in
cm/s
(Column 2)
AXB
(Column
3)
Factor D
Fibre Length
in mm
(Column 5)
BXD
(Column 6)
AXD
(Column 8)
Response
Abrasive
wear in g
S/N
Ratio
1 10 15 1 1 1 1 0.090 20.9151
2 10 15 2 5 2 2 0.040 27.9588
3 10 15 3 10 3 3 0.078 22.1581
4 10 30 1 5 3 3 0.057 24.8825
5 10 30 2 10 1 1 0.030 30.4576
6 10 30 3 1 2 2 0.044 27.1309
7 10 45 1 10 2 2 0.072 22.8534
8 10 45 2 1 3 3 0.048 26.3752
9 10 45 3 5 1 1 0.032 29.8970
10 15 15 3 1 2 3 0.096 20.3546
11 15 15 1 5 3 1 0.056 25.0362
12 15 15 2 10 1 2 0.073 22.7335
13 15 30 3 5 1 2 0.070 23.0980
14 15 30 1 10 2 3 0.049 26.1961
15 15 30 2 1 3 1 0.064 23.8764
16 15 45 3 10 3 1 0.082 21.7237
17 15 45 1 1 1 2 0.061 24.2934
18 15 45 2 5 2 3 0.060 24.4370
19 20 15 2 1 3 2 0.092 20.7242
20 20 15 3 5 1 3 0.055 25.1927
21 20 15 1 10 2 1 0.061 24.2934
22 20 30 2 5 2 1 0.098 20.1755
23 20 30 3 10 3 2 0.057 24.8825
24 20 30 1 1 1 3 0.090 20.9151
25 20 45 2 10 1 3 0.084 21.5144
26 20 45 3 1 2 1 0.070 23.0980
27 20 45 1 5 3 2 0.080 21.9382
Figure 5 clearly shows that weight loss is highly sensitive to load (factor A). Concrete visualization of impact of various factors
and their interactions can be easily made with the help analysis of means as shown in the response table (Table 3). The last row
indicates the order of significance of factors and their interactions. It can be observed that the load, factor A ( Rank 1), fibre length,
factor D (rank 2) and sliding velocity, factor B (rank 3) have great influence on weight loss. Before setting the optimal levels of
factors for minimum weight loss, the interaction effects of factors need to be analyzed. As such, the interactions do not have much
influence in comparison to factors as evident from the response table. The interaction graphs are shown in Figures 6-8. The
interaction between load (factor A) and fibre length (factor D) ranks highest among all interactions. The interactions BXD and
AXB are ranked next to AXD interaction in the order of significance. Analysis of factors and their interactions helps to set the
optimal process parameters for minimum weight loss as A
1
B
2
D
2.
The factor plots suggest that weight loss increases monotonically
as the load (factor A) increases. As far as sliding velocity is concerned, the weight loss is more at low level and high level of
velocity (15 cm/s and 45 cm/s respectively) but comparatively less weight loss is observed at medium level of velocity i.e. 30 cm/s.
Similarly, weight loss first increases at low level of fibre length and then decreases at medium level of fibre length (5 mm) and
again increases at high level of 10 mm.
Mahapatra and Chaturvedi / International Journal of Engineering, Science and Technology,
Vol. 1, No. 1, 2009, pp. 123-135
130
Table 3. Response table for smaller is better characteristic
Level A B AXB D BXD AXD
1 25.85 23.26 23.48 23.08 24.34 24.39
2 23.53 24.62 24.25 24.74 24.06 23.96
3 22.53 24.01 24.17 24.09 23.51 23.56
Delta 3.32 1.36 0.77 1.66 0.82 0.83
Rank 1 3 6 2 5 4
A
S
N
r
a
t
i
o
s
20 15 10
28
27
26
25
24
23
22
B
45
15
30
I nt er act i on Pl ot ( dat a means) f or SN r at i os
Signal-to-noise: Smaller is better
B
S
N
r
a
t
i
o
s
45 30 15
28
27
26
25
24
23
22
21
20
D
10
1
5
I nt er act i on Pl ot ( dat a means) f or SN r at i os
Signal-to-noise: Smaller is better
Figure 6. Interaction plot of AXB Figure 7. Interaction plot of BXD
A
S
N
r
a
t
i
o
s
20 15 10
28
27
26
25
24
23
22
21
D
10
1
5
I nt er act i on Pl ot ( dat a means) f or SN r at i os
Signal-to-noise: Smaller is better
Figure 8. Interaction plot of AXD
After analyzing the factor and interaction effects and setting optimal process parameter levels, it is customary to develop a linear
predictive equation for confirmation and verification tests. Therefore, the predictive equation for optimal factor settings, A
1
B
2
D
2
can be represented as follows:
)] ( ) ( ) [( )] ( ) ( ) [(
) ( )] ( ) ( ) [( ) ( ) (
2 2 2 2 2 1 2 1
2 2 1 2 1 2 1
T D T B T D B T D T A T D A
T D T B T A T B A T B T A T
+
+ + + + + =
(5)
Predicted response
T Overall experimental mean
2 2 2 1 2 2 1 2 1 , , , , , D B D A D B A B A Mean values of factors and interactions at designated level
Similar equations can be developed for different levels of factor combination and prediction of weight loss can be made. Such a
prediction is known as Taguchi prediction.
The predicted S/N Ratio for factor setting A
3
B
3
D
2
is found to be 21.4451 and the experimental result with this setting is 21.9382.
Since the error between the predicted and experimental result is less than 1%, it is confirmed that Taguchi optimal setting is good
enough to be used for setting parameters for minimum weight loss. In order to predict the weight loss correctly, artificial neural
network (ANN) model based on back propagation algorithm has been implemented. The network architecture has three neurons at
the input layer to feed the normalized data of process parameters such as load, sliding velocity and fibre length. The output layer
contains only one node to predict the normalized weight loss. The number of neurons at the hidden layer, the momentum and
learning parameters are varied until the network converges to minimum root mean square (RMS) error. Out of 27 experimental
Mahapatra and Chaturvedi / International Journal of Engineering, Science and Technology,
Vol. 1, No. 1, 2009, pp. 123-135
131
data, fifteen are used for training and twelve are tested. The momentum and learning parameters are set at 0.20 and 0.30
respectively. Number of neurons at the hidden layer is 4. The number of epochs needed for training is 25805 when root mean
square (RMS) error is 0.01. The popular neural network software Neunet Pro 2.3 is used for neural network implementation. The
neural network prediction is shown in Figure 9. It is evident that the maximum error between experimental data and neural
network prediction is around 0.7%. Therefore, it can be concluded that neural network prediction has proceeded in correct manner.
.
Figure 9. Neural network prediction of experimental data
Figure 10. Estimation of optimal fibre length via neural network
The trained neural network is used to predict the optimal fibre length for minimum weight loss in abrasive wear. It has been
demonstrated that weight loss depends on the factors viz., the load, fibre length, and sliding velocity in the order of importance.
Among them, sliding velocity has least influence on weight loss. The factors such as load and sliding velocity are to be maintained
Mahapatra and Chaturvedi / International Journal of Engineering, Science and Technology,
Vol. 1, No. 1, 2009, pp. 123-135
132
at low and medium level i.e 10 N and 30 cm/s respectively. Therefore, neural network is tested with varying fibre length from 1
mm to 10 mm keeping the setting for load and sliding velocity at 10 N and 30 cm/s respectively. The predicted weight loss as the
fibre length varies from 1 mm to 10 mm is shown in Figure 10. It is found that the fibre length must be maintained at 8 mm to
obtain minimum weight loss in abrasion test.
A more rigorous optimization technique based on PSO is used to find out the optimal fibre length in abrasive wear. The
experimental data is used to develop an equation relating to weight loss with process parameters such as load, sliding velocity and
fibre length. Non-linear regression equation is developed using software known as Systat 7.0 as shown in Equation 6. The
coefficient of determination (r
2
) is found to be 0.994. The rationale behind using PSO to determine optimal fibre length lies on the
fact that it is one of the popular soft-computing techniques that generate optimal values with less computational efforts and
controlling few parameters.
ln y = ln0.024+0.559lnA-0.095lnB-0.149lnD (6)
where y is the response and A, B and D represents three process parameters.
The parameters of PSO are set as recommended by Kennedy et al. (2001). Both the cognitive (c
1
) and social parameter (c
2
)
parameters are set at 2.0. Inertia weight (w) and decrement factor () are taken 0.8 and 0.9 respectively. The algorithm is coded in
Turbo C
++
and run on Pentium IV desktop PC. The program was run 50 iterations and the results are noted down. The
mathematical formulation of the optimization problem is given as:
Maximize
Z = -f(x)
subjected to constraints:
A
min
< A < A
max
(7)
B
min
< B < B
max
(8)
D
min
< D < D
max
(9)
Z represents objective function value evaluated using Equation (14). The min and max in Equations 7-9 shows the lowest and
highest control factor settings used in this study (Table 1).
The optimal parameter settings are found to be load (A) equals 10.023 N, sliding velocity at 29.300 cm/s and fibre length 7.004
mm. It is to be noted that there is slight discrepancy in the order of 1 mm on optimum fibre length resulting from ANN prediction
and PSO results. Therefore, it is suggested that optimum fibre length varies from 7 to 8 mm beyond which the wear performance
of composites degrades. The convergence curve is shown in Figure 11.
2.92
2.94
2.96
2.98
3.00
3.02
3.04
3.06
3.08
1 4 5 8 31 40 50
Number of iterations
O
b
j
e
c
t
i
v
e
v
a
l
u
e
Figure 11. The convergence curve
4 Conclusions
Abrasive wear performance and mechanical property characterization of sugarcane fibre reinforced composites were determined
experimentally and the following conclusions were drawn:
Composites using chopped sugarcane fibre as reinforcement in polyester resin matrix have been successfully developed.
Mahapatra and Chaturvedi / International Journal of Engineering, Science and Technology,
Vol. 1, No. 1, 2009, pp. 123-135
133
Results of composites under consideration revealed that abrasive wear (weight loss) increases with increasing load
because high temperature is generated at higher load causing degradation of the composite surface. The factor, sliding
velocity, does not have significant contribution on weight loss. However, the size of the chopped sugarcane fibre plays an
important role on surface damage by abrasive wear.
Minimum resistance to wear is observed at lowest size of the fibre i.e 1mm. Again, at highest length of fibre i.e 10 mm,
the wear (weight loss) increases. The minimum wear is observed at fibre length of 5 mm. The reasons for such behaviour
may be explained as that the fibres do not get support from the matrix at lower length because of poor interface adhesion.
Therefore, the fibre has been pulled out easily from the matrix by the rubbing surface. As the fibre length becomes too
large, anisotropy of the composite might increase due to random orientation of the fibre causing increase in weight loss. If
the long fibre would have been used, the composite might have been behaved differently. In this study, untreated fibre has
been used but chemical treatment of the fibre may give rise to different result.
Taguchis experimental design is a simple and systematic way of analyzing a complex process with less experimental
runs. The optimum abrasive process parameters as suggested by Taguchi analysis is estimated as A
1
B
2
D
2
.
Although fibre length of 5mm gives rise to minimum weight loss, it may not be a global optimum point because Taguchi
method tests only few discrete points in the landscape of optimal surface. Therefore, a global optimum strategy known as
particle swarm optimization (PSO), a search procedure, has been implemented after developing a predictive equation
relating weight loss with the process parameters using non-linear regression method. The validity of the equation is tested
using coefficient of determination. The optimal parameter settings as found by PSO are given as: load equals 10.023 N,
sliding velocity at 29.300 cm/s and fibre length 7.004 mm.
A neural network has been implemented to predict the correct optimum fibre size and it is found to be 8 mm. The
discrepancy between the results of PSO and neural network may be attributed to difference in principle of prediction
method by two techniques rather than implementation aspects. Therefore, it is suggested that best size of fibre length for
minimum abrasive wear should be 7 to 8 mm.
The hardness of the composite monotonically decreases as the fibre length increases but tensile strength first increases
and then decreases as length of the fibre is increased. In contrary to common belief that hardness and tensile strength
improve wear resistance, it has been observed that parameters encountered in wear process strongly influence wear
resistance.
In future, the study can be extended to other natural fibres to find out the optimum fibre length. The abrasive wear
behaviour of chemically treated sugarcane fibre and aging effects of the fibre on abrasive behaviour of the composite can
be studied.
Notations
LLB
Loss function for lower-the-better quality characteristic
w
y Abrasive wear response
n Number of experiments
A Load in N
B Sliding velocity in cm/s
D Fibre length in mm
v
ij
t
Velocity of particle i at iteration t with respect to j
th
dimension
p
ij
t
Personal best position value of the i
th
particle with respect to the j
th
dimension
x
ij
t
Position value of the i
th
particle with respect to j
th
dimension at iteration t
g
j
t
Global best position value of the swarm with respect to j
th
dimension at iteration t
c
1
Cognitive parameter
c
2
Social parameter
r
1
and r
2
Uniform random numbers between (0,1)
w Intertia weight
Decrement factor
Predicted response
T Overall experimental mean
2 2 2 1 2 2 1 2 1 , , , , , D B D A D B A B A Mean values of factors and interactions at designated level
Z Objective function value
Mahapatra and Chaturvedi / International Journal of Engineering, Science and Technology,
Vol. 1, No. 1, 2009, pp. 123-135
134
References
Aggarwal, L.K., 1995. Bagasse-reinforced Cement Composites. Cement and Concrete Composites, Vol. 17, pp 107-112.
Arzondo, L.M., Perez, C.J., Carella, J.M., 2005. Injection molding of long sisal fiber-reinforced polypropylene: effects of
compatibilizer concentration and viscosity on fiber adhesion and thermal degradation. Polymer Engineering and Science, Vol.
45, pp 613-621.
Brahmakumar, M., Pavithran, C., Pillai, R.M., 2005. Coconut fibre reinforced polyethylene composites: effect of natural waxy
surface layer of the fibre on fibre/matrix interfacial bonding and strength of composites. Composite Science and Technology,
Vol. 65, pp 563-569.
Chand, N., Dwivedi, U.K., 2007. High stress abrasive study on bamboo. Journal of Material Processing Technology, Vol. 183, pp
155-159.
Chand, N., Dwivedi, U.K., Acharya, S.K., 2007. Anisotropic abrasive wear behaviour of bamboo (Dentrocalamus strictus). Wear
2007, Vol. 262, pp 1031-1037.
Choudhury, A., 2008. Isothermal crystallization and mechanical behavior of ionomer treated Sisal/HDPE composites. Material
Science and Engineering, Vol. 491, pp 492-500.
Dany, L., Adi, R.K., Szab, O.Z., Nagy, G., Mocz, J., Pukanszky, B., 1996. Wood flour filled PP composites: Adhesion,
deformation, failure. Polymers for Advanced Technologies, Vol. 17, pp 967-974.
El-Tayeb, N.S.M., Yousif, B.F., 2007. Evaluation of glass fibre reinforced polyester composite for multi-pass abrasive wear
applications. Wear, Vol. 262, pp 1140-1151.
El-Tayeb, N.S,M., 2008a. Two-body abrasive behaviour of untreated SC and R-G fibres polyester composites. Wear,
doi:10.1016/j.wear.2008.06.018.
El-Tayeb, N.S.M., 2008b. A study on the potential of sugarcane fibers/polyester composite for tribological applications. Wear,
Vol. 265, pp 223-235.
Glen, S.P., 1993. Taguchi Methods: A Hands on Approach. Addison-Wesley, New York.
Hepworth, D.G., Bruce, D.M., Vincent, J.F.V., Jeronimidis, G., 2000a. The manufacture and mechanical testing of thermosetting
natural fibre composites. Material Science, Vol. 35, pp 293-298.
Hepworth, D.G., Hobson, R.N., Bruce, D.M., Farrent, J.W., 2000b. The use of untreated hemp in composite manufacture.
Composites Part A, Vol. 31, pp 1279-1283.
Joseph, K., Thomas, S., Pavithran, C., 1996. Effect of chemical treatment on the tensile properties of short sisal fiber-reinforced
polyethylene composite. Polymer, Vol. 37, pp 5139-5145.
Joseph, S., Sreekalab, M.S., Oommena, Z., Koshyc, P., Thomas, S.A., 2002a. Comparison of the mechanical properties of phenol
formaldehyde composites reinforced with banana fibres and glass fibres. Composite Science and Technology, Vol. 62, pp 1857-
1868.
Joseph, P.V., Kuruvilla, J., Thomas, S., 2002b. Short sisal fibre reinforced polypropylene composite: the role of interface
modification on ultimate properties. Composite Interfaces, Vol. 9, No. 2, pp 171-205.
Kennedy, J., Eberhart, R.C., Shi, Y., 2001. Swarm Intelligence. Morgan Kaufmann, San Mateo, CA, USA.
Kumar, A.P., Singh, R.P., Sarwade, B.D., 2005. Stabilization of -sterilized biomedical polyolefins by synergistic mixtures of
oligomeric stabilizers. Part II. Journal of Material Chemistry and Physics, Vol. 92, 458-469.
Mohanty, S., Verma, S.K., Nayak, S.K., 2006. Dynamic mechanical and thermal properties of MAPE treated jute/HDPE
composites. Composite Science and Technology, Vol. 66, pp 538-547.
Monteiro, S.N., Rodriquez, R.J.S., De Souza, M.V., 1998. Sugarcane bagasse waste as reinforcement in low cost composites.
Advanced Performance Materials. Vol. 5, pp 183-191.
Mukhopadhyay, S., Srikanta, R., 2008. Effect of ageing of sisal fibres on properties of sisal-polypropylene composites. Polymer
Degradation and Stability, doi:10.1016/j.polymdegradstab.2008.02.018.
Pracella, M., Chionna, D., Anguillesi, I., Kulinski, Z. Piorkowska, E., 2006. Property improvement and degradability control of
polyolefin based composites containing natural fibres and mineral fillers. Composite Science and Technology, Vol. 66, pp 2218-
2230.
Pothana, L.A., Oommenb, Z., Thomas, S., 2003. Dynamic mechanical analysis of banana fiber reinforced polyester composites.
Composite Science and Technology, Vol. 63, No. 2, pp 283-293.
Rana, A.K., Mandal, A., Bandyopadhyay, S., 2003. Study on the interface modification of bagasse fibre and the mechanical
properties of its composite with PVC. Composite Science and Technology, Vol. 63, pp 801-806.
Satyanarayana, K.G., Pai, B.C., Sukumaran, K., Pillai, S.G.K., 1990a. Handbook of Ceramics and Composites, Marcel Decker,
New York.
Satyanarayana, K.G., Sukumaran, K., Mukherjee P.S., Pavithran, C., Pillai, S.G.K., 1990b. Natural Fibers-Polymer Composites,
Cement and Concrete Composites. Haworth Press, London.
Tong, J., Arnell, R.D., Ren, L., 1998. Dry sliding wear behaviour of Bamboo. Wear, Vol. 221, pp 37-46.
Mahapatra and Chaturvedi / International Journal of Engineering, Science and Technology,
Vol. 1, No. 1, 2009, pp. 123-135
135
Tong, J., Ma, Y., Chen, D., Sun, J., Ren, L. 2005. Effects of vascular fiber content on abrasive wear of bamboo. Wea,r Vol. 259,
pp. 37-46.
Varma, I.K., Ananthakrishnan, S.R.,Krishnamoorthi S., 1989. Comparison of glass/modified jute fabric and unsaturated polyester.
Journal of Composite Materials, Vol. 20, pp 383-390.
Biographical notes
Dr. S.S. Mahapatra is Professor in the Department of Mechanical Engineering, National Institute of Technology Rourkela, India. He has more
than 20 years of experience in teaching and research. His current area of research includes Multi-criteria Decision-Making, Quality Engineering,
Assembly Line Balancing, Group Technology, Neural Networks, and Non-traditional Optimization and Simulation. He has published more than
fifty papers in referred international journals. He has also presented more than one hundred research articles in national and international
conferences. He has written few books related to his research work. He is currently dealing with few projects sponsored by government of India.
Mr. Vedansh Chaturvedi is a final year student of stream M.Tech. (Production) in Mechanical Engineering, National Institute of Technology
Rourkela, India. He is pursuing his thesis work on parametric appraisal of fused deposition modeling process. His current area of research
includes DoE, Multivariate Analysis and Optimization Techniques. He has published three journal papers and presented five papers in
international conferences.
Received September 2009
Accepted November 2009
Final acceptance in revised form November 2009
MultiCraft
International Journal of Engineering, Science and Technology
Vol. 1, No. 1, 2009, pp. 136-147
INTERNATIONAL
JOURNAL OF
ENGINEERING,
SCIENCE AND
TECHNOLOGY
www.ijest-ng.com
2009 MultiCraft Limited. All rights reserved
Solving the K-of-N Lifetime Problem by PSO
Tzung-Pei Hong
1, 2
and Guo-Neng Shiu
3
1
Department of Computer Science and Information Engineering, National University of Kaohsiung, Kaohsiung, 811, Taiwan
2
Department of Computer Science and Engineering, National Sun Yat-sen University, Kaohsiung, 804, Taiwan
3
Department of Electrical Engineering, National University of Kaohsiung, Kaohsiung, 811, Taiwan
e-mail: tphong@nuk.edu.tw, lunasea.deen@msa.hinet.net
Abstract
In wireless sensor networks, application nodes (ANs) may be produced by different manufacturers and may own different data
transmission rates, initial energies and parameter values. When different kinds of ANs exist in a wireless network, it is hard to
find the optimal base-station (BS) locations for minimizing power consumption and prolonging network lifetime. In this paper, a
heuristic algorithm based on PSO is thus proposed to solve the K-of-N lifetime problem under general power-consumption
constraints in wireless sensor networks. Give N ANs, the network survives as long as there are at least K ANs alive. The
proposed approach can search for nearly optimal BS locations in heterogeneous sensor networks, where application nodes may
own different data transmission rates, initial energies and parameter values. Experimental results also show the good
performance of the proposed PSO approach and the effects of the parameters on the results. The proposed algorithm can thus
help maximize the K-of-N network lifetime in wireless sensor networks.
Keywords: wireless sensor network, network lifetime, energy consumption, particle swarm optimization
1. Introduction
In a wireless sensor network, sensor nodes (SNs) are usually small, low-cost and disposable, and do not communicate with other
sensor nodes. They are usually deployed in clusters around interesting areas. Each cluster of sensor nodes is allocated with at least
one application node. Application nodes (ANs) possess longer-range transmission, higher-speed computation, and more energy
than sensor nodes. The raw data obtained from sensor nodes are first transmitted to their corresponding application nodes. After
receiving the raw data from all its sensor nodes, an application node conducts data fusion within each cluster. It then transmits the
aggregated data directly to the base station or via multi-hop communication. The base station (BS) is usually assumed to have
unlimited energy and powerful processing capability. It also serves as a gateway for wireless sensor networks to exchange data and
information to other networks. Wireless sensor networks usually have some assumptions for SNs and ANs. For instance, each AN
may be aware of its own location through receiving GPS signals (Niculescu and Nath, 2003) and its own energy.
A fundamental problem in wireless sensor networks is to prolong the system lifetime under some given constraints. In the past,
many approaches were proposed to efficiently utilize energy in wireless networks. For example, appropriate transmission ways
were designed to save energy for multi-hop communication in ad-hoc networks (Heinzelman et al., 1999, 2000; Ramanathan and
Hain, 2000; Intanagonwiwat et al., 2000; Lee et al., 2001; Ye et al., 2002; Tian and Georganas, 2003). Good algorithms for
allocation of base stations and sensors nodes were also proposed to reduce power consumption (Rodoplu and Meng, 1999;
Ramanathan and Hain, 2000; Pan et al., 2003; Kawadia and Kumar, 2003; Li et al., 2003). Pan et al. proposed several algorithms
to find the optimal locations of base stations in two-tiered wireless sensor networks (Pan et al., 2005). Their approaches assumed
the initial energy and the energy-consumption parameters were the same for all ANs. If any of the above parameters were not the
same, their approaches could not work.
Recently, the technique of particle swarm optimization (PSO) technique has been widely used in finding nearly optimal
solutions for optimization problems (Eberhart and Kennedy, 1995a, 1995b; del Valle et al., 2008). Finding BS locations is by
nature very similar to finding the food locations originated from PSO. In the past, we proposed a PSO-based approach to allocate
Hong and Shiu / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 136-147
137
base stations for the N-of-N problem in which given N ANs, the mission fails when any of the N ANs runs out of energy (Hong
and Shiu, 2006). In this paper, we extend it to solve the K-of-N lifetime problem under general power-consumption constraints in
wireless sensor networks. Give N ANs, the network survives as long as there are at least K ANs alive. The proposed approach can
search for nearly optimal BS locations in heterogeneous sensor networks, where application nodes may own different data
transmission rates, initial energies and parameter values. Pan et al.s extension from solving the N-of-N problem to solving the K-
of-N problem was a little complicated. The extension by the proposed PSO approach is, however, very easy. Besides, the proposed
approach can handle different initial energy and energy-consumption parameters, which cannot be processed in Pan et al.s
approach. Extensive experiments are also made to show the performance of the proposed PSO approach for the K-of-N lifetime
problem and the effects of the parameters on the results.
The remaining parts of this paper are organized as follows. Some related works about finding the locations of base stations in a
two-tiered wireless networks is reviewed in Section 2. The PSO approach is introduced in Section 3. An algorithm based on PSO
to solve the K-of-N lifetime problem in a two-tiered wireless network is proposed in Section 4. A small example to illustrate the
approach is given in Section 5. Experimental results for demonstrating the performance of the algorithm and the effects of the
parameters are described in Section 6. Conclusions are stated in Section 7.
2. Review of related works
As mentioned above, a fundamental problem in wireless sensor networks is to maximize the system lifetime under some given
constraints. Pan et al. (2005) first proposed an algorithm to find the optimal locations of base stations in two-tiered wireless
homogeneous sensor networks for the N-of N problem. Give N ANs, the mission failed when any of the N ANs ran out of energy.
Let d be the Euclidean distance from an AN to a BS, and r be the data transmission rate. Pan et al. adopted the following formula
to calculate the energy consumption per unit time:
), ( ) , (
2 1
n
d r d r p + =
where
1
was a distance-independent parameter and
2
was a distance-dependent parameter. The energy consumption thus related
to Euclidean distances and data transmission rates. Pan et al. assumed each AN had the same
1
,
2
and initial energy. For
homogenous ANs, they showed that the center of the minimal circle covering all the ANs was the optimal BS location (with the
maximum lifetime) for the N-of-N condition.
Pan et al. (2005) then proposed an approach to find the optimal locations of base stations for the K-of-N problem. Give N ANs,
the mission survived as long as there were at least K ANs alive, or the mission failed when N-K+1 ANs ran out of energy. Pan et
al.s approach for the K-of-N problem was a little complicated when compared to that for the N-of-N problem. It first used the N-
of-N approach to find a minimal circle covering all the ANs. Each minimum circle would have at least two ANs (or three if the
ANs were not on the diameter) on the circumference according to the property of circles. The approach then found the (N-1)-of-N
lifetime by removing any one point on the circumference. If there were R points on the circumference, then R possible circles
would be found, each as a possible (N-1)-of-N result. The approach then found the (N-2)-of-N lifetime by removing one more ANs
from the (N-1)-of-N circles. The same procedure was repeated until only N-K ANs remained in the network. The approach then
selected the minimum circle. A smaller circle represented that the system had a larger lifetime.
Pan et al. (2005) also extended their approaches to find the optimal BS location for ANs with different transmission rates by
using stacked planes. But if the application nodes own different data transmission rates, initial energies and parameter values, their
approaches can not work.
3. Review of particle swarm optimization
The PSO technique was proposed by Eberhart and Kennedy (1995a, b) and has been widely used in finding solutions for
optimization problems. Some related researches about its improvement and applications has also been proposed (Shi and Eberhart,
1998; Gaing, 2003; Stacey et al., 2003; Qin et al., 2004). It maintains several particles (each represents a solution) and simulates
the behavior of bird flocking to find the final solutions. All the particles continuously move in the search space, tending to better
solutions, until the termination criteria are met. After a lot of generations, the optimal solution or an approximate optimal solution
is expected to be found.
Suppose there is a group of birds searching for food in an area. Also assume each bird has the instinct of remembering its own
location with the most food in its experience. Besides, all the birds can exchange their best experiences to each other and thus learn
the location with the most food found so far in the search space. Each bird thus knows both the best location in its experience and
the best one from the flock of birds, and will tend to move according to the two locations. The PSO approach thus simulates the
mentioned behaviors of bird flocking to find solutions of optimization problems.
When the PSO technique is used to solve a problem, each possible solution in the search space is called a particle, which is
similar to a bird mentioned above. All the particles are evaluated by a fitness function, with the values representing the goodness
degrees of the solutions. The solution with the best fitness value for a particle can be regarded as the local optimal solution found
Hong and Shiu / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 136-147
138
so far and is stored as the pBest solution for the particle. The best one among all the pBest solutions is regarded as the global
optimal solution found so far for the whole set of particles, and is called the gBest solution. In addition, each particle moves with a
velocity, which will dynamically change according to pBest and gBest. After finding the two best values, a particle updates its
velocity by the following equation:
), ( ) ( ) ( ) (
2 2 id d id id 1 1
old
id
new
id
x gBest Rand c x pBest Rand c V w V + + =
where the terms are explained below.
1.
new
id
V
: the velocity of the i-th particle in the d-th dimension in the next generation;
2.
old
id
V : the velocity of the i-th particle in the d-th dimension in the current generation;
3. pBest
id
: the current pBest value of the i-th particle in the d-th dimension;
4. gbest
d
: the current gBest value of the whole set of particles in the d-th dimension;
5. x
id
: the current position of the i-th particle in the d-th dimension;
6. w: the inertial weight, generally set at 1 (Shi and Eberhart, 1998);
7. c
1
: the acceleration constant for a particle to move to its pBest, generally set at 2 (Eberhart and Kennedy, 1995);
8. c
2
: the acceleration constant for a particle to move to the gBest, generally set at 2 (Eberhart and Kennedy, 1995);
9. Rand
1
(), Rand
2
(): two random numbers within 0 to 1.
After the new velocity is found, the new position for a particle can then be obtained by the following formula:
.
new
id id id
V x x + =
All the particles thus continuously move in the search space, tending to better solutions, until the termination criteria are met.
After a lot of generations, an approximate optimal solution is expected to be found. The PSO algorithm is stated below.
4. Using the PSO algorithm to solve the K-of-N lifetime problem
The ANs produced by different manufacturers may own different data transmission rates, initial energies and parameter values.
When different kinds of ANs exist in a wireless network, it is hard to find the optimal BS location. In this section, a heuristic
algorithm based on PSO to solve the K-of-N lifetime problem under general constraints is proposed. An initial set of particles is
first randomly generated, with each particle representing the coordinate of a possible BS location. Each particle is also allocated an
initial velocity for changing its state. Let e
j
(0) be the initial energy, r
j
be the data transmission rate,
j1
be the distance-independent
parameter, and
j2
be the distance-dependent parameter of the j-th AN. The lifetime l
ij
of an application node AN
j
for the i-th
particle is calculated by the following formula:
, ) ( ) 0 (
n
ij j j j j ij
d r e l + =
where
n
ij
d is the n-order Euclidian distance from the j-th AN to the i-th particle. In the K-of-N lifetime problem, given N ANs, the
system survives as long as at least K ANs are alive. Restated, the system fails when at least N-K+1 ANs run out of energy. The
fitness function used for evaluating each particle is thus shown below:
}. { ) (
1
1
ij K N
N
j
l Min i fitness
+
=
=
That is, the fitness of the i-th particle is its (N-K+1)-th minimal lifetime among all the ANs. A larger fitness value thus denotes a
better solution quality for the K-of-N lifetime problem, meaning the corresponding BS location is better. The fitness value of each
particle is then compared with that of its corresponding pBest. If the fitness value of the i-th particle is larger than that of pBest
i
,
pBest
i
is replaced with the i-th particle. The best pBest
i
among all the particles is chosen as the gBest. Besides, each particle has a
velocity, which is used to change the current position. All particles thus continuously move in the search space. When the
termination conditions are achieved, the final gBest will be output as the location of the base station. The proposed algorithm is
stated below.
Hong and Shiu / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 136-147
139
The proposed PSO algorithm for the K-of-N lifetime problem
Input: A set of N ANs, each AN
j
with its location (x
j
, y
j
), data transmission rate r
j
, initial energy e
j
(0), parameter
j1
, parameter
j2
,
and the allowed number K of alive ANs.
Output: A BS location that will cause a nearly maximal lifetime for the K-of-N lifetime problem.
Step 1: Initialize the fitness values of all pBests and the gBest to zero.
Step 2: Randomly generate a group of n particles, each representing the coordinate of a possible BS location. Locations may be
two-dimensional or three-dimensional, depending on the problems to be solved.
Step 3: Randomly generate an initial velocity for each particle.
Step 4: Calculate the lifetime l
ij
of the j-th AN for the i-th particle by the following formula:
), ( ) 0 (
2 1
n
ij j j j j ij
d r e l + =
where e
j
(0) is the initial energy, r
j
is the data transmission rate,
j1
is a distance-independent parameter,
j2
is a distance-
dependent parameter of the j-th AN, and
n
ij
d
is the n-order Euclidean distance from the i-th particle (BS) to the j-th AN.
Step 5: Calculate the (N-K+1)-th minimal lifetime among all the ANs for the i-th particle as its fitness value (fitness(i)) by the
following formula:
}, { ) (
1
1
ij K N
N
j
l Min i fitness
+
=
=
where N is number of ANs, K is the allowed number of alive ANs, i = 1 to n, and j = 1 to N.
Step 6: Set pBest
i
as the current i-th particle if the value of fitness(i) is larger than the current fitness value of pBest
i
.
Step 7: Set gBest as the best pBest among all the particles. That is, let:
fitness of pBest
k
=
n
i
max
1 =
fitness of pBest
i
,
and set gBest=pBest
k
.
Step 8: Update the velocity of the i-th particle as:
), ( ) ( ) ( ) (
2 2 id d id id 1 1
old
id
new
id
x gBest Rand c x pBest Rand c V w V + + =
where
new
id
V is the new velocity of the i-th particle at the d-th dimension,
old
id
V
is the current velocity of the i-th particle at
the d-th dimension, w is the inertial weight, c
1
is the acceleration constant for particles moving to pBest, c
2
is the
acceleration constant for particles moving to gBest, Rand
1
( ) and Rand
2
( ) are two random numbers among 0 to 1, x
id
is the
current position of the i-th particle at the d-th dimension, pBest
id
is the value of pBest
i
at the d-th dimension, and gBest
d
is
the value of gBest at the d-th dimension.
Step 9: Update the position of the i-th particle as:
,
new
id
old
id
new
id
V x x + =
where
new
id
x
and
old
id
x
are respectively the new position and the current position of the i-th particle at the d-th dimension.
Step 10: Repeat Steps 4 to 9 until the termination conditions are satisfied.
In Step 10, the termination conditions may be predefined execution time, a fixed number of generations or when the particles
have converged to a certain threshold.
5. A small example
In this section, a simple example in a two-dimensional space is given to explain how the PSO approach can be used to find the
best BS location for the K-of-N lifetime problem. Assume there are totally four ANs in the example and their initial parameters are
shown in Table 1, where Location represents the two-dimensional coordinate position of an AN, Rate represents the data
transmission rate, and Power represents the initially allocated energy. All
j1
s are set at 0 and all
j2
s at 1 for simplicity. Also
assume the allowed number K of alive ANs is 3.
Hong and Shiu / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 136-147
140
Table 1: The initial parameters of ANs in the example
Assume 3 particles are used for simplicity. Also assume the three initial particles randomly generated are located at (4, 7), (9, 5)
and (6, 4). An initial velocity is randomly generated for each particle. In this example assume the initial velocity is set at zero for
simplicity. The lifetime of each AN for a particle is then calculated. Take the first AN for the first particle as an example. Its
lifetime is calculated as follows:
18 . 185 ] ) 10 7 ( ) 1 4 [( 3 10000
2 2
11
= + = l .
The lifetimes of all ANs for all particles are shown in Table 2.
Table 2: The lifetimes of all ANs for all particles
1(1, 10) 2(11, 0) 3(8, 7) 4(4, 3)
1(4, 7) 185.18 20.40 100 100
2(9, 5) 37.45 68.96 320 55.17
3(6, 4) 54.64 48.78 123.07 320
The (N-K+1)-th minimal lifetime among all the ANs for each particle is calculated as its fitness value. The results are 100, 55.17
and 54.64 for the three particles. The fitness value of each particle is then compared with that of its corresponding pBest. In the
first generation, the fitness values of all the pBests are zero, smaller than those of the particles. The particles are then stored as the
new pBests. The best pBest among all the particles is chosen as the gBest. In this example, the first particle is set as the gBest. The
new velocity and the position of each particle are then updated. Steps 4 to 9 are then repeated until the termination conditions are
satisfied. In this example, the solutions converge after about 70 generations. The solution was located at (4.5515, 6.419) and its
lifetime was 130.8344 time units.
6. Experimental results
Experiments were made to show the performance of the proposed PSO algorithm on finding optimal base stations for the K-of-N
lifetime problem. They were performed in C language on an AMD PC with a 2.0GHz processor and 1G main memory and running
the Microsoft Window XP operating system. The simulation was done in a two-dimensional real-number space of 1000*1000.
That is, the ranges for both x and y axes were between 0 to 1000. The data transmission rate was limited between 1 to 10 and the
range of initial energy was limited between 100000000 to 999999999. The data of all ANs, each with its own location, data
transmission rate and initial energy, were randomly generated.
Experiments were first made to show the convergence of the proposed PSO algorithm when the acceleration constant (c
1
) for a
particle moving to its pBest was set at 2, the acceleration constant (c
2
) for a particle moving to its gBest was set at 2, the inertial
weight (w) was set at 0.6, the distance-independent parameter (
j1
) was set at zero, and the distance-dependent parameter (
j2
) was
set at one. Assume totally there were 50 ANs in the system and the allowed number (K) of alive ANs was set at 40. The
experimental results of the resulting lifetime along with different generations for 10 particles in each generation are shown in
Figure 1.
It is easily seen from Figure 1 that the proposed PSO algorithm could converge very fast (at about 50 generations). Next,
experiments were made to show the effects of different parameters on the K-of-N lifetime problem. The influence of the
acceleration constant (c
1
) for a particle moving to its pBest on the proposed algorithm was first considered. The process was
terminated at 300 generations. When w = 1 and c
2
= 2, the lifetimes along with different acceleration constants (c
1
) are shown in
Figure 2.
It can be observed from Figure 2 that the lifetime first increased and then decreased along with the increase of the acceleration
constant (c
1
). It was because when the value of the acceleration constant (c
1
) was small, the velocity update of each particle was
also small, causing the convergence speed slow. The proposed PSO algorithm might thus not get the optimal solution after the
predefined number of generations. On the contrary, when the value of the acceleration constant (c
1
) was large, the velocity change
would be large as well, causing the particles to move fast. It was then hard to converge. In the experiments, the optimal c
1
value
was about 2.
AN Location Rate Power
1 (1, 10) 3 10000
2 (11, 0) 5 10000
3 (8, 7) 4 6400
4 (4, 3) 4 6400
Particle
AN
Hong and Shiu / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 136-147
141
0
50
100
150
200
250
0 50 100 150 200 250 300
Generations
L
i
f
e
t
i
m
e
Figure 1. The lifetime for the 40-of-50 lifetime problem and10 particles
200
201
202
203
204
205
206
0 1 2 3 4
c
1
L
i
f
e
t
i
m
e
Figure 2. The lifetimes along with different acceleration constants (c
1
)
Next, experiments were made to show the effects of the acceleration constant (c
2
) for a particle moving to its gBest on the
proposed algorithm. When w = 1 and c
1
= 2, the experimental results are shown in Figure 3.
201
202
203
204
205
206
0 0.5 1 1.5 2 2.5 3 3.5 4
c
2
L
i
f
e
t
i
m
e
Figure 3. The lifetimes along with different acceleration constants (c
2
)
Hong and Shiu / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 136-147
142
It can be observed from Figure 3 that the lifetime first increased and then decreased along with the increase of the acceleration
constant (c
2
). The reason was the same as above. In the experiments, the optimal c
2
value was about 2. Next, experiments were
made to show the effects of the inertial weight (w) on the proposed algorithm. When c
1
= 2 and c
2
= 2, the experimental results are
shown in Figure 4.
185
190
195
200
205
210
215
0 0.5 1 1.5 2 2.5 3
w
L
i
f
e
t
i
m
e
Figure 4. The lifetimes along with different inertial weights (w)
It can be observed from Figure 4 that the proposed algorithm could get better lifetime when the inertial weight (w) was smaller
than 0.6. The lifetime decreased along with the increase of the inertial weight (w) when w was bigger than 0.6. This was because
when the value of the inertial weight was large, the particles would move fast due to the multiple of the old velocity. It was then
hard to converge. Experiments were then made to show the effects of the distance-independent parameter (
1
) on the lifetime. In
the experiments, all ANs had the same value of the distance-independent parameter. The experimental results are shown in Figure
5.
0
50
100
150
200
250
0 200000 400000 600000 800000 1000000
1
L
i
f
e
t
i
m
e
Figure 5. The lifetimes along with different values of the distance-independent parameter (
1
)
It can be observed from Figure 5 that the lifetime decreased along with the increase of the value of the distance-independent
parameter (
1
). The relation was nearly linear. It was consistent with the formula of energy consumption. Next, experiments were
made to show the effects of the distance-dependent parameter (
2
) on the lifetime. The experimental results are shown in Figure 6.
Hong and Shiu / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 136-147
143
0
50
100
150
200
250
0 1 2 3 4 5
2
L
i
f
e
t
i
m
e
Figure 6. The lifetimes along with different values of the distance-dependent parameter (
2
)
It can be observed from Figure 6 that the lifetime decreased along with the increase of the distance-dependent parameter (
2
). It
was also consistent with the formula of energy consumption. Besides, the relation between the lifetime and the value of the
distance-dependent parameter presented an approximately inverse proportion. Next, experiments were made to show the relation
between lifetimes and numbers of ANs. The experimental results are shown in Figure 7.
0
100
200
300
400
500
600
700
800
40 50 60 70 80 90 100
ANs
L
i
f
e
t
i
m
e
Figure 7. The lifetimes along with different numbers of ANs
0
0.04
0.08
0.12
0.16
0.2
40 50 60 70 80 90 100
ANs
T
i
m
e
(
s
e
c
.
)
Figure 8. The execution time along with different numbers of ANs
Hong and Shiu / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 136-147
144
It can be seen from Figure 7 that the lifetime increased along with the increase of the number of ANs. The execution time along
with different numbers of ANs is shown in Figure 8.
It can be observed from Figure 8 that the execution time increased along with the increase of numbers of ANs. The relation was
nearly linear. Next, experiments were then made to show the relation between lifetimes and numbers of particles for 50ANs. The
experimental results for w = 1 are shown in Figure 9.
206.5
207
207.5
208
208.5
209
209.5
210
210.5
211
211.5
0 20 40 60 80 100
Particles
L
i
f
e
t
i
m
e
Figure 9. The lifetimes along with different numbers of particles
It can be seen from Figure 9 that the lifetime increased along with the increase of the number of particles. The execution time
along with different numbers of particles is shown in Figure 10.
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
0 20 40 60 80 100
Particles
T
i
m
e
(
s
e
c
.
)
Figure 10. The execution time along with different numbers of particles for 50 ANs
It can be observed from Figure 10 that the execution time increased along with the increase of numbers of particles. The relation
would be nearly linear. This was reasonable since the execution time would be approximately proportional to the number of
particles.
When implementing the particle swarm algorithm, several considerations must be taken into account to facilitate the
convergence and prevent an explosion of the swarm. These considerations include limiting the maximum velocity, selecting
acceleration constants, the constriction factor, or the inertia weight. A detailed discussion can be found in del Valle et al. (2008).
Note that no optimal solutions can be found in a finite amount of time since the problem is NP-hard. For a comparison, an
exhaustive search using grids was used to find nearly optimal solutions. The approach found the lifetime of the system when a BS
Hong and Shiu / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 136-147
145
was allocated at any cross-point of the grids. The cross-point with the maximum lifetime was then output as the solution. A
lifetime comparison of the PSO approach and the exhaustive search with different grid sizes are shown in Table 3.
It can be observed from Table 3 that the lifetime obtained by the proposed PSO algorithm was better than those by the
exhaustive grid search with grid sizes larger than 0.01. The lifetime by the proposed PSO algorithm was 212.4404, and was
212.0781, 212.4158 and 212.4379 by the exhaustive search for the grid size respectively set at 1, 0.1 and 0.01. For the exhaustive
grid search, the smaller the grid size, the better the results. The execution time by the two approaches is shown in Table 4.
Table 3. A lifetime comparison of the PSO approach and the exhaustive grid search
Method Lifetime
The proposed PSO algorithm 212.4404
The exhaustive grid search
(grid size = 1)
212.0781
The exhaustive grid search
(grid size = 0.1)
212.4158
The exhaustive grid search
(grid size = 0.01)
212.4379
Table 4. A comparison of execution time by the two approaches
Method Time (sec.)
The proposed PSO algorithm 0.06
The exhaustive grid search
(grid size = 1)
36.563
The exhaustive grid search
(grid size = 0.1)
2480.718
The exhaustive grid search
(grid size = 0.01)
170871.8558
It can be seen from Table 4 that the execution time of the exhaustive grid search increased along with the decrease of grid sizes.
It was very consistent with the processing way of the exhaustive grid search. Besides, the exhaustive grid search spent much more
execution time than the proposed PSO algorithm, especially when the grid size was small. The advantage of the proposed PSO
approach for solving the problem lies in that it can reduce the computation time and keep the same good quality.
7. Conclusion and future works
In wireless sensor networks, minimizing power consumption to prolong network lifetime has always been very crucial. In this
paper, we have thus extended our previous approach to solve the K-of-N lifetime problem under general power-consumption
constraints in wireless sensor networks. Finding BS locations is by nature very similar to finding the food locations originated
from PSO. It is thus easy to model such a problem by the proposed algorithm based on PSO. The extension from solving the N-of-
N problem to solving the K-of-N problem by the proposed PSO approach is also very simple when compared to Pan et al.s
extension.
The proposed approach can solve the K-of-N lifetime problem in heterogeneous sensor networks, where application nodes may
own different data transmission rates, initial energies and parameter values. Extensive experiments have also been made to show
the performance of the proposed PSO approach for the K-of-N lifetime problem and the effects of the parameters on the results.
From the experimental results, it can be easily concluded that the proposed PSO algorithm converges very fast. Besides, the
lifetime first increases and then decreases along with the increase of the acceleration constants c
1
, c
2
and w. It also increases along
with the increase of ANs, decreases along with the increase of the parameters
1
and
2
, and decreases along with the decrease of
particles. The execution time, however, increases in a nearly linear way along with the increase of ANs and particles. Note that if
the problem to be solved is the ANs with the same initial energy and the same energy-consumption parameters, it is fine to use Pan
et al.s approach. Their approach spends polynomial time and can get the optimal solution. In the future, we will attempt to
combine other approaches to automatically adjust the parameters.
Acknowledgement
This research was supported by the National Science Council of the Republic of China under contract NSC 98-2221-E-390 -033.
Hong and Shiu / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 136-147
146
References
del Valle Y., G. K. Venayagamoorthy, S. Mohagheghi, J. C. Hernandez and R. G. Harley, 2008. Particle swarm optimization:
basic concepts, variants and applications in power systems, IEEE Transactions on Evolutionary Computation, Vol. 12, No. 2,
pp. 171-195.
Eberhart R.C. and J. Kennedy, 1995a. A new optimizer using particles swarm theory, The Sixth International Symposium on
Micro Machine and Human Science, pp. 39-43.
Eberhart R.C. and J. Kennedy, 1995b. Particle swarm optimization, The IEEE International Conference on Neural Networks,
Vol. 4, pp. 1942-1948.
Gaing Z.L., 2003. Discrete particle swarm optimization algorithm for unit commitment, The IEEE Power Engineering Society
General Meeting.
Hong T.P. and G.N. Shiu, 2006. A PSO-based heuristic algorithm for base-station locations, International Conference on Soft
Computing and International Symposium on advanced Intelligent Systems (SCIS&ISIS 2006).
Heinzelman W., J. Kulik, and H. Balakrishnan, 1999. Adaptive protocols for information dissemination in wireless sensor
networks, The Fifth ACM International Conference on Mobile Computing and Networking, pp. 174-185.
Heinzelman W., A. Chandrakasan, and H. Balakrishnan, 2000. Energy-efficient communication protocols for wireless
microsensor networks, The Hawaiian International Conference on Systems Science.
Intanagonwiwat C., R. Govindan and D. Estrin, 2000. Directed diffusion: a scalable and robust communication paradigm for
sensor networks, The ACM International Conference on Mobile Computing and Networking.
Kawadia V. and P. R. Kumar, 2003. Power control and clustering in ad hoc networks, The 22nd IEEE Conference on Computer
Communications (INFOCOM).
Li N., J. C. Hou and L. Sha, 2003. Design and analysis of an mst-based topology control algorithm, The 22nd IEEE Conference
on Computer Communications (INFOCOM).
Lee S., W. Su and M. Gerla, 2001. Wireless ad hoc multicast routing with mobility prediction, Mobile Networks and
Applications, Vol. 6, No. 4, pp. 351-360.
Niculescu D. and B. Nath, 2003. Ad hoc positioning system (APS) using AoA, The 22nd IEEE Conference on Computer
Communications (INFOCOM), pp. 1734-1743.
Pan J., Y. Hou, L. Cai, Y. Shi and X. Shen, 2003. Topology control for wireless sensor networks, The Ninth ACM International
Conference on Mobile Computing and Networking, pp. 286-299.
Pan J., L. Cai, Y. T. Hou, Y. Shi and S. X. Shen, 2005. Optimal base-station locations in two-tiered wireless sensor networks,
IEEE Transactions on Mobile Computing, Vol. 4, No. 5, pp. 458-473.
Qin Y.Q., D. B. Sun, N. Li and Y. G. Cen, 1999. Path planning for mobile robot using the particle swarm optimization with
mutation operator, The IEEE Third International Conference on Machine Learning and Cybernetics, pp. 2473-2478.
Rodoplu V. and T. H. Meng, 1999. Minimum energy mobile wireless networks, IEEE Journal on Selected Areas in
Communications, Vol. 17, No. 8.
Ramanathan R. and R. Hain, 2000. Topology control of multihop wireless networks using transmit power adjustment, The 19th
IEEE Conference on Computer Communications (INFOCOM).
Shi Y. and R. C. Eberhart, 1998. A modified particle swarm optimizer, The IEEE International Conference on Evolutionary
Programming, pp. 69-73.
Stacey A., M. Jancic and I. Grundy, 2003. Particle swarm optimization with mutation, The IEEE Congress on Evolutionary
Computation, pp. 1425-1430.
Tian D. and N. Georganas, 2003. Energy efficient Routing with guaranteed delivery in wireless sensor networks, The IEEE
Wireless Communication & Networking Conference.
Ye F., H. Luo, J. Cheng, S. Lu and L. Zhang, 2002. A two-tier data dissemination model for large scale wireless sensor networks,
The ACM International Conference on Mobile Computing and Networking.
Biographical notes
Tzung-Pei Hong received his B.S. degree in chemical engineering from National Taiwan University in 1985, and his Ph.D. degree in computer
science and information engineering from National Chiao-Tung University in 1992. He was in charge of the whole computerization and library
planning for National University of Kaohsiung in Preparation from 1997 to 2000 and served as the first director of the library and computer
center in National University of Kaohsiung from 2000 to 2001, as the Dean of Academic Affairs from 2003 to 2006 and as the Vice President
from 2007 to 2008. He is currently a distinguished professor at the Department of Computer Science and Information Engineering and at the
Department of Electrical Engineering. He has published more than 350 research papers in international/national journals and conferences and has
planned more than fifty information systems. He is also the board member of more than thirty journals and the program committee member of
more than one hundred and thirty conferences. His current research interests include parallel processing, machine learning, data mining, soft
computing, management information systems, and www applications.
Hong and Shiu / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 136-147
147
Gou-Neng Shiu received his B.S. degree in Applied Mathematics in 2004 and his M.S. degree in Electrical Engineering in 2006 from National
University of Kaohsiung. His research interests include particle swarm optimization, genetic algorithms, wireless sensor network and fuzzy
theory.
Received October 2009
Accepted November 2009
Final acceptance in revised form November 2009
MultiCraft
International Journal of Engineering, Science and Technology
Vol. 1, No. 1, 2009, pp. 148-171
INTERNATIONAL
JOURNAL OF
ENGINEERING,
SCIENCE AND
TECHNOLOGY
www.ijest-ng.com
2009 MultiCraft Limited. All rights reserved
Design aids for fixed support reinforced concrete cylindrical shells under
uniformly distributed loads
Srinivasan Chandrasekaran
1*
,
S.K.Gupta
2
, Federico Carannante
3
1*
Dept. of Ocean Engineering, Indian Institute of Technology Madras, Chennai 600 036, India
2
Dept. of Civil Engineering, Institute of Technology, Banaras Hindu University, Varanasi, India
3
Dept. of Structural Engineering, University of Naples Federico II, Naples, Italy
*
Corresponding Author: e-mail:drsekaran@iitm.ac.in; Fax: +91-44-22574802
Abstract
Shells are objects considered as materialization of the curved surface. Despite structural advantages and architectural
aesthetics possessed by shells, relative degree of unacquaintance with shell behavior and design is high. Thin shells are examples
of strength through form as opposed to strength through mass; their thin cross-section makes them economical due to low
consumption of cement and steel as compared to other roof coverings such as slabs. Current study presents design curves for
reinforced concrete open barrel cylindrical shells for different geometric parameters. The analysis is done in two parts namely: i)
RC shell subjected to uniformly distributed load that remain constant along its length and curvature of the shell surface; and ii)
RC shell subjected to uniformly distributed load varying sinusoidally along its length in addition to different symmetric edge
loads present along its longitudinal boundaries. Design charts are proposed for easier solution of shell constants after due
verification of results obtained from finite element analysis. Expressions for stress resultants proposed in closed form make the
design more simple and straightforward; stress resultants plotted at closer intervals of can be useful for detailing of
reinforcement layout in RC shells. Axial force-bending moment yield interaction studied on shells under uniformly distributed
loads show compression failure, initiating crushing of concrete.
Keywords: cylindrical shells; design curves; open barrel; reinforced concrete; stress resultants
1. Introduction
Shells are skin structures by virtue of their geometry and shell action is essentially more towards transmitting the load by direct
stresses with relatively small bending stresses. Presence of significant shear makes their behavior different from other roof
covering structures; they have remarkable reserve strength with a greater degree-of-freedom in structure layout, shape and
architecture, making them virtually impossible to collapse though their supporting structure may collapse (Rericha 1996; Ha-Wong
Song et al. 2002). A number of procedures are proposed in the literature (see, for example, Flugge 1967) for analyzing the various
types of thin shells for ultimate load but experimental evidences, however are not abundant. In many of these procedures, bending
stresses are stated to be negligibly small justifying the application of membrane theory for analyzing thin shells, where normal
forces are not necessarily tensile. Though behavior of shells to external loads is generally quite complex, thanks to second order
equations developed
that are sufficient to model the behavior of most civil engineering shell structures (Jacques, 1977). Forces in
thin shells are, in a certain sense, statically determinant as they can be determined without referring to elastic property of material
used. Linear elastic behavior is applied to provide a direct relationship between stress and strains by which the equilibrium of
stress resultant and stress couples could be established. However mathematical models of thin RC shells developed on the
interpretation of their physical behavior are difficult to present rationally as the rigorous analysis is extraordinarily complex,
resolving to simplification (see, for example, Billington, 1965). These simplifications need extra caution for the design to become
conservative. Also, interestingly, load carrying capacity of thin curved shells often exceeds the prediction of even most-refined
available analysis. It is therefore commonly accepted that shells (reinforced concrete shells, in particular) are designed on the basis
of approximate analysis only, which of course exhibits defects. For example, aircraft impact on containment structures is studied
Chandrasekaran et al. / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 148-171
149
by employing modal analysis to estimate their elastic response (Prabhakar, 2003). The study approximates shell with uniform
thickness in longitudinal and circumferential directions and orthotropic properties are obtained by scaling the Youngs modulus
values. Wheen and Wheen (2003) studied design and construction principles of a pre-stressed conical concrete tent roof
,
idealizing
it as thin pre-tensioned radial circles suspended from a temporary central tower and anchored at the lower ends to a pre-stressed
ring beam in a catenary form. They showed that the curved concrete surface, usually a felicitous combination of sound structural
form and method of construction leading also to low cost, does not crack when suspended because of radial pre-stressing. Concrete
domes, in particular, possess limited membrane stress variations due to soundness in structural form and therefore ideal as all
stresses are compressive during construction and in final stages as well (Wheen and Anatharaman, 1998). Example structures
presented above for designing/constructing RC shell structures prove the confidence level of researchers in employing relevant
approximate theories for analyzing shell structures. This leads to the desire of simplifying the analysis procedure and adding to
architectural aesthetics and structural advantages shell possess; the objective is to simplify analysis procedure by means of ready to
use design curves and tables. This shall not only promote more construction of such advantageous and safe structures but also
improve the confidence level in design offices. However, other studies conducted by the authors emphasize verification of stress
resultants at closer intervals of geometric angles at boundaries, in particular, for appropriate detailing of reinforcement in RC shells
(see, for example, Chandrasekara et al. 2005; Chandrasekaran and Srivastava, 2006). Ramaswamy (1968) discussed limitations
and precision of various theories on shell structures and concluded that Flgges theory is acceptable and can be treated as a bench
mark to assess the accuracy of other theories.
2. Objectives and structural idealization
Based on the critical review of literature, main objective of the current study is to develop design curves based on the classical
flexural theory for single barrel open cylindrical shells with different geometric parameters; the outcome is to derive the stress
resultant estimates that can be readily determined without undergoing detailed computations. Though theory employed is not new
but the attempt made to simplify the analysis procedure is relatively new and the analytical solutions proposed in the paper are not
present in the literature. As the solution procedure of generic case of shell geometry shall involve complex mathematics, objective
of simplifying the design procedure through design curves shall not be fulfilled; hence a particular geometric case of cylindrical
shells is only addressed in the present study. Please note that cylindrical shells are commonly used geometry in structural/civil
engg applications. With the use of proposed design curves, stress resultants and stress couples can easily be readily determined at
the valley, crown or at any desired section for various type of edge loads. This shall be helpful to be employed in design offices
while encouraging such new RC shell structures in future and also enhances the confidence of the structural designer. The study
employs few structural idealizations namely: i) stress normal to the surface is neglected as thin shell is a curved slab whose
thickness is relatively small compared to its other dimensions and radius of curvature; ii) deflections under load are small enough
so that its static equilibrium remain affected due to changes in geometry, vide small deflection theorem; iii) linear elastic behavior
is assumed in the analysis enabling a direct relationship between stress and strain; iv) for all kinematic relations, any point on the
mid surface is considered as unaffected by the shell deformation; as well as v) all points lying normal to the middle surface before
deformation lie on the normal after deformation.
3. Mathematical development
Let us consider a cylindrical coordinate system ( ) r , , x and a cylindrical shell with thickness t. The cylindrical shell has the
directive along the x-axis and we denote
r
u , u , u
x
as the displacements along longitudinal (x), circumferential () and radial
directions (r), respectively. It is important to note that the displacements, strains and stress functions depend only on two variables
namely ( x , ). Figure.1 shows the overall geometry of the shell considered for the study while Figure.2 shows the differential
element considered for the analysis. The three strain components
,
x
and
x
of the middle surface and the three curvature
changes
,
x
and
x
can be expressed in terms of the displacement components as given below:
x
u
x
u
R
1
,
u
u
R
1
,
x
u
,
x
u
u
R
1
, u
u
R
1
,
x
u
r
2
x
2
r
2
2 2
r
2
x
x
x r
x
x
(1)
Corresponding quantities associated with the strain components and curvatures are the membrane resultant forces
x x
N , N , N , and
resultant moments
x x
M , M , M , respectively. In addition, resultant shear forces Q
x
, Q
M
+ d
M
M
x
d +
x M
x
M
x
M
M
x dx
x
M
x
+
x
M
M
x dx
x
Mx
+
Q
x+
x
Q
x
dx
Q d
Q
+
Q
x
Q
r
x
r
p
x
p
p
N
x +
x
N
x
dx
x
d
N
+ N
x
N
x
x
N
+
N
d N
dx
x
N
x
+
N
x
x
N
N
x
r
Figure 2. Membrane stresses, moment and shear forces acting on infinitesimal element of shell
Relationships between static quantities and strain/curvature given by the Hookes law are shown below:
( ) ( )
( )
( )
( )
( )
( )
( )
+
= = +
= +
=
+
= +
= +
=
x
3
x x x
2
3
x
2
3
x
x x x
2
x
2
x
1 12
t E
M M ,
1 12
t E
M ,
1 12
t E
M
,
1 2
t E
N ,
1
t E
N ,
1
t E
N
(2)
By substituting Eq. (1) in Eq. (2), force-displacement relationships are derived as given below:
( ) ( )
( )
( ) ( )
( )
x
r
2 3
x
2
r
2
2 2
r
2
2
3
2
r
2
2 2
r
2
2
3
x
x
x r
x
2
r
x
2
x
M
x
u
x
u
R 1 12
t E
M
,
u
u
R
1
x
u
1 12
t E
M ,
u
u
R
v
x
u
1 12
t E
M
,
u
R
1
x
u
1 2
t E
N , u
u
x
u
R
1 R
t E
N , u
u
x
u
R
1 R
t E
N
+
=
+
=
=
(3)
For a general case in three dimensions, we can write three equation of equilibrium along x,y and z direction and three equations of
moments with respect to the x,y, and z axes. In total there are six equations of equilibrium available for solving eight unknown
Chandrasekaran et al. / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 148-171
151
functions. If
x x
N , N , N are small in comparison with their critical values at which lateral buckling of the shell may occur, their
effect on bending becomes negligible; all terms containing the products of the resultant forces or resultant moments with the
derivatives of the small displacements can be neglected in the equilibrium equations (Timoshenko and Krieger, 1959). With this
hypothesis, equilibrium equations are now reduced to five as given below:
0 R
x
Q
M
R
x
M
0 R Q
M
R
x
M
0 R p N
Q
R
x
Q
0 R p Q
N
R
x
N
0 R p
N
R
x
N
x
x
x
r
x
x
x
x
x
= +
= + +
= +
= +
(4)
where
r x
p p p , ,
are the load for unit surface along x-axis, circumferential and radial direction, respectively. Solving last two
equations with respect to
Q ,
x
Q and substituting in remain three equations, we finally obtain the three following relationship:
0 R p N
M
R
1
x
M
R
x
M
x
M
0 R p
M
R
1
-
x
M
R
x
N N
0 R p
N
R
x
N
r
2
2
x
2
2
x
2
x
2
x x
x
x
x
= + +
= +
= +
(5)
The shear functions have the following expression:
( )
( )
( )
3
r
3
2
2
r
3
2
2 2
3
2
2
2
r
3
2
3
r
3
2
2
2 3
3
x
u
R
x
u
x
u
1 R 12
t E
x
Q
x
u
1
x
u
R
u
u
1 R 12
t E
Q
(6)
Substituting the equations (3) in equations (5), we obtain the equilibrium equation in terms of the unknown displacement functions
r x
u , u , u
as given below:
( )
( )
+
+
=
+
+
0
t E
- 1
R p
u
R
1
x
u
R
2 u
R
1
x
u
R
2
x
u
R
12
t
R
u
u
R
1
x
u
0
t E
- 1
R p
u
R
1
x
u
u
R
1
x
u
1
R 12
t u
u
R
1
x
u
R
2
1
x
u
2
1
0
t E
- 1 p
x
u
R x
u
R 2
1 u
R 2
1
x
u
2
r
3
3
3 2
3
4
r
4
3 2 2
r
4
4
r
4 2
r x
2
3
r
3
2 2
r
3
2
2
2 2
2
2
r
2
2
2
2
x
2
2
x r
2
2
x
2
2
x
2
(7)
3.1 Analytical solution in closed form for shells under uniform distributed load
Let us consider an open barrel cylindrical shell loaded by uniform distributed load, as seen in Figure 3. The boundaries of the shell
along its length are considered as fixed support in the analysis. Stress functions will depend only on one variable, namely leading
to the following differential equations system:
Chandrasekaran et al. / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 148-171
152
k
=45
u =
= 0
=
k
2
for
R/t=100
Rck=25N/mm
2
p
Surface load
0
0 = u
x
r
u = 0
x
0 =
u
r
z
=constant
fixed edge
fixed edge
k
-
p
z
R
p
p
r
Figure 3. Shell under uniform distributed load
0 0
x
x
x
0
x
0
x
x
0
x
x
0
x
0
x x
0
x
x x
0
x
r
2
r
2
r
r
2
r
2
2
r
2
2
r
2
2
2
3
2
x
2
3
2 3
3
2
3
2 3
3
x
x
2
x
2
x
2
x x
x
u
x
u M
0
u
u
R
1
x
u
M
u
u
R
v
x
u M
0
u
R
1
x
u N
u
u
x
u
R
N
0
u
u
x
u
R
N
+
=
0
u
u
x
0,
u
u
x
0
x
u
0
x
u
R
u
x
0 u
u
x
0
x
u
r
2
2
r
3
r
3
x
r
2
x
2
(8)
From the first and fourth equations of the above set of equations, we can derive the following:
( ) ( ) ( )
( ) ( ) ( ) ( )
2
2 1 0 r
1 0 x
x q x q q x, u
x p p x, u
+ + =
+ =
(9)
where ( ) ( ) ( ) ( ) ( )
2 1 0 1 0
q q q p p , , , , are unknown functions of variable .
By substituting Eq. (9) in the last equation of Eq. (8), we obtain the function ( )
x, u as given below:
( ) ( ) ( )
( )
2 2
1 0
x
d
q d
x h h x, u
+ =
(10)
where ( ) ( )
1 0
h h , are unknown functions of variable .
By substituting Eqs (9-10) in Eq. (8), we obtain the following relationships:
( )
( )
( )
( ) ( )
( ) ( )
6 2 1 1
8 7 3 4 0
5 4 3 1
2 1 2
3 4 1
C sin C cos C r 2 p
C C sin C cos C r p
C sin C cos C h
sin C cos C q
sin C cos C q
+ + =
+ + =
+ + =
+ =
=
(11)
By substituting in Eqs. (9-10), the displacements functions are obtained as given below:
Chandrasekaran et al. / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 148-171
153
( ) ( ) ( )
( ) ( ) ( ) ( )
( ) ( ) ( ) ( ) x C C xsin C x C xcos q x, u
C x C C xcos C x C xsin h x, u
x 2C C rsin x 2C C rcos C x C C x, u
2 3 4 1 0 r
5 2 3 4 1 0
2 3 1 4 8 6 7 x
+ + =
+ + + + =
+ + + + =
(12)
The uniform distributed load applied on surface of the shell is resolved into components along ( ) r , , x as given below:
( )
( )
=
=
=
k r
k
x
cos p p
sin p p
0 p
(13)
where
k
is total shell angle in radians. By substituting Eq.(12) in equilibrium Eq. (7), the first condition is verified. Further, by
substituting the function ( )
0
q with its first derivative, the third equation of Eq. (12) becomes:
( )
( )
( ) ( ) x C C xsin C x C xcos
d
q d
x, u
2 3 4 1
0
r
+ +
= (14)
Now, the second and third equations of Eq. (7) become:
( )
( )
( ) ( )
( )( ) ( ) ( ) 0 sin 1 12pR sin C cos C t 12R t 2ER
d
q d
12R
d
h d
t 12R t
d
q d
t E
k
2 4
1 2
2 2 2
2
0
2
2
2
0
2
2 2 2
4
0
4
= + +
+ +
(15)
( ) ( ) ( ) ( )
( )( ) [ ] ( ) ( ) 0 cos 1 12pR sin C cos C t 12R 6RC t R 2E
d
q d
d
h d
t
d
h d
d
q d
t E 12R
k
2 4
1 2
2 2
6
2
5
0
5
3
0
3
2 0 0 2
= + +
(16)
By integrating Eq. (15) twice we get:
( ) ( )
( ) ( )
( )
( )
( )
( )
10 9
2
0
2
2
0
2
2 2
2 2
k
2 4
1 2
2
0
C C
d
q d
t q 12R
t 12R
1
t 12R t E
sin 1 R 12p
sin C cos C 2R h
+ +
+
+
+
+ =
(17)
By substituting in Eq. (16) we get the following differential equation in ( )
0
q :
( ) ( ) ( )
( ) ( ) ( ) ( ) [ ] 0 cos 1 pR 2 R C C t E t R 12
d
q d
d
q d
2
d
q d
t E
k
2 2
6 9
2 2
5
0
5
3
0
3
0 3
= + + +
(18)
By solving, function ( )
0
q is obtained and is given as below:
( )
( )( ) ( ) [ ] ( )( ) ( )
( )( )( ) [ ]
( )
15 14 13 12
3
k
2 2 2 2 2
2
12 14 11 6 9
2 2
3
k
2 2 2 2 2
0
C cos C C C
t 8E
sin sin 8 cos 11 2 1 t 12R pR
t
sin C C C R C C t 12R
t 8E
cos sin 11 2 cos 8 1 t 12R pR
q
+ +
+
+
+ + + + +
+
+ +
=
(19)
Now the hypothesis is verified to satisfy the equilibrium and compatibility conditions. By substituting Eq. (19) and Eq. (17) in Eq.
(12), displacement functions
r x
u , u , u
are obtained in a closed form. The integration constants
14 ,...., 2 , 1 i
C
=
(the constant
15
C is not
present in displacement function as the function
r
u is considered as a derivative of is present the derivate of function ( )
0
q given
by Eq. (14)) are determined by imposing the boundary conditions at fixed supports and using polynomial identity rule. The
following sets of algebraic equations are obtained.
Chandrasekaran et al. / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 148-171
154
( ) ( )
( ) ( )
( ) ( )
( )
( ) ( )
( )
( ) ( ) ( )
( ) ( ) ( )
( ) ( ) ( )
( ) ( ) ( )
= =
= =
= =
= =
= =
= =
= =
= =
= = = = = = = =
= = = = = = = =
=
=
= = = = = =
=
=
= = = = = =
= = = = = =
= = = = = =
0 2
d
dq
0; 2
d
dq
0; ) 2 (
d
dq
x 0 2 x,
d
u d
0 0
d
dq
0; 0
d
dq
0; 0) (
d
dq
x 0 0 x,
d
u d
0 2 q 0; 2 q 0; ) 2 ( q x 0 2 x, u
0 0 q 0; 0 q 0; 0) ( q x 0 0 x, u
0
d
2 q d
0; 2 h 0; ) 2 ( h x 0 2 x, u
0
d
0 q d
0; 0 h 0; 0) ( h x 0 0 x, u
0 2 p 0; ) 2 ( p x 0 2 x, u
0 0 p 0; 0) ( p x 0 0 x, u
k
2
k
1
k
0
k
r
2 1 0 r
k 2 k 1 k 0 k r
2 1 0 r
k 2
k 1 k 0 k
2
1 0
k 1 k 0 k x
1 0 x
(20)
It can be easily seen from the above that there are only 14 independent equations (as same as the integration constants to be
determined); the remaining are linearly dependent ones.
( ) ( ) ( ) ( )( ) [ ]
( )
2 2
k k k
2 2
k
2
k
2 2
k
2 2 2
9 1,2,...8 i
t 12R
sin 4 cos3 t 4R 3 cos 1 8 12R 3 8 t R t 48
C 0, C
+
+ + +
= =
=
(21)
( ) ( ) ( ) ( ) ( )( ) [ ]
( )
2 2
k k k
4 2 2 4
k
2
k
2 2
k
2 2 2
k
2
10
t 12R
cos3 sin 4 t t 16R 48R 3 cos 1 8 12R 3 8 t t 12R R 48
C
+
+ + + +
= (22)
( ) ( ) ( )
( ) ( )
+ + + +
+ +
=
2 2 4 4 2 2
k k k k
2
2
k
2 2 2
k
4 2
k
4
k
11
t 72R 240R t t 12R cos2 sin cos3 12R
9 28 t 2R 5 22 24R t 2cos
6 C (23)
( ) ( ) ( )
( ) ( ) ( )
+ + + + +
+ +
=
2 2 4
k
2 2 2 2
k
2
2 2
k
k k
2
k k
2
2 2
12
t 5R 36R 6 cos2 5t 36R 6R t 12R 4sin
cos3 cos 4 1 t 12R
4 C (24)
( ) ( )
( ) ( ) ( )
+ +
+ +
+ + =
k
2 2 2 2 2 2 2
k
2
2 2
k
k
2
k
2
2 2
k k 13
sin3 t 3 20R 6R t 3 20R 18R t 12R sin 12
3cos2 8 3 t 12R cos 2 C
(25)
( )
( ) ( )
( )
( )
k k
2 2 2
k
2 2
2
k
2 2
k
2
2
2 2
k k 14
sin sin2 5t 36R 96R cos2
t 12R
1 2 t 5 2 12R
t 12R sin 8 C + +
+
+
+ = (26)
where
( )
( ) ( ) [ ]
k k
2 2
k
4 2
k
2 2
k
2 3
2 2
sin2 t 12R cos2 24R 1 24R t 2 t 16E
1 pR
+ + + +
=
Substituting the integration constants in the displacement function, we get the following displacement functions:
0 u
x
= (27)
( ) ( )( ) [ ] ( ) ( )
( ) ( ) ( ) [ ]
( ) ( ) [ ] ( )
( )
2 2 3
k
2 2 2 2 2 4 4 2 2
2 2
2 2
12 14 11
2
14 11
2
3
k
2 4
2 2
14 13
2 2 2 2
12
2
2 2
9 10
t 12R t E 8
sin t 12R 2 t 216R 1584R t 1 pR
t 12R
sin t 12R C C C t C C 12R
t E
cos 1 12pR
t 12R
cos C C t 12R 12R t C
t
t 12R
C C u
+
+ +
+
+
+ + + +
+
+
+
+ + +
+
+ =
(28)
Chandrasekaran et al. / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 148-171
155
( )( )( ) [ ]
( ) ( ) ( ) [ ] ( ) ( ) + + + + + +
+
+
+
=
sin C C cos C C sin 3 2 cos 4 sin 1 pR C
t
t 12R
t E 8
cos sin 4 cos 3 2 1 t 12R pR
u
14 13 12 11
2
k
2 2
9
2
2 2
3
k
2 2 2 2 2
r
(29)
The stress functions are given by:
( ) ( )
( ) ( ) ( )
( ) ( ) ( )
( ) ( )
( ) ( ) ( ) ( ) ( ) ( ) ( ) [ ]
( ) ( )
( ) ( ) ( )
( ) ( ) ( )
+
+ +
+ + =
+ + + + + +
+ + =
+ + +
+
+ =
= = = = = =
sin t R 12 2 cos t R 12 sin
sin t R 12 cos t R 12 2 cos
1 pR sin C cos C t E 4 Q
sin t 12R 2 cos t 4R 3 cos sin t 4R 3 cos t 12R 2 sin 1 R p
sin C cos C t 24R t 12R C
R 6
t E
M
sin sin t 12R cos t 12R 2
sin t 12R 2 cos t 12R cos
1 R p sin C cos C t E 4 N
M M , N N 0, Q 0, N 0, M 0, M
2 2 2 2
k
2 2 2 2
k 2 2
14 12
3
2 2 2 2
k
2 2 2 2
k
2 3
12 14
2 2
2
2 2
9
k
2 2 2 2
2 2 2 2
k 2 2
12 14
3
x x x x x x
(30)
where
( )( ) 1 t R 12 R 2
1
2 2 2
+
=
3.2 Analytical solution in closed form for shells under edge loads
The open barrel RC cylindrical shell in now analyzed under uniformly distributed surface load varying sinusoidal in addition to the
edge loads. Bellington (1965) presented the general equations of equilibrium under certain assumptions to simplify the stress
functions. Poissons modulus is neglected in this treatment to get the following relationships:
x
x
x
x
u
R
1
x
u
2
t E
N , u
u
R
t E
N ,
x
u
R
t E
N
r
(31a)
= =
r
3
x x
2
r
2
2
3
2
r
2 3
x
u
u
x R 12
t E
M M ,
u
u
12R
t E
M ,
x
u
12
t E
M (31b)
=
2
r
2
2 r
2
3
r 2 r
3
3
x
u
R
u
u
R 12
t E
x
Q , u
u
R
u
u
R 12
t E
Q
x
x
2
2
2
2
(31c)
In equations (31b), terms of radial displacement
u are neglected and
+ =
2
2
2 2
2
2 2
x
3
3
z
4
4
r
4
2
r
8
3
p
r
1
x
p
r
1
x
p
r
1
p
x
u
r
t E
u
12
t E
(32)
Where the differential operator
r
8
u and
r
4
p are given by the following expressions:
4
r
4
4 2 2
r
4
2 4
r
4
r
4
8
r
8
8 6 2
r
8
6 4 4
r
8
4 2 6
r
8
2 8
r
8
r
8
p
r
1
x
p
r
2
x
p
p
u
r
1
x
u
r
4
x
u
r
6
x
u
r
4
x
u
u
=
(33)
Let the shell be subjected to surface load p that is distributed uniformly over the shell surface and also varying sinusoidal along
its length. The load can be resolved into its components along longitudinal, circumferential and radial directions as given below:
Chandrasekaran et al. / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 148-171
156
( )
( )
=
=
L
x
sin cos p p
L
x
sin sin p p
0 p
k r
k
x
(34)
where, p = load per unit surface area of the shell surface and L is the span of shell. Equilibrium equations given by Eq. (4) can be
satisfied by a particular integral of the form:
( ) ( )
=
L
x
sin cos C x, u
k 0 r
(35)
By substituting, the value of constant C
0
that depends on the geometric parameters of the shell is determined as:
( )
( )
+ +
+ +
=
6 4 4
4
2 2 2 2
4 4 2 2 2 4 4 4
0
R L 12 R L t t E
R R L 4 L 2 R p L 12
C
(36)
Further by substituting C
0
at any desired section, Eq. (35) gives the displacement in terms of load and geometric characteristics of
the reinforced concrete shell. However, particular integral of Eq. (35) does not provide the complete solution and hence
complimentary function of Eq. (32) is now considered. Equating the RHS of Eq. (32) to zero, the complimentary function assumes
the following form:
( )
=
=1,3,5.. n
M
m r
L
x
sin e A x, u (37)
where, A
m
represents the eight arbitrary constants depending on the longitudinal boundary conditions and M (in e
M
) represents
the corresponding roots. Variation of ( ) x, u
r
along the longitudinal boundary can be described by Fourier series considering only
the first term of the series. By substituting Eq. (37) in homogenous form of Eq. (32), roots M are determined. On substitution, we
get:
0
L t R
12
k
R
M
4 2 2
4
4
2
2
2
= +
(38)
where
L
k
= . Eq. (38) in now expressed as:
( ) 0 Q 4 R k M
8
4
2 2 2
= + (39)
where,
4
3
t
3 r
L
Q ,
= =
2
6 4
8
t
R k 3
Q (40)
By introducing the constant
Q
R k
2 2
= , Eq. (39) is rewritten as:
( ) i 2 Q M
2 2
= (41)
By applying the Eulers equation, Eq. (41) becomes:
( ) i 1 Q M
2 2
= (42)
By solving, we obtain the follows roots:
) ( M , ) ( M
2 2 2 1 1 1
= = (43)
where,
,
2
) (1 1 ) (1
Q ,
2
) (1 1 ) (1
Q
,
2
) (1 1 ) (1
Q ,
2
) (1 1 ) (1
Q
2
2
2
1
2
2
2
1
+ +
=
+ + +
=
+
=
+ + + +
=
(44)
Now, the complimentary function of Eq. (32) is given by:
Chandrasekaran et al. / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 148-171
157
+ + + +
+ + + +
=
+ +
+ +
L
x
sin
e B e A e B e A
e B e A e B e A
) (x, u
) i (
4
) i (
4
) i (
3
) i (
3
) i (
2
) i (
2
) i (
1
) i (
1
r
2 2 2 2 1 1 1 1
2 2 2 2 1 1 1 1
(45)
For ( ) x, u
r
to be real, the integration constants 1,2,3,4) (i B , A
i i
= must be complex; by recalling the integration constants, the
displacement function ( ) x, u
r
is written as:
( ) ( )
( ) ( )
+ + + +
+ + +
=
L
x
sin
e sin D cos D e sin D cos D
e sin D cos D e sin D cos D
u
2 2
1 1
2 8 2 7
2 6 2 5
1 4 1 3
1 2 1 1
r
(46)
where 6,7,8) 1,2,3,4,5, (i D
i
= are integration constants. The final solution is the sum of particular integral given by Eq. (35) and
complementary function given by Eq. (46). By substituting in Eq. (31) and utilizing the equilibrium conditions given by Eq. (4),
we obtain the stress functions as follows:
R p
u
R
1
x
u
R
2
x
u
R
12
t E
N
z
4
r
4
3 2 2
r
4
4
r
4 3
(47)
=
dx
p
r
x
p
u
R
1
x
u
R
2
x
u
12
t E
N
5
r
4
4 3 2
r
5
2 4
r
5 3
(48)
=
dxdx
x R
p
R
p
R
p u
R
1
x
u
R
2
x
u
12R
t E
N
x
2
r
2
6
r
6
4 4 2
r
6
2 2 4
r
6 3
x
(49)
=
2
r
3
2 3
r
3 3
x
x
u
R
1
x
u
12
Et
Q (50)
2
r
3
3
r
3
2
3
x
u u
R
1
R 12
Et
Q (51)
By introducing the following vector, displacement function can be written in more compact form:
2
2
1
1
-
2
2
-
1
1
e
e
e
e
where
( )
( )
( )
( )
=
2
2
2
1
1
1
sin
cos
,
sin
cos
(52)
( ) ( ) [ ]
+ =
L
x
sin D cos C x, u
k 0 r
(53)
where [ ]
T
8 2 1
D ,....., D , D D = . In explicit form the stress functions are given by:
( ) [ ] ( ) [ ]
( ) [ ] [ ] ( ) [ ]
( ) [ ] ( ) [ ]
( ) [ ] [ ] ( ) [ ]
+ =
+ =
+ =
+ =
L
x
sin D sin H
12
Et
Q
L
x
cos D cos H
12
Et
Q
L
x
cos D sin H
12
Et
N , M
L
x
sin D cos H
12
Et
N , N , M , M
k 0
3
k 0
3
x
k 0
3
x x
k 0
3
x x
(54)
where H
0
assumes different values for different stress functions as given below:
Chandrasekaran et al. / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 148-171
158
( )
( )
( )
( )
( )
+
+
+
=
Q for
R L
C R L
Q for
R L
C R L
N for
t E L
24p
R L
C L R
M for
L R
C
N for
Et
12pR
R L
C R L
N for
Et
24pL
R L
C R L
M for
R
C
M for
L
C
H
3 2
0
2 2 2
x
2 3
0
2 2 2
x
3 3 4 3
0
2
2 2 2
x
0
3 3 4
0
2
2 2 2
x
3
2
5 2 2
0
2
2 2 2
2
0
x
2
0
2
0
R
2
(55)
The matrices [ ] [ ] , are given by:
[ ]
[ ] [ ] [ ] [ ]
[ ] [ ] [ ] [ ]
[ ] [ ] [ ] [ ]
[ ] [ ] [ ] [ ]
=
T
(2)
(2)
T
(1)
(1)
0 0 0
0 0 0
0 0 0
0 0 0
where
=
(2)
11
(2)
12
(2)
12
(2)
11
(2)
(1)
11
(1)
12
(1)
12
(1)
11
(1)
] [ ,
] [ (56)
[ ]
[ ] [ ] [ ] [ ]
[ ] [ ] [ ] [ ]
[ ] [ ] [ ] [ ]
[ ] [ ] [ ] [ ]
=
(0)
(0)
(0)
(0)
0 0 0
0 0 0
0 0 0
0 0 0
where
=
1 0
0 1
] [
(0)
(57)
( ) ( )
( ) ( ) ( ) [ ]
( ) [ ] ( ) [ ]
( ) ( ) ( ) [ ]
( ) [ ] ( ) [ ]
( ) ( ) [ ] ( ) [ ]
( ) ( ) [ ] ( ) [ ]
+
=
+ + +
=
+
=
+ + +
=
+
=
+ + + +
=
+
=
+ + + +
=
= = = =
= = = =
N for
R L
R L 4
,
R L
6 L - R 2L R
,
R L
R L 4
,
R L
6 L - R 2L R
N for
R L
L R L - R 2
,
R L
15 L 6 R 2L - R
,
R L
L R L - R 2
,
R L
15 L 6 R 2L - R
M for
R
2
,
R
-
,
R
2
,
R
-
M for 0 ,
L
3 4
2 2 2
2
2
2
2
2 2 (2)
12
3 4
4
2
2
2
2
2
4
2
4 2
2
2
2
2 2 2 4 4
(2)
11
3 4
2 2 2
1
2
1
2
1 1 (1)
12
3 4
4
1
2
1
2
1
4
1
4 2
1
2
1
2 2 2 4 4
(1)
11
5 2 2
2
2
2
2
2 2 2 2
2
2
2
2 2 2
2 2 (2)
12
5 2 2
6
2
4
2
2
2
2
2
4
2
6
2
4 4
2
2
2
2
2
4
2
2 2 2 2
2
2
2
4 4
(2)
11
5 2 2
2
1
2
1
2 2 2 2
1
2
1
2 2 2
1 1 (1)
12
5 2 2
6
1
4
1
2
1
2
1
4
1
6
1
4 4
1
2
1
2
1
4
1
2 2 2 2
1
2
1
4 4
(1)
11
2
2 2 (2)
12
2
2
2
2
2 (2)
11
2
1 1 (1)
12
2
2
1
2
1 (1)
11
x
(2)
12
(1)
12
2
2
(2)
11
(1)
11
x
3 3
15
3 3
15
(58)
Chandrasekaran et al. / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 148-171
159
( ) ( ) [ ]
( ) ( ) [ ]
( ) ( ) [ ]
( ) ( ) [ ]
+ + +
=
+ +
=
+ + +
=
+ +
=
= = = =
x
4 3
4
2
2
2
2
2
4
2
4 2
2
2
2
2 2 2 4 4
2 (2)
12
4 3
4
2
2
2
2
2
4
2
4 2
2
2
2
2 2 2 4 4
2 (2)
11
4 3
4
1
2
1
2
1
4
1
4 2
1
2
1
2 2 2 4 4
1 (1)
12
4 3
4
1
2
1
2
1
4
1
4 2
1
2
1
2 2 2 4 4
1 (1)
11
x
2 (2)
12
2 (2)
11
1 (1)
12
1 (1)
11
N for
R L
10 5 L 3 R 2L R
R L
5 10 L 3 R 2L R
R L
10 5 L 3 R 2L R
R L
5 10 L 3 R 2L R
M for
RL
,
RL
,
RL
,
RL
(59)
( ) [ ]
( ) [ ]
( ) [ ] ( ) [ ]
( ) [ ] ( ) [ ]
+
=
=
+
=
=
=
+
=
=
+
=
Q for
R L
3 L R
,
R L
3 L R
R L
3 L R
,
R L
3 L R
Q for
LR
2
,
R L
L R
LR
2
,
R L
L R
3 2
2
2
2
2
2 2 2
2 (2)
12
3 2
2
2
2
2
2 2 2
2 (2)
11
3 2
2
1
2
1
2 2 2
1 (1)
12
3 2
2
1
2
1
2 2 2
1 (1)
11
x
2
2 2 (2)
12
2 3
2
2
2
2
2 2 2
(2)
11
2
1 1 (1)
12
2 3
2
1
2
1
2 2 2
(1)
11
(60)
An open barrel reinforced concrete cylindrical shell with geometric properties given in Table 1 is analyzed under uniform surface
load varying sinusoidally in addition to the sinusoidal symmetric edge loads namely: i) fixed unitary moment; ii) unit axial edge
load; and iii) unit shear edge load, considered to act one-by-one successively along the longitudinal boundaries. Figures 4-5 show
the stress resultants obtained from the analysis using proposed close form expressions, for different values of . It can be seen
from the figures that stress resultants are symmetric with respect to the crown of the shell and qualitatively similar under different
edge loads. It is also seen that the obtained results on the basis of proposed expressions closely agree with those obtained from the
detailed FEM analysis based on the procedures discussed by Stanley and Huges (1984). The stress resultants obtained from
detailed FEM analysis are shown in Figure 6, for comparison.
Table 1 Geometric properties of the RC shell (case i)
Description values
k
45
L 26.67m
R 8m
t 80mm
Concrete Mix and steel grade M
25
and Fe 415 steel
Dead load 25 x 0.08 = 2.0 kN/m
2
Live load for inspection 0.5 kN/m
2
Load due to finishes 0.75 kN/m
2
Total load 3.25 kN/m
2
Youngs modulus of concrete, E 25 x 10
6
kN/m
2
Poissons ratio of concrete, 0.20
Chandrasekaran et al. / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 148-171
160
-0.15
-0.10
-0.05
0.00
0.05
0.10
0.15
0.20
0.25
0 10 20 30 40 50 60 70 80 90
Analytical
FEM
-1.00
-0.50
0.00
0.50
1.00
1.50
0 10 20 30 40 50 60 70 80 90
-6.00
-5.00
-4.00
-3.00
-2.00
-1.00
0.00
0 10 20 30 40 50 60 70 80 90
-35.00
-30.00
-25.00
-20.00
-15.00
-10.00
-5.00
0.00
0 10 20 30 40 50 60 70 80 90
-3.00
-2.00
-1.00
0.00
1.00
2.00
3.00
0 10 20 30 40 50 60 70 80 90
-1.5
-1.0
-0.5
0.0
0.5
1.0
0 10 20 30 40 50 60 70 80 90
Analytical
FEM
Analytical
FEM
Analytical
FEM
Analytical
FEM
Analytical
FEM
(deg)
(deg)
(deg)
(deg)
(deg)
M
x
(
k
N
-
m
/
m
)
M
(
k
N
-
m
/
m
)
(deg)
N
x
(
k
N
/
m
)
N
(
k
N
/
m
)
Q
(
k
N
/
m
)
u
z
(
m
m
)
Figure 4. Stress resultants for RC shell at = 45
Chandrasekaran et al. / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 148-171
161
-0.08
-0.06
-0.04
-0.02
0.00
0.02
0.04
0.06
0.08
0.10
0 10 20 30 40 50 60 70
-0.40
-0.30
-0.20
-0.10
0.00
0.10
0.20
0.30
0.40
0.50
0.60
0 10 20 30 40 50 60 70
-6.60
-6.40
-6.20
-6.00
-5.80
-5.60
-5.40
-5.20
0 10 20 30 40 50 60 70
-41.00
-40.00
-39.00
-38.00
-37.00
-36.00
-35.00
-34.00
-33.00
0 10 20 30 40 50 60 70
-2.00
-1.50
-1.00
-0.50
0.00
0.50
1.00
1.50
2.00
0 10 20 30 40 50 60 70
-0.00030
-0.00025
-0.00020
-0.00015
-0.00010
-0.00005
0.00000
0.00005
0 10 20 30 40 50 60 70
Analytical
FEM
Analytical
FEM
Analytical
FEM
Analytical
FEM
Analytical
FEM
Analytical
FEM
(deg) (deg)
(deg)
(deg)
(deg)
(deg)
M
x
(
k
N
-
m
/
m
)
N
x
(
k
N
/
m
)
Q
(
k
N
/
m
)
M
(
k
N
-
m
/
m
)
N
(
k
N
/
m
)
u
z
(
m
m
)
Figure 5. Stress resultants for RC shell at = 35
Chandrasekaran et al. / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 148-171
162
N
x
(kN/m) N
(kN/m) M
x
(kN-m/m)
M
(kN-m/m) Q
(kN/m) Displacement u
z
(mm)
Figure 6. Stress resultants and displacement of RC shell obtained from FEM at = 45
3.3 Design charts
After satisfactory comparison of stress resultants and displacement of RC shells obtained by employing the proposed expressions
with that of detailed FEM, RC shell is now analyzed for different R/L ratios and angles of curvature to obtain Design charts for
closer intervals of while Table 2 shows the stress resultants for edge loads. Based on classical flexural theory, using the proposed
expressions in close form, Design charts are developed for computing the stress resultants in open cylindrical shells with varying
geometric parameters. Figures 7-9 show the proposed Design charts plotted for different R/L ratios for practical cases of R/t and
angles those can be readily used for analyzing the shells without undergoing detailed computations. The expressions are also
proposed in closed form for analysis of RC shells under sinusoidal surface load in addition to unit edge loads of different types. It
is seen from the design curves that they follow the same trend for all edge load cases namely: i) fixed moment couple along
longitudinal edge; ii) axial load along the longitudinal edge; as well as iii) shear edge load. The design curves can be readily used
to obtain the stress resultants for RC open barrel cylindrical shell for varying R/L values. The stress resultants can also be obtained
for any desired values of that can be necessary for detailing of reinforcement, particularly near the crown and the valley. The
numerical problem solved using the design charts show computation of final stress resultants at closer intervals of . Results
obtained closely agree with that of the results obtained through finite element analysis. The proposed design curves shall be readily
used in design offices that do not have access to high end software. Further, they shall also serve as an easy checking tool for
assessing the correctness of the results given by the software. The proposed design curves shall enable the design engineers to
understand the qualitative range of deviation, if any, from the results given by the software.
Table 2 Stress resultants of RC shell under sinusoidal surface load and different edge loads (case ii)
Shells under uniform distribued load
Surface load + moment edge load Surface load + axial edge load Surface load + shear edge load
f N N M Q
(deg) (kN/m) (kN/m) (kN-m/m) (kN-m/m) (kN/m) (kN/m) (kN/m) (kN-m/m) (kN-m/m) (kN/m) (kN/m) (kN/m) (kN-m/m) (kN-m/m) (kN/m) (kN/m) (kN/m) (kN-m/m) (kN-m/m) (kN/m)
0 -4.988 -31.177 0.231 1.447 -2.298 1283.62 0.000 2.237 1.000 0.000 1372.76 1.00 2.52 0.00 0.00 1482.95 0.00 2.98 0.00 -1.00
10 -4.560 -28.499 -0.070 -0.439 -0.566 181.59 -8.464 1.506 -0.445 -2.947 176.91 -8.13 1.67 -1.44 -3.09 153.08 -9.82 1.93 -2.50 -3.87
20 -4.199 -26.247 -0.088 -0.549 0.278 -218.88 -24.066 0.795 -4.875 -4.042 -241.31 -24.44 0.86 -5.99 -4.19 -277.95 -26.46 0.96 -7.76 -4.66
30 -3.940 -24.625 0.004 0.023 0.449 -261.97 -35.258 0.216 -9.511 -3.164 -271.41 -35.80 0.22 -10.73 -3.25 -274.24 -37.25 0.20 -12.83 -3.44
40 -3.804 -23.778 0.081 0.507 0.197 -211.41 -40.322 -0.110 -12.285 -1.164 -206.95 -40.77 -0.14 -13.55 -1.19 -180.18 -41.57 -0.22 -15.74 -1.23
50 -3.804 -23.778 0.081 0.507 -0.197 -211.41 -40.322 -0.110 -12.285 1.164 -206.95 -40.77 -0.14 -13.55 1.19 -180.18 -41.57 -0.22 -15.74 1.23
60 -3.940 -24.625 0.004 0.023 -0.449 -261.97 -35.258 0.216 -9.511 3.164 -271.41 -35.80 0.22 -10.73 3.25 -274.24 -37.25 0.20 -12.83 3.44
70 -4.199 -26.247 -0.088 -0.549 -0.278 -218.88 -24.066 0.795 -4.875 4.042 -241.31 -24.44 0.86 -5.99 4.19 -277.95 -26.46 0.96 -7.76 4.66
80 -4.560 -28.499 -0.070 -0.439 0.566 181.59 -8.464 1.506 -0.445 2.947 176.91 -8.13 1.67 -1.44 3.09 153.08 -9.82 1.93 -2.50 3.87
90 -4.988 -31.176 0.231 1.447 2.298 1283.62 0.000 2.237 1.000 0.000 1372.76 1.00 2.52 0.00 0.00 1482.95 0.00 2.98 0.00 1.00
f x
M
x
f f N N M Q f x
M
x
f f N N M Q f x
M
x
f f N N M Q f x
M
x
f f
k
=45
p(x) sin(
x
L
(
p.
=
M
M
k
=45
k
=45
N N
p(x) sin(
x
L
(
p.
=
Q
Q
p(x) sin(
x
L
(
p.
=
k
=45
p
z
=constant
fixed edge
fixed edge
R
90
0.5
1.0
1.5
2.0
2.5
-12
-10
-8
-6
-4
-2
-4
-2
2
4
(deg)
M
x
(
k
N
-
m
/
m
)
M
(
k
N
-
m
/
m
)
-0.5
0.0
-40
-30
-20
-10
2
0
-14
(deg)
Q
(
k
N
/
m
)
N
(
k
N
/
m
)
N
x
(
k
N
/
m
)
(deg)
(deg)
3
1
0
-1
-3
-5
-15
0
-25
-35
-45
500
1000
1500
0
-500
1250
750
250
-250
(deg)
10 80 70 60 50 40 30 20
90 10 80 70 60 50 40 30 20
90 10 80 70 60 50 40 30 20
90 10 80 70 60 50 40 30 20
-5
R/L
R/L =0.3,0.35,0.4,0.45,0.5
R/L=0.3
R/L=0.35
R/L=0.4
R/L=0.45
R/L=0.5
R/L=0.3
R/L=0.35
R/L=0.5
R/L=0.3
R/L=0.35
R/L=0.4
R/L=0.45
R/L=0.5
R
/L
=
0
.3
5
R
/L
=
0
.4
R
/L
=
0
.4
5
R
/L
=
0
.5
R
/L
=
0
.3
k
=45
M sin(
x
L
(
1
.
=
= 0
=
k
2
N 0 =
Q 0 =
N 0 = x
for
R/t=100
Rck=25N/mm
2
p(x) sin(
x
L
(
p
.
=
Surface load
Edge load
LOAD CASE=Surface load + Edge load
M
M
Figure 7. Design charts for RC shells under surface load and moment edge loads
Chandrasekaran et al. / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 148-171
165
45
0.5
1.0
1.5
2.0
2.5
-
14
-
12
-
10
- 8
- 6
- 4
- 2
-4
-2
2
4
-0.5
-1
-3
0
1
3
3.0
0.0
-1.0
M
x
(
k
N
-
m
/
m
)
Q
(
k
N
/
m
)
M
(
k
N
-
m
/
m
)
500
1000
1500
-40
-30
-20
-10
5
-5
-500
0
(deg)
(deg)
(deg)
-5
-15
-25
-35
-45
N
(
k
N
/
m
)
(deg)
(deg)
90 10 80 70 60 50 40 30 20
90 10 80 70 60 50 40 30 20
90 10 80 70 60 50 40 30 20
90 10 80 70 60 50 40 30 20
90
10 80 70 60 50 40 30 20
R/L
R/L =0.3,0.35,0.4,0.45,0.5
R/L=0.3
R/L=0.35
R/L=0.4
R/L=0.45
R/L=0.5
R
/L
=
0
.3
5
R
/L
=
0
.4
R
/L
=
0
.4
5
R
/L
=
0
.5
R
/L
=
0
.3
R/L=0.3
R/L=0.35
R/L=0.5
R/L=0.3
R/L=0.35
R/L=0.4
R/L=0.45
R/L=0.5
k
=45
M
sin(
x
L
(
1
.
=
= 0
=
k
2
N
0
=
Q 0 =
N 0 = x
for
R/t=100
Rck=25N/mm
2
N N
N
x
(
k
N
/
m
)
p(x) sin(
x
L
(
p
.
=
Surface load
Edge load
LOAD CASE=Surface load + Edge load
Figure 8. Design charts for RC shells under surface load and axial edge loads
Chandrasekaran et al. / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 148-171
166
1.0
2.0
3.0
-15.0
-10.0
-5.0
-4
-2
2
4
M
x
(
k
N
-
m
/
m
)
0.5
1.5
2.5
0.0
-0.5
3.5
M
(
k
N
-
m
/
m
)
-1
-3
1
3
0
5
-5
0
-2.5
-7.5
-12.5
-17.5
Q
(
k
N
/
m
)
500
1000
1500
-40
-30
-20
-10
-5
-15
-25
-35
-45
(deg)
(deg)
(deg)
(deg)
(deg)
-500
1250
750
250
0
-250
-1.0
90 10 80 70 60 50 40 30 20
90 10 80 70 60 50 40 30 20
90 80 70 60 50 40 30 20
90 10 80 70 60 50 40 30 20
90 10 80 70 60 50 40 30 20
R/L
R/L =0.3,0.35,0.4,0.45,0.5
R/L=0.3
R/L=0.35
R/L=0.4
R/L=0.45
R/L=0.5
R
/L
=
0
.3
5
R
/L
=
0
.4
R
/L
=
0
.4
5
R
/L
=
0
.5
R
/L
=
0
.3
R/L=0.3
R/L=0.35
R/L=0.5
R/L=0.3
R/L=0.35
R/L=0.4
R/L=0.45
R/L=0.5
k
=45
M
sin(
x
L
(
1
.
=
= 0
=
k
2
N 0 =
Q =
N 0 = x
for
R/t=100
Rck=25N/mm
2
0
N
x
(
k
N
/
m
)
p(x) sin(
x
L
(
p
.
=
LOAD CASE=Surface load + Edge load
Surface load
Edge load
Q
Q
Figure 9. Design charts for RC shells under surface load and shear edge loads
Chandrasekaran et al. / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 148-171
167
4. Axial force-bending moment interaction
Concrete under multi-axial compressive stress state exhibits significant nonlinearity that can be successfully represented by
nonlinear constitutive models capable of handling inelastic deformations and cyclic loading. Further, it is necessary to describe a
suitable failure criterion for complete description of ultimate strength. A reinforced concrete element of rectangular cross-section
is studied for axial force-bending moment yield interaction behaviour for different percentage of tension and compression
reinforcements (Chandrasekaran et al. 2009). Bernoulli hypothesis of linear strain over the cross-section, both for elastic and
elastic-plastic responses of the element, under bending moment combined with axial force was assumed. The interaction behaviour
becomes critical when one of the following conditions apply namely: i) strain in reinforcing steel in tension reaches ultimate limit;
ii) strain in concrete in extreme compression fibre reaches ultimate limit; as well as iii) maximum strain in concrete in compression
reaches elastic limit under only axial compression. Axial force-bending moment limit domain consisting of six sub-domains are
described; collapse in sub-domains (1) and (2) is caused by yielding of steel whereas for sub-domains (3) to (6), the collapse is
caused by crushing of concrete. Table 3 gives the resume of expressions for various sub-domains, as reported by the researchers; in
general terms, axial force-bending moment interaction can be expressed
as below:
( ) ( ) ( )
( ) ( ) ( ) ( )
+ +
=
+ =
d
2
D
A A dy y
2
D
y b M
A A dy y b N
sc sc st st c c u
sc sc st st c c u
c
c
A
c
A
c
x
x
(61)
where A
c
is the area of concrete in compression; if the section is in full tension, the integral vanishes as the neutral axis is negative.
For the neutral axis greater than the depth of the section, say in this case, thickness of the shell, the integral shall extend for the
total thickness of the shell. In general, as the shell remains in compression, the depth of neutral axis is always greater than the
thickness of the shell. Figure 10 shows the cross-section element of the shell; axial force normal to the section, N
and moment
about the section, M
are now considered for studying the interaction. The stress resultants (N
e,
M
e
) are determined from the
proposed equations (case i) presented above and the values are plotted in the P-M domain. Points A to E correspond to shells with
different values of
k
. It can be seen that the stress resultants lie in the sub-domain (3) for all values of
k
indicating a compression
failure initiated by crushing of concrete. For the failure point on the P-M boundary, for example, A , following equation holds
good:
0 ) (x M
M
N
) (x N
c u
e
e
c u
=
(62)
By iteration, Eq. (62) is solved for x
c
and further by substitution, (N
u,
M
u
) are determined using Eq. (30). Depths of neutral axis
thus obtained for different angles of curvature and the corresponding points on the P-M boundary namely: A to E are shown in
Figure 9. Factor of safety (F.S) is obtained from the following relationship:
AO
O A
F.S
= (63)
Using the above equation, factor of safety for shells with different
k
are computed and shown in Figure 10. It can be seen from the
figure that factor of safety reduces with increase in angle of curvature of the shell. Also, strain in concrete, for all cases, reaches
ultimate limit prompting a compression failure while those in tension and compression steel are within the yield limits. The failure
points interpreted on the P-M domain show compression failure of shell section initiating crushing of concrete.
Chandrasekaran et al. / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 148-171
168
Table 3. Resume of expressions for P-M interaction (Chandrasekaran et. al. 2009)
s
u
b
-
c
x ( )
c
x q ( )
c max , c
x ( )
c st
x ( )
c sc
x
( )
c u
x P
( )
c u
x M
(1)
, ]
0
d x D
x
c
c su
su
d x D
d x
c
c
su
st , u
P
st , u
M
(2a)
]
0
d x D
x
c
c su
su
d x D
d x
c
c
su
( ) 0 q P P
con , 1 u st , u
= + ( ) 0 q M M
con , 1 u st , u
= +
(2b)
] ( ) d x D x
c
su
0 c
c
d x D
x
c
c su
su
d x D
d x
c
c
su con , 1 u st , u
P P +
con , 1 u st , u
M M +
(3)
c
cu
0 c cu
x
cu
( )
c
c cu
x
d x D ( )
c
c cu
x
d x
con , 1 u st , u
P P +
con , 1 u st , u
M M +
(4)
c
cu
0 c cu
x
cu
( )
c
c cu
x
d x D ( )
c
c cu
x
d x
con , 1 u st , u
P P +
con , 1 u st , u
M M +
(5)
]
c
cu
0 c cu
x
cu
( )
c
c cu
x
d x D ( )
c
c cu
x
d x
con , 1 u st , u
P P +
con , 1 u st , u
M M +
(6) D
cu
0 c cu
( ) 0 c cu
c cu
c 0 c cu
D x
x
( )
( )
0 c cu c cu
c 0 c cu
D x
d x D
( )
( )
0 c cu c cu
c 0 c cu
D x
d x
con , 2 u con , 1 u st , u
P P P + +
con , 2 u con , 1 u st , u
M M M + +
]
x
'
c
0
0, [
[ x
'
c ,
x
"
c
] [ x
"
c
,
x
"'
c
] [x
"'
c,
D-d
[D-d,
D
[D,
+ [
d
o
m
a
i
n
( ) ( ) 0 y , x x
c c c max , c
= = ( )( )
st t sc c st st sc sc st , u
p p d D b A A P = =
( ) ( )( )
= d
2
D
p p d D b d
2
D
A A M
st t sc c st st sc sc st , u
( )
( ) [ ]
c max , c c 0 c
2
0 c
2
c
2
c max , c 0 c
0 c con , 1 u
x q x 3
x 3
x q b
bq P +
+ =
( )
( )
( ) ( )( ) [ ] D 2 q 3 x x q D 3 x 2 q 4 x 2
x 12
x q b
q D
2
bq
M
c c max , c c c 0 c
2
0 c
2
c
2
c max , c 0 c 0 c
con , 1 u
+ + +
=
( ) [ ]( )
2
0 c
2
c
2
c c max , c c 0 c max , c 0 c
con , 2 u
x 3
x D x D x 3 b
P
+
=
( ) ( ) ( ) [ ]
2
0 c
2
c
2
c
2
max , c c c 0 c
2
c max , c 0 c
con , 2 u
x 12
x D x 2 D x 2 x D b
M
+ +
=
Point N M F.S.
(deg) (kN/m) (kN-m/m) (kN/m) (kN-m/m)
(m) (kN/m )
A 30 87.162 0.628 862.85 6.22 9.90
0.079 0.003500 11023 -0.00083 -174581 0.0026 360870
B 35 89.881 1.477 631.24 10.38 7.02
0.059 0.003500 11023 0.00006 12243 0.0023 360870
C 40 92.801 2.708 421.08 12.29 4.54
0.044 0.003500 11023 0.00127 266762 0.0019 360870
D 45 95.928 4.451 239.47 11.11 2.50
0.030 0.003500 11023 0.00339 360870 0.0012 252626
E 50 99.236 6.849 132.09 9.12 1.33
0.023 0.003500 11023 0.00549 360870 0.0005 105996
N M f f f
k u u e e f f
x
c
e
c,max
s
c,max
e
st
s
st
s
sc
e
sc
-250
0
250
500
750
1000
1250
-12 -10 -8 -6 -4 -2 2 4 6 8 10 12
A'
B'
C'
D'
E'
2a
2b
3
4
5
6
M
u
(kN-m)
P
u
(kN-m)
Point
A'
B'
C'
D'
E'
k
=45
p
z
=constant
fixed edge
fixed edge
R
2
(kN/m )
2
(kN/m )
2
Point
A'
B'
C'
D'
E'
A
B
C D
E
1.00m
8
0
m
m
p
z
=3.25 kN/m
2
p
t
= p
c
= 0.5%
R
ck
= 25N/mm
2
f
y
= 415N/mm
2
R/L=
R/t=
0.3
100
R=8m Nf
Mf
Figure 10. Axial force-bending moment interaction of RC shells under uniformly distributed surface loads (P-M interaction)
5. Conclusions
Design curves for stress resultants based on the classical flexural theory for single barrel open cylindrical shells are presented
with analytical expressions in closed form. The effect of variation of angle of curvature of shell geometry on the final stress
resultants is demonstrated using the proposed design curves. Plots of axial force-bending moment values on the P-M domain
are
illustrated to show the compression failure in RC shells initiated by crushing of concrete. As the behavior of the plots for various
values of R/L is nonlinear, interpolations for still closer values of fraction of R/L is not recommended.
Common theories adopted for analysis of cylindrical shells interpret the design behavior in different ways. Therefore every
design procedure developed without undergoing rigorous analysis leads to simple and close form solution but of course with some
defects. Despite shell structures possess inherent advantages gained by its structural form, they are not very common due to their
analysis complexities. Current study attempts to make shell analysis easy, popular and develops confidence in design offices,
encouraging the designer to handle a complex problem through a simple graphical tool. With the use of proposed Design charts,
stress resultants can be readily determined at the valley, crown or at any desired section for various type of edge loads. In the
present context of design offices highly influenced by use of software, this study is an attempt in the direction of developing
simple graphical tools based on well established theories so that the designer is sure to know the error, if committed both
qualitatively and quantitatively.
Nomenclature
R radius (m) L length (m)
k
angle of the shell (deg) t thickness of the shell (mm)
E Youngs modulus of concrete (kN/m
2
) Poissons ratio of concrete
generic angle in shell (deg)
x
M Bending moment in plane x-z
x
shear strain in plane x-
x
curvature in plane x-z
curvature in plane r-
x
curvature in plane x-z
x
p surface load on the shell in x-direction (kN/m
2
)
c
strain in generic fibre of concrete
c,max
maximum strain in concrete
c0
elastic limit strain in concrete
cu
ultimate limit strain in concrete
st
strain in tensile reinforcement
sc
strain in compression reinforcement
s0
elastic limit strain in reinforcement
su
ultimate limit strain in reinforcement
E
s
modulus of elasticity in steel (kN/m
2
) R
ck
compressive cube strength of concrete (kN/m
2
)
c
stress in generic fibre of concrete (kN/m
2
)
c,max
maximum stress in concrete (kN/m
2
)
c0
design ultimate stress in concrete in compression (kN/m
2
)
Chandrasekaran et al. / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 148-171
171
y
yield strength of steel (kN/m
2
)
s0
design ultimate stress in steel (kN/m
2
)
st
stress in tensile reinforcement (kN/m
2
)
sc
stress in compression reinforcement (kN/m
2
)
p
t
percentage of tensile reinforcement p
c
percentage of compression reinforcement
A
st
area of tension reinforcement (m
2
) A
sc
area of compression reinforcement (m
2
)
c
A area of concrete in compression F.S. safety of safety in P-M interaction
e
N Axial force in -direction in elastic range (kN)
u
N
=
K
x
rx
dt
dx
1 (1)
where r is the intrinsic growth rate, K is the environmental carrying capacity & x is the biomass. Equation (1) implies a parabolic
growth curve, where the logistic function is strictly concave from below and exhibits positive growth for all positive values
of . K x < The logistic growth function is assumed to describe a process of feedback which controls the growth of the population
as its level increases. The logistic growth curve is symmetrical round its point of inflexion. This simple model has been used
widely both from a theoretical standpoint and as a convenient empirical curve.
2.2. Gompertz growth model
The Gompertz growth model proposed by Gompertz, is given by
=
x
K
rx
dt
dx
e
log (2)
The basic difference between the logistic and Gompertz curves is that the logistic is symmetrical while the Gompertz is
asymmetrical (inflecting at K 0.368 K/e x = = ). For a given stock size at time 0 t = and a given carrying capacity, the Gompertz
model may expect a faster growth of the stock size over time than the Logistic model.
These population growth models implies that an increase in the stock biomass leads to an increase in the catch at the same rate,
keeping the fishing effort unchanged.
2.3. Harvest function
The harvesting function ) (t H is derived using the following two assumptions (Clark, 1990): (a) catch per unit effort is directly
proportional to the density of fish, and (b) the density of fish is directly proportional to the abundance ). (t x
Thus we assume the following production function, which is commonly used in fishery management as
qEx t H = ) ( , (3)
where q is the catchability coefficient, E is the fishing effort. Similarly the biomass variable x is the quantity of harvested fish in a
fishing year (to be measured in Bushel).
If the population described by the logistic equation (1) is subject to harvesting at a rate ) (t H , then equation (1) becomes
. 1 qEx
K
x
rx
dt
dx
= (4)
At equilibrium, from equation (4) we get the fish biomass
L
x as
Kar
and Chakraborty / International Journal of Engineering, Science and Technology,
Vol. 1, No. 1, 2009, pp. 172-189
175
). ( qE r
r
K
x
L
= (5)
Thus from equation (3) harvesting at equilibrium can be obtained as
). ( qE r
r
K
qE H
L
= (6)
Again equation (2) subject to the harvesting function (3) is reduced to
. log qEx
x
K
rx
dt
dx
e
= (7)
At equilibrium, from equation (7) we get the fish biomass
G
x as
.
r
qE
G
Ke x
= (8)
Thus from equation (3) harvesting at equilibrium can be obtained as
.
r
qE
G
qKEe H
= (9)
2.4 Economic model
To attain efficiency in the economic sense, we need to take into account the costs of fishing and revenues that we get from
selling the harvested fish. It is necessary to use catch-effort relationship to define revenues and costs as a function of fishing effort.
Assuming a constant price, , p per unit of fish harvested, the total revenue, , TR will be given by; ). ( ) ( E pH E TR =
And assuming a constant cost, , c per unit of effort, the total costs, , TC of fishing is given by; . ) ( cE E TC =
The difference between the total revenue of the fishery and the total fishing costs is known as the sustainable economic rent
(profit) provided by the fishery resource at any given level of fishing effort E thus sustainable economic rent can be defined as
). ( ) ( ) ( E TC E TR E = (10)
3. Reference points
In this section, we derive the analytical expressions of maximum economic yield (MEY), open-access equilibrium (OAE) and
the biological equilibrium (MSY) in terms of biological parameters along with economic variables separately for logistic and
Gompertz models. These reference points are analyzed for the future management policies of a fishery and sustainable
development of ecosystem.
3.1 Maximum sustainable yield (MSY)
The equilibrium level of the fishing effort that produces the maximum sustainable yield, found by differentiating equations (6)
and (9) with respect to E and setting the result equal to zero gives us maximum sustainable yield. For logistic model
q
r
E
MSY
2
= (11)
and
4
Kr
H
MSY
= . (12)
For the Gompertz model we get,
q
r
E
MSY
= (13)
Kar
and Chakraborty / International Journal of Engineering, Science and Technology,
Vol. 1, No. 1, 2009, pp. 172-189
176
and
.
e
Kr
H
MSY
= (14)
3.2 Open access equilibrium (OAE)
The bionomic or open access equilibrium (OAE) can be obtained from the fishery when no economic rent is obtained from the
fishery or profit is zero, i.e., ) ( ) ( E TR E TC = which gives . ) ( cE E pH = For logistic model,
. 1
=
pqK
c
q
r
E
OAE
(15)
Now to get the yield level at open access let us substitute the effort of open access in the equation (6), which gives
. 1
=
pqK
c
pq
rc
H
OAE
(16)
The expressions for Gompertz model are
, log
=
c
pqK
q
r
E
e OAE
(17)
and
. log
=
c
pqK
pq
rc
H
e OAE
(18)
3.3 Maximum economic yield (MEY)
The level of harvesting which maximize the profits to the fishery is determined by maximum economic yield (MEY). Maximum
economic yield can be obtained from the fishery when the difference between total revenue earned by the fishery and total cost
used to harvest is at a maximum. The equilibrium level of fishing effort that produces the maximum economic, rent can be derived
by differentiating equation (10) with respect to fishing effort ). (E Thus for logistic growth model we get
=
pqK
c
q
r
E
MEY
1
2
(19)
and
.
4
2 2
2
=
K q p
c
K
r
H
MEY
(20)
Again, for the Gompertz growth model we have got the following expressions:
q
r
E
MEY
) 1 (
= where
Kpq
ce
e =
(21)
and
.
) 1 (
) 1 (
=
e
rK
H
MEY
(22)
Kar
and Chakraborty / International Journal of Engineering, Science and Technology,
Vol. 1, No. 1, 2009, pp. 172-189
177
4. Optimal sustainable yield (OSY)
The equation that maximize the present value (PV) of the fishery can be expressed as
=
0
) (
) (
) ( max dt t H
t qx
c
t pH e PV
t
, (23)
subject to ) (
) (
1 ) ( t H
K
t x
t rx
dt
dx
+ = ) (
) (
1 ) ( ) (
) (
) ( t H
K
t x
t rx t H
t qx
c
t pH L (24)
where is the adjoint variable. The Hamiltonian must be maximized for ] , 0 [
max
H H . Assuming that the control constraints are
not binding (i.e. the optimal solution does not occur at 0 or
max
H ) and = qx c p t / ) ( is called the switching function (see
Clark, 1990).
Since Hamiltonian L is linear in the control variable, the optimal control will be a combination of extreme controls and the
singular control. The optimal control ) (t H that maximizes L must satisfy the following conditions:
max
H H = , when 0 ) ( > t , i.e. qx c p / < ,
and
0 = H , when 0 ) ( < t , i.e. qx c p / > .
The above condition simply states that the resource should be harvested if and only if the net revenue per unit harvest exceeds
the shadow price of the resource stock.
When 0 ) ( = t , the Hamiltonian L becomes independent of the control variable H, i.e., 0 / = H L . This is the necessary
condition for the singular control ) (
*
t H to be optimal over the control set . 0
max
*
H H < <
Thus the optimal harvesting policy is
=
<
>
=
0 ) (
0 ) ( , 0
0 ) (
) (
*
max
t when H
t when
t when H
t H
Again the optimal stock level is such that the rent earned by the stock equals the social rate of discount. The optimal stock
solution is given by
Kq
c rp
rp
K
qK
cr
r p
rp
K
pq
c
r
K
x
8
4
) (
4 4
1
4
2 2 2
*
+ +
= .
Proceeding in the similar manner it is possible to obtain the optimal stock corresponding to Gompertz model by solving the
following equation:
. 0 1 1 log =
pqx
c
r x
K
e
It is noted that the optimal stock solution for both the models depends only on the bioeconomic parameters c, p, q, , r and K.
Kar
and Chakraborty / International Journal of Engineering, Science and Technology,
Vol. 1, No. 1, 2009, pp. 172-189
178
Consequently, the optimal yield and optimal effort are respectively obtained as
=
K
x
rx H
*
* *
1 and
*
*
*
qx
H
E = . (25)
Thus the optimal profit of Marylands Chesapeake Bay oyster fishery is given by
*
*
* *
) , , , , , ( H
qx
c
pH x c r K q p = . (26)
The present value of profits of Marylands Chesapeake Bay oyster fishery is reduced to
1
) (
0
= =
dt e PV
t
. (27)
5. Tax policies to achieve reference points
Management of fisheries has the potential to yield a socially optimal outcome through proper utilization of different types of tax
policies. These include maximization of fishing production or revenues; minimize catch fluctuations, to avoid the risk of collapse
of the resource, etc. The manager has two kinds of tools available: first one is technical tool which includes effort or selective
regulations; second is economic tool which includes subsidies and taxes. The effectiveness of each of these tools varies in case of
achieving the purpose of the fishery managers and has significant implications for the control of the stock. We can take taxes in the
form of license fees, and are levied prior to any harvest and effort being expended, or in the form of a per unit effort tax. Again, we
are concerned about several problems that are imposed for fishery managers due to input taxes such as they may not have a
predicable effect on the level of effort expended in the fishery. If the tax changes the relative cost of inputs, fishers will substitute
higher cost for lower cost inputs. This result will have an effect on the efficiency of fishers as now they will be using inputs in non-
optimal proportions, that is, they become inefficient transferring a larger amount of fishery rent to cost.
We can consider the outcome of open access fisheries in a case of market failure which ultimately sends wrong message to the
economy. Thus tax policies are considered to be a suitable measure to manage fishery. It can be argued that the market failure is a
result of ill defined property rights to the resources. Therefore introduction of suitable tax policies are of great concern to
overcome property rights problems. It may be assumed that fishery managers should set taxes on landings, or effort, such that
perceived bionomic equilibrium for the fleet will be the optimal stock. Otherwise if the resource stock is driven below the optimal
stock then the revenue earned by the fishery will not exceed the cost used for fishing. In order to apply taxes precisely, we have to
be conscious about the continuous adjustments required to implement the policies due to repeated fluctuations in resource and
economic conditions. Additionally, as the fisheries are not identical in order to achieve the optimum outcome we should have
separate tax for each fishery which is practically impossible due to the insufficient information obtained from the fishery. Again,
the control through the use of taxes is politically unacceptable. Hence more gradual tax programmes, as a means of control in
fisheries, are appropriate, in which the taxes are initially imposed at a low level, and in which there are successive turns of the tax
rotation.
Let us now assume that the Marylands Chesapeake Bay oyster fishery is in open access situation and the controlling agency of
Marylands Chesapeake Bay oyster fishery levis a tax ) 0 (> T in order to achieve
MSY
H ,
OSY
H or
MEY
H by incorporating the
fishing effort equally
MSY
E ,
OSY
E or
MEY
E . Here we derive the expressions to calculate different type of tax policies
using
MSY
H and
MSY
E only but we evaluate the following three types of tax to achieve MSY, OSY and MEY for the Marylands
Chesapeake Bay oyster fishery.
Landing tax
If ) 0 (> T is the landing tax that needs to be imposed to achieve ,
MSY
H
OSY
H or
MEY
H by incorporating the fishing effort equally
,
MSY
E
OSY
E or
MEY
E then we get the following equation:
MSY MSY
cE H T p = ) ( which gives
MSY MSY
H cE p T / = (28)
The intercept between H T p TR ) (
1
= and cE TC = will give us ) ) ( , (
MSY MSY
H T p E .
Kar
and Chakraborty / International Journal of Engineering, Science and Technology,
Vol. 1, No. 1, 2009, pp. 172-189
179
Effort tax
If ) 0 (> T is the effort tax that needs to be imposed to achieve ,
MSY
H
OSY
H or
MEY
H by incorporating the fishing effort equally
MSY
E
OSY
E or
MEY
E then we get the following equation:
MSY MSY
E T c pH ) ( + = which gives c E pH T
MSY MSY
= ] / [ (29)
The intercept between pH TR = and E T c TC ) (
1
+ = will give us ) , (
MSY MSY
pH E .
Entry tax
If ) 0 (> T is the entry tax that needs to be imposed to achieve ,
MSY
H
OSY
H or
MEY
H by incorporating the fishing effort
equally ,
MSY
E
OSY
E or
MEY
E then we get the following equation:
T cE pH
MSY MSY
+ = which gives
MSY MSY
cE pH T = (30)
The intercept between pH TR = and T cE TC + =
1
will give us ) , (
MSY MSY
pH E .
The result of different kind of taxes to achieve MSY, OSY and MEY are given in Table no. 7, 10, 13 and 16. Following figures
are drawn to achieve MEY using the tax values given in Table no. 7&10. It is clearly observed from the figures that maximum
economic yield 4358 bushel for logistic growth model and 4225.64 bushel for Gompertz growth model are achieved through
incorporating tax values in fishing effort (considered as boat days by gear type).
0 1000 2000 3000 4000 5000 6000 7000 8000 9000
0
0.5
1
1.5
2
2.5
3
x 10
6
Fishing effort
T
R
,
T
C
E
MEY
(p-T)*H
MEY
TR=p*H
TC=c*E
TR
1
=(p-T)*H
=TR-TC
Figure 1. Landing tax to achieve MEY corresponding to the Logistic growth model using the effort data as boat days by gear type
of Marylands Chesapeake Bay oyster fishery
Kar
and Chakraborty / International Journal of Engineering, Science and Technology,
Vol. 1, No. 1, 2009, pp. 172-189
180
0 1000 2000 3000 4000 5000 6000 7000 8000 9000
0
0.5
1
1.5
2
2.5
3
x 10
6
Fishing effort
T
R
,
T
C
E
MEY
p*H
MEY
TR=p*H
TC=c*E
TC
1
=(c+T)*E
=TR-TC
Figure 2. Effort tax to achieve MEY corresponding to the Logistic growth model using the effort data as boat days by gear type of
Marylands Chesapeake Bay oyster fishery
0 1000 2000 3000 4000 5000 6000 7000 8000 9000
0
0.5
1
1.5
2
2.5
3
3.5
4
x 10
6
Fishing effort
T
R
,
T
C
E
MEY
p*H
MEY
TR=p*H
TC=c*E
TC
1
=c*E+T
=TR-TC
Figure 3. Entry tax to achieve MEY corresponding to the Logistic growth model using the effort data as boat days by gear type of
Marylands Chesapeake Bay oyster fishery
Kar
and Chakraborty / International Journal of Engineering, Science and Technology,
Vol. 1, No. 1, 2009, pp. 172-189
181
0 1000 2000 3000 4000 5000 6000 7000 8000 9000
0
0.5
1
1.5
2
2.5
3
x 10
6
Fishing effort
T
R
,
T
C
E
MEY
(p-T)*H
MEY
TR=p*H
TC=c*E
TR
1
=(p-T)*H
=TR-TC
Figure 4. Landing tax to achieve MEY corresponding to the Gompertz growth model using the effort data as boat days by gear
type of Marylands Chesapeake Bay oyster fishery
0 1000 2000 3000 4000 5000 6000 7000 8000 9000
0
0.5
1
1.5
2
2.5
3
x 10
6
Fishing effort
T
R
,
T
C
E
MEY
p*H
MEY
TR=p*H
TC=c*E
TC
1
=(c+T)*E
=TR-TC
Figure 5. Effort tax to achieve MEY corresponding to the Gompertz growth model using the effort data as boat days by gear type
of Marylands Chesapeake Bay oyster fishery
Kar
and Chakraborty / International Journal of Engineering, Science and Technology,
Vol. 1, No. 1, 2009, pp. 172-189
182
0 1000 2000 3000 4000 5000 6000 7000 8000 9000
0
0.5
1
1.5
2
2.5
3
3.5
4
x 10
6
Fishing effort
T
R
,
T
C
E
MEY
p*H
MEY
TR=p*H
TC=c*E
TC
1
=c*E+T
=TR-TC
Figure 6. Entry tax to achieve MEY corresponding to the Gompertz growth model using the effort data as boat days by gear type
of Marylands Chesapeake Bay oyster fishery
6. The empirical model
The regression analysis results were obtained using the empirical model described in this section. The parameters of the
biological growth models are estimated using the time series catch and effort data of Marylands Chesapeake Bay. Here in this
model we consider time series data of Marylands Chesapeake Bay from the year 1989 to 2006. Generally we consider catch per
unit effort is directly proportional to the stock of biomass i.e., , x S where S denotes the catch per unit effort and x denotes the
stock of the population at time . t
Using the equation (1) we have the following equation,
. qE S
qK
r
r
S
S
=
&
(31)
From the finite difference approximation S
&
can be written as
.
2
1 1 +
t t
S S
dt
dS
(32)
Thus, .
2
1 1
t t
t
t t
qE S
qK
r
r
S
S S
=
+
It is noted that we can easily calculate CPUE using the time series data of catch and effort, consequently a linear regression of
t t t
S S S Y 2 / ) (
1 1 +
= can be obtained using two independent variables S and E which yield the following equation with the error
term
1
committed to the model during the linear regression analysis using logistic growth model,
,
1 3 2 1
+ + + =
t t
E S Y (33)
where ,
1
r =
,
2
qK
r
=
Kar
and Chakraborty / International Journal of Engineering, Science and Technology,
Vol. 1, No. 1, 2009, pp. 172-189
183
.
3
q =
Thus it is possible to calculate value of the biological parameters of Logistic growth model using the estimated value of
2 1
, and .
3
Again using the equation (2) we can write
. log log qE r r
S
S
S
e
qK
e
=
&
(34)
Thus, for the Gompertz growth model we have the following equation for estimation,
,
2 3 2 1
+ + + =
t t
E S Y (35)
where , log
1
qK
e
r =
,
2
r =
,
3
q =
and
2
be the error committed to the model during the linear regression analysis using Gompertz growth model. Consequently it is
possible to calculate value of the biological parameters of Gompertz growth model using the estimated value of
2 1
, and .
3
Prices are considered as the interaction between the demand and supply of the resource. It is possible to examine the relative
profit through allocating resources between various outputs and allow an optimal decision regarding to the relative production
quantities for the fishery managers. Prices also reflect the relative quantities in which consumers would like to consume resources.
It is assumed that demand is linear in price and stationary, the inverse demand or willing to pay (WTP) function of harvest can be
expressed as
). (
2 1
t H d d p + = (36)
7.1 Results of estimation
In order to estimate the biological parameters of Logistic and Gompertz models we mainly use time series harvest and effort data
of Marylands Chesapeake Bay oyster fishery (Wieland, 2007) from1989 to 2006. For the purpose of estimation we have two
different types of effort data namely boat days by gear type and man days by gear type.
We divide this section in two parts. In the first part we consider effort data as boat days by gear type and estimation is done for
both the biological growth models. Man days by gear type is considered as effort data for the second part and both the biological
models are considered for estimation. The estimation has been done using ordinary least squares method and results are as follows:
Table1. Estimation of the biological parameters of Logistic growth model using the effort data as boat days by gear type of
Marylands Chesapeake Bay oyster fishery
Variable Estimate Std. error t-Statistics P-value
1
0.0725736 0.209626 0.346206 0.734727
2
-0.00244829 0.0173103 -0.141435 0.889694
3
-6
10 1.27039 -
-6
10 2.81883
-0.450681 0.659644
Rsquared = 0.0161406,
AdjustedRSquared = -0.135222,
EstimatedVariance = 0.0167142.
Thus the consequent value of biological parameters is as follows:
0.0725736, = r , 10 33334 . 2
7
= K . 10 1.27039 q
-6
=
Kar
and Chakraborty / International Journal of Engineering, Science and Technology,
Vol. 1, No. 1, 2009, pp. 172-189
184
Table2. Estimation of the biological parameters of Gompertz growth model using the effort data as boat days by gear type of
Marylands Chesapeake Bay oyster fishery
Variable Estimate Std. error t-Statistics P-value
1
0.159956 0.493311 0.324251 0.750906
2
-0.0474837 0.200823 -0.236446 0.816771
3
-6
10 1.29835 -
-6
10 2.81639
-0.460999 0.652417
Rsquared = 0.0188461,
AdjustedRSquared = -0.132101,
EstimatedVariance = 0.0166683.
Thus the consequent value of biological parameters is as follows:
0.0474837, = r , 10 23665 . 2
7
= K . 10 1.29835 q
-6
=
Table3. Estimation of the biological parameters of Logistic growth model using the effort data as man days by gear type of
Marylands Chesapeake Bay oyster fishery
Variable Estimate Std. error t-Statistics P-value
1
0.073271 0.176661 0.414755 0.685084
2
-0.00210312 0.0255024 -0.0824677 0.935531
3
-7
10 8.93172 -
-6
10 16437 . 1
-0.767089 0.456734
Rsquared = 0.0458702,
AdjustedRSquared = -0.100919,
EstimatedVariance = 0.0095219.
Thus the consequent value of biological parameters is as follows:
0.073271, = r , 10 90061 . 3
7
= K . 10 93172 . 8 q
-7
=
Table 4. Estimation of the biological parameters of Gompertz growth model using the effort data as man days by gear type of
Marylands Chesapeake Bay oyster fishery
Variable Estimate Std. error t-Statistics P-value
1
0.119009 0.296561 0.401296 0.694721
2
-0.0315917 0.154788 -0.204096 0.841439
3
-7
10 9.42475 -
-6
10 16451 . 1
-0.809329 0.43289
Rsquared = 0.0484201,
AdjustedRSquared = -0.0979768,
EstimatedVariance = 0.00949645.
Thus the consequent value of biological parameters is as follows:
0.0315917, = r , 10 58942 . 4
7
= K . 10 42475 . 9 q
-7
=
It is possible to estimate inverse demand function by using price and harvest data through ordinary least squares method.
Table 5. Estimation of price of oyster fish using the harvest and price data of Marylands Chesapeake Bay oyster fishery
Variable Estimate Std. error t-Statistics P-value
1
d 23.1238 1.85659 12.455
9
10 19776 . 1
2
d
6
10 00178 . 8
6
10 56569 . 7
-1.05764 0.305927
Rsquared = 0.0653443,
AdjustedRSquared = 0.00692834,
EstimatedVariance = 16.0392.
Kar
and Chakraborty / International Journal of Engineering, Science and Technology,
Vol. 1, No. 1, 2009, pp. 172-189
185
Thus the price ) ( p of oyster fish of the Marylands Chesapeake Bay oyster fishery used in our study is , 1015 . 22 $ = p estimated
price of oyster fish corresponding to the year 2006, ). 127761 ) 2006 ( ( = H
Generally the unit cost of fishing ) (c composed of labour costs, capital, materials, energy, fuel costs and indirect cost. Again
indirect costs incorporate reported maintenance, repair and replacement costs. But these data are not available from the source.
Thus the unit cost of fishing is estimated based on the assumption that Marylands Chesapeake Bay oyster fishery is in open access
equilibrium and we consider that the effort in open access equilibrium is assumed to be equal with the effort level in 2006. If we
consider fishing effort as boat days by gear type then corresponding to the Logistic and Gompertz model respective unit cost of
fishing are $555.187 and $505.719. Again for fishing effort as man days by gear type the unit cost of fishing for Logistic and
Gompertz models are respectively $630.02 and $635.855. The price of oyster fishery and unit cost of fishing are assumed to be
constant over the years.
7.2. Results of bioeconomic model
Static biological equilibrium (MSY), economic optimum (MEY) and economic equilibrium for open-access fisheries (OAE)
evaluated for both the models are presented in the following tables, using the estimated biological parameters of Marylands
Chesapeake Bay oyster fishery along with the price and harvesting cost of oyster fish. Total revenue of the Marylands Chesapeake
Bay oyster fishery, total operational cost of oyster fishing and economic rent of the fishery are also presented in the tables.
Table 6. MSY, MEY, OAE levels and economic rents for Marylands Chesapeake Bay oyster fishery using Logistic growth
model, taking effort data as boat days by gear type
Harvest condition MSY MEY OAE
Fishing effort (boat days
by gear type)
28563.5 4358 8716
Harvest (measured in
Bushel)
423347 119327 218945
Total revenue (TR)
(measured in dollar)
9356610 2637310 4839010
Total operational cost
(TC) (measured in
dollar)
15858100 2419510 4839010
Economic rent (TR-TC)
of the fishery.
-6501490 217806 0
Table 7. Tax policies to achieve MSY, OSY (considering 1% discount rate) and MEY levels for Marylands Chesapeake Bay
oyster fishery using Logistic growth model, taking effort data as boat days by gear type
Landing tax Effort tax Entry tax
Tax to achieve MSY -15.3573 -227.615 -6501490
Tax to achieve OSY 1.70775 46.4908 216745
Tax to achieve MEY 1.82528 49.9784 217806
Table8. Optimal values of biomass, effort, harvest and profit of Marylands Chesapeake Bay oyster fishery using Logistic growth
model, taking effort data as boat days by gear type for different discounting rates
*
x
*
H
*
E
Profits
0
7
10 15534 . 2
119327 4358 217806
0.01
7
10 14292 . 2
126919 4662.11 216745
0.05
7
10 10617 . 2
148814 5561.76 201188
0.1
7
10 07769 . 2
165206 6259.03 176361
7
10 97734 . 1
218945 8716 0
Kar
and Chakraborty / International Journal of Engineering, Science and Technology,
Vol. 1, No. 1, 2009, pp. 172-189
186
Table 9. MSY, MEY, OAE levels and economic rents for Marylands Chesapeake Bay oyster fishery using Gompertz growth
model, taking effort data as boat days by gear type
Harvest condition MSY MEY OAE
Fishing effort (boat days
by gear type)
36572.3 4225.64 8716
Harvest (measured in
Bushel)
390704 109321 199437
Total revenue (TR)
(measured in dollar)
8635150 2416160 4407850
Total operational cost
(TC) (measured in
dollar)
18495300 2136990 4407850
Economic rent (TR-TC)
of the fishery.
-9860190 279170 0
Table10. Tax policies to achieve MSY, OSY (considering 1% discount rate) and MEY levels for Marylands Chesapeake Bay
oyster fishery using Gompertz growth model, taking effort data as boat days by gear type
Landing tax Effort tax Entry tax
Tax to achieve MSY -25.237 -269.608 -9860190
Tax to achieve OSY 2.32329 59.4055 276479
Tax to achieve MEY 2.55367 66.0658 279170
Table11. Optimal values of biomass, effort, harvest and profit of Marylands Chesapeake Bay oyster fishery using Gompertz
growth model, taking effort data as boat days by gear type for different discounting rates
*
x
*
H
*
E
Profits
0
7
10 99259 . 1
109321 4225.64 279167
0.01
7
10 96939 . 1
119003 4654.1 276479
0.05
7
10 90946 . 1
143398 5784.14 244153
0.1
7
10 87013 . 1
158929 6545.47 202402
7
10 76237 . 1
199437 8716 0
Table12.MSY, MEY, OAE levels and economic rents for Marylands Chesapeake Bay oyster fishery using Logistic growth model
taking, effort data as man days by gear type
Harvest condition MSY MEY OAE
Fishing effort (man days
by gear type)
41017.3 7456.5 14913
Harvest (measured in
Bushel)
714504 236166 425107
Total revenue (TR)
(measured in dollar)
15791600 5219620 9395490
Total operational cost
(TC) (measured in
dollar)
25841700 4697750 9395490
Economic rent (TR-TC)
of the fishery.
-10050100 521870 0
Table13. Tax policies to achieve MSY, OSY (considering 1% discount rate) and MEY levels for Marylands Chesapeake Bay
oyster fishery using Logistic growth model taking, effort data as man days by gear type
Landing tax Effort tax Entry tax
Tax to achieve MSY -14.0659 -245.022 -10050100
Tax to achieve OSY 2.069 65.0697 519292
Tax to achieve MEY 2.20976 69.9886 521870
Kar
and Chakraborty / International Journal of Engineering, Science and Technology,
Vol. 1, No. 1, 2009, pp. 172-189
187
Table14. Optimal values of biomass, effort, harvest and profit of Marylands Chesapeake Bay oyster fishery using Logistic growth
model taking, effort data as man days by gear type for different discounting rates
*
x
*
H
*
E
Profits
0
7
10 54607 . 3
236166 7456.5 521870
0.01
7
10 52115 . 3
250988 7980.55 519292
0.05
7
10 44731 . 3
293537 9533.38 481383
0.1
7
10 39013 . 3
325084 10736 420917
7
10 19152 . 3
425107 14913 0
Table15.MSY, MEY, OAE levels and economic rents for Marylands Chesapeake Bay oyster fishery using Gompertz growth
model, taking effort data as man days by gear type
Harvest condition MSY MEY OAE
Fishing effort (man days
by gear type)
33519.9 6474.37 13668.3
Harvest (measured in
Bushel)
533379 230856 393235
Total revenue (TR)
(measured in dollar)
11788500 5102260 8691080
Total operational cost
(TC) (measured in
dollar)
21313800 4116760 8691080
Economic rent (TR-TC)
of the fishery.
-9525340 985501 0
Table16.Tax policies to achieve MSY, OSY (considering 1% discount rate) and MEY levels for Marylands Chesapeake Bay
oyster fishery using Gompertz growth model, taking effort data as man days by gear type
Landing tax Effort tax Entry tax
Tax to achieve MSY -17.8585 -284.17 -9525340
Tax to achieve OSY 3.73644 129.367 965155
Tax to achieve MEY 4.2689 152.216 985501
Table17. Optimal values of biomass, effort, harvest and profit of Marylands Chesapeake Bay oyster fishery using Gompertz
growth model, taking effort data as man days by gear type for different discounting rates
*
x
*
H
*
E
Profits
0
7
10 78333 . 3
230856 6474.37 985502
0.01
7
10 67364 . 3
258309 7460.59 965155
0.05
7
10 4373 . 3
313899 9689.54 776496
0.1
7
10 31196 . 3
341320 10934.7 590787
7
10 05258 . 3
393235 13668.3 0
8. Conclusion
In this paper we intend to provide a framework for conservation of the resource and economic viability of the Marylands
Chesapeake Bay oyster fishery. In general the framework consists of several management plans which are broad and
comprehensive in scope. Management plans include all relevant aspects of the fishery for biological as well as economical
perspective. It is clear from our study that the oyster stock is not sustainable with respect to all reference points and for
conservation, restoration, and enhancement of oyster resource it is necessary to enforce sustainability with respect to the reference
points. We have used Verhulst-Schaefer and Gompertz-Fox models to describe the development of the fishery. The biological and
economical parameters of the bioeconomic models are estimated through two different types of effort data. To maximize the yield
and resource rent optimal sustainable strategy is used and all the results are obtained in numerical with proper units. It is evident
from the obtained results that the existing fishing effort of the oyster fishery should be reduced to archive MSY otherwise total
operational cost of the fishery will exceed total revenue earned by the fishery consequently the fishery will not be able to fulfil
Kar
and Chakraborty / International Journal of Engineering, Science and Technology,
Vol. 1, No. 1, 2009, pp. 172-189
188
commercial purpose. From the estimated results it is possible to obtain a rough idea of fishing effort to achieve MEY, in this
regard it should be remembered that at the present scenario maximum economic yield (MEY) is globally accepted fishery
management tool. Tax policies are discussed and it is found that subsidies should be given to the harvest effort to achieve MSY.
Here it is noted that for landing, effort and entry tax subsidy is necessary to achieve MSY. Therefore, it is not only feasible to
assess the biological, social, and economic impacts of existing oyster resource but also provide appropriate measures to maintain
long run sustainability through the obtained results.
It is observed from Tables 6&8 and 9&11 that the values of
MEY
H and fishing effort applied to achieve MEY for Marylands
Chesapeake Bay oyster fishery obtained using dynamic model with zero discounting rate are equal to the values of
MEY
H and
fishing effort applied to achieve MEY obtained by means of the static model for the same fishery. Again it is noted that the values
of
OAE
H and fishing effort applied to achieve OAE for Marylands Chesapeake Bay oyster fishery obtained using dynamic model
with infinite discounting rate are equal to the values of
OAE
H and fishing effort applied to achieve OAE obtained by means of the
static model for the same fishery. It is also observed from table 9&11 that when rate of discounting is gradually increased up to
1 . 0 = , corresponding harvest of the resource increases and the consequent fishing effort is simultaneously increased. As a result
the stock of the resource is gradually decreases as the rate of discounting increases. It depicts from table 9&11 that the fishing
effort tends to its maximum level when rate of discounting is considered to be infinite where as for zero discounting rate, the
fishing effort is at its maximum value as expected. It is interesting to see that the revenue earned from the fishery is gradually
decreased as the rate of discounting increased and at the infinite discounting rate i.e., at the open access equilibrium, revenue
earned from the fishery is tend to zero where as at the zero discounting rate the revenue earned from the fishery takes its maximum
level. Again, it should be mentioned that it is not possible to impose tax regulations in open access equilibrium situation of the
fishery since individual fishers attempt to maximize their income using maximum level of fishing effort. In this situation, it is
necessary to enforce strong regulatory agency to reduce wasteful harvesting practices.
We concentrate mainly on the harvesting problem in a deterministic framework. The evolution of natural stocks, however, is
seldom deterministic; it is subject to stochastic perturbations due to environmental and other factors. Thus a future research
problem would be considered in stochastic environment. Another area to explore is to determine optimal harvesting strategies
using game theory.
Acknowledgement
Acknowledgement: Research of T. K. Kar is supported by the Council of Scientific and Industrial Research (C S I R), India
(Grant no. 25(0160)/ 08 / EMR-II dated 17.01.08)
References
Bjorndal T., 1987. Production economics and optimal stock size in a North Atlantic fishery, Scand. J. of Economics, Vol. 89,
No. 2, pp. 145-164.
Chaudhuri K. S. and Johnson T., 1990. Bioeconomic dynamics of a fishery modelled as an S-system, Math. Biosci., Vol. 99,
pp. 231-249.
Chaudhuri. K. S. and Saha Ray S., 1996. On the combined harvesting of a prey-predator system, J. Biol. Syst., Vol. 4, No. 3,
pp. 373-389.
Clark C.W. and Munro G.R., 1975. Economics of fishing and modern capital theory: a simplified approach, Journal of
Environmental Economics and Management, Vol. 2, pp. 92-106.
Clark C.W., 1985. Bioeconomic modelling and fisheries management, John Wiley & Sons, New York.
Clark C.W., 1990. Mathematical bioeconomics: the optimal management of renewable resources, John Wiley and Sons, New
York.
Hannesson R., 1993. Bioeconomic analysis of fisheries, Fishing News Books: Blackwell Science, Oxford.
Jennings S., Kaiser M.J. & Reynolds J.D., 2001. Marine fisheries ecology, Blackwell Science, Oxford.
Kar T. K. and Chaudhuri K. S., 2003. Regulation of a prey-predator fishery by taxation: a dynamic reaction model. Journal of
Biological System, Vol. 11, No. 2, pp. 173-187.
Kar T. K. and Chaudhuri K. S., 2004. Harvesting in a two-prey one predator fishery: a bioeconomic model, ANZIAM J. Vol. 45,
pp. 443-456.
Kar T. K., Misra S. and Mukhopadhyay B., 2006. A bioeconomic model of a ratio-dependent predator-prey system and optimal
harvesting, J. Appl. Math. Comp., Vol. 22, No. 1/2, pp. 387-401.
Kar, T.K. and Matsuda H, 2008. A bioeconomic model of a single species fishery with marine reserve, Journal of Environmental
Management, Vol. 86, No. 1, pp. 171-180.
Kar T. K. and Chakraborty K., 2009. Marine reserves and its consequences as a fisheries management tool, World Journal of
Modelling and Simulation, Vol. 5, No. 2, pp. 83-95.
Kizner Z., 1997. Stability properties of discrete stock-production models, Sci. Mar., Vol. 61, No. 2, pp. 195-201.
Kronbak L. G., 2005. The dynamics of an open-access fishery: Baltic Sea Cod, Marine Resource Economics, Vol. 19, pp. 459-479.
Kar
and Chakraborty / International Journal of Engineering, Science and Technology,
Vol. 1, No. 1, 2009, pp. 172-189
189
Loehle C., 2006. Control theory and the management of ecosystems, Journal of Applied Ecology, Vol. 43, pp. 957966.
Mikkelsen E., 2007. Aquaculture-Fisheries Interactions, Marine Resource Economics, Vol. 22, pp. 287-303.
Patra B., Maiti A. and Samanta G.P., 2009. Effect of time-delay on a ratio-dependent food chain model, Nonlinear Analysis:
Modelling and Control, Vol. 14, No. 2, pp. 199-216.
Pitcher T.J., Kalikoski D., Short K., Varkey D. and Pramoda G., 2009. An evaluation of progress in implementing ecosystem-
based management of fisheries in 33 countries, Marine Policy, Vol. 33, pp. 223 232.
Pomeroy R., Nguyen K.A.T. and Thong H. X., 2009. Small-scale marine fisheries policy in Vietnam, Marine Policy, Vol. 33,
pp. 419428.
Thorpe A., Whitmarsh D., Ndomahina E., Baio A., Kemokai M., and Lebbie T., 2009. Fisheries and failing states: The case of
Sierra Leone, Marine Policy, Vol. 33, pp. 393400.
Wang L., Xu R. and Tian X., 2009. Global stability of a predator-prey model with stage structure for the Predator, World Journal
of Modelling and Simulation, Vol. 5, No. 3, pp. 192-202.
Wieland R., 2007. Managing oyster harvests in Marylands Chesapeake Bay, NOAA Chesapeake Bay Office, Non-native Oyster
Research Program.
Biographical notes
Dr. T. K. Kar is an Associate Professor at the Department of Mathematics, Bengal Engineering and Science University, Shibpur, in India. His
research interests include Dynamical systems, stability and bifurcation theory, population dynamics, mathematical modeling in ecology and
epidemiology, management and conservation of fisheries, bioeconomic modeling of renewable resources. He wrote around 50 academic papers
on those topics. He also supervised several students of master and doctor degree.
Kunal Chakraborty is a Lecturer at the Department of Mathematics, MCKV Institute of Engineering, Howrah, in India. He is currently doing
his Ph.D. under the guidance of Dr. T. K. Kar in the Department of Mathematics, Bengal Engineering and Science University, Shibpur, India. His
research topic is Bio-economic modelling and development of solution techniques for the management and conservation of fisheries. He has
obtained his post graduate degree in Mathematics from the University of Burdwan in 2006.
Received October 2009
Accepted November 2009
Final acceptance in revised form December 2009
MultiCraft
International Journal of Engineering, Science and Technology
Vol. 1, No. 1, 2009, pp. 190-204
INTERNATIONAL
JOURNAL OF
ENGINEERING,
SCIENCE AND
TECHNOLOGY
www.ijest-ng.com
2009 Multicraft Limited. All rights reserved
A parametric study on the growth of yield front in rotating annular disks
Shubhankar Bhowmick
1
, Dipten Misra
2
and Kashi Nath Saha
3
1
Department of Mechanical Engineering, National Institute of Technology, Raipur 492010, India
2, 3
Department of Mechanical Engineering, Jadavpur University, Kolkata 700032, India
Email: kashinathsaha@gmail.com (telefax: +91 33 2414 6890)
Abstract
The growth of elastic-plastic front under post-elastic conditions in high speed rotating annular disks having exponential and
parabolic geometry variation is investigated in the present study. The problem is addressed in post-elastic regime through an
extension of variational method based on von-Mises yield criterion and linear strain hardening material behavior. Assuming a
series expression of the unknown variable, the solution of the governing equation is obtained using Galerkins principle. The
approximate solution is obtained using an iterative method. The von-Mises stress distribution in the disk is reported in waterfall
plots at various load steps corresponding to angular speed at the initiation of yielding till the attainment of fully plastic state.
Effect of geometry parameters on the stress state of the disk is studied and the relevant results are reported in dimensionless
form.
Keywords: Variational method, von-Mises stress, plastic front, limit angular speed, annular disk
1. Introduction
Due to its widespread applications in industry, the analysis of rotating disk has been of great interest to many researchers across
the world. The study has become much more feasible in the past few decades due to the intensified application of numerical
methods as well as advent of computational machines. It has been observed that thickness variation of disks results in higher
operating speeds and reduced weight and rotary inertia as compared to those of uniform disks. Further, post-elastic analysis of
rotating disks leads to efficient utilization the disk material. The present paper investigates the stress and deformation states of
such annular disks with parabolic and exponential thickness variation.
The analysis of admissible angular speed of rotation for annular and solid disks of hyperbolically varying thickness, made of
incompressible material, using large elastic deformation theory has been reported by Chaudhary and Gupta (1992). Gven (1992)
investigated the effect of radial density gradient on elastic-plastic stresses and radial displacement of rotating annular disks with
variable thickness and reported the influence of density gradient on angular velocity and elastic plastic interface radius. Eraslan
and Argeso (2002) reported the elastic and plastic limit angular speeds for disks of variable thickness in power function. Based on
von-Mises yield criterion and its flow rule the plastic limit angular speed is calculated for disks under linear as well as non-linear
strain hardening behaviour. Results indicate increase in limit angular speed with reduction in thickness at the edge of the disk and
reduction in disk mass due to shape of the profile. In another paper, analytical methods based on laws of equilibrium have been
developed and reported by Eraslan and Apatay (2003) to study the elastic deformation and establish limit angular speeds of disks
of parabolic thickness variation. The stresses and deformations of rotating constant and linearly varying thickness solid and
annular disks under unified yield criterion were studied by Ma et al. (2001). In this criterion, with suitable adjustment of weighing
coefficients one of Trescas, von Mises or Yu criterion could be obtained. A semi analytical solution for an elastic-perfectly plastic
annular rotating disk and its associated flow rule was presented and the effect of yield criterion on the size of the plastic zone was
reported by Alexandrova and Alexandrov (2004). In another work, Alexendrova et al. (2004) developed a semi analytical method
based on von Mises yield criterion and its associated flow rule to determine displacement field and strain distribution in a thin
annular rotating disk of constant thickness. In a recent work Alexandrova and Villa Real (2007) reported the influences of
rotational speed and thickness variation on the plastic solution of hyperbolic annular disks in terms of von-Mises yield criterion
Bhowmick et al. / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 190-204 191
and its associated flow rule. Bhowmick et al. (2008) proposed a generalized variational method to study the behavior of externally
loaded rotating solid disks of variable thickness in the elastic regime. Based on deformation theory of plasticity, the method has
also been used into elasto-plastic domain under the assumption of von-Mises failure criterion with linear strain hardening theory
(Bhowmick et al., 2006).
In the present study, a numerical method based on variational principle for elasto-plastic analysis of rotating annular disks having
parabolic and exponential thickness variation has been proposed. A solution algorithm using von Mises yield criterion has been
developed to obtain an approximate solution of the unknown displacement field. The governing set of equations is solved in an
iterative manner and the validation of the proposed method with ANSYS is also presented. The coordinate functions used to
approximate the radial displacement field are generated numerically following Gram-schimdt orthogonalization scheme. New
results indicating the growth of the yield front in annular non-uniform disks have been presented for few different types of
thickness variation.
2. Mathematical formulation
The rotational symmetry of the loading, geometry and boundary condition of a disk made of homogeneous and isotropic material
makes the analysis axisymmetric. In the present study exponential and parabolic disk thickness variation of annular disks is
considered. The disk subjected to rotational loading experiences radial and tangential strain. At a certain speed, known as elastic
limit angular speed, the stress field of the disk attains the yield limit value at a certain location thus giving rise to a plastic front.
The present method proposes a suitable iterative algorithm which can capture the location of plastic front numerically and take up
the stress-strain analysis in post-elastic regime. Stress-strain relation of the disk is considered linear elastic followed by linear
strain hardening (Eraslan and Argeso (2002) as shown in Figure 1. In the present paper, von Mises theory is taken as the failure
criteria and under plane stress condition, the general condition of yielding is given by,
2 2
r r t t y
+ =
(1)
where the left hand side of the equation gives von-Mises stress,
vm
. The advantage in using von Mises criterion lies in the single
equation based formulation that takes care of the whole plastic region. Whenever the von Mises stress at a particular radial location
of a rotating disk reaches the uniaxial yield stress value, the plastic front initiates at that location and the corresponding rotational
speed is termed as elastic limit angular speed (
1
). On further increase in rotational speed, a certain region of the disk attains post-
elastic state and when this region encompasses the entire disk we get plastic limit speed or collapse speed (
2
). It is important to
note that the initiation of plastic front gives rise to non-homogeneity due to presence of separate regions with different flow
characteristics depending on elasticity modulus in elastic region and tangent modulus in post-elastic region, however the
assumption of isotropy remains valid for the entire disk. The limit elastic state followed by the advancement of the yield front till
limit plastic state is reached is shown in Figure 2 (a-c). At elastic limit speed the von-Mises stress attains uniaxial yield stress value
at the inner boundary of rotating annular disk (Figure 2 (a)) and as a consequence, the yield front starts propagating towards the
outer edge with increase in rotational speed (Figure 2 (b)) until it attains a fully plastic state (Figure 2 (c)). The domain of the disk
beyond elastic limit speed gets divided into two (inner plastic-outer elastic) regions, an inner plastic region (r = a to r = r
y
) and an
outer elastic region (r = r
y
to r = b). However for certain thickness variation based on the selected geometry parameters and
loading conditions, an annular disk may also start yielding at an intermediate radial location but such cases has not been
considered for the present study.
The variational form of total potential energy states that the extremum of the summation of strain energy (U) and work potential
(V) is conserved for a system in equilibrium, i.e.,
0 (U V ) + = (2)
The total strain energy U consists of outer elastic
( )
e
U
and inner plastic
( )
p
U
part and V is the potential of the external forces.
The strain energy in the disk is obtained by evaluating the integral
1
( )
2
ij ij
dv
.
Expression of the elastic part of strain energy for the outer region (for
y
r r = to r b = ) is given by (Bhowmick et al., 2008),
2
2
2
2
1 r
y
b
e
r
E u du du
U u r hdr
r d dr
= + +
. (3)
Bhowmick et al. / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 190-204 192
Figure 1. Linear strain hardening material behavior.
Figure 2 Propagation of elastic-plastic interface with increasing angular speed in annular disks:
(a) initiation at root (b) at intermediate location and (c) collapse at the outer radius
In arriving at these equations, the following strain-displacement and stress-strain relations have been used,
/
r
du dr = , /
t
u r =
and
( )
[ ]
( )
[ ]
2 2
,
1 1
r r t t t r
E E
= + = +
.
Here u is the displacement field in radial direction. Post-elastic part of strain energy,
p
U
(for r a = to
y
r r =
), is obtained
from the integral (
vol
area under radial ' - ' curve + area under tangential ' - ' curve) dv, i.e,
p
U
=
( )
r t
vol
dU dU dv +
, where,
r
U
and
t
U
are the contributions coming from radial and tangential stresses and strains. To develop the mathematical expression of
these energies, we note that the strain-displacement and stress-strain relations in the post-elastic region becomes,
0 p
r r
du
dr
= +
,
0 p
t t
u
r
= +
and using deformation theory of plasticity and linear strain hardening behaviour the following relations are obtained,
( ) ( )
0 0 0 0 0 0
2 2
, .
1 1
r r t t t r
E E
= + = +
( ) ( )
1 1
2 2
, .
1 1
p p p p p p
r r p t t t p r
p p
E E
= + = +
Here,
0 0
,
r t
and
0 0
,
r t
are the strains and stresses in radial and tangential direction in yield condition respectively. With proper
substitutions the following relations for
r
dU and
t
dU are obtained.
Bhowmick et al. / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 190-204 193
( ) ( ) ( )
( )
( )
( )
( )
( )
2
0 0 0 0 1 1
2 2 2
2 2
0 0 0 0 0 0 0 0 1 1
2 2 2
1
2 1 1 1
1 1 1
2 2 2 1 1 1
r p r t r p t
p p
r p r t r r t p t r
p p
E E du du u E du du
dU
dr dr r dr dr
E E E du u
dr r
= + + + +
+ + + +
and
( ) ( ) ( )
( )
( )
( )
( )
( )
2
0 0 0 0 1
2 2
2 2
0 0 0 0 0 0 0 0 1 1
2 2
1
2 1 1 1
1 1 1
2 2 2 1 1 1
1
t p t r t p r
2
p p
t p r t t r t p r t
2
p p
E E u du u E u u
dU = + + + - +
r dr r r r - - -
E E E u du
+ + + + - .
r dr - - -
Thus the strain energy in the inner post-elastic region is obtained from the expression
( ) ( )
( ) ( )
( )
( ) ( )
( )
( ) ( )
{ }
2
2
0 0 0 0 1
2 2
2 2
0 0 0 0 0 0 0 0 1
2 2
1
2
2
1
2 2
2 1 1
2 2
1 2 1
2 1
y y
y y
r r
p p r t t r
a a p
r r
r p t t p r r t t r
a a p
hdr
hdr
E u du du E du
U r u hdr r u
r dr dr dr
E du E
r u r r r hdr
dr
E
= + + + + + +
+ + + + +
+
( )
( ) ( )
{ }
2 2
0 0 0 0
2
2 2 .
y
r
r t p t r
a p
r r r hdr + +
(4)
The expression for work potential is as follows
2 2
2 ( )
b
a
V r uh dr =
(5)
In the mathematical expressions, the normalization of the energy expressions is carried out with three parameters
1
,
and
2
b a = ,
1 y
r a =
and
2 y
b r =
.
Accordingly three normalized coordinates
1
, and
2
are obtained as shown below
( ) r a = ,
1 1
( ) r a = and
( )
2
2
y
r r =
.
Substituting the normalized expressions of the energies (Eq. 3, 4 and 5) in Eq. (2), the governing equilibrium equation becomes,
( )
( )
( )
1
2 2
2 2 2 2
2 2 2 2 2 0 2 2 2
1
y
y
r
E u u du du du du
u u h d
d d d d r
+
+ + +
+
( )
( )
( )
( )
1
1 1 1
1 1 2 2
1 1 1 1 1 1 1 0 1
1
p
p
E du du u u du du
u u h d
d d d d
+ + + +
( )
( ) ( )
1
1 1 1 1 0 0
1 1
2
1 1 1 1 0
1
r t
E du du
u u h d
d d
+ + + +
( )
( ) ( )
1
1 1 1 1 0 0 1
1 1
2
1 1 1 1 0
1
r p t p
p
E du du
u u h d
d d
+ + +
( )
{ }
1
2
2
0
0 h u d =
(6)
Bhowmick et al. / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 190-204 194
The global displacement field
( ) u
in the above equation is approximated by
( )
i i
u c
, where
i
is the set of orthogonal
functions developed by satisfying the relevant boundary conditions (
( )
0
r a
=
and
( )
0
r b
=
) of an annular rotating disk in
elastic regime. For the purpose of computation, displacement functions in the post-elastic and elastic regions are expressed as
p
i i
c u ) (
1
and
e
i i
c u ) (
2
respectively, where
p
i
and
e
i
are portions of the global function
i
in post-elastic and elastic
region. Substituting these assumed displacement functions and replacing operator by
/
j
c
the governing equation is obtained
in matrix form as follows:
1
' ' ' ' 2 2
2 2 2 2
1 1 2 2 2 2 0
( )
( )
1 ( ) ( )
e e
nf nf
i j e e e e e e c
i i j i j i j
j i c
r E
c h d
r
= =
+
+ + +
+
( )
( )
{ }
1
' ' ' ' 1 1 1
1 1 2 2
1 1 1 1 1 1 0
( )
( ) 1
p p
nf nf
i j p p p p p p p
i i j i j i j
j i
p
E
c h d
= =
+ + + +
{ }
1
2 2
0
1
( )
nf
j
j
h d
=
=
( )
( ) ( )
1
1 1 1 1 0 ' 0 '
1 1
2
1 1 1 0
1
nf
p p p p
r j j t j j
j
E
h d
=
+ + +
( )
( ) ( )
1
1 1 1 1 0 ' 0 ' 1
1 1
2
1 1 1 0
1
nf
p p p p
r p j j t j p j
j
p
E
h d
=
+ + + +
(7)
In this equation ( ) indicates differentiation with respect to normalized coordinates.
2.1 Solution algorithm: The governing equation can be expressed in matrix form as,
[ ]{ } { } K c f =
and the required solution of
unknown coefficients
{ } c
is obtained numerically by using an iterative scheme involving two loops. The rotational speed is
incremented through a suitable step size (
1
are necessary to
generate the right hand side of Eq.(7). To this end, it is assumed that the ratio of tangential and radial stress when yielding occurs
at each point in the plastic region in the present load step is same as the ratio of tangential and radial stress obtained in the
preceding solved load step. For each load step, the location of plastic front is given a small increment starting from its previous
location. From this defined limits of elastic and plastic zones the solution of the displacement field is obtained and the stress values
are computed at the assumed elasto-plastic interface. The value of von Mises stress at the assumed plastic front location is checked
for the present load step and equality of the value with that of yield stress gives the required solution for that load step. All relevant
data are stored at this instant and it is followed by a load increment by returning to outer loop till the disk attains a fully plastic
state. A brief description of the solution algorithm which presupposes solution of limit angular speed
1
and yield front location r
y
is provided below.
1. Two loops are commenced with r
y
initiated to the value of a.
2. The rotational speed is incremented through a suitable step size (
1
and
0
t
in the elastic region is constant for a particular radial location. That ratio is stored during
elastic analysis. Whenever yield front reaches a particular radial location, its von-Mises stress reaches the yield stress.
Corresponding to this value of von-Mises stress , the values of
0
r
and
0
t
=
5. The displacement field is then obtained using the elastic-plastic formulation of Eq. (7).
6. A check is performed on von Mises stress and if it attains the value of yield stress at the yield front location then current
position is the exact yield front location. This ensures solution for the current load step, when the relevant results are
Bhowmick et al. / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 190-204 195
noted and followed by a load increment by returning to outer loop mentioned in step 2. Otherwise the program returns to
step 3 so long the plastic region covers whole of the disk.
2.2. Determination of plastic strain: At
1
> , radial and tangential strains at each coordinate inside the post-elastic region of the
disk can be decomposed into elastic and plastic parts as given below.
e p
r r r
= + (9a)
e p
t t t
= + (9b)
where the superscript e and p denote the elastic and plastic part of the strains respectively.
The relations between stresses and elastic strains can be derived from generalized Hookes law. For a state of plane stress
( 0
z
= ), assuming Henckys deformation theory of plasticity, the plastic strain components are given as
( )
2
p
p vm t
r r
vm
=
(10a)
( )
2
p
p vm r
t t
vm
=
(10b)
In Eq. (10a, b)
p
vm
is the equivalent plastic strain and
vm
is the equivalent or von-Mises stress. Based on linear strain
hardening material behavior of the disk material, the relation between equivalent plastic strain and von-Mises stress can be
expressed as
( )
vm y p
vm
H
=
(11)
where H is the hardening parameter and is given by
1
1
EE
H
E E
=
(12)
3. Results and discussions
The results are presented in terms of the following dimensionless and normalized variables: angular velocity =
/
y
b
,
stress
/
y
=
, displacement
/
y
u u E b =
and hardening parameter
( )
1 1
/ H E E E =
. Results for post-elastic behaviour of
linear strain hardening material have been generated using the values H =0.5, a = 0.2, b = 1.0 m, = 7850 kg/m
3
, E=210 GPa and
y
1
,
2
Elastic and fully plastic limit angular speed
,
r t
Strains in radial and tangential direction
0 0
,
r t
Strains in radial and tangential direction in yield condition
,
p p
r t
Part of radial and tangential strain above yield strain
,
r t
Radial and tangential stresses
0 0
,
r t
Stresses in radial and tangential direction in yield condition
,
p p
r t
Part of radial and tangential stress above yield strain
vm
+ + = + +
(2)
energy equation:
( )
2
2 2
2
2
;
p p p
T T T T u B
u v u U
t x y c c y c
y
+ + = + +
(3)
and corresponding boundary and initial conditions:
( ) 0 0
w
u , v , T T t = = = for 0 y = ; (4)
( ) u U x,t , T T
for y ; (5)
( )
0 0
u u ( x, y ) , T T x, y = = for
0
t t = ; (6)
( )
1 1
u u ( t , y ) , T T t , y = = for
0
x x = . (7)
For further consideration stream function, ( ) x, y,t is introduced with following relations:
, v u
x y
= =
; (8)
which satisfies equation (1) identically and transform moment equation (2) into equation:
2 2 2 3 2
2 3
;
U U B
U U
t y y x y x t x y
y y
+ = + +
(9)
and energy equation into equation:
Nikodijevic et al./ International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp.206-215
208
2
2
2 2 2
2 2
p p p
T T T T B
U .
t y x x y c c c y y y
+ = + +
(10)
Boundary and initial conditions are transformed into equations:
0 0
w
, ; T T ( t )
y
= = =
for 0 y = ; (11)
( ) U x,t ; T T
y
for y ; (12)
( )
0 0
u ( x, y ) , T T x, y
y
= =
for
0
t t = ; (13)
( )
1 1
u ( t , y ) , T T t , y
y
= =
for
0
x x = . (14)
In the case of flow around a body boundary conditions for third order equation (9) are conditions (11) and (12). First two
conditions ( 0 0 , y = = ) means: line 0 y = coincides with zero streamline ( 0 = ); longitudinal velocity is equal to zero
on the body ( 0 u y = = ). Third boundary condition ( ) y U x,t means asymptotic approaching of longitudinal
velocity to the velocity on outer edge of boundary layer. Condition 0 v y = = for 0 y = is consequence of equality 0 =
on body surface. Equation (9) does not depend from equation (10and it can be solved independently. Solution of equation (9) is
used for solving of equation (10).
3. Universal equations
Obtained partial differential equations can be solved for every particular case using modern numerical methods and computer. In
this paper, quite different approach is used based on ideas in papers (Skadov, 1963; Loicijanski, 1965; Saljnikov, 1978; Busmarin
et al., 1974), which is extended in papers (Boricic et al., 2005; Obrovic et al., 2005; Nikodijevic et al., 2009). Essence of this
approach is in introducing adequate transformations and sets of parameters in starting equations, which transform the equations
system and corresponding boundary conditions into form unique for all particular problems and this form is considered as
universal. Solution of universal equations can be used to derive general conclusions about developing of temperature MHD
boundary layer and also for special cases calculation. Integration of obtained universal equations is performed once for all. In order
to solve particular problems it is necessary to determine impulse equation using obtained universal solutions. In order to realize
described procedure following new variables are introduced:
( )
, , ;
D y
x x t t
h x,t
= = =
( )
( )
( ) ( )
( ) , ;
w
w
D x, y,t T T
x,t , x,t ,
U x,t h x,t T T
= =
(15)
where D is normalizing constant, and ( ) h x,t is characteristic linear scale of transversal coordinate in boundary layer. According
to introduced variables, equation (9) is transformed in new form:
( ) ( )
2
3 2 2 2
2
1 0 0 1 1 0
3 2 2
1
1 1
2
, , ,
D f f g F g z UzX( ; x )
t
+ + + + + + = +
; (16)
where for the sake of shorter expression, the notations are introduced:
Nikodijevic et al./ International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp.206-215
209
2
h
z
= ;
z
g
t
;
2
B
N
= ;
1 0 ,
g Nz = ;
z
F U
x
;
1 0 ,
U
f z
x
;
0 1 ,
z U
f
U t
;
( )
2 2
1 2
1 2 2 1
X x ; x
x x x x
=
; (17)
and equation (10) now have the form:
( )
( )
2
2
2 2 2
2
1 0 1
2 2
1 0
1 1
1 1
2 ;
2 2
c c ,
r
,
D
D E E g l
P
g F f z UzY( x; )
t
+ +
+ + + =
(18)
where the following notations have been used for the sake of shorter statement:
p
c
Pr
= -Prandtl number;
( )
2
c
p w
U
E
c T T
-Eckert number;
1
w
w
dT z
l
T T dt
;
( )
1 2
1 2 2 1
Y x ; x
x x x x
=
. (19)
Now we introduce sets of parameters: dynamical
k ,n
f , magnetic
k ,n
g , temperature
k
l , and constant parameter g , as new
independent variables:
1
0 1 2 0 ;
k n
k k n
k ,n
k n
U
f U z ( k,n , , ,...;k n )
x t
+
+
= =
(20)
1
1
1
0 1 2 0 ;
k n
k k n
k ,n
k n
N
g U z ( k,n , , ,...,;k )
x t
+
+
= =
(21)
1
k
k
k
k
q
l z
q x
1 2 ( k , ,...) = ; (22)
z
g const.
t
= =
; (23)
where:
w
q T T
= . (24)
Nikodijevic et al./ International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp.206-215
210
Introduced sets of parameters reflect the nature of velocity change on outer edge of boundary layer, alteration characteristic of
variable N and temperature change on body surface, and a part from that, in the integral form (by means of z and z / t ) pre-
history of flow in boundary layer. These parameters enable transformations of differential equations (16) and (18) into universal
form in sense that neither equations nor boundary conditions explicitly depends from values that characterized particular problems.
Procedure for obtaining universal equations has following steps. First, we find the derivates in Eqs. (16) and (18) using the
operators given in Eqs. (25)-(26), and then transform these equations with introduced independent variables
k ,n k ,n k ,n
; f ; g ;l . For
derivate along longitudinal coordinate x and for time derivate next operators are used:
0 1
0 0
1
0
;
k ,n k ,n
k
k ,n k ,n
k ,n k
k n n k
k
, for
f g
l
x x f x g
, for
x l
= =
=
=
= + +
(25)
0 1
0 0
1
0
k ,n k ,n
k
k ,n k ,n
k ,n k
k n n k
k
, for
f g
l
t t f t g
, for
t l
= =
=
=
= + +
. (26)
Parameter derivates along coordinate x and time t are obtained by differentiation of Eqs. (20)-(22):
( ) { ( ) }
1 0 1
1 1
1 ;
k ,n
, k ,n k ,n k ,n k ,n
f
k f f k n Ff f Q
x Uz Uz
+
= + + + =
(27)
( ) { ( ) }
0 1 1
1 1
1 ;
k ,n
, k ,n k ,n k ,n k ,n
f
k f f k n gf f E
t z z
+
= + + + =
(28)
( ) { ( ) }
1 0 1
1 1
1 ;
k ,n
, k ,n k ,n k ,n k ,n
g
k f g k n Fg g K
x Uz Uz
+
= + + + =
(29)
( ) { ( ) }
0 1 1
1 1
1 ;
k ,n
, k ,n k ,n k ,n k ,n
g
k f g k n gg g L
t z z
+
= + + + =
(30)
{ }
1 1
;
k
k k
l
kFl M
x Uz Uz
= =
(31)
( ) { }
1 1
1 1
k
k k k
l
kg l l l N
t z z
+
= + =
; (32)
where
k ,n k ,n k ,n k ,n k k
Q ; E ; K ; L ; M ; N are terms in curly brackets in obtained equations. It is important to notice
k ,n k ,n k
Q ; K ;M
beside the parameters depend on value U z x F = . Using parameters (20)-(23) instead coordinate x and time t , operators (25)
and (26), and terms (27)-(32) Eqs. (16) and (18) are transformed into equations:
( ) ( )
2 2
1
0 1
0 0
;
k ,n k ,n k ,n k ,n k ,n k ,n
k ,n k ,n
k ,n k
k n n
Q X ; f E K X ; g L
f g
= =
=
= + + +
(33)
( ) ( )
( )
2
0 1
0
1
0
;
k ,n k ,n k ,n k k k
k ,n k
k ,n k
k n
k ,n k ,n k ,n
k ,n
k
n
Q Y ; f E M Y ;l N
f l
K Y ; g L
g
= =
=
=
= + + + +
+ +
(34)
Nikodijevic et al./ International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp.206-215
211
where the following markings have been used for shorter statement:
1
-left side of equation (16),
2
- left side of equation (18).
From the Eqs. (33) and (34) can be seen, the characteristics of external flow dominate by the means of function F which appears
in terms for
k ,n k ,n k
Q ; K ; M . In order the equations (33) and (34) to be independent of the outer flow characteristics i.e. to be
universal, it is necessary to show the existence of the equality
( ) ( ) ( )
k ,n k ,n k
F F f , g , l
=
.
So as to show the existence of such equality, we start from the momentum equation:
( ) ( )
2
0
* ** w
U
U U U N
t x x
+ + + =
; (35)
in which:
( )
0
1
*
u
x,t dy
U
-friction stress on the body. (38)
Introducing dimensionless characteristic functions:
( )
*
H x,t
h
= ;
( ) H x,t
h
= ;
( )
w
h
x,t
U
= ; (39)
which, according to Eqs. (15) and (36)-(38), can be expressed in the following form:
( )
0
1
1 H x,t d
D
;
( )
0
1
1 H x,t d
D
;
( )
2
2
0
x,t D
. (40)
After transition to new independent variables values H
, H
+
=
+
=
=
= + + + + +
+ +
(42)
( ) ( )
0 1
0 0
1
2
k ,n k ,n
k ,n k ,n
k ,n k
k n n
H H
Q H k n f k n g .
f g
= =
=
= + + + +
(43)
Last two equations define function F in terms of values, which depends only from introduced parameters. Eqs. (33) and (34) with
respect to (41)-(43) are now universal equations of described problem. Since the system of Eqs. (33) and (34) do not contain
explicitly: the free stream velocity distribution, outer magnetic field, body temperature and the temperature of the free stream
which defines each particular case of considered flow, these equations are considered as universal (independent from outer flow
characteristic). Boundary conditions, also universal, are given with terms:
0 0 0 , ,
= = =
for 0 = ; (44)
1 1 , for ; (45)
( ) ( )
0 0
, = = for
( )
( )
( )
0 0 1 2 0
0 0 1 2 0
0 1 2
0
k ,n
k ,n
k
f , k,n , , ,...k n
g k,n , , ,...,k
l k , ,...
g
= =
= =
= =
=
; (46)
where ( )
0
is Blasius solution for stationary boundary layer on the plate, ( )
0
is solution of following equation:
2
2 2 2
2 0 0 0 0
0
2 2
0
d d d D
D Ec .
Pr d d d H
+ =
(47)
Universal Eqs. (33) and (34) with boundary conditions (44)-(46) are exact for wide class of problems in which ( ) z At C x = + ,
where A is arbitrary constant and ( ) C x some function of longitudinal coordinate. For other problems this equations are
approximated universal equations.
Eqs. (33) and (34) are integrated in m-parametric approximation once for good and all. Obtained characteristic function can be
used to yield general conclusions about development of described boundary layer and to solve any particular problem. Before
integration for scale of transversal coordinate in boundary layer ( ) h x,t some characteristic value is chosen. In this case h
=
and Eq. (41) now have form:
Nikodijevic et al./ International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp.206-215
213
( )
1 0 0 1 1 0
0 1
0 0
1
2 2
2
, , , k ,n k ,n
k ,n k ,n
k ,n k
k n n
H H
F f H f g g H E L .
f g
= =
=
= + + +
(48)
Taking parameters 0
k ,n
f = , 0
k ,n
g = , 0 g = Eq. (33) is simplified into form:
3 2
0 0 0
0
3 2 2
0
d d
d D d
+ = ; (49)
which describe boundary layer flow on flat plate. Previous equation must coincide with well known Blasisus equation, which
repute for exact equation and according to previous statement it is necessary to chose
2
0
D = . Using exact Blasisus solution
(Schlichting, 1982)
0
0 2209 . = for further investigations normalizing constant 0 47 D . = must be chosen (Loicijanski, 1965,
White, 1994). For selected value h Eq. (47) for determining variable
0
became:
2
2 2
0 0 0
0
2 2
1
0
d d d
Ec
Pr d d d
+ =
. (50)
Eqs. (33) and (34) contain on the right sides infinite sums. In order to solve these equations it is necessary to limit number of
independent variables. This leads to the implementation of snipping method, which implies equalization of variables with zero
starting form some index. In this way approximated equations are derived. For higher number of detained parameters-variables,
approximated equations are more precise. In this paper approximations of Eqs. (33) and (34) are given in which influence of
parameters
1 0 ,
f ,
0 1 ,
f ,
1 0 ,
g ,
1
l , and g are detained and influence of parameters
0 1 ,
f ,
1
l derivatives are disregarded.
Approximation of Eqs. (33) and (34) are given with following terms:
( ) ( )
2 2
1 1 0 1 0 1 0 1 0 1 0 1 0
1 0 1 0
;
, , , , , ,
, ,
Ff X ; f gf Fg X ; g gg
f g
= + + +
(51)
( ) ( )
2 1 0 1 0 1 0 1 0 1 0 1 0
1 0 1 0
, , , , , ,
, ,
Ff Y ; f gf Fg Y ; g gg ;
f g
= + + +
(52)
and function F is obtained from Eq. (48) in same approximation:
( )
1 0 0 1 1 0 1 0 1 0
1 0 1 0
1
2 2
2
, , , , ,
, ,
H H
F f H f g g H gg gf .
g f
= + + +
(53)
Boundary conditions which coincide to system of Eqs. (51) and (52) are conditions (44), (45) and condition:
( ) ( )
0 0
, = = for
1 0
0
,
f = ,
0 1
0
,
f = ,
1 0
0
,
g = ,
1
0 l = , 0; g = (54)
which is obtained from condition (46), using same simplifications like as equations. Universal equations need to be solved with
corresponding boundary conditions (44), (45) and (54) using three-diagonal method, known in Russian literature as the "progonka"
method. Solving of these equations is subject of further research. Obtained universal solutions can be used for general conclusions
about boundary layer development and also for calculations for every particular problem.
4. Summary
In this paper unsteady thermal two-dimensional MHD boundary layer is considered. This problem can be analyzed for every
particular case i.e. for given outer flow characteristics. Here is used quite different approach in order to use benefits of multi-
parametric method and universal equations of observed problem are derived. The independence of the obtained system of
equations and boundary conditions of particular velocity and temperature distribution at the edge of boundary layer; of outer
Nikodijevic et al./ International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp.206-215
214
magnetic field and of body temperature enable the integration of these equations, once for good and all. In the procedure of
integration, the velocity and temperature dimensionless profiles , in cross-stream sections of boundary layer, friction
coefficient , and also the characteristic functions H
, H
P
Ux
Uz
Sensor
Mild steel
Beam
Figure 1. Finite Element discretization of the mild steel beam with sensor bonded at the clamped end.
For a plane stress problem the stress components
2 4 6
( ) ( ) ( ) 0
yy yz xy
= = = , electric displacement component
and magnetic induction component
2
( ) 0
y
D D =
2
( ) 0
y
B B = . The strain displacement relation for a 2D plane stress problem can be
written as
1
x
xx
u
S S
x
= =
;
2
z
zz
u
S S
z
= =
;
5
x z
xz
u u
S S
z x
= = +
(6)
The relation of electric field to the electric potential can be expressed as
1 x
E E
x
= =
;
3 z
E E
z
= =
(7)
The relation of magnetic field to the magnetic potential can be expressed as
1 x
H H
x
= =
;
3 z
H H
z
= =
(8)
Jiang et al. (2004) have derived the reduced material constants under a plane stress assumption, which are used to evaluate the
stiffness matrices. A finite element formulation for the fully coupled magneto-electro-elastic continuum can be expressed as
[ ]{ } [ ] { } { } { } {
i t
uu uu u u
}
M u K u K K Fe
+ + + =
(9a)
{ } { } { } 0
T
u
K u K K
=
=
(9b)
(9c)
{ } { } { } 0
T T
u
K u K K
where
i t
Fe
is the harmonically applied mechanical load which is varying with the circular frequency . Since no electric charge
or current density is applied, the right hand side of equation 9b and 9c is zero. Elemental mass and stiffness matrices
corresponding to the above equations are derived as follows.
[ ] [ ] [ ][ ]
T
uu u u
v
K B C B dV =
(10a)
[ ] [ ]
T
u u
v
K B e B dV
=
(10b)
[ ] [ ]
T
u u
v
K B q B
=
dV
(10c)
Biju et al. / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 216-227 219
dV [ ]
T
v
K B B
=
(10d)
[ ]
T
v
K B B
=
dV (10e)
[ ]
T
v
K B m B d
=
V (10f)
[ ] [ ] [ ][ ]
T
uu
v
M N N d =
V (10g)
[ ]
u
B
,
B
and
B
represent the shape function derivative matrices for strain-displacement, electric field-electric potential and
magnetic field-magnetic potential respectively. Three point Gauss quadrature is used to evaluate the integrals.
Using standard condensation techniques, the electric potential and magnetic potential are eliminated from equation (9) which can
be written as
[ ]{ } { } { }
i t
uu eq
M u K u Fe
+ =
(11)
where is the equivalent stiffness matrix.
eq
K
[ ] [ ] [ ] [ ] [ ]
1
eq uu u II I u IV III
K K K K K K K K
= + +
1
(12)
The component matrices in equation 11 can be written as
[ ]
1 T T
I u u
K K K K K
=
(13a)
[ ]
1 T
II
K K K K K
=
(13b)
(13c)
[ ]
1 T T
III u u
K K K K K
=
T
[ ]
1 T
IV
K K K K K
=
(13d)
The Eigen vectors corresponding to and is given by
{ } [ ] [ ]{ }
1
II I
K K u
= (14a)
{ } [ ] [ ]{ }
1
IV III
K K u
= (14b)
4. Validation
A computer code has been developed to study the response of magneto-electro-elastic sensor bonded on the top surface of a
cantilever beam while the beam is excited harmonically. Ansys 10.1 used for validation studies cannot model fully coupled
magneto-electro-elastic materials but it can model piezoelectric materials. The present code is validated using piezoelectric
material PZT-5 whose material properties (Chen et al., 2007) are given in Table 1. The dimensions of mild steel beam are as
follows: length=0.3m, thickness=0.02m. The dimensions of the sensor patch are length=0.015m, thickness=0.002m. The PZT
sensor is placed at the clamped end of the beam and a point load of 1N is applied at the free end in the negative z-direction.
Finite element mesh for the mild steel beam and the sensor patch are 40 x 2 and 2 x 2 respectively. Full method is used as the
solution technique and a constant damping ratio of 0.01 is used. Response is taken from that node of the sensor which shows
maximum value of electric potential. The absolute value of transverse displacement ( ) and the electric potential ( )
corresponding to the above said node are obtained for the first two natural frequencies of the structure. Log scale plots of the
responses using code and Ansys are shown in Figure 2(a) and 2(b), which are in good agreement. Hence the above computer code
is used to study the response of magneto-electro-elastic sensors under harmonic mechanical loading.
z
u
Biju et al. / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 216-227 220
Table1. Material properties of PZT-5 and different volume fraction v
f
of multiphase
magneto-electro-elastic BaTiO
3-
CoFe
2
O
4
(Aboudi, 2001)
0.0 v
f
0.2 v
f
0.4 v
f
0.6 v
f
0.8 vf 1.0 v
f
PZT-5
Elastic constants
C
11
=C
22
286 250 225 200 175 166 99.2
C
12
173 146 125 110 100 77 54.01
C
13
=C
23
170 145 125 110 100 78 50.77
C
33
269.5 240 220 190 170 162 86.85
C
44
=C
55
45.3 45 45 45 50 43 21.1
Piezoelectric constants
e
31
=e
32
0 -2 -3.0 -3.5 -4.0 -4.4 -7.20
e
33
0 4 7.0 11.0 14.0 18.6 15.11
e
24
=e
15
0 0 0 0 0 11.6 12.32
Dielectric constants
11 22
=
0.08 0.33 0.8 0.9 1.0 11.2 1.53
33
0.093 2.5 5.0 7.5 10 12.6 1.5
Magnetic permeability
constants
11 22
=
-5.9 -3.9 -2.5 -1.5 -0.8 0.05 0
33
1.57 1.33 1.0 0.75 0.5 0.1 0
Piezomagnetic constants
q
31
=q
32
580 410 300 200 100 0 0
q
33
700 550 380 260 120 0 0
q
24
=q
15
560 340 220 180 80 0 0
Magnetoelectric constants
m
11
=m
22
0 2.8 4.8 6.0 6.8 0 0
m
33
0 2000 2750 2500 1500 0 0
Density
5300 5400 5500 5600 5700 5800 7750
C
ij
in 10
9
N/m
2
, e
ij
in C/m
2
, in 10
lk
-9
C/Vm, q
ij in
N/Am
, lk
in 10
-4
Ns
2
/C
2
, m
lk
in 10-
12
Ns/VC, inkg/m
3
0 200 400 600 800 1000 1200 1400 1600
1E-10
1E-9
1E-8
Ansys
Code
U
z
(
m
)
f (Hz)
0 200 400 600 800 1000 1200 1400 1600
0.01
0.1
1
(
v
o
l
t
)
f (Hz)
Ansys
Code
(a) (b)
Figure 2 (a) Transverse displacement and (b) Electric potential ( ) at the top surface of the PZT-5 sensor
Biju et al. / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 216-227 221
5. Results and discussion
Response of magneto-electro-elastic sensors bonded on the top surface of a cantilever beam with clamped-free boundary
condition under harmonic mechanical loading is discussed below. Numerical results are presented for clamped free boundary
condition for the first three modes of the structure.
Two types of loadings are used. (i) A point load of 1N at the tip of the beam in the negative z-direction (ii) Uniformly distributed
load of 1N at the bottom surface of the beam in the negative z-direction. The location of the sensor is varied along the length of the
beam i.e. at the clamped end, at the centre and at the tip of the beam. The harmonic response of the system is analysed with the
base beam subjected to clamped-free boundary condition with
x
u = = = =0 at the clamped end. The electric potential ( )
at the interface between base beam and sensor is also assumed to be zero. The material properties of multiphase MEE material
with volume fraction of BaTiO
z
u
3
varying from 0.0 to 1.0 in a composite of BaTiO
3
-CoFe
2
O
4
are listed in Table 1. Volume fraction
of 0.0 represents pure CoFe
2
O
4
and volume fraction of 1.0 represents pure BaTiO
3
. The material properties of mild steel beam are
as follows: Youngs modulus (E) = 210GPa, Poisson ratio ( ) = 0.3, Magnetic permeability constant ( ) = 2.51e-4, density ( )
= 7800 kg/m
3
. The sensor response during the first three modes for different volume fraction of the MEE composite is plotted
graphically.
5.1 A point load at the tip of the beam
The beam is clamped at one end and response of the sensor is studied when the beam is subjected to a harmonic excitation of 1N
in the negative z-direction at the free end of the beam.
5.1.1 Sensor patch located at the clamped end
Figure 3 shows the transverse displacement ( , electric potential ( ) and magnetic potential )
z
u ( ) at the top surface of the
MEE sensor when the sensor is located at the clamped end of the mild steel beam. The transverse displacement is found maximum
at the first mode. It goes on increasing as the volume fraction increases attributing to the fact that the elastic constants decreases as
the volume fraction increases. The electric potential ( ) is found maximum for 0.2 v
f
of the composite and the electric potential
decreases considerably at higher modes. Volume fraction 1.0
which corresponds to pure BaTiO
3
also gives electric potential
comparable to 0.2 v
f.
The magnetic potential ( ) is found maximum for 0.6 v
f
of the composite and the magnetic potential also
decreases at higher modes. Volume fraction 0.0
which corresponds to pure CoFe
2
O
4
gives less comparable magnetic potential out
put with 0.6 v
f.
0 500 1000 1500 2000 2500 3000 3500
1E-10
1E-9
1E-8
U
z
(
m
)
f (Hz)
Vf=0.0
Vf=0.2
Vf=0.4
Vf=0.6
Vf=0.8
Vf=1.0
0 500 1000 1500 2000 2500 3000 3500
0.01
0.1
1
(
v
o
l
t
)
f (Hz)
Vf=0.2
Vf=0.4
Vf=0.6
Vf=0.8
Vf=1.0
(a) (b)
Fig 3 (a) Transverse displacement ( , (b) Electric potential ( ) )
z
u
Biju et al. / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 216-227 222
0 500 1000 1500 2000 2500 3000 3500
1E-5
1E-4
1E-3
0.01
(
A
)
f (Hz)
Vf=0.0
Vf=0.2
Vf=0.4
Vf=0.6
Vf=0.8
(c)
Fig 3 (c) Magnetic potential ( ) at the top surface of the MEE sensor for point load
5.1.2 Sensor patch located at the middle
0 500 1000 1500 2000 2500 3000 3500
1E-10
1E-9
1E-8
1E-7
1E-6
U
z
(
m
)
f (Hz)
Vf=0.0
Vf=0.2
Vf=0.4
Vf=0.6
Vf=0.8
Vf=1.0
0 500 1000 1500 2000 2500 3000 3500
1E-4
1E-3
0.01
0.1
(
v
o
l
t
)
f (Hz)
Vf=0.2
Vf=0.4
Vf=0.6
Vf=0.8
Vf=1.0
(a) (b)
0 500 1000 1500 2000 2500 3000 3500
1E-7
1E-6
1E-5
1E-4
1E-3
(
A
)
f (Hz)
Vf=0.0
Vf=0.2
Vf=0.4
Vf=0.6
Vf=0.8
(c)
Figure 4 (a) Transverse displacement , (b) Electric potential ( ) and (c) Magnetic potential ( )
z
u ( ) at the top surface of the
MEE sensor for point load
The transverse displacement ( , electric potential ( ) and magnetic potential )
z
u ( ) at the top surface of the MEE sensor when
the sensor is located at the middle of the mild steel beam is shown in Figure 4. The sensor output i.e. electric potential ( ) and
Biju et al. / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 216-227 223
magnetic potential ( ) decreases considerably when the sensor is placed at centre. This is because the sensor is less strained as it
is moved away from the clamped end. The absolute values of sensor out put during second mode of vibration of the beam are very
much close to that of first mode when the sensor is placed at the middle.
5.1.3 Sensor patch located at the tip
Figure 5 shows the transverse displacement ( , electric potential ( ) and magnetic potential )
z
u ( ) at the top surface of the
MEE sensor when the sensor is located at the free end of the mild steel beam. The third mode of vibration becomes dominant when
the sensor is placed at the free end. The sensor is strained very little that the absolute values of sensor out put decreases
considerably. Volume fraction 1.0
which corresponds to pure BaTiO
3
gives maximum electric potential ( ). It is also seen that
volume fraction 0.0
which corresponds to pure CoFe
2
O
4
gives maximum magnetic potential ( ) .
0 500 1000 1500 2000 2500 3000 3500
1E-10
1E-9
1E-8
1E-7
1E-6
U
z
(
m
)
f (Hz)
Vf=0.0
Vf=0.2
Vf=0.4
Vf=0.6
Vf=0.8
Vf=1.0
0 500 1000 1500 2000 2500 3000 3500
1E-6
1E-5
1E-4
1E-3
(
v
o
l
t
)
f (Hz)
Vf=0.2
Vf=0.4
Vf=0.6
Vf=0.8
Vf=1.0
(a) (b)
0 500 1000 1500 2000 2500 3000 3500
1E-7
1E-6
1E-5
(
A
)
f (Hz)
Vf=0.0
Vf=0.2
Vf=0.4
Vf=0.6
Vf=0.8
(c)
Figure 5 (a) Transverse displacement ( , (b) Electric potential ( ) and (c) Magnetic potential )
z
u ( ) at the top surface of the
MEE sensor for point load
5.2 Uniformly distributed load at the bottom surface of the beam
The beam is clamped at one end and response of the sensor is studied when the beam is subjected to a harmonic excitation of 1N in
the negative z-direction along the entire bottom surface of the beam.
5.2.1 Sensor patch located at the clamped end
Figure 6 shows the transverse displacement ( , electric potential ( ) and magnetic potential )
z
u ( ) at the top surface of the
MEE sensor when the sensor is located at the clamped end of the mild steel beam. The sensor output i.e. electric potential ( ) and
magnetic potential ( ) is high compared with point loading, as expected. It is also noticed that over a wide range of frequencies
the sensor shows significantly large response when the applied load is uniformly distributed compared to point load.
Biju et al. / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 216-227 224
0 500 1000 1500 2000 2500 3000 3500
1E-11
1E-10
1E-9
1E-8
1E-7
U
z
(
m
)
f (Hz)
Vf=0.0
Vf=0.2
Vf=0.4
Vf=0.6
Vf=0.8
Vf=1.0
0 500 1000 1500 2000 2500 3000 3500
1E-3
0.01
0.1
1
10
(
v
o
l
t
)
f (Hz)
Vf=0.2
Vf=0.4
Vf=0.6
Vf=0.8
Vf=1.0
(a) (b)
0 500 1000 1500 2000 2500 3000 3500
1E-6
1E-5
1E-4
1E-3
0.01
0.1
(
A
)
f (Hz)
Vf=0.0
Vf=0.2
Vf=0.4
Vf=0.6
Vf=0.8
(c)
Fig 6 (a) Transverse displacement , (b) Electric potential ( ) and (c) Magnetic potential ( )
z
u ( ) at the top surface of the MEE
sensor for UDL
5.2.2 Sensor patch located at the middle
0 500 1000 1500 2000 2500 3000 3500
1E-9
1E-8
1E-7
1E-6
1E-5
U
z
(
m
)
f (Hz)
Vf=0.0
Vf=0.2
Vf=0.4
Vf=0.6
Vf=0.8
Vf=1.0
0 500 1000 1500 2000 2500 3000 3500
1E-3
0.01
0.1
1
(
v
o
l
t
)
f (Hz)
Vf=0.2
Vf=0.4
Vf=0.6
Vf=0.8
Vf=1.0
(a) (b)
Figure 7 (a) Transverse displacement ( , (b) Electric potential ( ) )
z
u
Biju et al. / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 216-227 225
0 500 1000 1500 2000 2500 3000 3500
1E-6
1E-5
1E-4
1E-3
0.01
(
A
)
f (Hz)
Vf=0.0
Vf=0.2
Vf=0.4
Vf=0.6
Vf=0.8
(c)
Figure 7 (c) Magnetic potential ( ) at the top surface of the MEE sensor for UDL
The transverse displacement ( , electric potential ( ) and magnetic potential )
z
u ( ) at the top surface of the MEE sensor when
the sensor is located at the middle of the mild steel beam is shown in Figure 7. The second mode domination for response is less
when the sensor is placed at the middle and the applied load is distributed uniformly. This is because strains are developed in the
sensor during first mode of vibration also. The magnetic potential ( ) response become out of phase during the third mode.
5.2.3 Sensor patch located at the tip
0 500 1000 1500 2000 2500 3000 3500
1E-10
1E-9
1E-8
1E-7
1E-6
1E-5
U
z
(
m
)
f (Hz)
Vf=0.0
Vf=0.2
Vf=0.4
Vf=0.6
Vf=0.8
Vf=1.0
0 500 1000 1500 2000 2500 3000 3500
1E-5
1E-4
1E-3
0.01
(
v
o
l
t
)
f (Hz)
Vf=0.2
Vf=0.4
Vf=0.6
Vf=0.8
Vf=1.0
(a) (b)
0 500 1000 1500 2000 2500 3000 3500
1E-7
1E-6
1E-5
1E-4
(
A
)
f (Hz)
Vf=0.0
Vf=0.2
Vf=0.4
Vf=0.6
Vf=0.8
(c)
Fig 8 (a) Transverse displacement , (b) Electric potential ( ) and (c) Magnetic potential ( )
z
u ( ) at the top surface of the MEE
sensor for UDL
Figure 8 shows the transverse displacement ( , electric potential ( ) and magnetic potential )
z
u ( ) at the top surface of the
MEE sensor when the sensor is located at the free end of the mild steel beam. When the sensor is placed at the free end, v
f
1.0
Biju et al. / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 216-227 226
which corresponds to pure BaTiO
3
gives maximum electric potential ( ) and volume fraction 0.0
which corresponds to pure
CoFe
2
O
4
gives maximum magnetic potential ( ) .
6. Conclusions
The finite element formulation for coupled magneto-electro-elastic sensor bonded to a mild steel beam with plane stress
assumption is presented in this paper. The location of the sensor is varied along the length of the beam to study the effect on sensor
response. Two types of harmonic loading i.e. a point load and a uniformly distributed load is applied harmonically to observe the
sensor response at various locations.
The sensor out put viz. electric potential ( ) and magnetic potential ( ) reduces considerably as the sensor is placed away from
the fixed end. This is due to the fact that the first natural frequency is the dominant mode and the beam is strained maximum at the
fixed end. The electric potential ( ) is found maximum for 0.2 v
f
of the composite where as the magnetic potential ( ) is found
maximum for 0.6 v
f
of the composite. The electric and magnetic potential decreases at higher modes. It is beneficial to place the
sensor at the fixed end because it will be protected from damage as well as maximum sensor out put will be obtained. When the
sensor is placed at the free end of the cantilever beam, the third mode dominates the response of the sensor.
References
Aboudi J., 2001. Micromechanical analysis of fully coupled electro-magneto-thermo-elastic multiphase composites, Smart
Materials and Structures, Vol. 10, pp. 867-877.
Aimin J. and Ding H.J., 2004. Analytical solutions to magneto-electro-elastic beams, Structural Engineering and Mechanics, Vol.
18, No. 2, pp. 195-209.
Annigeri A.R., Ganesan N. and Swarnamani S., 2007. Free vibration behavior of multiphase and layered magneto-electro-elastic
beams, Journal of Sound and Vibration, Vol. 299, pp. 44-63.
Atul D., Ganesan N. and Shankar K., 2009. Behaviour of magneto-electro-elastic sensors under transient mechanical loading,
Sensors and Actuators A:Physical, Vol. 150, pp. 46-55.
Bakhashwain J.M., Sunar M., Hyder S.J., 2004. Finite element modeling of magnetostrictive smart structures, The Arabic journal
of Science and Engineering, Vol. 29, pp. 27-138
Buchanan, G. R., 2003. Free vibration of an infinite magneto-electro-elastic cylinder, Journal of Sound and Vibration, Vol. 268,
pp. 413-426.
Bhangale R.K. and Ganesan N., 2005. Free vibration studies of simply supported non-homogeneous functionally graded
magneto-electro-elastic finite cylindrical shells, Journal of Sound and Vibration, Vol. 288, No. 1-2, pp. 412-422.
Buchanan G.R., 2004. Layered versus multiphase magneto-electro-elastic composites, Composites Part B, Vol. 35, pp. 413-420.
Dai H.l. and Wang X., 2006. Magneto-thermo-electro-elastic transient response in a piezoelectric hollow cylinder subjected to
complex loading, International Journal of Solids and Structures, Vol. 43, pp. 5628-5646.
Ghosh, D. P. and Gopalakrishnan S., 2005. Coupled analysis of composite laminate with embedded magnetostrictive patches
Smart Materials and Structures, Vol. 14, pp. 1462-1473.
Hou P.-F. and Leung A.Y.T., 2004. The transient responses of magneto-electro-elastic hollow cylinders, Smart Material systems
and Structures, Vol. 13, pp. 762-776.
Marek P., 2008. Piezoelectric control of composite plate vibration: Effect of electric potential distribution, Computers and
Structures, Vol. 86, pp. 948-954.
Nocolas G., Xavier M., Frederic B. and Laurent D., 2008. IEEE transactions on Magnetics, 44(2008) 834-837.
Ramirez F., Heyliger P.R. and Pan E., 2006. Free vibration response of two-dimensional magneto-electro-elastic laminated plates.
Journal of Sound and Vibration, Vol. 292, pp. 626-644.
Sirohi J. and Copra I., 2000. Fundamental understanding of piezoelectric strain sensors, Journal of Intelligent Systems and
Structures, Vol. 11, pp. 246-257.
Sunar, M., Ahmed Z. Al-Garni, M. H. Ali and R. Kahraman, 2002. Finite element modeling of thermopiezomagnetic smart
structures, AIAA Journal, Vol. 40, pp. 1846-1851.
Wang, X. and Z. Zhong, 2003a. A finitely long circular cylindrical shell of piezoelectric/ piezomagnetic composite under
pressuring and temperature change, International Journal of Engineering Science, Vol. 41, pp. 2429-2445.
Zienkiewicz O.C., Lyness John and Owen D.R.J., 1977. Three dimensional magnetic field determination using a scalar potential-
A finite element solution, IEEE transactions on Magnetics, Vol. 13, pp. 1649-1656.
Biju et al. / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 216-227 227
Biographical notes
B. Biju has received his B.Tech Degree in Mechanical Engineering from Mahatma Gandhi University, Kottayam, Kerala, India in 1992 and
M.Tech Degree in Machine Dynamics from Indian Institute of Technology Madras, Chennai, India in 2004. He is currently a Ph.D research
scholar in Machine Design Section at Indian Institute of Technology Madras, Chennai, India. His research interest includes modeling and
analysis of smart materials and structures.
N. Ganesan was born in India in1945. He did his graduation in 1967, post graduation in 1971 as well as Ph.D in 1976 from Indian Institute of
Technology Madras, Chennai, India. He is working as Professor in the department of Mechanical Engineering at Indian Institute of Technology
Madras, India. His present fields of interest are finite element modeling of smart materials and structures, sandwich structures as well as
acoustics.
K. Shankar was born in India in1967. He did his graduation in 1989 from College of Engineering, Trivandrum, India. He obtained his Ph.D in
1996 from University of Oxford, England. He is working as Associate Professor in the department of Mechanical Engineering at Indian Institute
of Technology Madras, India. His present fields of interest include inverse problems in vibration, system identification and design optimization.
Received November 2009
Accepted December 2009
Final acceptance in revised form December 2009
MultiCraft
International Journal of Engineering, Science and Technology
Vol. 1, No. 1, 2009, pp. 228-244
INTERNATIONAL
JOURNAL OF
ENGINEERING,
SCIENCE AND
TECHNOLOGY
www.ijest-ng.com
2009 MultiCraft Limited. All rights reserved
Propagation of shear waves in an irregular magnetoelastic monoclinic layer
sandwiched between two isotropic half-spaces
A. Chattopadhyay, S. Gupta, A.K. Singh and S.A. Sahu
Department of Applied Mathematics, Indian School of Mines (ISM),Dhanbad-826004, Jharkhand, India
E-mail: (amares.c@gmail.com (A. Chattopadhyay),
*
Corresponding author)
Abstract
In the present paper we study the propagation of horizontally polarised shear waves (SH waves) in an internal magnetoelastic
monoclinic stratum with rectangular irregularity in lower interface and is sandwiched between two semi-infinite isotropic elastic
media. The dispersion equation has been obtained. It is observed that the dispersion equation is in assertion with the classical
Love-type wave equation, for isotropic layer sandwiched between two isotropic half spaces, in the absence of magnetic field and
irregularity. The effects of depth of irregularity and monoclinic-magnetoelastic coupling parameters on dispersion curves are
depicted by means of graphs. This study shows the remarkable effect of wave number, size of irregularity and magnetic field on
the phase velocity.
Keywords: SH wave, magnetoelastic monoclinic, irregularity, dispersion equation, perturbation
1. Introduction
Many results of theoretical and experimental studies revealed that a real earth is considerably more complicated than the models
presented earlier. This has led to a need for more realistic representation as a medium through which seismic waves propagate. The
study of wave propagation in elastic medium with different boundaries is of great importance to seismologists as well as to
geophysicists to understand and predict the seismic behaviour at the different margins of earth.
The propagation of Love waves has been studied by many authors with assuming different forms of irregularities at the interface.
Bhattacharya (1962) discussed the dispersion curves for Love wave propagation in a transversely isotropic crustal layer with an
irregularity in the interface. Chattopadhyay (1975) studied the Love wave propagation due to irregularity in the thickness of non-
homogeneous crustal layer. The wave propagation in crystalline media plays a very interesting role in geophysics and also in
ultrasonic and signal processing. Chattopadhyay and Bandyopadhyay (1986) studied the propagation of shear waves in an infinite
monoclinic crystal plate as well as in infinite anisotropic non-homogeneous monoclinic plate. The study of wave propagation in
irregular boundaries is also of equal importance to seismologists and geophysicists. Ge and Chen (2007) proposed a direct
boundary element method for calculating elastic wave propagation in two-dimensional irregularly
stratified models. Further a more
efficient approach was developed by same authors (Ge and Chen, 2008) for seismic-wave
simulation in a stratified medium with
irregular interfaces. In particular different authors have done the study of SH-wave propagation in different irregular mediums.
Chattopadhyay et al. (1983) studied the propagation of SH guided wave in an internal stratum with parabolic irregularity in the
lower interface. Campillo (2006) studied the case of a dome in a stratified medium for modelling of SH-wave propagation in an
irregularly layered medium. Ding and Dravinski (1998) studied the scattering of SH-waves in irregular interfaces. Reflection and
transmission of SH-wave at a corrugated interface has been studied before. Kaur et al. (2005) obtained the Reflection and
transmission coefficients due to incident plane SH-waves at a corrugated interface between two isotropic, laterally and vertically
heterogeneous visco-elastic solid half spaces. Tomar and Kaur (2007) studied the reflection and transmission of SH-waves at a
corrugated interface between anisotropic elastic and viscoelastic solid half-spaces.
Keeping in the mind the characteristic and structure of earth, the medium of monoclinic is of core interest for seismologists to
study the seismic wave propagation. Singh and Tomar (2007) studied the propagation of quasi-P wave at the interface between two
monoclinic half-spaces. Kalyani et al. (2008) studied the propagation of SH waves in the plane of mirror symmetry of a
Chattopadhyay et al. / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 228-244 229
monoclinic multilayered medium with displacement normal to the plane. Recently, the propagation of SH waves in an irregular
monoclinic crustal layer has been studied by Chattopadhyay et al. (2008). In this paper we have discussed the propagation of SH
waves in an irregular magnetoelastic monoclinic layer lying in between two isotropic half-spaces. The irregularity is in the form of
rectangle. The perturbation technique indicated by Eringen and Samuels (1959) has been used. It is observed that phase velocity is
strongly influenced by the depth of irregularity and magnetic field. These important peculiarities are highlighted by means of
graphs.
2. Formulation of the problem
1
M Isotropic Medium
y H =
2
M Magnetoelastic Monoclinic layer H
0 y = / 2 s O / 2 s z
H
3
M d
Isotropic Medium
S
y
Figure 1: Geometry of the problem
We consider the model, which is comprised of perfectly conducting magnetoelastic monoclinic irregular internal stratum,
2
M sandwiched between two isotropic elastic semi-infinite media
1
M (upper) and
3
M (lower) as shown in Figure 1. The
irregularity is considered in the form of rectangle on the lower interface of the internal stratum. We assume the rectangular
irregularity with the length s and depth H
. Let us choose the rectangular coordinate system with the origin O at the middle point
of the span, , s of the irregularity. We have taken y-axis vertically downwards and z axis along the regular interface between the
lower semi-infinite medium and the monoclinic magnetoelastic layer. Let H be the thickness of the internal stratum except for the
irregular region and S be the position of the source on the y-axis at a depth d below the origin ( d H
> ).
The equation of the interface between layer and lower half space is defined as
( ) y h z = (1)
where
0 for ,
2 2
( )
( ) for
2 2
s s
z z
h z
s s
f z z
Chattopadhyay et al. / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 228-244 230
where
H
s
= and 1 << .
Let us consider , , ( 1, 2, 3)
r r r
u r = as the rigidities, densities and displacements of the upper half space, sandwiched layer, lower
half space respectively.
For upper (
1
M ) and lower (
3
M ) semi-infinite medium,
2 2 2
1 1 1
2 2 2 2
1
1 u u u
y z t
+ =
(2)
and
2 2 2
3 3 3
2 2 2 2
3
1 u u u
y z t
+ =
(3)
where
1
1
1
= and
3
3
3
+ + + =
+ + + =
+ + + =
(6)
where
2
is the density of the layer.
For SH wave propagating in the z- direction and causing displacement in the x- direction only, we shall assume that
( )
2 2 2 2
, , , 0 u u y z t v w = = = and 0.
x
(7)
Using Eqs. (4) and (7), the stress-strain relation (5) becomes
Chattopadhyay et al. / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 228-244 231
1 2 3 4
2 2
5 55 56
2 2
6 56 66
0,
,
.
T T T T
u u
T C C
z y
u u
T C C
z y
= = = =
= +
= +
(8)
Using Eq. (8) in Eq. (6), we have
( )
2 2 2 2
2 2 2 2
66 56 55 2 2 2 2
2 .
x
u u u u
C C C J B
y y z z t
+ + + =
(9)
The well known Maxwells equations governing the electromagnetic field are . 0, ,
t
= = =
B
B E H J with
,
i
e
u
t
= = +
B H J E B (10)
where E is the induced electric field, J is the current density vector and magnetic field H includes both primary and induced
magnetic fields.
e
and are the induced permeability and conduction coefficient respectively.
The liberalized Maxwells stress tensor
( )
0 x
M
ij
due to the magnetic field is given by
( ) ( )
0 x
M
ij e i j j i k k ij
H h H h H h = + .
Let ( )
1 2 3
, , H H H = H and ( )
1 2 3
, ,
i
h h h = h where
i
h is the change in the magnetic field. In writing the above equations, we have
neglected the displacement current.
From Eq. (10), we get
{ }
2
.
i
e
u
t t
= +
H
H H (11)
In component form, Eq. (11) can be written as
2 2
2 3
2 1
1
2 2
2
2 3
3
1
,
1
,
1
.
e
e
e
u u
H H
H
t t
H
t y z
H
H
t
H
H
t
= + +
(12)
For perfectly conducting medium, the equation (12) become
3 2
0,
H H
t t
= =
(13)
and
2 2
2 3
1
.
u u
H H
H
t t
t y z
= +
(14)
It is clear from Eq. (13) that there is no perturbation in
2
H and
3
H , however from Eq. (14) there may be perturbation in
1
H .
Therefore, taking small perturbation, say
1
h in
1
H , we have
1 01 1 2 02
, H H h H H = + = and
3 03
H H = , where ( )
01 02 03
, , H H H are components of the initial magnetic field
0
H .
We can write ( )
0 0 0
0, cos , sin H H = H , where
0 0
H = H and is the angle at which the wave crosses the magnetic field.
Thus we have
( )
1 0 0
, cos , sin h H H = H . (15)
We shall take initial value of
1
h as
1
0 h = . Using Eq. (15) in Eq. (14), we get
Chattopadhyay et al. / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 228-244 232
2 2
0 0
1
cos sin
.
u u
H H
h
t t
t y z
= +
(16)
Integrating with respect to t , we get
2 2
1 0 0
cos sin .
u u
h H H
y z
= +
(17)
Considering ( ) ( )
2
.
2
H
= +
H H H H and Eqs (10), we get
( )
2
. .
2
e
H
= +
J B H H (18)
In the component form Eq. (18) can be written as
( ) ( )
( )
2 2 2
2 2 2 2 2 2
0 2 2
0
and
cos sin 2 sin .
y z
e x
u u u
H
y y z z
= =
= + +
J B J B
J B
(19)
Using Eqs. (9) and (19), we find the equation of motion for the magnetoelastic monoclinic layer in the form
2 2 2 2
2 2 2 2
66 56 55 2 2 2 2
2 .
u u u u
M M M
y y z z t
+ + =
(20)
where
( )
2
66 66
2 55
55 66
66
56
56 66
66
1 cos ,
sin ,
cos sin .
H
H
H
M C m
C
M C m
C
C
M C m
C
= +
= +
= +
(21)
Where
2
0
66
e
H
H
m
C
= is monoclinic-magnetoelastic coupling parameter. Here in equations (2), (3) and (20) suffixes 1, 2 and
3 are used to represent the upper semi-infinite medium (
1
M ), monoclinic layer (
2
M ) and lower semi-infinite medium (
3
M )
respectively.
The boundary conditions are as follows:
(i) The stresses are continuous at the interface of
1
M and
2
M ,
1 2 2
1 56 66
u u u
M M at y H
y z y
= + =
. (22)
(ii) The Displacements are continuous at the interface of
1
M and
2
M ,
1 2
u u at y H = = . (23)
(iii) The stresses are continuous at the interface of
2
M and
3
M ,
( ) ( ) ( )
3 3 2 2
56 55 66 56 3
u u u u
M M h M M h h at y h z
z y y z
+ = =
(24)
where
( ) dh z
h
dz
= .
(iv) The Displacements are continuous at the interface of
2
M and
3
M ,
( )
2 3
u u at y h z = = . (25)
Chattopadhyay et al. / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 228-244 233
3. Solution of the problem
Let us assume ( ) ( , ) 1, 2, 3
i t
i i
u U y z e i
+ + =
(26)
2 2 2
3 3
3 2 2 2
3
0
U U
U
y z
+ + =
(27)
and
2 2 2
2 2 2 2
66 56 55 2 2 2 2
2 0.
U U U
M M M U
y y z z
+ + + =
(28)
where is the circular frequency.
Defining the Fourier transform ( , )
r
U y of ( , ) ( 1, 2, 3)
r
U y z r = as
( , ) ( , ) ,
i z
r r
U y U y z e dz
=
We find the inverse Fourier transform as
1
( , ) ( , ) .
2
i z
r r
U y z U y e d
=
Using Fourier transform in equations (26), (27) and (28), we have
2
2 1
1 1 2
0,
d U
p U
dy
= (29)
2
2 3
3 3 2
0
d U
p U
dy
= (30)
and
2
2 2 2
2 2 2
0
d U dU
a p U
dy dy
+ + = (31)
where
2 2 2
2 2 2 2 2 2 55 56
1 2 3 2 2 2
1 2 66 3 66
, , and 2 .
M M
p p p a i
M M
= = = =
The appropriate solutions of Eqs. (29) and (30) can be taken as
1
1
p y
U Ae = (32)
and
3
3
.
p y
U E e
= (33)
The solution of Eq. (31) is
2
2
( cos sin )
a
y
U e B py D py
= + (34)
where
1/ 2
2 2 2
2 2 2 55 56
2 2 2
2 66 66
4
M M a
p p
M M
= = +
.
The constants , , and A B D E are functions of .
The displacements in
1 2
M , M and
3
M are
1
1
1
( , )
2
p y i z
U y z Ae e d
=
, (35)
Chattopadhyay et al. / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 228-244 234
( )
2
2
1
( , ) cos sin ,
2
a
y
i z
U y z e B py D py e d
= +
(36)
and
3 3 3
3
3
1 2
( , )
2
p y p y p d i z
U y z E e e e e d
p
= +
. (37)
where the second term in the integrand of
3
U is introduced due to the effect of source at S in
3
M (Willis, 1948). We set the
following approximations due to small value of
0 1 0 1 0 1
, , . B B B D D D E E E + + + (38)
Since the boundary is not uniform at the interface of
2
M and
3
M so the term , and B D E appearing in Eq.(38) are also function
of , expanding these terms in ascending powers of and keeping in view that is small and so retaining the term up to the first
order of , hence approximated as in Eq.(38). These assumptions are justified in the real earth model where the depth H
of the
irregular boundary is too small with respect to the length of the boundary s .
Also for small , following approximations can be accepted
1 , cos 1, sin
h
e h p h p h p h
.
where is any quantity.
Using the boundary conditions (i), (ii) and (iv), we have the following equations
( ) ( ) { }( ) ( ) {
( ) } ( ) ( ) { }( )
1
2
56 0 1 0 1 66 0 1
2 2
0 1 66 0 1 0 1
1 1
cos sin cos
sin sin cos
2
,
aH
aH aH
p H
M B B pH D D pH i e M B B pH
a
D D pH e M B B pH D D pH p e
Ap e
+ + +
+ + + + +
=
(39)
( ) ( ) { }
1 2
0 1 0 1
cos sin
aH
p H
B B pH D D pH e Ae
+ + = (40)
and
3
3
0 0 3 0
0 0 1 1
3
2 ( )
2
2
( ) .
p d i z
p d i z
a
B pD p E e h z e d
B E e B E e d
p
+
= +
(41)
Now applying the Fourier transform of ( ) h z defined as
( ) ( )
i z
h h z e dz
=
(42)
so that
1
( ) ( )
2
i z
h z h e d
=
(43)
and
( ) ( )
2
i z
i
h z h e d
=
. (44)
Now applying the transform (43) and (44) on (41), we have
( )
3
3
0 0 3 0
0 0 1 1
3
2 ( )
2 2
2
( ) .
p d i z
p d i z
a
B pD p E e h e d d
B E e B E e d
p
+
= +
(45)
Chattopadhyay et al. / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 228-244 235
Taking k + = for the inner integral in the left-hand side of the Eq. (45), so that may be treated as a constant such that
d dk = , replacing by k in the right-hand side of the Eq.(45), and finally after taking Fourier transform as defined above we
can have
3
0 0 1 1 1
3
2
( ) ( )
p d
B E e B E R k
p
+ =
(46)
where
3
1 0 0 3 0
1
( ) 2 ( )
2 2
k
p d
a
R k B pD p E e h d
= +
(47)
From the boundary condition (iii), we obtain
{ }
( )
3
66 0 0 1 1 56 0 1 3 3 0 3 1
2
( ) ( 2 )
2 2
( )
p d
a a
M B pD B pD iM k B B p E e p E
R k
+ + + +
=
(48)
where
( )
3
3
2
2 2
2 66 0 0 0 56 0 0 3 3 0 3
56 0 0 55 0 3 0
3
1
( ) 2
2 4 2
2
( ) .
2
p d
k
p d
a a
R k M B p B apD M B pD ik p E p e
a
iM D p B M kB k E e h d
p
= + +
+ + +
(49)
Equating the absolute term ( term not containing ) and the coefficient of from Eqs. (39),
(40), (46) and (48) we get
1
2
0 56 66 66
2
0 56 66 66 1 1
cos cos sin
2
sin sin cos ,
2
aH
aH
p H
a
B i M pH M pH pM pH e
a
D i M pH M pH pM pH e Ap e
+
+ + + =
(50)
2
1 56 66 66
2
1 56 66 66
cos cos sin
2
sin sin cos 0,
2
aH
aH
a
B i M pH M pH pM pH e
a
D i M pH M pH pM pH e
+
+ + + =
(51)
( )
1 2
0 0
cos sin ,
aH
p H
B pH D pH e Ae
= (52)
( )
2
1 1
cos sin 0,
aH
B pH D pH e = (53)
3
0 0
3
2
0,
p d
B E e
p
= (54)
1 1 1
( ), B E R k = (55)
( ) ( )
3
0 66 56 66 0 3 3 0
2 0,
2
p d
a
B M ikM pM D p E e
+ =
(56)
and
( )
1 66 56 66 1 3 3 1 2
( ) ( ).
2
a
B M ikM pM D p E R k + =
(57)
Solving the above eight equations, we get
Chattopadhyay et al. / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 228-244 236
( )
( )
( )
3
3
3
3
0 56 66 66 1 1
3
0 56 66 66 1 1
0 2 3
3
8
2 tan tan 2 2 tan ,
( )
8
2 2 tan 2 ,
( )
2
,
( )
p d
p d
p d
e
B ikM pH aM pH pM p pH
G k
e
D ikM aM pM pH p
G k
e
E E E
p G k
= + + +
= + +
= +
( )
( )
( )
( )
( )
( )
( ) 1
3 3 1 2
1 56 66 66 1 1
3 3 1 2
1 56 66 66 1 1
4 5
1
2
0 0
2
2 tan tan 2 2 tan ,
( )
2
2 2 tan 2 ,
( )
,
( )
cos sin ,
a
p H
p R R
B ikM pH aM pH pM p pH
G k
p R R
D ikM aM pM pH p
G k
E E
E
G k
A B pH D pH e
+
= + + +
= + +
+
=
=
2 3 3 56 3 3 66 3 3 66 3 3 1 1
4 tan 2 tan 4 4 tan , E ikp M pH ap M pH p pM p p pH = + + +
2 2 2 2
3 66 1 1 66 56 66 66
2 2
1 1 66 56 1 1 56
4 tan 4 4 tan tan
2 tan 4 tan 4 tan ,
E p M pH pp M ikaM M pH a M pH
ap M pH k M pH ikp M pH
= + +
+ +
4 2 56 2 66 2 66 1 1 2 1 56 66
1 1 1 56
4 tan 2 tan 4 4 tan 4 tan
4 tan ,
E ikR M pH aR M pH pR M p R pH ikaR M M pH
ikR p M pH
= + + +
2 2 2 2 2 2
5 1 66 1 1 1 66 1 66 1 1 1 66 1 56
4 tan 4 tan 2 tan 4 tan , E p R M pH pp R M a R M pH ap R M pH k R M pH = + +
2 2 2 2
66 1 1 66 56 66 66 1 1 66
2 2
56 1 1 56 3 3 56 3 3 66
3 3 66 3 3 1 1
( ) 4 tan 4 4 tan tan 2 tan
4 tan 4 tan 4 tan 2 tan
4 4 tan .
G k p M pH pp M ikaM M pH a M pH ap M pH
k M pH ikp M pH ikp M pH p aM pH
p pM p p pH
= + + +
+
With the help of above obtained values the displacement in the monoclinic layer is given by
( ) {
( ) }
3
3
2
3 3 3 1 2
2
3
56 66 66 1 1
56 66 66 1 1
8 ( ) 1
1
2 ( ) 4
2 tan tan 2 2 tan cos
2 2 tan 2 sin .
a
y
p d
p d
ikz
e e p R R
U
G k e
ikM pH aM pH pM p pH py
ikM aM pM pH p py e dk
= +
+ + +
+ + +
(58)
From Eq. (1), the interface for the rectangular irregularity gives
2
( ) sin .
2
s s
h
=
(59)
Further on simplification, we get
3
3 3 1 2
2 1
{ ( ) ( )} sin ,
2
s s
p R R k k d
= + +
(60)
where
( )
3
2 3 4 5 6 7 8 9 10 11 12
3
( )
( )
k
p d
e
k B B B B B B B B B B B
p G k
=
= + + + + + + + + + +
and
Chattopadhyay et al. / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 228-244 237
2 2 2 2 3 2 2 2
2 3 3 66 3 66 3 66 1 3 1 66 3
2 2 2 2 2
3 3 3 66 3 3 66 3 66
2 2 3 2
4 3 3 66 3 56 66 3 66 3 56 66
2
5 1 3 1
2 tan 2 8 8 tan 2 ( ),
4 tan 8 tan 4 tan ,
8 4 tan tan 8 ,
4
B a p M pH a pp M p p M p p p M pH p E k
B iak p M pH p p M pH ap p M pH
B ikpp M ika p M M pH a p M pH iakpp M M
B ap p
= +
= +
= + +
=
2 2
3 1 3 1 3 1 3 1 66 1 3 1 66
2 2 2
6 3 56 3 56 3 55 56 3 55 66
tan 8 2 tan 8 ,
8 4 tan 8 tan 8 ,
pH pp p a p p M pH app p M
B kpp M akp M pH i k p M M pH kpp M M
+ +
= + + +
2
7 3 3 56 3 3 66 3 55 66
2 2 2
8 66 56 66 3 3 66
3 2 2 2 2 2
9 56 66 3 56 1 1 66
2 2
10 3 56 66
4 tan 2 tan 4 tan ,
4 tan 4 tan 4 ( ),
4 tan tan 4 tan 4 ,
8 tan 2
B i k p M pH a k p M pH a kp M M pH
B kp M pH ia k M M pH kpp M kE k
B k M pH a kM pH ak p M pH kpp M
B i p p M M pH ia
= +
= +
=
=
2 2
3 56 66 3 56
11 1 3 1 56 1 3 1 56 1 3 1 55
2
12 1 1 66 1 1 56 1 3 1 3
tan 8 ,
8 4 tan 8 tan ,
2 tan 4 tan 4 tan .
p M M pH k pp M
B i pp p M ia p p M pH kp p M pH
B a kp M pH i k p M pH kp p pH
+
= +
= +
Here the argument of ( ) k is because of k + = . Following the asymptotic formula of Willis (1948) and neglecting the
terms containing 2/ s and higher powers of 2/ s for large s , we get
1
{ ( ) ( )} sin 2 ( ) ( ),
2 2
s
k k d k k
+ + =
= = (62)
Therefore, in view of Eq. (62) the displacement in the magnetoelastic monoclinic layer is
{
( ) }
3
3
2
3
2 56 66 66
1 1 56 66 66 1 1
8 1
(2 tan tan 2
2
( ) 1 ( )
2
2 tan ) cos 2 2 tan 2 sin .
a
y
p d
p d
ikz
e e
U ikM pH aM pH pM
H
G k k e
p pH py ikM aM pM pH p py e dk
= + +
+ + + +
Since the value of this integral depends entirely on the contribution of the poles of the Integrand, the dispersion equation for SH
waves is given by
3
( ) 1 ( ) 0.
2
p d
H
G k k e
=
(63)
The poles are located at the roots of above equation which has been examined in our study of shear waves (cf. Achenbach, 1975).
Eq. (63) may be rewritten as
2 3 4 5 6 7 8
( ) 0
2
H
D D D D D D D
+ + + + + = (64)
2 2 2 2
2 66 1 1 66 56 66 66
2 2
3 1 1 66 56 1 1 56 3 3 56
4 3 3 66 3 3 66 3 3 1 1
2 2 2 2 2
5 3 3 66 3 66 1 3
4 tan 4 4 tan tan ,
2 tan 4 tan 4 tan 4 tan ,
2 tan 4 4 tan ,
2 tan 2 8
D p M pH pp M ikaM M pH a M pH
D ap M pH k M pH ikp M pH ikp M pH
D p aM pH p pM p p pH
D a p M pH a pp M p p p
= + +
= +
=
=
2
1 66 3
2 2 2 2 2
6 3 3 66 3 3 66 3 66
2 2 3 2
7 3 3 66 3 56 66 3 66
2 2 2 2 3 2
8 3 56 3 56 3 56 66 3 66
tan 2 ( ),
4 tan 8 tan 4 tan ,
8 4 tan tan ,
4 tan 8 8 8 .
M pH p E k
D iak p M pH p p M pH ap p M pH
D ikpp M ika p M M pH a p M pH
D ak p M pH k pp M iakpp M M p p M
+
= +
= + +
= +
Chattopadhyay et al. / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 228-244 238
Let us assume that the wave is propagating along the surface with c as the common wave velocity, and then we can set the
following
1 1 3 3
, , and p kP p kP p kP a kA = = = =
1/ 2 1/ 2 1/ 2
2 2 2 2
55 56 56
1 3 2 2 2 2
2 66 66 1 3 66
where , 1 , 1 , 2 .
M M M c c c
P P P A i
M M M
= + = = =
In the view of above setting, Eq. (64) gives
2 2
55 56 1 2
2 2
2 66 66 1 2
tan
M M J J c
kH i
M M L L
+ = +
(65)
where
( )
( )
( )
3 2 2
1 1 1 66 3 3 66 1 3 1 3 1 3 1 66 66 3 3 66
2 2 2 2 2 2 2
66 1 3 1 3 3 3 66 3 66 1 1 66 3 1 1 3
2
3 3 56 1 1 56 56 66 1 3 1 3
8 8 8 8 8 8
8 8 8 8 8 8
8 8 8 8
J PP M PP M H kPPP H kPPP M H kP M H kPP M
P M PP H kP P M H kP PM H kP P M H kP P
H kPP M H kPP M H kP M M H kPP
= + +
+ + +
+ +
( )
( )
( )
( )(
56 66
2
2 2 2 2 2 2 2
1 66 1 3 1 3 3 3 66 3 66 1 1 66 3 1 1 3
2
2
56 66 1 3 1 3 56 66
2 2 2 2 2
2 3 3 56 1 1 56 66 1 3 1 3 3 3 66 3 66
2
1
/ ,
8 8 8 8 8 8
8 8 / ,
8 8 8 8 8 8
8
M M
L P M PP H kP P M H kP PM H kP P M H kP P
H kP M M H kPP M M
J H kPP M H kPP M P M PP H kP P M H kP PM
H kP P
= + + +
+
= + +
+
) ( )
( )
2 2
1 66 3 1 1 3 56 66 1 3 1 3 56 66
3 2 2
1 1 66 3 3 66 1 3 1 3 1 3 1 66 66 3 3 66
2 2 2 2 2 2 2
2 66 1 3 1 3 3 3 66 3 66 1 1 66 3 1 1
8 8 8 /
8 8 8 8 8 8 ,
8 8 8 8 8 8
M H kP P H kP M M H kPP M M
PP M PP M H kPPP H kPPP M H kP M H kPP M
L P M PP H kP P M H kP PM H kP P M H kP P
+
+ +
= + + +
( )
( )
2
3
2
2
56 66 1 3 1 3 56 66
8 8 / . H kP M M H kPP M M
+
The real part of the Eq. (65) gives the dispersion equation of SH waves. The dispersion relation for the SH waves is
2 2
55 56 1
2 2
2 66 66 1
tan
M M J c
kH
M M L
+ =
. (66)
4. Particular cases
Case 1:
When 0
H
m = the dispersion relation (66) reduces to
2 2
55 56 3
2 2
2 66 66 3
tan
C C J c
kH
C C L
+ =
where
( )
( )
( )
3 2 2
3 1 1 66 3 3 66 1 3 1 3 1 3 1 66 66 3 3 66
2 2 2 2 2 2 2
66 1 3 1 3 3 3 66 3 66 1 1 66 3 1 1 3
2
3 3 56 1 1 56 56 66 1 3 1 3
8 8 8 8 8 8
8 8 8 8 8 8
8 8 8 8
J PP C PP C H kPPP H kPPP C H kP C H kPP C
P C PP H kP P C H kP PC H kP P C H kP P
H kPP C H kPP C H kP C C H kPP
= + +
+ + +
+ +
( )
( )
( )
56 66
2
2 2 2 2 2 2 2
3 66 1 3 1 3 3 3 66 3 66 1 1 66 3 1 1 3
2
2
56 66 1 3 1 3 56 66
/ ,
8 8 8 8 8 8
8 8 / .
C C
L P C PP H kP P C H kP PC H kP P C H kP P
H kP C C H kPP C C
= + + +
+
which is the dispersion relation for SH waves propagating in an irregular monoclinic layer sandwiched between two isotropic half
spaces.
Chattopadhyay et al. / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 228-244 239
Case 2:
When
1
0 and 0
H
m = = the dispersion relation (66) reduces to
2 2
55 56 4
2 2
2 66 66 4
tan
C C J c
kH
C C L
+ =
where
( )
( )
( )
2 2 2
4 66 3 3 66 3 3 66
2 2 2 2 2 2 2 3 2
3 66 66 3 3 66 3 3 56 66
2
2 2 2 2 2 2 2 4 2 2
4 3 66 66 3 3 66 56 66
2 4 4 4
8 8 8 64 ,
8 8 8 64 .
J P H kP C H kP C P C
H kP PC P C H kP P C H k P P C C
L H kP PC P C H kP P C H k P C C
= +
+
= +
Which is the result obtained by Chattopadhyay et al. (2008) for SH waves propagating in an irregular monoclinic layer lying over
an isotropic half space.
Case 3:
When 0
H
m = and 0 H
which is the dispersion relation for SH waves propagating in a regular monoclinic layer sandwiched between two isotropic half
spaces.
Case 4:
When
66 55 2 56
, 0, 0
H
C C C m = = = = and 0 H
where
1/ 2
2
2 2
2
1
c
P
=
This is the dispersion relation for SH waves propagating in a regular isotropic layer sandwiched between two isotropic half spaces.
Case 5:
When
66 55 2 56 1
, 0, 0 and 0
H
C C C m = = = = = the dispersion relation (66) reduces to
2
5
2
2 5
tan 1
J c
kH
L
=
where
( )
( )
( )
1/ 2
2
2 2
2
2 2 2
5 2 2 3 3 2 3 3 2
2 2 2 2 2
2 3 2 2 2 2 3 3 2
2
2 2 2 2 2
5 2 3 2 2 2 2 3 3 2
1 ,
2 4 4 4
8 8 8 ,
8 8 8 .
c
P
J P H kP H kP P
H kP P P H kP P
L H kP P P H kP P
=
= +
=
This is the result obtained by Chattopadhyay (1975) for SH waves propagating in an irregular isotropic layer lying over an
isotropic half space.
Chattopadhyay et al. / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 228-244 240
Case 6:
When
1
0, 0
H
m = = and 0 H
+ =
which is the result obtained by Chattopadhyay and Pal (1983) for SH waves propagating in a regular monoclinic layer lying over
an isotropic half space.
5. Numerical examples
For the case of an irregular monoclinic magnetoelastic layer between two isotropic half spaces we take the following data
(i) For upper isotropic half space,
1
M (Gubbins, 1990)
10 2
1
3
1
7.10 10 N/m ,
3, 321Kg/m .
=
=
(ii) For monoclinic layer,
2
M (Tiersten, 1969)
11 2 11 2
55 56
11 2 3
66 1
0.94 10 N/m , 0.11 10 N/m ,
0.93 10 N/m , 7, 450 Kg/m .
C C
C
= =
= =
(iii) For lower isotropic half space,
3
M (Gubbins, 1990)
10 2
3
3
1
6.77 10 N/m ,
3, 323Kg/m .
=
=
Moreover, the following data are used
2
0
66
0.0, 0.35, 0.65, 0.4, 0.8; 10 .
e
H
H
m
C
= = =
The effect of magnetic field and irregular boundary on the propagation of plane SH waves propagating in perfectly conducting
monoclinic magnetoelastic layer with rectangular irregularity sandwiched between two isotropic elastic half spaces has been
depicted by means of graphs. Figure 2 to 6 gives the variation of non-dimensional phase velocity ( )
2
/ c with respect to non-
dimensional wave number kH for different values of monoclinic-magnetoelastic coupling parameters ( )
H
m . The small change in
the non-dimensional wave number produces substantial change in non-dimensional phase velocity. In each of these figures graphs
are drawn for different sizes of irregularity. It is observed that as the size of irregularity ( ) / H H increases, the phase velocity
decreases. Also with the increase of parameter ( )
H
m , the phase velocity decreases.
Chattopadhyay et al. / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 228-244 241
0 0.5 1 1.5 2 2.5 3 3.5
1.05
1.1
1.15
1.2
1.25
1.3
1.35
1.4
kH
c
/
2
1
2
3
1 for H
'
/H=0
2 for H
'
/H=0.15
3 for H
'
/H=0.30
m
H
=0.0
Figure 2: Dimensionless phase velocity against dimensionless wave number for 0.0
H
m = .
0 0.5 1 1.5 2 2.5 3 3.5
1.05
1.1
1.15
1.2
1.25
1.3
1.35
1.4
kH
c
/
2
1
2
3
1 for H'/H=0.0
2 for H'/H=0.15
3 for H'/H=0.30
m
H
=0.035
Figure 3: Dimensionless phase velocity against dimensionless wave number for 0.035
H
m = .
Chattopadhyay et al. / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 228-244 242
0 0.5 1 1.5 2 2.5 3
1.05
1.1
1.15
1.2
1.25
1.3
1.35
1.4
kH
c
/
2
1
2
3
1 for H'/H=0.0
2 for H'/H=0.15
3 for H'/H=0.30
m
H
=0.065
Figure 4: Dimensionless phase velocity against dimensionless wave number for 0.065
H
m = .
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
1.05
1.1
1.15
1.2
1.25
1.3
1.35
1.4
kH
c
/
2
1
2
3
1 for H'/H=0.0
2 for H'/H=0.15
3 for H'/H=0.30
m
H
=0.4
Figure 5: Dimensionless phase velocity against dimensionless wave number for 0.4
H
m = .
Chattopadhyay et al. / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 228-244 243
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45
1.05
1.1
1.15
1.2
1.25
1.3
1.35
1.4
kH
c
/
2
1
2
3
1 for H'/H=0.0
2 for H'/H=0.15
3 for H'/H=0.30
m
H
=0.8
Figure 6: Dimensionless phase velocity against dimensionless wave number for 0.8
H
m = .
6. Conclusion
In this paper the dispersion relation has been obtained in the closed form. The effect of size of irregularity and magnetoelastic
monoclinic parameter on the phase velocity of SH waves has been depicted and shown by means of graphs. From the above
studies it can be concluded that in a monoclinic medium, with the increase in the size of irregularity the phase velocity decreases.
Also, with the increase in the magnetoelastic monoclinic parameter the phase velocity retains the same nature. In the isotropic
case, when irregularity and magnetic field are absent, the dispersion equation is matched with the classical SH wave equation. The
present study has its application specially to the problem of waves and vibrations where the wave signals have to travel through
different layers of different material properties and containing irregularities due to continental margin, mountain roots etc. These
results can also be utilized in the interpretation and analysis of data of geophysical studies. The findings will be useful in
forecasting formation details at greater depth through signal processing and seismic data analysis. The present study may be
effectively utilized to generate initial data prior to exploitation of the formation. This study may be useful to geophysicist and
metallurgist for analysis of rock and material structures through Non-Destructive Testing (NDT).
Acknowledgement
The authors convey their sincere thanks to Indian School of Mines University, Dhanbad for providing JRF to Mr. Abhishek Kr.
Singh and also facilitating us with its best facility. Acknowledgement is also due to DST, New Delhi for the providing financial
support through Project No. SR/S4/MS: 436/07, Project title: Wave propagation in anisotropic media.
References
Achenbach J.D., 1975. Wave propagation in elastic solids, North Holland Pub. Comp., New York.
Bhattacharya J., 1962. On the dispersion curve for Love wave due to irregularity in the thickness of the transversely isotropic
crustal layer, Gerlands Beitrage zur Geophysik, Vol. 6, pp. 324-334.
Campillo M., 2006. Modeling of SH-wave propagation in an irregularly layered mediumapplication to seismic profiles near a
dome, Geophysical prospecting, Vol. 35, No. 3, pp. 236-249.
Chattopadhyay A., 1975. On the dispersion equation for Love wave due to irregularity in the thickness of non-homogeneous
crustal layer, Acta Geophysica Polonica, Vol. 23, pp. 307-317.
Chattopadhyay A., Bandyopadhyay U.K., 1986. Shear waves in an infinite monoclinic crystal plate, International Journal of
Engineering Science, Vol. 24, No. 10, pp. 1587-1596.
Chattopadhyay A., Chakraborty M., Pal A.K., 1983. Effects of irregularity on the propagation of guided SH waves, Journal de
Mecanique, Vol. 2, No. 2, pp. 215-225.
Chattopadhyay et al. / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 228-244 244
Chattopadhyay A., Gupta S., Sharma V.K., Kumari Pato, 2008. Propagation of SH waves in an irregular monoclinic crustal layer,
Archive of Applied Mechanics, Vol. 78, No. 12, pp. 989-999.
Chattopadhyay A., Pal A.K., 1983. Dispersion curves of SH waves caused by irregularity in the prestressed internal stratum, Acta
Geophysica Polonica, Vol. 31, No. 1, pp. 37-49.
Ding G., Darvinski M., 1998. Scattering of SH-waves in multilayered media with irregular interfaces, Earthquake Engineering &
Structural Dynamics, Vol. 25, No. 12, pp. 1391-1404.
Eringen A.C., Samuels C.J., 1959. Impact and moving loads on a slightly curved elastic half space, Journal of Applied Mechanics,
Vol. 26, pp. 491-498.
Ge Z., Chen X., 2007. Wave propagation in irregularly layered elastic models: a boundary element approach with a global
reflection/transmission matrix propagator, Bulletin of the Seismological Society of America, Vol. 97, No. 3, pp. 1025-1031.
Ge Z., Chen X., 2008. An efficient approach for simulating wave propagation with the boundary element method in multilayered
media with irregular interfaces, Bulletin of the Seismological Society of America, Vol. 98, No. 6, pp. 3007-3016.
Gubbins D., 1990. Seismology and Plate Tectonics, Cambridge University press, Cambridge.
Kalyani V.K., Sinha A., Pallavika, Chakraborty S.K., Mahanti N. C., 2008. Finite difference modelling of seismic wave
propagation in monoclinic media, Acta Geophysica, Vol. 56, No. 4, pp. 1074-1089.
Kaur J., Tomar S.K., Kaushik V.P., 2005. Reflection and refraction of SH-waves at a corrugated interface between two laterally
and vertically heterogeneous viscoelastic solid half-space, International Journal of Solids and Structures, Vol. 42, No. 13, pp.
3621-3643.
Singh S.S., Tomar S.K., 2007. Quasi-P waves at a corrugated interface between two dissimilar monoclinic elastic half-spaces,
International Journal of Solids and Structures, Vol. 44, No. 1, pp. 197-228.
Tiersten H.F., 1969. Linear Piezoelectric plate Vibrations, Plenum Press, New York.
Tomar S.K., Kaur J., 2007. Reflection and transmission of SH-waves at a corrugated interface between anisotropic elastic and
viscoelastic solid half-spaces J. Seismology, Vol. 3, pp. 235-258.
Tranter C.J., 1966. Integral Transform in Mathematical Physics, Methuen and Co. Ltd. London.
Willis H.F., 1948. A formula for expanding an integral as series, Philosophical Magazine, Vol. 39, pp. 455-459.
Biographical notes
Professor Amares Chattopadhyay is working in Indian School of Mines since 1976. He was Awarded Atomic Energy Fellowship by the
Government of India during 1973-76 for Ph.D work, Post Doctoral Fellowship for research work in Pierre et Marie Curie University, Paris,
France during 1982-1983, and INSA-Royal Society Fellowship in 1999 for research work in UK. He worked as a Visiting Professor in the TU,
Vienna, Austria and also in the University of Kaiserslautern, Germany during 2002-2003. He received Michael Madhusudan Award in 2004 and
Subarna and Sashibhusan Memorial Award, in 2005. Recently he became Fellow (FICDM) of a learned society.
Dr. Shishir Gupta is an Associate professor in the Department of Applied mathematics, Indian School of Mines, Dhanbad. A Gold Medalist
from Ranchi University, he has had a brilliant career. He has more than 21 years of teaching experience at undergraduate and postgraduate levels
in Indian School of Mines, Dhanbad. He possesses experience of guiding students of MPhil and PhD. He has published more than 45 papers in
International/National journals/Proceedings. He has served as reviewer in renowned International/ National books and journals. He has also
carried out several sponsored research projects.
Mr. Abhishek Kumar Singh is a Junior Research Fellow (JRF) in the Department of Applied Mathematics, Indian School of Mines (ISM),
Dhanbad, Jharkhand, India. He is pursuing his PhD under the supervision of Prof. (Dr.) Amares Chattopadhyay in the field of Elastodynamics.
He did his B.Sc. from University of Allahabad, Allahabad in 2005 and then switched to ISM to earn M.Sc. in Maths & Computing (2007) and
MPhil in Applied Mathematics (2008).
Mr. Sanjeev Anand Sahu is a DST-JRF (Project No. DST(39)/2008-09/223/AM) in the Department of Applied Mathematics, Indian School of
Mines (ISM), Dhanbad, Jharkhand, India. He is pursuing his Ph. D. under the supervision of Prof. (Dr.) Amares Chattopadhyay in the field of
Elastodynamics. He did his B.Sc. and M.Sc. from Guru Ghansidas University, Bilaspur in 2003 and 2005 respectively, and then switched to ISM
for Project as well as PhD.
Received November 2009
Accepted December 2009
Final acceptance in revised form December 2009
MultiCraft
International Journal of Engineering, Science and Technology
Vol. 1, No. 1, 2009, pp. 245-253
INTERNATIONAL
JOURNAL OF
ENGINEERING,
SCIENCE AND
TECHNOLOGY
www.ijest-ng.com
2009 MultiCraft Limited. All rights reserved
Effects of superficial gas velocity and fluid property on the hydrodynamic
performance of an airlift column with alcohol solution
V. Sivasubramanian
a
* and B.S. Naveen Prasad
b
a
Department of Chemical Engineering, National Institute of Technology Calicut, Kerala 673601, India
b
Department of Chemical Engineering, Sathyabama University, Chennai, India
E-mail: (siva@nitc.ac.in (V. Sivasubramanian),
*
Corresponding author), Tel: 0495-2286525, 2286544
Abstract
In the present study, the influence of superficial gas velocity and fluid properties on gas holdup and liquid circulation velocity
in a three-phase external loop airlift column using polystyrene (0.0036 m diameter and 1025.55 kg/m
3
density) and nylon-6
(0.0035 m diameter and 1084.24 kg/m
3
density) particles with aqueous solutions of alcohols (isoamyl alcohol and propanol) as
liquids were investigated. The column was constructed using acrylic tube of inner diameter 0.084m and 2.6m in height. The gas
holdup in the riser increased with increase in superficial gas velocity for air-alcohol-solid system. The presence of alcohol
surfactants increased the gas holdup in the riser. It was also found that an increase in the superficial gas velocity increased the
liquid circulation velocity
for air-alcohol-solid system. Correlations were proposed for the prediction of gas holdup and liquid
circulation velocity.
Keywords: External loop airlift bioreactor, three-phase, effect of additives, hydrodynamics
1. Introduction
Airlift reactors have several advantages: simple design concept, high mixing performance, high mass transfer ability, good heat
transfer and low energy consumption. Airlift loop reactors find extensive applications in many areas of chemical engineering,
especially for homogeneous as well as heterogeneous single and multiphase systems due to their simple construction and
operation, directed circulation flow, good mixing and favorable ratio of interfacial area to energy dissipation rate per unit volume,
low investment, operational costs and relatively lower power requirements (Merchuk and Siegel, 1988). Based on their
configurations, airlift reactors can be classified into two groups: internal-loop and external-loop airlift reactors. An internal-loop
airlift reactor contains a vertical baffle or a draft tube by which a loop channel for fluid circulation is formed in the reactor. An
external-loop airlift reactor consists of two vertical tubes (a riser and a downcomer) which are connected by horizontal connections
at the top and bottom. A distractive difference between the two groups is the presence of the horizontal connections (Choi, 2001).
Herskowitz and Merchuk (1986) reported the influence of solid particles on the gas hold and liquid superficial velocity in a novel
three phase fluidized bed reactor. Posarac and Petrovic (1988) found that the minimum fluidization velocity of 3-6 mm glass
spheres was found to increase with solids loading. Kochbeck et al. (1992) reported the hydrodynamics and local parameters in
three-phase-flow in airlift-loop reactors of different scale. Douek et al. (1994) observed that gas holdup and the liquid circulation
velocity in the riser increased with increase in gas superficial velocity. A new method for the measurement of solids holdup in gas-
liquid-solid three-phase systems was proposed by Wenge et al. (1995). Guo et al. (1997) reported the influence of hydrodynamics
and mass transfer studies in a novel external-loop airlift reactor. Freitas et al. (1997) proposed a new sampling device for
measuring solids holdup in a three-phase system. Freitas and Teixeira (1998) reported on solid-phase distribution in an airlift
reactor with an enlarged degassing zone.
Freitas et al. (2000) reported the influence of the diameter of the distributor orifice, superficial gas velocity, solid loading and
solid density on the hydrodynamics of a three-phase external-loop airlift bioreactor. The development and use of simple method
for regime identification and flow characterization in bubble columns and airlift reactors was reported by Vial et al. (2001).
Shimizu et al. (2001) developed the phenomenological simulation model for prediction of gas hold-up and volumetric mass
Sivasubramanian and Prasad / International Journal of Engineering, Science and Technology,
Vol. 1, No. 1, 2009, pp. 245-253
246
transfer coefficients in external-loop airlift reactors. Meng et al. (2002) reported the use of a modified volume expansion (inclined
tube) method for determination of overall gas holdup in an external loop airlift bubble column. Wang et al. (2003) reported the
determination of gas holdup, liquid circulating velocity and mass transfer coefficient in a mini-scale external- loop airlift bubble
column. Blazej et al. (2004) reported the simulation of gas-liquid flow in an airlift reactor using commercially available software
Fluent.
From the literature, it is observed that most of the studies used pure water as liquid while all applications in wastewater treatment
or biochemical reactors use complex aqueous solutions. Studies conducted with regular fluidized beds have shown that the
presence of minor liquid impurities can greatly affect the bed hydrodynamics (Saberian-Broujenni et al., 1984; Nacef et al., 1988;
Song et al., 1989; Prakash, 1991; Del Pozo et al., 1994). However, little attention has been focused on the effect of additives in the
hydrodynamic performance studies in external loop airlift bioreactors. The main objective of the present study is to establish
whether changes in liquid coalescing properties caused by minute concentrations of additives can greatly affect the hydrodynamic
properties of a three-phase external loop airlift column and to investigate the influence of superficial gas velocity and fluid
properties on gas holdup and liquid circulation velocity.
2. Material and methods
The schematic diagram of the external loop airlift column used in the present investigation is shown (Figure 1). The column was
constructed using acrylic tube of inner diameter 0.084 m and 2.6 m in height. The column consisted of four sections namely riser,
downcomer, base and gas-liquid separator. The riser and downcomer were connected with an inclined angle 45
o
at the top and the
bottom section of the column. In order to avoid solids settling, the connection pipes between the riser and downcomer were
inclined. The external loop airlift column had a perforated plate gas sparger with 243 holes of 1mm diameter on a triangular pitch
placed at the base of the column. Air from the compressor was sparged to the column through an air filter, pressure regulator and
calibrated rotameter. Provision was made to insert the dissolved oxygen meter probe above the gas sparger. All runs were made at
room temperature (252C).
The solid particles used are polystyrene of 0.0036m diameter (1025.55 kg/m
3
density) and nylon-6 of 0.0035m diameter
(1084.24 kg/m
3
density). The physical properties of aqueous solutions of alcohols were given in Table 1. The gas holdup in the
riser in three-phase system was determined by manometric technique and the liquid circulation velocity in the downcomer of the
column was determined by tracer technique using 10 mm polystyrene spherical particle of density of 1012 kg/m
3
.
Table 1. Physical properties of aqueous solutions of isoamyl alcohol and propanol
Additive Concentrations
Density,
l
(kg/m
3
)
Viscosity,
l
(mPa.s)
Surface tension,
l
(N/m)
Isoamyl alcohol 20 mg/l 998 1.0 0.072
Isoamyl alcohol 40 mg/l 998 1.0 0.071
Isoamyl alcohol 60 mg/l 998 0.98 0.069
Isoamyl alcohol 80 mg/l 998 0.97 0.069
Isoamyl alcohol 100 mg/l 997 0.97 0.067
Propanol 0.1 wt % 998 1.0 0.072
Propanol 0.2 wt % 998 1.0 0.072
Propanol 0.3 wt % 998 0.98 0.070
Propanol 0.4 wt % 998 0.97 0.069
Propanol 0.5 wt % 998 0.97 0.068
2.1 Measurement of gas holdup in three-phase system
In the three-phase systems, two different types of solid particles Polystyrene and Nylon-6 at various solids loading were used and
the solid loadings were varied from 1000-3000 ml (4.25-12.765% (Vol.)). The gas holdup in the riser in three-phase system was
determined by manometric technique and are related to solid holdup,
s
by
+
+
=
) ( ) (
) / ( ) (
g l s l s
l s l s
gr
z h
(1)
Sivasubramanian and Prasad / International Journal of Engineering, Science and Technology,
Vol. 1, No. 1, 2009, pp. 245-253
247
1
2
3
4
5
6
7
7
7
7
9
10
8
1. Compressor
2. AirFilter
3. Pressure Regulator
4. Rotameter
5. Calm Section
6. Sparger
7. Manometer Tappings
8. Downcomer
9. Gas-liquid Separator
10. Riser
11. DO Probe
11
Figure 1. Schematic diagram of external loop airlift column
Sivasubramanian and Prasad / International Journal of Engineering, Science and Technology,
Vol. 1, No. 1, 2009, pp. 245-253
248
( ) z
h
l s
l
s
(2)
2.2 Measurement of liquid circulation velocity
The liquid circulation velocity in the downcomer of the column was determined by tracer technique using 10 mm polystyrene
spherical particle of density of 1012 kg/m
3
. The time required for the particle to travel between two ports on the wall of the
downcomer was observed to calculate the liquid circulation velocity in the downcomer (U
ld
). Thirty measurements were taken for
each calculation of liquid velocity in the downcomer. The measurement of tracer particle traveling close to the column wall was
excluded from the data (Bello et al., 1984; Kawase, 1990; Kemblowski et al., 1993; Wang et al., 2003).
The liquid circulation velocity in the riser and downcomer were obtained by
c c ld
t x U / =
(3)
d ld r lr
A U A U =
(4)
) / (
d r ld lr
A A U U =
(5)
3. Results and discussion
3.1 Variation of gas holdup in the riser with superficial gas velocity for air-alcohol-solid systems
The variation of gas holdup in the riser,
gr
with superficial gas velocity for air-alcohol-solid systems is shown in Figure 2. The
experiments were carried out with an optimized solids loading of 2000 ml. It was observed that the
gr
increased linearly with
increase in superficial gas velocity. The presence of alcohol surfactants increased the
gr
. This was mainly due to the suppression of
bubble coalescence i.e. number of small bubbles produced in the riser had an insufficient bubble rise velocity to escape from the
liquid. A similar trend was observed by Koide et al. (1985), Nicol and Davidson (1988) and Al-Masry and Dukkan (1997).
3.2 Effect of alcohol concentrations on
gr
The effects of propanol and isoamyl alcohol concentrations on
gr
are shown in Figures 3. The addition of propanol and isoamyl
alcohol was found to increase the
gr
. The bubble diameter and bubble rise-velocity were the functional parameters, which govern
the riser gas holdup for alcohol surfactants. The increase in concentration of surfactants decreased the bubble diameter and rise
velocity, which increased the
gr
. This may be due to the decrease in the surface tension of the liquid phase.
3.3 Variation of liquid circulation velocity in the riser with superficial gas velocity for air-alcohol-solid systems
The variation of liquid circulation velocity, U
lr
with superficial gas velocity for three-phase systems is shown in Figure 4. It was
found that an increase in the superficial gas velocity increased the U
lr
. The addition of alcohol surfactants decreased the U
lr
. It was
also observed that the concentrations of alcohol surfactants did not have any marginal effects on the U
lr
, which was obtained by
Al-Masry and Dukkan.
3.4 Effect of alcohol concentrations on U
lr
The effect of propanol and isoamyl alcohol concentrations on U
lr
is shown in Figures 5 for 2000 ml solids loading. It was observed
that the addition of solids and an increase in alcohol concentration decreased the U
lr
. The alcohol surfactants did not have any
marginal effects on the U
lr
. A same trend was reported by Al-Masry and Dukkan.
Sivasubramanian and Prasad / International Journal of Engineering, Science and Technology,
Vol. 1, No. 1, 2009, pp. 245-253
249
Figure 2. Variation of gas holdup in the riser with superficial gas velocity for (a) Air-propanol-polystyrene system; (b) Air-
propanol-nylon-6-system; (c) Air-IAA-polystyrene system; (d) Air-IAA-nylon-6 system
a b
c d
Sivasubramanian and Prasad / International Journal of Engineering, Science and Technology,
Vol. 1, No. 1, 2009, pp. 245-253
250
Figure 3. Effect of alcohol concentrations on
gr
for (a) Air-propanol-polystyrene system; (b) Air-propanol-nylon-6-system;
(c) Air-IAA-polystyrene system; (d) Air-IAA-nylon-6 system
a b
c d
Sivasubramanian and Prasad / International Journal of Engineering, Science and Technology,
Vol. 1, No. 1, 2009, pp. 245-253
251
Figure 4. Variation of liquid circulation velocity in the riser with superficial gas velocity for air-propanol-polystyrene system
Figure 5. Effect of alcohol concentration on liquid circulation velocity in the riser for (a) air-propanol-polystyrene system;
(b) air-IAA-polystyrene system
3.5 Correlation for gas holdup in the riser and liquid circulation velocity for air-alcohol-solids system
The value of gas holdup in the riser and liquid circulation velocity in the riser for three-phase system have been correlated as
996 . 0 038 . 0 05 . 0 ) 10 16 . 1 984 . 0 (
4
577 . 18
l l
C
sgr gr
s
U
+
= R
2
= 0.93 (6)
997 . 0 0034 . 0 0096 . 0 ) 10 52 . 6 384 . 0 (
5
727 . 2
l l
C
sgr lr ld
s
U U U
+
= = R
2
= 1 (7)
0 U
sgr
0.033
Parity plots between experimental and predicted values of
gr
and U
lr
based on the proposed empirical equations 6 and 7 for air-
alcohol-solids system are shown in Figures 6a and b respectively. The average RMS error for
gr
and U
lr
are 11.6 and 3.8%
respectively. It was found that the proposed equations fitted the experimental data well within 30 and 14% respectively.
a b
Sivasubramanian and Prasad / International Journal of Engineering, Science and Technology,
Vol. 1, No. 1, 2009, pp. 245-253
252
Figure 6. Parity diagram of predicted and experimental values for (a)
gr
estimated from empirical Eqn 6 for air-alcohol-solid
system; (b) U
lr
estimated from empirical Eqn 7 for air-alcohol-solid system.
4. Conclusions
The gas holdup in riser,
gr
increased linearly with increase in superficial gas velocity. The addition of propanol and isoamyl
alcohol was found to increase the
gr
. An increase in the superficial gas velocity increased the liquid circulation velocity in the
riser, U
lr
. It was observed that the addition of solids and an increase in alcohol concentration decreased the U
lr
. The proposed
correlations for gas holdup in the riser and liquid circulation velocity predicted the experimental data well for air-alcohol (propanol
and isoamyl alcohol)-solid system.
Nomenclature
A
d
downcomer cross sectional area, m
2
A
r
riser cross sectional area, m
2
C
s
solids loading, w/v
d diameter, m
t
c
time taken between two peaks, s
U
sgr
superficial gas velocity in the riser, m/s
U
lr
,U
ld
liquid
circulation velocity in the riser and the downcomer, m/s
x
c
distance between the two adjacent peaks, m
IAA iso amyl alcohol
z distance between two ports, m
h manometric reading, m
Greek symbols
g
gas holdup
volume fraction of solids in liquid - solid
two-phase slurry
surface tension, mN/m
viscosity, mPas
density, kg/m
3
Subscripts
d downcomer
g gas
l liquid
r riser
s solid
a b
Sivasubramanian and Prasad / International Journal of Engineering, Science and Technology,
Vol. 1, No. 1, 2009, pp. 245-253
253
References
Al-Masry W.A., and Dukkan A.R. 1997. Chem. Eng. J., Vol. 65, pp. 263-271.
Bello R.A., Robinson C.W. and Moo-Young M. 1984, Can. J. Chem. Eng., Vol. 62, pp. 573-577.
Blazej M., Cartland Glover G.M., Generalis S.C. and Markeos J. 2004. Chem. Eng. Proc., Vol. 43, pp. 137-144.
Choi K.H. 2001. Korean J. Chem. Eng., Vol. 18, No. 2, pp. 240-246.
Del Pozo M., Briens C.L. and Wild G. 1994. Chem. Eng. J., Vol. 55, pp. 1-14.
Douek R.S., Livingston A.G., Johansson A.C., and Hewitt G.F. 1994. Chem. Eng. Sci., Vol. 49, No. 22, pp. 3719-3737.
Freitas C., and Teixeira J.A. .1998. Biotechnol. Techn., Vol. 12, No. 3, pp. 219-224.
Freitas C., Fialova M., Zahradnik J., and Teixeira J.A. .2000. Chem. Eng. Sci., Vol. 55, pp. 4961-4972.
Guo Y.X., Rathor M.N., and Ti H.C. .1997. Chem. Eng. J., Vol. 67, pp. 205-214.
Herskowitz M., and Merchuk, C.J. .1986. Can. J. Chem. Eng., Vol. 64, pp. 57-61.
Kawase Y. 1990. Biotechnol. Bioeng., Vol. 35, pp. 540-546.
Kemblowski Z., Przywarski J. and Diab A. 1993. Chem. Eng. Sci., Vol. 48, No. 23, pp. 4023-4035.
Kochbeck B., Lindert M., and Hempel D.C. 1992. Chem. Eng. Sci., Vol. 47, No. 13/14, pp. 3443-3450.
Koide K., Horibe K., Kawabata H., and Ito S. 1985. J. Chem. Eng. Japan, Vol. 18, No. 3, pp. 248-254.
Meng A.X., Hill G.A. and Dalai A.K. 2002. Can. J. Chem. Eng., Vol. 80, pp. 194-199.
Merchuk J. C. and Siegel, H. 1988. Chem. Teeh. Biotechnol., Vol. 41, p. 105.
Nacef S., Wild G., Laurent A. and Kim S.D. 1988. Entropie, Vol. 143-144, pp. 83-101.
Nicol R.S., and Davidson J.F. 1988. Chem. Eng. Res. Dev., Vol. 66, pp. 159-164.
Posarac D., and Petrovic D. 1988. Chem. Eng. Sci., Vol. 43, No. 5, pp. 1161-1165.
Prakash A. 1991. Enhancement of gas-liquid and particle-liquid mass transfer in bubble columns and three-phase fluidized beds.
Ph.D. Thesis. The University of Western Ontario, London, Canada.
Saberian-Broujenni M., Wild G., Charpentier, J.C., Fortin, Y., Enzen, J.P. and Patoux, R., 1984. Entropie, Vol. 120, pp. 30-44.
Shimizu K., Takada S., Takahadhi T. and Kawase Y. 2001. Chem. Eng. J., Vol. 84, pp. 599-603.
Song, G.H., Bavarian, F. and Fan, L.S. 1989. Can. J. Chem. Eng., Vol. 67, pp. 256-275.
Vial C., Poncin S., Wild G. and Midoux N. 2001. Chem. Eng. Proc., Vol. 40, pp. 130-151.
Wang S., Arimatsu Y., Koumatsu K., Furumoto K., Yashimoto M., Fukunaga K. and Nakao K. 2003. Chem. Eng. Sci., Vol. 58, pp.
3353-3360.
Wenge F.W., Chisti. M.Y., and Moo-Young M. 1995. Ind. Eng. Chem. Res., Vol. 34, pp. 928-935.
Biographical notes
Dr. V.Sivasubramanian is Associate Professor in Chemical Engineering at National Institute of Technology Calicut, India. His research has
focused on Wastewater Treatment using Inverse Fluidized Bed Reactor and Treatment of Effluent using Biomembrane Techniques. He is the
author of over 20 articles which have appeared in journals such as Asia-Pacific J Chem Eng., J Chem Technol Biotechnol., J Chem Eng Japan,
IETECH J Mech Design, J Indian Chem Engr and Rubber India. He is also author of a book titled, Process Economics and Industrial
Management. He is the recipient of the award and first prize for an oral presentation from Prof. M. M. Sharma in CHEMCON 2004, Mumbai,
India. He is Editor in Intern J Eng., Sci Technol, Nigeria and Medwell Journals. He is also reviewer in Clean, Elsevier, USA, Asia-Pacific J
Chem Eng, Wiley Interscience, Australia, Chem Product and Process Modeling, The Berkeley Electronic Press, USA, Chem Eng Sci, Adv
Powder Technol, Elsevier, Desalination Journal, Elsevier, USA, Intern J Biotechnol., Inderscience Publishers, Intern J Chem Reactor Eng, The
Berkeley Electronic Press, USA, Intern J Eng, Science and Technology, Nigeria and Indian J Sci and Technol.
Mr. B.S. Naveen Prasad is Faculty in Chemical Engineering Department at Sathyabama University, Chennai, India. He did his graduation in
Chemical Engineering at Kongu Engineering College, Erode. Bharathiar University, Coimbatore. He completed his M.S. (By Research)
programme in the Department of Chemical Engineering, A. C. College of Technology, Anna University, Chennai. He registered for his Doctoral
programme in Biotechnology Department at Sathyabama University, Chennai.
MultiCraft
International Journal of Engineering, Science and Technology
Vol. 1, No. 1, 2009, pp. 254-271
INTERNATIONAL
JOURNAL OF
ENGINEERING,
SCIENCE AND
TECHNOLOGY
www.ijest-ng.com
2009 MultiCraft Limited. All rights reserved
A combined approach of complex eigenvalue analysis and design of
experiments (DOE) to study disc brake squeal
M. Nouby
1*
, D. Mathivanan
2*
, K. Srinivasan
1
1
AU/FRG Institute for CAD/CAM, Anna University, Chennai-600025, India
2
Director of CAE Infotech, Chennai-600020, India
E-mails:(noubyluxor@gmail.com (M. Nouby),mathvan@yahoo.com (D. Mathivanan);
*
Corresponding authors)
Abstract
This paper proposes an approach to investigate the influencing factors of the brake pad on the disc brake squeal by integrating
finite element simulations with statistical regression techniques. Complex eigenvalue analysis (CEA) has been widely used to
predict unstable frequencies in brake systems models. The finite element model is correlated with experimental modal test. The
input-output relationship between the brake squeal and the brake pad geometry is constructed for possible prediction of the
squeal using various geometrical configurations of the disc brake. Influences of the various factors namely; Youngs modulus of
back plate, back plate thickness, chamfer, distance between two slots, slot width and angle of slot are investigated using design
of experiments (DOE) technique. A mathematical prediction model has been developed based on the most influencing factors
and the validation simulation experiments proved its adequacy. The predicted results show that brake squeal propensity can be
reduced by increasing Youngs modulus of the back plate and modifying the shape of friction material by adding chamfer on
both sides of friction material and by introducing slot configurations. The combined approach of modeling brake squeal using
CEA and DOE is found to be statistically adequate through verification trials. This combined approach will be useful in the
design stage of the disc brake.
Keywords: Disc brake squeal, finite element analysis, experimental modal analysis, design of experiments
1. Introduction
Brake squeal is a noise problem caused by vibrations induced by friction forces that can induce a dynamic instability (Akay,
2002). During the braking operation, the friction between the pad and the disc can induce a dynamic instability in the system.
Usually brake squeal occurs in the frequency range between 1 and 20 kHz. Squeal is a complex phenomenon, partly because of its
strong dependence on many parameters and, partly, because of the mechanical interactions in the brake system. The mechanical
interactions are considered to be very complicated because of nonlinear contact effects at the friction interface. The occurrence of
squeal is intermittent or even random. Under certain conditions, even when the vehicle is brand new, it often generates squeal
noise, which has been extensively studied with the goal of eliminating the noise. However, mechanistic details of squeal noise are
not yet fully understood (Joe et al., 2008). Several theories have been formulated to explain the mechanisms of brake squeal, and
numerous studies have been made with varied success to apply them to the dynamics of disc brakes (Kinkaid et al., 2003).
The reason for the onset of instability has been attributed to different reasons. Some of the major reasons are the change of the
friction characteristic with the speed of the contact points (Ibrahim, 1994; Ouyang et al., 1998; Shin et al., 2002) the change of the
relative orientation of the disk and the friction pads leading to a modification of the friction force (Millner, 1978), and a flutter
instability which is found even with a constant friction coefficient (Chowdhary et al., 2001; Chakrabotry et al., 2002; Von Wagner
et al., 2003; Von Wagner et al., 2004). In fact, recent literature reviews (Kinkaid et al., 2003; Papinniemi et al., 2002) have
reported on the complexity and lack of understanding of the brake squeal problem.
Though much work was done on the issue of squeal, it requires continuous study and investigation to refine the prediction
accuracy of finite element models of brake assemblies to give brake design engineers appropriate tools to design quiet brakes.
There are two main categories of numerical methods that are used to study this problem: (1) transient dynamic analysis (Hu et al.,
Nouby et al. / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 254-271
255
1999; AbuBakar et al., 2006) and (2) complex eigenvalue analysis. Currently, the complex eigenvalue method is preferred and
widely used (Liles, 1989; Lee et al., 1998; Blaschke et al., 2000; Bajer et al., 2003; AbuBakar et al., 2006; Liu et al., 2007; Mario
et al., 2008; Dai et al., 2008) in predicting the squeal propensity of the brake system including damping and contact due to the
quickness with which it can be analysed and its usefulness in providing design guidance by analysing with different operating
parameters virtually.
Many researchers in their studies on the dynamics of brake system tried to reduce squeal by changing the factors associated with
the brake squeal. For example (Liles, 1989) found that shorter pads, damping, softer disc and stiffer back plate could reduce squeal
whilst in contrast, higher friction coefficient and wear of the friction material were prone to squeal. (Lee et al., 1998) reported that
reducing back plate thickness led to less uniform of contact pressure distributions and consequently increasing the squeal
propensity. (Hu et al., 1999) based on the DOE analysis found that the optimal design was the one that used the original finger
length, the vertical slot, the chamfer pad, the 28mm thickness of disc, and the 10mm thickness of friction material. (Brooks et al.,
1993) found that by shifting the pistons away from the leading edge of the pads the system could destabilise. They also reported
that the predicted unstable system was due to the coupling of translational and rotational modes of the disc particularly at high
values of pad stiffness. From the sensitivity studies, they suggested that the effective half length of the pad, the piston masses, the
effective mass, inertia and grounding stiffness of the disc and the second circuit actuation stiffness also have potential on the disc
brake instability. (Shin et al., 2002) have shown that the damping of the pad and the disc were important in reducing instability.
Their analysis also has shown and confirmed that increasing damping of either the disc or the pad alone could potentially
destabilise the system. (Liu et al., 2007) found that the squeal can be reduced by decreasing the friction coefficient, increasing the
stiffness of the disc, using damping material on the back of the pads and modifying the shape of the brake pads. (Dai et al., 2008)
have shown that the design of the pads with a radial chamfer possesses the least number of unstable modes, which implies lesser
tendency towards squeal.
In the present study, an investigation of disc brake squeal is done by performing complex eigenvalue analysis using finite
element software ABAQUS /standard. A positive real part of a complex eigenvalue is being seen as an indication of instability.
Though the FE simulations can provide guidance, it will rather be trial-and-error approach to arrive at an optimal configuration and
also one may need to run many number of computationally intensive analyses to formulate the input-out relationships for
possible prediction. Hence, in the present investigation, a novel approach is proposed by integrating the complex eigenvalue FE
analyses with structured DOE. The proposed approach is aimed towards prediction of optimal pad design through the various
factors of the brake pad geometrical construction.
The paper is organised as follows, it presents a detailed literature survey in this field in the recent period. From the literature
survey the main objectives were formed. Methodology to develop FE model of the disc brake is presented and it was subsequently
validated using experimental modal analysis. The CEA approach is presented in order to predict brake squeal. Then DOE approach
for soft computing is presented. Also a methodology to test the adequacy of the developed statistical model is discussed.
2. Finite element model and component correlation
A disc brake system consists of a disc that rotates about the axis of a wheel, a caliperpiston assembly where the piston slides
inside the caliper that is mounted to the vehicle suspension system, and a pair of brake pads. When hydraulic pressure is applied,
the piston is pushed forward to press the inner pad against the disc and simultaneously the outer pad is pressed by the caliper
against the disc. Figure 1 (a) shows the finite element model of the car front brake under consideration, built using the ABAQUS
finite element software package. The brake model used in this study is a simplified model consisting of the two main components
contributing to squeal: the disc and the pad (Figure 1(b)).
(a) (b)
Figure 1. Finite element models of a (a) realistic (b) simplified disc brake model
Nouby et al. / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 254-271
256
A simplified model was used in this study for the following reasons:
1. For brake squeal analysis, the most important source of nonlinearity is the frictional sliding contact between the disc and
the pads.
2. The simulation includes geometry simplifications to reduce CPU time, allowing far more configurations to be computed.
The disc is made of cast iron. The pair of brake pads, which consist of friction material and back plates, are pressed against the
disc in order to generate a friction torque to slow the disc rotation. The friction material is made of an organic friction material and
the back plates are made of steel. The FE mesh is generated using 19,000 solid elements. The friction contact interactions are
defined between both sides of the disc and the friction material of the pads. A constant friction coefficient and a constant angular
velocity of the disc are used for simulation purposes. Figure 2 presents the constraints and loadings for the pads and disc assembly.
The disc is completely fixed at the four counter-bolt holes and the ears of the pads are constrained to allow only axial movements.
The caliperpiston assembly is not defined in the simplified model of the disc brake system, hence the hydraulic pressure is
directly applied to the back plates at the contact regions between the inner pad and the piston and between the outer pad and the
caliper, and it is assumed that an equal magnitude of force acts on each pad.
Figure 2. Constraints and loading of the simplified brake system
For the purpose of validation, the main brake components, Frequency Response Functions (FRFs) were measured at free-free
boundary conditions by exciting each component with a small impact hammer with sensitivity of 10mV/N and a hard tip. The
acceleration response was measured with a light small accelerometer with sensitivity of 10mV/g through Dynamic Signal Analyzer
type DEWE-41-T-DSA. FRF measurements were recorded for each component using SISO configurations. Then, the FRFs were
processed using DEWE FRF software in order to identify the modal parameters, namely; resonance frequencies, modal shapes and
damping values. Figure 3 shows the experimental modal test components.
Figure 3. Experimental Modal Analysis Components
The frequencies measured on the disc and calculated by the simulated model for modes with free-free boundary conditions are
shown in table 1. It can be observed that the measured and simulated frequencies are in good agreement. Figure 4 shows the mode
shapes of rotor with nodal diameters. In a similar way, the parameters for the pads are estimated based on the measured data
indicated in Table 2. The measured and simulated frequencies are in good agreement. Figure 5 shows the mode shapes of the pad.
Nouby et al. / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 254-271
257
Table 1. Modal results of the rotor at free-free boundary conditions
Mode shape
Experimental
Frequency (Hz)
FEA
Frequency(Hz)
Differences (%)
2
nd
bending 1220 1303 6.8
3
rd
bending 2551 2636 3.3
4
th
bending 4003 4108 2.6
5
th
bending 5774 5591 -3.1
6
th
bending 7873 7790 -1
7
th
bending 9008 9209 2.2
Table 2. Modal results of the pad at free-free boundary conditions
Mode shape
Experimental
Frequency (Hz)
FEA
Frequency(Hz)
Differences (%)
1
st
bending 3051 3231 5.8
2
nd
bending 8459 8381 -1
2
nd
Nodal Diameter Mode (1303Hz) 3
rd
Nodal Diameter Mode (2636 Hz)
4
th
Nodal Diameter Mode (4108 Hz) 5
th
Nodal Diameter Mode (5591 Hz)
Figure 4. Mode shapes of the rotor at free-free boundary conditions
Nouby et al. / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 254-271
258
6
th
Nodal Diameter Mode (7790 Hz) 7
th
Nodal Diameter Mode (9209 Hz)
Figure 4 (contd). Mode shapes of the rotor at free-free boundary conditions
1
st
Bending Mode (3231 Hz) 2
nd
Bending Mode (8381 Hz)
Figure 5. Mode shapes of the pad at free-free boundary conditions.
3. Complex eigenvalue analysis
The complex eigenvalue analysis (CEA) has been widely used by researchers. It deals with computation of system eigenvalues
which are prove to be complex valuated functions in the general case because friction causes the stiffness matrix to be asymmetric.
The real and imaginary parts of the complex eigenvalues are, respectively, responsible for the stability and for the frequency of the
corresponding modes. This method was first used on lumped models (Kinkaid et al., 2003; Ibrahim, 1994). Then, improvements in
computer systems have made it possible to perform analyses on finite element (FE) models (Liles, 1989; Lee et al., 1998; Blaschke
et al., 2000; Bajer et al., 2003; AbuBakar et al., 2006; Liu et al., 2007; Mario et al., 2008; Dai et al., 2008).
In order to perform the complex eigenvalue analysis using ABAQUS (Bajer et al, 2003), four main steps are required as follows:
(1) nonlinear static analysis for the application of brake pressure; (2) nonlinear static analysis to impose a rotational velocity on the
disc; (3) normal mode analysis to extract the natural frequency to find the projection subspace; and (4) complex eigenvalue
analysis to incorporate the effect of friction coupling.
The governing equation of the system is:
0 = + + Ku u C u M (1)
Where M, C and K are respectively the mass, damping and stiffness matrices, and u is the displacement vector. Because of friction,
the stiffness matrix has specific properties:
F S
K K K + = (2)
Where
S
K is the structural stiffness matrix,
F
K the asymmetrical friction induced stiffness matrix and the friction coefficient.
The governing equation can be rewritten as:
0 ) (
2
= + + K C M (3)
Nouby et al. / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 254-271
259
Where, is the eigenvalue and is the corresponding eigenvector. Both eigenvalues and eigenvectors may be complex. In order
to solve the complex eigenproblem, this system is symmetrized by ignoring the damping matrix C and the asymmetric
contributions to the stiffness matrix K. Then this symmetric eigenvalue problem is solved to find the projection subspace. The N
eigenvectors obtained from the symmetric eigenvalue problem are expressed in a matrix as ] ,........ , [
2 1 N
. Next, the original
matrices are projected onto the subspace of N eigenvectors
], ,......... , [ ] ,......... , [
2 1 2 1
*
N
T
N
M M = (4a)
], ,......... , [ ] ,......... , [
2 1 2 1
*
N
T
N
C C = (4b)
and
], ,......... , [ ] ,......... , [
2 1 2 1
*
N
T
N
K K = (4c)
Then the projected complex eigenproblem becomes
0 ) (
* * * * 2
= + + K C M (5)
Finally, the complex eigenvectors of the original system can be obtained by
k
N
k *
2 1
] ,......... , [ = (6)
Detailed description of the formulation and the algorithm are presented by Sinou et al. (2003). The eigenvalues and the
eigenvectors of Eq. (3) may be complex, consisting of both a real and imaginary part. For under damped systems the eigenvalues
always occur in complex conjugate pairs. For a particular mode the eigenvalue pair is
i i i
i =
2 , 1
(7)
Where,
i
and
i
are the damping coefficient (the real part) and damped natural frequency (the imaginary part) describing
damped sinusoidal motion. The motion for each mode can be described in terms of the complex conjugate eigenvalue and
eigenvector. A positive damping coefficient causes the amplitude of oscillations to increase with time. Therefore the system is not
stable when the damping coefficient is positive. By examining the real part of the system eigenvalues the modes that are unstable
and likely to produce squeal are revealed. An extra term, damping ratio, is defined as / 2 . If the damping ratio is negative,
the system becomes unstable, and vice versa.
3.1. Complex eigenvalue analysis (CEA) results
Since friction is the main cause of instability, which causes the stiffness matrix in Eq. (3) to be asymmetric, complex eigenvalue
analysis has been undertaken to assess the brake stability as the friction coefficient values. It was observed that high values for this
parameter tend to facilitate two modes merging to form an unstable complex mode. In addition, an increase in the friction
coefficient leads to an increase in the unstable frequency. Figure 6 shows the results of a complex eigenvalue analysis with
variation of the friction coefficient () between 0.2 and 0.6. As predicted in the complex eigenvalue analysis, as the friction
coefficient further increases, real parts of eigenvalues, the values that can be used to gauge the degree of instability of a complex
mode, increase further, as well, and more unstable modes may emerge. This is because the higher coefficient of friction causes the
variable frictional forces to be higher resulting in the tendency to excite greater number of unstable modes. In the past, a friction
coefficient of 0.35 was typical. However, brake compounds today possess coefficient of friction that is 0.45 or higher, which
increases the likelihood of squeal. This poses a greater challenge for brake designer to develop a quiet brake system.
In an earlier work (Nouby et al., 2009), an attempt was made using parametric study to reduce squeal at 12 kHz. Based on the
earlier study, a decision was taken to understand the effects of influencing variables of squeal at 6.2 kHz using DOE.
Nouby et al. / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 254-271
260
Figure 6. Results of CEA with variation of , and unstable mode at interested Frequency (6.2 kHz)
4. Methodology
Human made products or processes can be treated like a system, if it produces a set of responses for a given set of inputs. Disc
brake system can also be treated like a system as shown in Figure 7. Some systems like disc brake system produce unwanted
outputs namely squeal for a set of inputs parameters. The present study was aimed at establishing the input-output relationships for
prediction of brake squeal.
Disc brake system has numerous variables. In order to arrive at the most influential variables and its effects a two phase
strategies were proposed. In the first phase, initial screening with various variables was taken up. Fractional factorial design (FFD)
of experiments was conducted to identify the most influential variables. Subsequently, in the second phase, central composite
design (CCD) based Response surface methodology (RSM) was deployed to develop a non-linear model for prediction of disk
brake squeal.
Figure 7. Disc brake squeal system
4.1 First phase: screening FFD of experiments
If any experiment involves the study of the effects of two or more factors, then factorial designs are more efficient than one-
factor-at-a-time experiments. Furthermore, a factorial design is necessary when interactions may be present to avoid misleading
Nouby et al. / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 254-271
261
conclusions (Montgomery, 2005). Hence, in the first phase, an L
8
FFD array was used for screening. Parameters used for the initial
screening and its levels are indicated in Table 3. Detailed discussions on the selection of parameters are included in section 4.3.
Table 3. Initial screening parameters and their levels
Factor Level
Coded Uncoded Units Low High
A Back plate Youngs modulus GPa 190 230
B Back plate thickness mm 4 6
C Friction material Chamfer mm 5 9
D Slot width mm 1 3
E Distance between slots mm 30 50
F Slot Angle Deg. 5 15
4.2 Second phase: RSM -CCD of experiments
CCD is one of the most important experimental designs used for optimizing parameters. CCD is nothing but 2
k
factorial
design augmented with centre points and axial points. CCD is far more efficient than running 3
k
factorial design with quantitative
factors (Montgomery, 2005). Hence CCD approach was selected for the present study. In light of the first phase, four variables out
of the six variables were selected for the second phase.
The CCD of experiments for four factors consists of the following:
1. Full factorial design for the four factors (2
4
)
2. 1 centre point runs
3. 8 Axial points with value 2. The value is nothing but the distance between centre point and some additional
axial points to fit a non-linear curve.
A second order polynomial equation was developed to study the effects of the variables on the negative damping ratio. The
equation is of the general form
= < =
+ + + =
k
i
i ii j i ij
j i
k
i
X A X X A AiXi A Y
1
2
1
0
(8)
The above equation contains linear terms, squared terms and interaction terms. Equation (8) can be rewritten in matrix form as
CX X b X A Y
T T
+ + =
0
(9)
Where X
T
= (X
1,
X
2,
. . . , X
k
), b = (A
1
, A
2,
. . . , A
k
) and C is the k x k symmetric matrix (3)
C =
kk k k
k
k
A A A
A A A
A A A
2 1
2 22 12
1 12 11
2
1
2
1
. . . . .
. . . . .
2
1
. .
2
1
2
1
. .
2
1
Equation (9) will be differentiated with respect to X and equated to zero in order to obtain different coefficients of the second-
order polynomial equation. Minitab software was used for calculation of coefficients and for analysis.
4.3 Control variables/factors
This work was directed to study and possibly prevent certain unstable modes which causes squeal at frequency 6.2 kHz. Disc
brake squeal is dependent upon numerous factors based on geometry, material characteristics and operating conditions. The factors
can be broadly classified into:
1. Fixed factors
2. Variable factors
Nouby et al. / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 254-271
262
4.3.1 Setting up of fixed processing factors
Squeal in brake system is relatively complex. Inputs (both fixed and variable factors) affect the output. The disk brake system in
abstract form is shown in Figure 7. For the purpose of the present work, factors like the hydraulic pressure, rotational velocity of
the disc, friction coefficient at the contact interactions between the pads and the disc and the stiffness of the disc and friction
material property of brake pads are fixed at specific values.
4.3.2 Selection of controllable processing factors
Based on the literature survey and discussions with service engineers, six controllable factors were taken up for initial screening.
They are listed in the Table 3.
4.3.2.1 Variation of Youngs modulus of back plate
The Youngs modulus is a critical variable and various researchers have attempted (Liu et al., 2007; Nouby et al., 2009)
variation in the Youngs modulus value. In this study, this factor was varied between 190 and 230 GPa.
4.3.2.2 Variation of back plates thickness.
Commonly used back plate thickness is 5mm for the type of the brake system under study. In order to understand the effects and
arrive at back plate thickness, thickness was varied between 4 and 6mm.
4.3.2.3 Variation of the chamfer size
The original pads for this commercial brake did not contain any chamfer. During the literature survey, this parameter was found
to contribute to a certain extent to the squeal (Dai et al., 2008; Dunlap et al., 1999). Hence an angle of 45 was provided on both
sides of the friction material and the size of chamfer was varied between 5 mm and 9mm.
4.3.2.4 Slots configurations on the pad
In this study, a double slot configuration was attempted Figure 8. This configuration has three variables. They are the slot width,
slot angle or orientation and the distance between the two slots. The slot width was varied between 1mm and 3 mm. the slot
angle/orientation was altered between 5 and 15 degrees, as shown in Figure 8, and distance between the slots was varied between
30 and 50 mm.
Figure 8. Various factors of the brake pad geometry
5. Experimental results and analysis
This section discusses about the three phases of experiments, its results, developed mathematical models of the system and its
adequacy.
5.1 First phase - initial screening FFD experiments
A total number of eight trials were conducted and the negative damping ratio was measured for each case. The recorded negative
damping ratio along with FFD array is tabulated in Table 4. Pareto chart of the effects is plotted to determine the magnitude and
the importance of the effects Figure 9. It displays the absolute values of the effects of each factor.
Main effects plot shown in Figure 10 also confirms the contributions of the factors considered.
Nouby et al. / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 254-271
263
Table 4. L
8
Initial FFD array and measured negative damping ratio (NDR)
Runs A B C D E F NDR
1 190 4 5 3 50 15 0.001
2 230 4 5 1 30 15 0.53
3 190 6 5 1 50 5 0.52
4 230 6 5 3 30 5 1.11
5 190 4 9 3 30 5 4.34
6 230 4 9 1 50 5 0.01
7 190 6 9 1 30 15 2.66
8 230 6 9 3 50 15 0.03
5.2 Second phase RSM -CCD of experiments
In light of the screening experiments, a decision was taken to study the effects of the top four factors, namely; Youngs
modulus of the pad back plate, chamfer, and slot angle and distance between two slots. The variables and their levels are listed in
Table 5.
Table 5. Coded levels of variable and actual values for CCD
Factor Level
Coded Uncoded Units Low High
A Back plate Youngs modulus GPa 190 230
C Friction material Chamfer mm 5 9
E Distance between slots mm 30 50
F Angle of slots Deg. 5 15
Figure 9. Pareto chart of effects of initial screening.
Nouby et al. / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 254-271
264
Figure 10. The main effects of factors
A total number of twenty five trials were conducted and a set of data was collected as per the structure of CCD of experiments.
Table 6 lists the experimental results.
Table 6. Experimental results of CCD
Runs A C E F NDR
1 190 5 30 5 3.53
2 230 5 30 5 0.57
3 190 9 30 5 3.55
4 230 9 30 5 0.09
5 190 5 50 5 1.89
6 230 5 50 5 0.12
7 190 9 50 5 0.12
8 230 9 50 5 0.01
9 190 5 30 15 3.21
10 230 5 30 15 0.29
11 190 9 30 15 2.74
12 230 9 30 15 0.01
13 190 5 50 15 0.37
14 230 5 50 15 0.17
15 190 9 50 15 1.10
16 230 9 50 15 0.19
17 170 7 40 10 3.90
18 250 7 40 10 0.01
19 210 3 40 10 0.84
20 210 11 40 10 0.10
21 210 7 20 10 2.67
22 210 7 60 10 0.05
23 210 7 40 0 1.52
24 210 7 40 20 0.01
25 210 7 40 10 0.38
Nouby et al. / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 254-271
265
A significance test was conducted to examine the effects of different process parameters and their interaction terms on the
negative damping ratio. Table 7 shows the estimated coefficients for the negative damping ratio.
Table 7. Estimated regression coefficients for the negative damping ratio
Term Coeff SE Coeff T P
Constant 0.380000 0.28501 1.3330 0.2090
A -0.951667 0.08228 -11.5670 0.0000
C -0.159167 0.08228 -1.9350 0.0790
E -0.635833 0.08228 -7.7280 0.0000
F -0.200833 0.08228 -2.4410 0.0330
A*A 0.391250 0.09648 4.0550 0.0020
C*C 0.020000 0.09648 0.2070 0.8400
E*E 0.242500 0.09648 2.5140 0.0290
F*F 0.093750 0.09648 0.9720 0.3520
A*C 0.040000 0.10077 0.3970 0.6990
A*E 0.567500 0.10077 5.6320 0.0000
A*F 0.096250 0.10077 0.9550 0.3600
C*E 0.005000 0.10077 0.0500 0.9610
C*F 0.146250 0.10077 1.4510 0.1750
E*F 0.073750 0.10077 0.7320 0.4800
Different terms used in the Table 6 are as follows. The term coeff stands for coefficients used in the regression equation for
representing relationships between the negative damping ratio and the various factors. The term SE coeff stands for the standard
error for the estimated coefficients, which measures the precision of the estimates. The smaller values of standard error indicate
that better precision of coefficient estimates. The term T values are calculated as the ratio of corresponding value under
coefficient and standard error. The T value of the independent variable can be used to test whether the predictor significantly
predicts the response. The P value determines the appropriateness of rejecting the null hypothesis in a hypothesis test. The P-
values range from 0 to 1. The smaller the P-value, the smaller the probability that rejecting the null hypothesis is a mistake. The P-
values less than 0.05 indicate that the corresponding model terms are significant. In this present study, A, E, F A
2
, E
2
and AE are
significant model terms to the response damping ratio. It can be observed that, though the term F is significant its square term is
found to be insignificant. It indicates that, Fs relationship could be linear in nature. Figure 11(a) to 11(f) show the 3D surface
plots for the response negative damping ratio (Y).
The following observations can be made from the surface plots:
1. From the surface plot, Figure 11(a) it is found that increased Youngs modulus (A) in combination with reduced angle of
slot (F) decrease negative damping ratio.
2. It is observed Figure 11(b) that, as the distance between two slots are closer with increased Youngs modulus reduces the
negative damping ratio. Surface plot of A and E interaction confirms its significance on the output.
3. It is observed Figure 11(c) that, as the chamfer increases the negative damping ratio reduces on the other hand as the
angle in combination with increasing chamfer causes the negative damping ratio to behave non-linearly. It is also noted
from the surface plot the near significance of C and F interaction.
4. From Figure 11(d) it can be observed that increased chamfer value does not have significant interaction with increase in
Youngs modulus.
5. It is observed from the Figure 11(e) that, increase in the distance between slots reduces the negative damping ratio and the
increase in the chamfer value does not have considerable effect in C and E interaction.
6. From the Figure 11(f) it is observed that increase in the distance between slots reduces the negative damping ratio and the
change in the angle does not have much influence in E and F interaction.
Nouby et al. / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 254-271
266
a. Surface plot of NDR (Y) vs. A,F
b. Surface plot of NDR (Y) vs. A,E
c. Surface plot of NDR (Y) vs. C,F
d. Surface plot of NDR (Y) vs. A,C
e. Surface plot of NDR (Y) vs. C,E
f. Surface plot of NDR (Y) vs. E,F
Figure 11. 3D Surface plots for the response negative damping ratio (Y)
From the above observations, it may be concluded that squeal propensity decreases with higher back plate Youngs modulus,
higher distance between two slots and higher chamfer in both sides of friction material. ANOVA was performed to test the
significance of the factors (Refer Table 8).
Nouby et al. / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 254-271
267
Table 8. ANOVA for negative damping ratio
Different terms used in the Table-8 are as follows. The term DF means degrees of freedom. The DF refers to the number of
terms that will contribute to the error prediction. The term Seq SS represents the sum of squares for each term, which measures
the variability in the data contributed by the term. The term Adj SS means adjusted sum of squares after removing insignificant
terms from the model.
The Model F-value of 18.64 implies the model is significant. There is only a 0.01% chance that a "Model F-Value" this large
could occur due to noise. Values of "Prob > F" less than 0.0500 indicate model terms are significant. In this case A, C, F, AE, A
2
,
and E
2
are significant model terms. Values greater than 0.1000 indicate the model terms are not significant.
Figure 12 shows the normal probability plot of residuals. It shows that there is no abnormality in the methodology adopted (R
2
=
0.9596). The R
2
analysis is tabulated in Table 9. The "Pred R-Squared" of 0.7671 is in reasonable agreement with the "Adj R-
Squared" of 0.9081. "Adeq Precision" measures the signal to noise ratio. Models ratio of 13.476 indicates an adequate signal.
The statistical analysis shows that, the developed non-linear model based on central composite design is statistically adequate and
can be used to navigate the design space.
0.50 0.25 0.00 - 0.25 - 0.50 - 0.75
99
95
90
80
70
60
50
40
30
20
10
5
1
Residual
P
e
r
c
e
n
t
Nor mal Pr obabi l i t y Pl ot
( r esponse is Y)
Figure 12. Normal probability plot for residuals
Table 9. R
2
analysis
Std. Dev. 0.40
Mean 1.07
C.V. % 37.67
PRESS 10.29
R-Squared 0.9596
Adj R-Squared 0.9081
Pred R-Squared 0.7671
Adeq Precision 13.476
Source DF Seq SS Adj SS Adj MS F P
Regression 14 42.40710 42.40710 3.02908 18.64 0.000
Linear 4 33.01490 33.01490 8.25373 50.8 0.000
Square 4 3.63580 3.63580 0.90895 5.59 0.010
Interaction 6 5.75640 5.75640 0.95940 5.91 0.006
Residual Error 11 1.78710 1.78710 0.16246
Lack-of-Fit 10 1.78710 1.78710 0.17871
Pure Error 1 0.00000 0.00000 0.00000
Total 25 44.19420
Nouby et al. / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 254-271
268
The response equation for negative damping ratio in coded form is given under:
Negative damping ratio = - [0.38 -0.951666667A -0.159166667 C -0.635833333 E -0.200833333 F +0.04 A C
+0.5675 A E +0.09625 A F +0.005 C E +0.14625 C F +0.07375 E F +0.39125 A^2 +0.02 C^2 +0.2425
E^2 +0.09375 F^2]
5.3 Testing of model
The model was statistically validated as explained above. Performance of the developed model was also tested on randomly
generated 22 test runs. Table 10 gives the factor settings, predicted responses, measured negative damping ratio and percentage
deviation for each run.
Table 10. Comparison between predicted versus actual simulation
Test A C E F Actual Predicted % Deviation
1 200 6 45 7 0.750 0.853 13.69
2 205 8 35 12 1.010 0.951 5.89
3 210 8 45 9 0.065 0.071 9.68
4 200 7 45 11 0.550 0.516 6.22
5 195 8 40 7 1.210 1.378 13.88
6 195 7 40 10 1.250 1.314 5.11
7 220 6 40 10 0.068 0.077 12.59
8 215 8 40 7 0.170 0.193 13.47
9 225 7 35 7 0.170 0.185 8.89
10 205 6 35 12 1.080 1.064 1.51
11 205 6 40 7 0.940 0.945 0.48
12 195 8 35 11 1.660 1.771 6.67
13 200 6 35 7 1.750 1.819 3.94
14 220 6 35 10 0.291 0.314 8.07
15 200 8 45 10 0.442 0.471 6.52
16 215 6 45 10 0.060 0.059 1.82
17 195 6 45 8 1.140 1.081 5.20
18 210 7 45 8 0.213 0.203 4.56
19 195 6 40 13 1.223 1.239 1.35
20 215 8 40 13 0.071 0.069 3.50
21 195 7 35 9 1.964 1.971 0.35
22 195 8 45 13 0.670 0.691 3.18
% Deviation =
100 value] actual value)/ predicted - value [(actual
Figure 13 shows the graphical presentation of predicted versus actual simulation. The best fit line plot (Figure 14) of the 22
points was found to be close to the ideal Y = X line. Predicted responses show good agreement with actual results. Figure 15
shows the percentage deviation plot. The average absolute percentage deviation was found to be 6.2. Actual results varied between
-13.69% and 6.21% from predicted responses. This indicates that designed model space can be navigated for prediction.
Nouby et al. / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 254-271
269
0.0
0.5
1.0
1.5
2.0
2.5
1 7 13 19
Test Runs
N
e
g
a
t
i
v
e
D
a
m
p
i
n
g
R
a
t
i
o
Test Run vs Actual
Test Run vs Predicted
Figure 13. Predicted versus actual simulation
2.0 1.5 1.0 0.5 0.0
2.0
1.5
1.0
0.5
0.0
Act ual
P
r
e
d
i
c
t
e
d
Figure 14. The best fit line plot
-15
-10
-5
0
5
10
1 3 5 7 9 11 13 15 17 19 21
Test Runs
%
D
e
v
i
a
t
i
o
n
Figure 15. Percentage deviation of experimental results
Nouby et al. / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 254-271
270
6. Conclusion
Non-linear model for the damping ratio based on Central composite design of experiments was successfully developed for
prediction of brake squeal. To validate the model, randomly generated twenty two test cases were examined. Deviations between
predicted and actual simulation results were found to be within 15%. It shows reasonable agreement and also the adequacy of the
developed model in prediction. Though these investigations are predominantly for the disc brake squeal, it can also be extended to
other issues. By applying this methodology, while designing brake system, corrective and iterative design steps can be initiated
and implemented for betterment of component design. The proposed methodology will also remove surprises during the system
deployment. Continued research in this direction can bring about more comprehensive and appropriate guide lines for designers.
The predicted results show that brake squeal propensity can be reduced by increasing Youngs modulus of the back plate and
modifying the shape of friction material by adding chamfer on both sides of friction material and by introducing slot
configurations. The combined approach of modeling brake squeal using CEA and DOE is found to be statistically adequate
through verification trials. This combined approach will be useful in the design stage of the disc brake. The target of future work
will be to develop the brake model such as considering full brake assembly, including effect of material damping, wear effect, and
pressure contact distribution.
Acknowledgments
We would like to thank Prof. (Dr.) A. Elayaperumal, Department of Mechanical Engineering, Anna University, for his helpful
comments and recommendations about frequency response measurements of the brake system. We would also like to extend our
sincere thanks to Prof. (Dr.) Latha Nagendran, Anna University and the reviewers of this paper for their valuable comments and
suggestions.
References
ABAQUS Analysis Users Manual, Version 6.6-1. ABAQUS, Inc. 2006.
AbuBakar, A.R., and Ouyang, H. 2006. Complex eigenvalue analysis and dynamic transient analysis in predicting disc brake
squeal. Int. J Vehicle Noise Vibration, Vol. 2, No. 2, pp. 143155.
Akay, A. 2002. Acoustic of friction. Journal of Acoustical Society of America, Vol. 111, No. 4, pp. 15251548.
Bajer, A., Belsky, V., and Zeng, L. J. 2003. Combining a nonlinear static analysis and complex eigenvalue extraction in brake
squeal simulation. S.A.E Technical Paper 2003-01-3349.
Blaschke P., Tan M., and Wang A. 2000. On the analysis of brake squeal propensity using finite element method. SAE Paper,
2000-01-2765.
Brooks, P. C., Crolla, D. A., Lang, A.M., Schafer, D. R. 1993. Eigenvalue sensitivity analysis applied to disc brake squeal. Proc.
IMechE, C444/004, pp. 135-143.
Chakrabotry, G., Jearsiripongkul, T., von Wagner, U., Hagedorn, P. A New model for a floating caliper disc-brake and active
squeal control. VDI-Bericht, Vol. 1736, pp. 93102.
Chowdhary, H. V., Bajaj, A. K. and Krousgrill, C. M. 2001. An analytical approach to model disc brake system for squeal
prediction. DETC2001/VIB-21560, ASME, pp. 110.
Dai Y. and Lim T.C.. 2008. Suppression of brake squeal noise applying finite element brake and pad model enhanced by spectral-
based assurance criteria. Applied Acoustics, Vol. 69, pp. 196214.
Dunlap, K.B., Riehle, M.A., and Longhouse, R.E. 1999. An investigative overview of automotive disc brake noise, SAE Paper:
1999-01-0142.
Hu, Y., Mahajan, S., and Zhang, K. 1999. Brake squeal DOE using nonlinear transient analysis. SAE Paper 1999-01-1737.
Ibrahim, R.A. 1994. Friction-induced vibration, chatter, squeal and chaos, part II: dynamics and modelling, ASME Applied
Mechanics Reviews, Vol. 47, pp. 227259.
Joe, Y.G., Cha, B.G., Sim, H.J., Lee, H.J., Oh, J.E. 2008. Analysis of disc brake instability due to friction-induced vibration using
a distributed parameter model, International Journal of Automotive Technology, Vol. 9, No. 2, pp. 161-171.
Kinkaid, N.M., Reilly, O.M., and Papadopoulos, P. 2003. Automotive disc brake squeal, Journal of Sound and Vibration, Vol.
267, pp. 105-166.
Lee, Y. S., Brooks, P. C., Barton, D. C., and Crolla, D. A. 1998. A study of disc brake squeal propensity using a parametric finite
element model, Proc. Instn Mech. Engrs, paper 98, pp. 191201.
Liles, G. D. 1989. Analysis of disc brake squeal using finite element methods, SAE paper 891150, pp. 11381146.
Liu, P., Zheng, H., Cai, C., Wang, Y.Y., Lu, C., Ang, K.H., Liu, G.R. 2007. Analysis of disc brake squeal using the complex
eigenvalue method, Applied Acoustics, Vol. 68, pp. 603615.
Mario, T.J., Samir, N.Y., and Roberto J. 2008. Analysis of brake squeal noise using the finite element method: A parametric study.
Applied Acoustics, Vol. 69, pp. 147162.
Millner, N. An analysis of disc brake squeal. SAE Technical Paper, 1978, 780332.
Montgomery, D.C. 2005. Design and analysis of experiments. 5
th
edition. Wiley, Singapore.
Nouby et al. / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 254-271
271
Nouby, M. and Srinivasan, K. 2009. Parametric studies of disc brake squeal using finite element approach, Jurnal Mekanikal,
(accepted for publication).
Nouby, M., and Srinivasan, K. 2009. Disc Brake squeal reduction through pad structural modifications. Proceedings of the
International Conference on Recent and Emerging Advanced Technologies in Engineering, Kuala Lumpur.
Ouyang, H. and Mottershead, J.E. 1998. Friction-induced parametric resonances in disc: effect of a negative frictionvelocity
relationship, Journal of Sound and Vibration, Vol. 209, No. 2, pp. 251264.
Papinniemi A, Lai JCS, Zhao, J. and Loader, L. 2002. Brake squeal: a literature review, Applied Acoustics, Vol. 63, pp. 391400.
Shin, K., Brennan, M.J., Oh, J.-E., Harris, C.J. 2002. Analysis of disk brake noise using a two-degree-of-freedom model. Journal
of Sound and Vibration, Vol. 254, No. 5, pp. 837848.
Sinou, J. J., Thouverez, F., Jezequel, L. 2003. Analysis of friction and instability by the center manifold theory for a non-linear
sprag-slip model, Journal of Sound and Vibration, Vol. 265, No. 3, pp. 527559.
Von Wagner, U., Jearsiripongkul, T., Vomstein, T., Chakrabotry, G., Hagedorn, P. 2003. Brake squeal: modelling and
experiments, VDIBericht, Vol. 1749, pp. 173186.
Von Wagner, U., Hochlenert, D., Jearsiripongkul, T., Hagedorn, P. 2004. Active control of brake squeal via Smart pads. SAE
Paper 2004-01-2773.
Biographical notes
M. Nouby is a research scholar in the Department of Mechanical Engineering, Anna University, India since 2007. He received his
B.Sc. and M.S. degrees from Department of Automotive and Tractor Engineering, Minia University, Egypt in 1999 and 2003
respectively. Prior to joining the faculty at Anna University, he worked as an assistant lecturer in Minia University from 2003 to
2007. His research interests are in the areas of Vehicle Dynamics, Finite Element Methods, Automotive Design, Noise and
Vibration.
Dr. D. Mathivanan is a practicing engineer in the field of mechanical engineering. He obtained his B.E. (Mechanical) from GCT,
Coimbatore, India in the year 1990. He received his M.E. (Manufacturing) and Ph.D. from Anna University, Chennai, India. He
has international and national journal papers to his credit. His research areas are CAD, CAM, CAE and Plastics.
Dr. K. Srinivasan is one of the senior professors in the Department of Mechanical Engineering of Anna University. He has many
international and national journal and conference papers to his credit. His current research interests are Tribology, Vibration and
CNC machining.
Received December 2009
Accepted December 2009
Final acceptance in revised form December 2009
MultiCraft
International Journal of Engineering, Science and Technology
Vol. 1, No. 1, 2009, pp. 272-282
INTERNATIONAL
JOURNAL OF
ENGINEERING,
SCIENCE AND
TECHNOLOGY
www.ijest-ng.com
2009 MultiCraft Limited. All rights reserved
Cardinal priority ranking based decision making for economic-emission
dispatch problem
Lakhwinder Singh
1*
and J.S. Dhillon
2
1
Department of Electrical Engineering, Baba Banda Singh Bahadur Engineering College, Fatehgarh Sahib -140 407, Punjab, India.
2
Department of Electrical & Instrumentation Engineering, Sant Longowal Institute of Engineering & Technology,
Longowal -148 106, Distt. Sangrur, Punjab, India.
E-mails:(lakhwinder1968@yahoo.com (Lakhwinder Singh),Corresponding author); jsdhillonp@yahoo.com (J.S. Dhillon)
Abstract
This paper deals with the economic emission dispatch (EED) problem relating to real and reactive power scheduling of
thermal power generating units. The formulated EED problem is solved using weighting method to generate non-inferior
solutions which allows explicit trade-offs between objective levels for each non-inferior solutions. Fuzzy decision making
methodology is exploited to decide the generation schedule. To access the indifference band, interaction with the decision maker
is obtained via cardinal priority ranking (CPR) of the objectives. The cardinal priority ranking is constructed in the functional
space and then transformed into the decision space, so the cardinal priority ranking of objectives relate the decision makers
preferences to non-inferior solutions through normalized weights. Regression analysis is performed between the cardinal priority
ranking and simulated weights to decide the best compromised solution. Decoupled load flow analysis is performed to find the
loss coefficients and transmission losses. The validity of the proposed method is demonstrated on IEEE 11-bus system which
comprises 3-generators.
Keywords: Economic emission dispatch; fuzzy decision making; weighting method, cardinal priority ranking
1. Introduction
In a large number of real-life decision problems, a decision maker is faced with difficulties to take a decision, especially when
multiple objectives are present in the decision space and these are to be achieved simultaneously. This problem becomes more
complicated when the objectives are conflicting, non-commensurable and imprecise in nature. Such types of problems are called
multiobjective optimization problems in which the goal is to maximize or minimize several objective functions, simultaneously.
For effective operations, the optimal power scheduling problem has mainly confined to minimize the generation cost regardless of
emission constraints. With the increase in the environmental awareness and the passage of environmental regulations, the clean air
act amendments of 1990 (El-Keib, 1994) has forced utilities to modify their operating strategies to reduce pollution and
atmospheric emissions of thermal power generation to meet environmental standards. So the environmental constraints have
become of vital concern to system operators to control emissions such as oxides of nitrogen (NO
x
), oxides of sulphur (SO
x
) and
oxides of carbon (CO
x
) from thermal plants, which are of greater concern to power utility and communities (Tsay, 2003). The
conventional optimization techniques are not suitable to obtain the optimal solution, which simultaneously optimizes a variety of
objectives. Multiobjective optimization methodology permits a better simulation of real word problems, often characterized by
contrasting/ conflicting goals, and gives the planner the capability of making the final decision by selecting, on the basis of his
individual point of view, the most trade-off solution in a wide range of suitable solutions. For the solution of such multiobjective
problems different techniques have been reported in literature pertaining to economic-emission dispatch problem (Bath et al.,
2004; Chaaban et al., 2004; Singh and Dhillon, 2008). Ramanathan (1994) has presented a methodology to include emission
constraints in classical economic dispatch, which contains an efficient weights estimation technique. Talaq et al. (1994) have given
a summary of work in the area of environmental/economic dispatch which includes several techniques intended to reduce
emissions into the atmosphere due to electric power generation. Hota et al. (2000) have solved the economic emission load
Singh and Dhillon / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 272-282
273
dispatch through an interactive fuzzy satisfying method. Basu (2002) has used the Hopfield neural networks to solve fuel
constrained economic emission load dispatch problem. Chen and Chen (2003) have presented a direct Newton-Raphson economic
emission dispatch method which considers the line flow constraints by evaluating the B-coefficients from the sensitivity factors
with dc load flow. Abido (2003) has presented a novel approach based on the strength of Pareto evolutionary algorithm to solve
environmental/economic power dispatch optimization problem. Fuzzy based mechanism is employed to extract the best
compromise solution over the trade-off curve. Brar et al. (2002) have used fuzzy logic based weightage pattern searching to obtain
the solution of multiobjective load dispatch problem. The evolutionary optimization technique has been employed in which the
preferred weightage pattern is searched to get the best optimal solution in non-inferior domain. An analytical solution technique
for combined economic emission dispatch problem has been presented by Palanichamy et al. (2008). The fuel cost as well as the
emission characteristics of generating units is represented by their respective equivalent characteristic in terms of power plant total
generations.
The intent of the paper is to solve EED problem in which four objectives like operation cost, NO
x
emission, SO
x
emission and
CO
x
emission are minimized simultaneously. The objectives are of conflicting nature and improvement in one objective can be
reached only by the reduction of other. The formulated EED problem is solved using weighting method to generate non-inferior
solutions which allows explicit trade-offs between objective levels for each non-inferior solution. Exploiting fuzzy decision
making theory, membership functions relating to objectives are defined those play a vital role to find the best alternative among
the non-inferior solutions. To access the indifference band, interaction with the decision maker is obtained via cardinal priority
ranking of the objectives. The cardinal priority ranking is constructed in the functional space and then transformed into the
decision space, so the cardinal priority ranking of objectives relate the decision makers preferences to non-inferior solutions
through normalized weights. Regression analysis is performed between the cardinal priority ranking and simulated weights to
decide the best compromised solution. Decoupled load flow (DLF) analysis is performed to find the loss coefficients, real and
reactive power losses. The effectiveness of the proposed method is demonstrated on IEEE 11-bus, 17-lines system, comprising 3-
generators.
2. EED problem formulation
The problem formulation treats EED problem in which the attempt is made to minimize conflicting objective functions
simultaneously, while satisfying equality and inequality constraints. Generally the problem is formulated as:
Minimize operating cost: ( )
=
+ + =
Ng
i
i Gi i Gi i Gi
c P b P a P F
1
2
1
) ( $/h (1)
Minimize NOx emission: ( )
=
+ + =
Ng
i
i Gi i Gi i Gi
f P e P d P F
1
2 2
2
2 2
) ( kg/h (2)
Minimize SOx emission: ( )
=
+ + =
Ng
i
i Gi i Gi i Gi
f P e P d P F
1
3 3
2
3 3
) ( kg/h (3)
Minimize CO
x
emission: ( )
=
+ + =
Ng
i
i Gi i Gi i Gi
f P e P d P F
1
4 4
2
4 4
) ( kg/h (4)
where a
i
, b
i
and
c
i
are the cost coefficients of i
th
generator. d
2i
, e
2i
and f
2i
are the NO
x
emission coefficients of i
th
generator. d
3i
, e
3i
and f
3i
are the SO
x
emission coefficients of i
th
generator. d
4i
, e
4i
and
f
4i
are the CO
x
emission coefficients of i
th
generator
Equality constraints
The total real and reactive powers generated must meet the total demand and losses in the system.
loss
Nb
i
Di
Ng
i
Gi
P P P + =
= = 1 1
(5)
loss
Nb
i
Di
Ng
i
Gi
Q Q Q + =
= = 1 1
(6)
where N
b
are the number of buses in the system.
Gi
P and
Gi
Q are the real and reactive powers of i
th
generator, respectively. P
Di
and Q
Di
are real and reactive demands at i
th
bus,
respectively
.
loss
P and
loss
Q are real and reactive power losses in the transmission
lines, respectively.
Singh and Dhillon / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 272-282
274
Inequality constraints
To ensure stable operation, each generating unit is restricted by its lower and upper limits of real and reactive power outputs.
max min
Gi Gi Gi
P P P ; i = 1, 2,, Ng (7)
max min
Gi Gi Gi
Q Q Q ; i =1, 2,, Ng (8)
where
min
Gi
P and
max
Gi
P are the minimum and maximum values of real power output of i
th
unit, respectively.
min
Gi
Q and
max
Gi
Q are
the minimum and maximum values of reactive power output of i
th
unit, respectively.
Power transmission losses
The real and reactive power transmission losses,
loss
P and
loss
Q are given by following equations:
[ ]
= =
+ + =
Nb
i
Nb
j
j i j i ij j i j i ij loss
Q P P Q B Q Q P P A P
1 1
) ( ) ( (9)
[ ]
= =
+ + =
Nb
i
Nb
j
j i j i ij j i j i ij loss
Q P P Q D Q Q P P C Q
1 1
) ( ) ( (10)
where ) ( ) (
Di Gi Di Gi i i
Q Q j P P jQ P + = +
) cos(
j i
j i
ij
ij
V V
R
A =
;
) sin(
j i
j i
ij
ij
V V
R
B =
) cos(
j i
j i
ij
ij
V V
X
C =
;
) sin(
j i
j i
ij
ij
V V
X
D =
with
ij
A ,
ij
B ,
ij
C and
ij
D are loss coefficients, and are evaluated from line data by performing DLF analysis.
i
and
j
are
load angles at
th
i
and
th
j
buses, respectively.
i
V
and
j
V are voltage magnitude at
th
i
and
th
j
buses, respectively.
ij
R is the
real component of impedance bus matrix.
ij
X is the reactive component of impedance bus matrix (Dhillon, 1993).
3. Solution procedure
To generate the non-inferior solutions, the multiobjective problem is converted into scalar optimization problem as:
Minimize
=
L
j
j j
F w
1
(11)
Subject to: i)
=
=
L
j
j
w
1
0 . 1 , 0 . 0
j
w (12)
ii) Eq. (5) to (8)
where w
j
are the levels of the weighting coefficients. L is the total number of objectives. The sum of all weights is equal to one. To
find the solution, constrained problem is converted into an unconstrained problem. Equality and inequality constraints are clubbed
with objective function to form generalized augmented function as:
Singh and Dhillon / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 272-282
275
=
= = = = =
loss
Nb
i
Di
Ng
i
Gi q loss
Nb
i
Di
Ng
i
Gi p
L
j
j j q p Gi Gi
Q Q Q P P P F w Q P L
1 1 1 1 1
) , , , (
QGi PGi
U U + +
(13)
where
) (
( )
>
<
=
=
max
1
2
max
1
max min
min
1
2
min
1
;
1
; 0
;
1
Gi Gi
Ng
i
Gi Gi
k
Gi Gi Gi
Gi Gi
Ng
i
Gi Gi
k
PGi
P P P P
r
P P P
P P P P
r
U
) (
( )
>
<
=
=
max
1
2
max
1
max min
min
1
2
min
1
;
1
; 0
;
1
Gi Gi
Ng
i
Gi Gi
k
Gi Gi Gi
Gi Gi
Ng
i
Gi Gi
k
QGi
Q Q Q Q
r
Q Q Q
Q Q Q Q
r
U
p
,
q
are lagrangian multipliers,
k
r
1
is penalty factor. The Newton-Raphson method is applied to obtain the non-inferior
solutions for simulated weight combinations, to achieve the necessary conditions.
4. Cardinal priority ranking
The fuzzy sets are defined by equations called membership functions, which represent the goals of each objective function. The
membership function represents the degree of achievement of the original objective function as a value between 0 and 1 with
( )
i
F =1 as completely satisfactory and ( )
i
F = 0 as unsatisfactory. Such a linear membership function represents the decision
makers fuzzy goal of achievement, and at the same time scales the original objective functions with different physical units into
measure of 0-1. By taking account of the minimum and maximum values of each objective function together with the rate of
increase of membership satisfaction, the decision maker must determine the membership function ( )
i
F in a subjective manner.
( )
F F ; 0
F F F ;
F F
F F
F F ; 1
F
max
i i
max
i i
min
i
min
i
max
i
i
max
i
min
i i
i
=
(14)
where
min
i
F
and
max
i
F are minimum and maximum values of i
th
objective function in which the solution is expected. The value of
the membership function indicates how much (in the scale from 0 to 1) a noninferior solution has satisfied the
i
F objective. The
sum of the membership function values ( ( )
i
F ; i = 1, 2, , L) for all the objectives can be computed in order to measure the
accomplishment of each solution in satisfying the objectives. The accomplishment of each non-dominated solution can be rated
with respect to all the M non-dominated solutions by normalizing its accomplishment over the sum of the accomplishment of
the M non-dominated solutions as follows:
= =
=
=
M
k
L
i
k
i
L
i
k
i
k
D
F
F
1 1
1
) (
) (
(15)
Singh and Dhillon / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 272-282
276
where
k
i
k
i
F F ) ( 1 ) ( =
The function
k
D
can be treated as an unsatisfied membership function for non-dominated solutions in a fuzzy set and
represented as a fuzzy cardinal priority ranking of the non-dominated solutions. The smaller the unsatisfied cardinal priority, the
better is the solution. The function
k
S is defined as:
} { ) (
min max min
S S S S
k
D
k
+ = (16)
where
min
S and
max
S are minimum and maximum values of scaling factor to map the function,
k
S in the required range to adjust
the cardinal priority. Regression analysis is performed between cardinal priority ranking and simulated weights,
i
w ; i = 1, 2, ,
L to achieve maximum satisfaction.
5. Flow chart
The economic emission dispatch problem is solved by various steps. The step wise procedure is depicted in flow chart given in
Figure 1.
6. Test system and results
The validity of the proposed method is illustrated on 11-bus, 17-lines IEEE system, comprising of three generators (Singh et al.,
2006). Minimum and maximum values of the objectives are obtained by performing minimum economic and emission dispatch
respectively. Minimum and maximum values of the objectives are shown in Table 1.
Table 1: Minimum and maximum values of objectives
=
min
1
F 4584.7830 $/h
max
1
F = 4742.0610 $/h
=
min
2
F 619.1288 kg/h =
max
2
F 953.5742 kg/h
=
min
3
F 2848.7130 kg/h
max
3
F = 6246.4340 kg/h
=
min
4
F 5.887253 kg/h
max
4
F = 15.05706 kg/h
To obtain the solution of EED problem, three different cases are considered in which weights are simulated with different step
sizes so that their sum remains equal to one and are as:
Case-I: Weights are simulated by giving variation in step of 0.05
Case-II: Weights are simulated by giving variation in step of 0.02
Case-III: Weights are simulated by giving variation in step of 0.01
Non-inferior solutions, corresponding membership functions for all the objectives along with the cardinal priority ranking for the
simulated weight combinations are shown in Table 2, Table 3 and Table 4 for all the above three cases.
Singh and Dhillon / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 272-282
277
No
No
No
Start
Yes
Read data
Compute initial values of
P
and
Gi
P
when
0 =
q
,
0 =
loss
P
,
0 =
loss
Q
and also calculate initial fuel cost,
prevs
Cost
F
Run load flow, Compute loss coefficients
Calculate real and reactive power losses using Eq.(9) and Eq.(10)
Apply Newton-Raphson Method to solve Eq.(13)
Modify values of
new
P
,
new
q
and
new
Gi
P
Yes
Simulate weight combinations
No
More non-inferior solutions
Stop
Yes
Compute
D
P and
D
Q
Err P P P
loss
Nb
i
Di
Ng
i
Gi
1 1
Err Q Q Q
loss
Nb
i
Di
Ng
i
Gi
1 1
ITI=ITI+1
Yes
Error F F
prevs
t
new
t
cos cos
IT=IT+1
Compute membership function of objectives using Eq.(14)
Calculate cardinal priority ranking and function
k
S
using Eq.(15) and Eq. (16), resp.
Best weights are evaluated
Write optimal values of
Gi
P
,
Gi
Q
and real & reactive losses
Figure 1: Flow chart for economic emission dispatch problem
Singh and Dhillon / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 272-282
278
Table 2: Non-inferior solutions, membership functions and cardinal priority ranking for case-I
Weights
Objectives and membership functions
Objectives
1
F $/h
2
F kg/h
3
F kg/h
4
F kg/h
1
w
2
w
3
w
4
w
Membership
1
( ) F
2
( ) F
3
( ) F
) (
4
F
k
S
Objectives 4701.1930 863.6878 2899.6690 9.214497
0.25 0.25 0.25 0.25
Membership 0.740153 0.731237 0.014997 0.362848
0.7900239
Objectives 4713.5030 885.0974 2874.9760 9.344164
0.20 0.20 0.30 0.30
Membership 0.818423 0.795253 0.007730 0.376988
0.8537466
Objectives 4685.3470 837.9921 2945.2890 9.039403
0.30 0.30 0.20 0.20
Membership 0.639403 0.654407 0.028424 0.343753
0.7117372
Objectives 4711.3310 866.9720 2892.6060 9.467503
0.20 0.30 0.30 0.20
Membership 0.804613 0.741057 0.012918 0.390439
0.8326569
Objectives 4689.8210 858.0472 2913.2680 8.939130
0.30 0.20 0.20 0.30
Membership 0.667847 0.714372 0.019000 0.332818
0.7408092
Objectives 4723.0240 903.7136 2861.3140 9.432228
0.15 0.15 0.35 0.35
Membership 0.878959 0.850916 0.003709 0.386592
0.9057738
Objectives 4706.1980 752.3051 3217.0410 10.69450
0.35 0.35 0.15 0.15
Membership 0.771978 0.398200 0.108404 0.524247
0.7701989
Objectives 4720.8640 868.8564 2889.2360 9.709036
0.15 0.35 0.35 0.15
Membership 0.865224 0.746692 0.011927 0.416779
0.8717871
Objectives 4676.0400 848.1931 2940.1950 8.621339
0.35 0.15 0.15 0.35
Membership 0.580227 0.684908 0.026925 0.298162
0.6793687
Objectives 4730.3710 920.6932 2853.7960 9.483382
0.10 0.10 0.40 0.40
Membership 0.925673 0.901685 0.001496 0.392171
0.9488586
Objectives 4730.1370 869.8763 2888.1770 9.945444
0.10 0.40 0.40 0.10
Membership 0.924186 0.749741 0.011615 0.442560
0.9091604
Objectives 4736.1320 937.0949 2849.9040 9.502926
0.05 0.05 0.45 0.45
Membership 0.962301 0.950726 0.000351 0.394302
0.9858791
By performing linear regression analysis, the obtained best values of weights and the corresponding values of fuel cost, NO
x
emission, SO
x
emission, CO
x
emission for all the three cases are compared and are shown in Table 5. It has been observed from
Case-I, Case-II and Case-III that the minimum value of the membership functions of the objectives is improved by decreasing the
step size of the weights. The membership function increases from 0.5165955 to 0.5241862 with the decrease in step size from
0.05 to 0.02 and further 0.5241862 to 0.5430062 with the decrease in step size from 0.02 to 0.01. The results obtained in the
proposed method are also compared with the results of Brar et al. (2002). It has been observed that, the solutions achieved in
proposed method have more membership satisfaction as compared to the results presented in Brar et al. (2002). It is clear from all
the three cases that by reducing the step size of the weights, better membership satisfaction is achieved and when the overall
outcome is less than some of the non-inferior solutions try with the reduced step size which will improve the overall outcome. The
best power generation schedule for all the three cases of proposed method is given in Table 6.
Singh and Dhillon / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 272-282
279
Table 3: Non-inferior solutions, membership functions and cardinal priority ranking for case-II
Weights
Objectives and membership functions
Objectives
1
F $/h
2
F kg/h
3
F kg/h
4
F kg/h
1
w
2
w
3
w
4
w
Membership
1
( ) F
2
( ) F
3
( ) F
) (
4
F
k
S
Objectives 4703.9120 868.2581 2893.4810 9.243966 0.24 0.24 0.26 0.26
Membership 0.757442 0.744903 0.013176 0.366062
0.08304875
Objectives 4698.3330 858.9479 2906.6590 9.183201 0.26 0.26 0.24 0.24
Membership 0.721970 0.717065 0.017054 0.359435
0.08013305
Objectives 4703.2930 864.4938 2897.8570 9.266447 0.24 0.26 0.26 0.24
Membership 0.753506 0.733647 0.014464 0.368513
0.08254325
Objectives 4699.0450 862.7900 2901.7310 9.161678 0.26 0.24 0.24 0.26
Membership 0.726496 0.728553 0.015604 0.357088
0.08067229
Objectives 4708.9530 876.9422 2883.1320 9.297540 0.22 0.22 0.28 0.28
Membership 0.789494 0.770868 0.010130 0.371904
0.08573294
Objectives 4707.3750 865.8672 2894.8760 9.368151 0.22 0.28 0.28 0.22
Membership 0.779460 0.737754 0.013586 0.379604
0.08432090
Objectives 4692.1590 848.9004 2923.5310 9.115059 0.28 0.28 0.22 0.22
Membership 0.682715 0.687023 0.022020 0.352004
0.07696564
Objectives 4713.5030 885.0973 2874.9760 9.344166 0.20 0.20 0.30 0.30
Membership 0.818423 0.795252 0.007730 0.376989
0.08820449
Objectives 4711.3300 866.9719 2892.6060 9.467499 0.20 0.30 0.30 0.20
Membership 0.804607 0.741057 0.012918 0.390438
0.08602533
Objectives 4717.6100 892.8093 2868.5340 9.384130 0.18 0.18 0.32 0.32
Membership 0.844535 0.818312 0.005834 0.381347
0.09048350
Objectives 4684.6950 854.7670 2921.8040 8.818712 0.32 0.18 0.18 0.32
Membership 0.635256 0.704564 0.021512 0.319686
0.07419626
Objectives 4715.1920 867.8572 2890.9120 9.565063 0.18 0.32 0.32 0.18
Membership 0.829161 0.743704 0.012420 0.401078
0.08767353
Singh and Dhillon / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 272-282
280
Table 4: Non-inferior solutions, membership functions and cardinal priority ranking for case-III
Weights
Objectives and membership functions
Objectives
1
F $/h
2
F kg/h
3
F kg/h
4
F kg/h
1
w
2
w
3
w
4
w
Membership
1
( ) F
2
( ) F
3
( ) F
) (
4
F
k
S
Objectives 4701.1930 863.6878 2899.6690 9.214497 0.25 0.25 0.25 0.25
Membership 0.740153 0.731237 0.014997 0.362848
0.08131214
Objectives 4698.3330 858.9479 2906.6590 9.183201 0.26 0.26 0.24 0.24
Membership 0.721970 0.717065 0.017054 0.359435
0.07982981
Objectives 4703.2930 864.4938 2897.8570 9.266446 0.24 0.26 0.26 0.24
Membership 0.753506 0.733647 0.014464 0.368513
0.08223089
Objectives 4706.4970 872.6721 2887.9970 9.271636 0.23 0.23 0.27 0.27
Membership 0.773878 0.758101 0.011562 0.369079
0.08409920
Objectives 4705.3530 865.2177 2896.2670 9.317632 0.23 0.27 0.27 0.23
Membership 0.766604 0.735812 0.013996 0.374095
0.08312687
Objectives 4695.3240 854.0240 2914.5690 9.150059 0.27 0.27 0.23 0.23
Membership 0.702839 0.702342 0.019382 0.355821
0.07828473
Objectives 4708.9530 876.9423 2883.1330 9.297540 0.22 0.22 0.28 0.28
Membership 0.789494 0.770869 0.010130 0.371904
0.08540853
Objectives 4692.1590 848.9002 2923.5310 9.115069 0.28 0.28 0.22 0.22
Membership 0.682715 0.687022 0.022020 0.352005
0.07667442
Objectives 4711.2870 881.0804 2878.8130 9.321701 0.21 0.21 0.29 0.29
Membership 0.804334 0.783242 0.008859 0.374539
0.08666504
Objectives 4711.3310 866.9720 2892.6060 9.467502 0.20 0.30 0.30 0.20
Membership 0.804613 0.741057 0.012918 0.390439
0.08570008
Objectives 4713.2720 867.4395 2891.6940 9.516477 0.19 0.31 0.31 0.19
Membership 0.816954 0.742455 0.012650 0.395780
0.08652723
Objectives 4717.6100 892.8094 2868.5340 9.384129 0.18 0.18 0.32 0.32
Membership 0.844535 0.818312 0.005834 0.381347
0.09014110
Table 5: Comparison of results
Cost
($\h)
NO
x
emission
(kg/h)
SO
x
emission
(kg/h)
CO
x
emission
(kg/h)
Weights
0.1355 0.4247 0.0753 0.3645
Objectives 4660.8120 744.4364 3278.2300 9.624076
Case-I
Membership 0.5165955 0.6253272 0.8735867 0.5924862
Weights
0.1737 0.3660 0.0854 0.3748
Objectives 4659.6180 765.2352 3177.3690 9.243322
Case-II
Membership 0.5241862 0.5631382 0.9032716 0.6340087
Weights
0.1872 0.4139 0.0861 0.3128
Objectives 4656.6580 756.0088 3223.5530 9.296052
Case-III
Membership 0.5430062 0.5907254 0.8896790 0.6282583
Weights
0.5400 0.2070 0.1270 0.1260
Objectives
4650.108 810.5679 3052.2760 8.243926
Brar et al.
(2002)
Membership
0.5846541 0.4275925 0.9400884 0.7429965
Singh and Dhillon / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 272-282
281
Table 6: Generation schedule corresponding to optimal solutions
Optimal values Case-I Case-II Case-III Brar et al. (2002)
1 G
P (p.u.)
2.4461 2.4235 2.4155 2.2664
2 G
P (p.u.)
0.6624 0.5897 0.6256 0.4888
3 G
P (p.u.)
1.1012 1.1973 1.1654 1.4428
1 G
Q (p.u.)
0.7217 0.7200 0.7171 0.6988
2 G
Q (p.u.)
0.7248 0.7351 0.7250 0.7390
3 G
Q (p.u.)
0.3942 0.3946 0.3916 0.3808
loss
P
0.3635 0.3647 0.3605 0.3618
loss
Q
0.5777 0.5872 0.5710 0.5762
7. Conclusions
The solution set of the problem is non-inferior due to conflicting nature of the objectives and has been obtained through weighting
method. The novel formulation as economic emission dispatch problem has made it possible to quantitatively grasp trade-off
relations among conflicting objectives. The trade-off approach is effective only for two objectives, as the number of objectives
increases the selection of best solution becomes cumbersome. Exploiting fuzzy set theory an interactive cardinal priority ranking
method has been applied to identify the best compromise solution for EED problem, when conflicting objectives are more than
two. The major characteristics and advantages of the cardinal priority ranking method are that the cardinal priority ranking
functions, which relate the decision makers preference to the non-inferior, solutions though the trade-off functions, are
constructed in the functional space and only then are transformed in to the decision space. The proposed method provides
interface between the decision maker and the mathematical model through cardinal priority ranking. It also allows explicit trade-
off between fuel cost of units with NO
x
emission, SO
x
emission and CO
x
emission levels, respectively. Results of the proposed
method are compared with Brar et al. (2002). The proposed method gives better results in terms of overall membership
satisfaction and real and reactive power losses. Study can be extended by adopting -constraint method or shifted min-max
method to generate the non-inferior solution surface. Generally, the weights are either simulated or searched in the non-inferior
domain. Evolutionary search technique may be implemented to search the preferred weightage pattern in the non-inferior
domain, which may correspond to the best compromised solution.
References
Abido M.A., 2003. Environmental/economic power dispatch using multiobjective evolutionary algorithms, IEEE Trans on Power
Systems, Vol. 18, No. 4, pp. 1529-1537.
Basu M., 2002. Fuel constrained economic emission load dispatch using Hopfield neural networks, Electric Power Systems
Research, Vol. 63, pp. 51-57.
Bath S.K, Dhillon J.S. and Kothari D.P., 2004. Fuzzy satisfying stochastic multi-objective generation scheduling by weightage
pattern search methods, Electric Power Systems Research, Vol. 69, pp. 311-320.
Brar Y.S., Dhillon J.S., and Kothari D.P., 2002. Multi-objective load dispatch by fuzzy logic searching weightage pattern, Electric
Power Systems Research, Vol. 63, pp. 149-160.
Chaaban F.B., Mezher T. and Ouwayjan M., 2004. Options for emissions reduction from power plants: an economic evaluation,
International Journal of Electrical Power & Energy Systems, Vol. 26, No. 1, pp. 57-63.
Chen S.D., and Chen J.F., 2003. A direct NewtonRaphson economic emission dispatch, International Journal of Electrical
Power & Energy Systems, Vol. 25, pp. 411-417.
El-Keib A.A., Ma H. and Hart J.L., 1994. Economic dispatch in view of clean air Act of 1990, IEEE Trans on Power Systems,
Vol. 9, No. 3, pp. 972-978.
Hota P.K., Chakrabarti R., and Chattopadhyay D.P.K., 2000. Economic emission load dispatch through an interactive fuzzy
satisfying method, Electric Power Systems Research, Vol. 54, pp.151-157.
Palanichamy C. and Babu N.S., 2008. Analytical solution for combined economic and emissions dispatch, Electric Power Systems
Research, Vol. 78, No. 7, pp. 1129-1137.
Ramanathan R., 1994. Emission constrained economic dispatch, IEEE Trans on Power Systems, Vol. 9, No. 4, pp. 1994-2000.
Singh and Dhillon / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 272-282
282
Talaq J.H., El-Hawary F., and El-Hawary M.E., 1994. A summary of environmental/ economic dispatch algorithms, IEEE Trans
on Power Systems, Vol. 9, No. 3, pp.1508-1516.
Singh L., Dhillon J.S., and Chauhan R.C., 2006. Evaluation of best weight pattern for multiple criteria load dispatch, Electric
Power Components and Systems, Vol. 34, No.1, pp. 21-35.
Singh L. and Dhillon J.S., 2008. Secure multiobjective real and reactive power allocation of thermal power units, International
Journal of Electrical Power & Energy Systems, Vol. 30, No. 10, pp. 594-602.
Tsay M.T., 2003. Applying the multi-objective approach for operation strategy of cogeneration systems under environment
constraints, International Journal of Electrical Power & Energy Systems, Vol. 25, pp. 219-226.
Biographical notes
Lakhwinder Singh is Associate Professor & Head, Department of Electrical Engineering, Baba Banda Singh Bahadur Engineering College,
Fatehgarh Sahib. He obtained his B.E. (Electrical) (1990) from Guru Nanak Dev Engineering College, Ludhiana and M.E. (Power Systems)
(1995) from Punjab University, Chandigarh. He is currently pursuing Ph.D. from Punjab Technical University, Jallandhar. He is a life member of
the Indian Society for Technical Education, and a member of the Institution of Engineers (India). He has published/presented 29 papers in
national and international conferences/ journals. His research interests include power system operation, optimization, neural networks and fuzzy
logic.
J.S.Dhillon is Professor, Electrical and Instrumentation Engineering Department, Sant Longowal Institute of Engineering and Technology,
Longowal, India (Deemed University). He received his B.E. (Electrical) (1983) from Guru Nanak Dev Engineering College, Ludhiana, M.E.
(Systems) (1987), Punjab Agricultural University, Ludhiana, Ph.D., (1996) Thapar Institute of Engineering & Technology, Patiala. His research
activities include Microprocessor applications, Multi-objective thermal dispatch, hydrothermal scheduling, Optimization, Neural Networks,
Fuzzy theory and soft computing applications in Power system. A member of The Institution of Engineers (India), Life Member: ISTE, Life
Member: SSI, has supervised 02 Ph.D, 13 M.E. scholars and supervising 06 Ph.D. Scholars. He has contributed in these areas as evidenced by 87
research papers published in various national and international Journals/ proceedings. He has co-authored 02 books "Power System
Optimization", PHI, Publication, "Principles of Electrical & Electronics Engineering, Kalyani Publication.
Received December 2009
Accepted December 2009
Final acceptance in revised form December 2009
MultiCraft
International Journal of Engineering, Science and Technology
Vol. 1, No. 1, 2009, pp. 283-290
INTERNATIONAL
JOURNAL OF
ENGINEERING,
SCIENCE AND
TECHNOLOGY
www.ijest-ng.com
2009 MultiCraft Limited. All rights reserved
Technical note
New table look-up lossless compression method based on binary
index archiving
R. Rdescu
Polytechnic University of Bucharest, 1-3, Iuliu Maniu Blvd, Sector 6, Bucharest, Romania.
E-mail: rradescu@atm.neuro.pub.ro, phone: +40-21-402-48-73; fax: +40-21-402-48-21
Abstract
This paper intends to present a common use archiver, made up following the dictionary technique and using the index
archiving method as a simple and original procedure. The original contribution of the paper consists in the structure of the
archived file and in the transformation of the dictionary codes into archived characters. This archiver is useful in order to
accomplish the lossless compression for any file types. The application can offer important conclusions regarding the
compression performances and the influence of the chosen dictionary over the parameters.
Keywords: Archive, data compression, dictionary codes, lossless algorithms
1. Introduction
The archivers, using dictionary techniques (Murgan, 1998; Rdescu, 2003a,b), can be very efficient, especially when using some
files that have different words which are very often repeated. This happens because of the fact that the archivers generate their
dictionaries during the archiving process, this way the program learns new words. Almost all the archiving programs, such as
Zip, PKZip, LHArc, ARJ, GZIP, RAR etc., make use of the LZ77 and LZSS algorithms or their variants (Cover, 1991; Storer,
1998; Rao and Yip, 2001; Salomon, 2007, 2008; van Lint, 1992; Sayood, 2005; Pu, 2006; Hankerson et al., 2003; Wayner, 1999,
Nelson, 1991) after the files are merged together in a reversible fashion. Because the application can make an archive that contains
more files, the archive has to be very well configured, so that during the unpacking of the files it can be separated with lossless
information.
2. Structure of the archived file
The archived file is composed from header, followed by archived data. The header is formed from general header, followed by n
archived file headers, where n is the number of files in the archive. For the beginning, the structure of the general header is
presented (an example could be given bit by bit):
3 bytes to store 3 letters (CBA). These letters are used as the identification of the archive. It is very important to
verify these characters in order not to let the archiver to try unpacking a file that is not a CBA archive.
2 bytes to store the maximum length of the dictionary.
2 bytes to store the minimum length of the dictionary.
1 byte to store the settings. This byte is used to store 3 binary validation variables:
o 1 bit if file has path;
o 1 bit if we keep the unpacked size of the file;
o 1 bit if file has password.
Rdescu / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 283-290
284
1 byte to store the length of the password (optional).
n bytes to store the password, where n is the length of the password (optional).
2 bytes to store the number of files.
After describing the general header, the archiver repeats the following sequence for every new added archive file (archived file
header):
2 bytes to store the length of the string that contains the path of the file (optional).
s bytes for the s characters of the string that contains the path of the file (optional).
1 byte for the length of the file name.
f bytes for the f characters of the file name.
4 bytes to store the unpacked size of the file (optional).
1 byte for the archiver type. Some files can only be copied in the archive, because their archiving would cause the
increasing of the file size.
4 bytes for the size of the packed file.
nr bytes for the nr characters of the packed file.
3. Packing and unpacking of the files
The process of packing and unpacking of files has 2 stages:
the transformation of the initial characters into dictionary codes;
the transformation of the dictionary codes into archive characters.
Transformation of the initial characters into dictionary codes
This stage is accomplished using the Lempel-Ziv-Welch (LZW) dictionary compression (Murgan, 1998; Rdescu, 2003a,b,c;
Rdescu and Olteanu, 2005; Rdescu and Ene, 2005). Initially, it begins with a 257-word dictionary, i.e., the 256 ASCII characters
and a special word that indicates the end of the file. The application allows a dictionary limitation. Therefore, all the values that
exceed the minimum size of the dictionary will be deleted every time the maximum size of the dictionary is obtained.
The user can set both the maximum and minimum values. According to the chosen values, the number of the packed files and
the compression time change. The choice of a too large maximum value of the dictionary results in a very long waiting time,
getting a too small dimension improvement. The optimal values for the two limitation dictionary variables are different from one
file to another.
Transformation of the dictionary codes into archive characters
This method relies on tackling from two different perspectives of two strings of numbers, having the same basic table.
Dictionary codes greater than 256 elements cannot be written in the archive using only one byte. Therefore, it is necessary to have
a 2 bytes space. This space is too large comparing it to the necessary one, especially in the initial phases, where the dictionary has
not a large size.
From the first steps, the dictionary has a maximum of 512 elements, and the dictionary code can be written on 9 bits from the 16
available bits. Hence, 7 out of the 16 available bits remain unused, meaning almost half of the overall space. Grouping 8
codewords, 8 9 bits = 72 bits are needed. It can be written on 72 bits / 8 bits = 9 bytes, comparing to the 8 codewords 2 bytes
(9 bits) = 16 bytes usually needed. Even for dictionary larger than 512 elements, this method will reduce the necessary code to
store dictionary codes. Table 1 refers to the transformation of codewords (the dictionary indexes) into archive characters.
Initially, it works with a 257-word dictionary (256 characters + 1 end of file control character). Table 1 is, in fact, an example in
which the dictionary has a number of words 512. On the first column (423, 137, 481, 45, ) there are the codewords (dictionary
indexes), which have values up to 512 that can be written on 9 bits. In order to have the certainty to obtain archive characters (8
bits), 8 codewords are used each time.
The codewords (423, 137, 481, 45, ) are binary written on the rows. This means that it will be 1byte (8bits) on each column of
the table (the archive characters):
(8 codewords) (9 bits) = (9 archive characters) (8 bits) (1)
The first row (161, 231, 44, 182, ) contains the character words obtained by transforming every column from binary to the
10
th
base. For example,
161 = 12
7
+02
6
+12
5
+02
4
+02
3
+02
2
+02
1
+12
0
(2)
It works similarly for the other values: 231, 44, 182,
Rdescu / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 283-290
285
If the dictionary has between 512 and 1024 words, the procedure is similar, Table I having 8 rows, but one extra column,
because every dictionary word needs 10 bits (10 columns):
(8 codewords) (10 bits) = (10 archive characters) (8 bits) (3)
Table 1. Transformation of dictionary codes into archive characters
ARCHIVE CHARACTERS
CODEWORDS
161 231 44 182 14 93 152 137 241
423 1 1 0 1 0 0 1 1 1
137 0 1 0 0 0 1 0 0 1
481 1 1 1 1 0 0 0 0 1
45 0 0 0 1 0 1 1 0 1
94 0 0 1 0 1 1 1 1 0
248 0 1 1 1 1 1 0 0 0
176 0 1 0 1 1 0 0 0 0
395 1 1 0 0 0 1 0 1 1
4. Experimental results
In order to test the application, different file types are used, so that one can remark the behavior of the archive (Rdescu and
Balasan, 2004; Rdescu and Popa, 2004; Rdescu and Harbatovschi, 2006; Rdescu and Balanescu, 2006; Rdescu and Bontas,
2008; Rdescu, 2009). The characteristics of the test files are presented in Table 2.
Table 2 Experimental files
File type File no.
Min. size
[KB]
Max. size
[KB]
Total size
[B]
Average size
[B]
XLS 6 1.08 84.5 224420 37403
DOC 3 44 77.5 199680 66560
PPS 2 111 179 296960 148480
PAS 6 0.53 1.81 6247 1041
EXE 6 11.4 83.8 195050 32508
RAR 3 16.7 100 163185 54395
BMP 5 1.24 47.5 120148 24030
WAV 4 1.16 78.9 97286 24322
DLL 6 7 69 147968 24661
MID 3 21.5 39.1 86425 28808
Next, the results of the compression are shown according to the maximum size of the dictionary. For the maximum size of 512
words and the minimum size of 256 words, the compression ratio and the packing time are shown in Table 3.
Rdescu / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 283-290
286
Table 3 Compression ratio and compression time for the parameters (256, 512)
File type
Size
[B]
Compression
ratio [%]
Compression
time [s]
Compression
speed [KB/s]
XLS 111924 49.87 42 5.22
DOC 82044 41.09 33 5.91
PPS 257445 86.69 70 4.14
PAS 3717 59.50 2 3.05
EXE 153063 78.47 45 4.23
RAR 163185 100.00 44 3.62
BMP 70209 58.44 24 4.89
WAV 96723 99.42 26 3.65
DLL 108927 73.62 33 4.38
MID 70956 82.10 20 4.22
For the maximum size of 640 words and the minimum size of 256 words, the compression ratio and time are shown in Table 4.
Table 4 Compression ratio and compression time for the parameters (256, 640)
File type
Size
[B]
Compression
ratio [%]
Compression
time [s]
Compression
speed [KB/s]
XLS 103614 46.17 42 5.22
DOC 79726 39.93 35 5.57
PPS 264199 88.97 73 3.97
PAS 3458 55.35 2 3.05
EXE 152520 78.20 45 4.23
RAR 163185 100.00 45 3.54
BMP 68126 56.70 24 4.89
WAV 96506 99.20 27 3.52
DLL 107743 72.82 34 4.25
MID 69431 80.34 20 4.22
For the maximum size of 768 words and the minimum size of 256 words, the compression ratio and compression time are
shown in Table 6.
Table 5 Compression ratio and compression time for the parameters (256, 768)
File type
Size
[B]
Compression
ratio [%]
Compression
time [s]
Compression
speed [KB/s]
XLS 97888 43.62 42 5.22
DOC 77860 38.99 36 5.42
PPS 266541 89.76 78 3.72
PAS 3414 54.65 1 6.10
EXE 151155 77.50 46 4.14
RAR 163185 100.00 48 3.32
BMP 66000 54.93 25 4.69
WAV 96292 98.98 28 3.39
DLL 106453 71.94 36 4.01
MID 67374 77.96 22 3.84
Rdescu / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 283-290
287
For the maximum size of 1024 words and the minimum size of 256 words, the compression ratio and compression time are shown
in Table 6.
Table 6 Compression ratio and compression time for the parameters (256, 1024)
File type
Size
[B]
Compression
ratio [%]
Compression
time [s]
Compression
speed [KB/s]
XLS 90852 40.48 45 4.87
DOC 76123 38.12 39 5.00
PPS 268006 90.25 87 3.33
PAS 3396 54.36 3 2.03
EXE 150764 77.30 48 3.97
RAR 163185 100.00 53 3.01
BMP 64273 53.49 27 4.35
WAV 96056 98.74 30 3.17
DLL 15846 10.71 39 3.71
MID 66594 77.05 23 3.67
The diagrams shown in Figures 14 are obtained from Tables 26.
Figure 1. Compression speed [KB/s] for different file types [extensions].
Rdescu / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 283-290
288
Figure 2. Compression ratio [%] for different file types [extensions].
Figure 3. Compression time [s] according to the maximum size of the dictionary [KB].
Rdescu / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 283-290
289
Figure 4. Compression ratio [%] according to the maximum size of the dictionary [KB].
5. Conclusion
The application described in this paper represents a good example of the way the archive performance and the waiting time are
determined in the case of the alternation of the dictionary, making thus easier to understand the dictionary-based lossless
compression. At the same time, the indexes archiving method can be very efficiently used not only by specialized archivers (Grupo
RAR, 2009), but also in programs that manage information. This method is also recommended for storing the information for long
time, where it is necessary only to check periodically the information, because of the good archiving speed.
References
Cover T.M., Thomas J.A., 1991. Elements of information theory, Wiley, New York.
Hankerson D., Harris G.A., Johnson P.D., Jr., 2003. Introduction to information theory and data compression, 2nd Ed.,
Chapman&Hall/CRC.
Nelson M., 1991. The Data compression book, IDG Books Worldwide, Inc. Foster City, CA, USA.
Murgan A.T., 1998. Principles of information theory in information engineering and communication engineering, Romanian
Academy Press, Bucharest, (in Romanian).
Pu I. M., 2006. Fundamental data compression, Elsevier.
Rdescu R., 2003. Digital information transmission practical works, Polytechnic Press, Bucharest, (in Romanian).
Rdescu R., 2003. Lossless compression methods and applications, Matrix Rom Press, Bucharest, (in Romanian).
Rdescu R., 2003. Integrated study system of lossless data compression, Symposium of Educational Technologies on Electronic
Platforms in Engineering Higher Education, Technical University of Civil Engineering of Bucharest, 9-10 May 2003, pp. 415-
422, Conspress Bucharest, (in Romanian).
Rdescu R., Blan I., 2004. Recent results in lossless text compression using the burrows-wheeler transform (BWT),
Proceedings of IEEE International Conference on Communications 2004 (COMM04), pp. 105-110, Bucharest, Romania, 3-5
June.
Rdescu / International Journal of Engineering, Science and Technology, Vol. 1, No. 1, 2009, pp. 283-290
290
Rdescu R., Popa R., 2004. On the performances of symbol ranking text compression method, Scientific Bulletin of the
Politehnica University of Timioara, Romania, Transactions on Electronics and Communications, special issue dedicated to
the Electronics and Telecommunications Symposium ETC 2004, 22-23 October 2004, Vol. 49 (63), No. 2, pp. 25-27.
Rdescu R., t. Olteanu, 2005. Text and image compression using derived LZW algorithms, EEA Revue of Electro-technique,
Electronics and Automatics, Vol. 53, No. 4, pp. 7-10, October-December (in Romanian).
Rdescu R., Ene Al., 2005. Interactive learning of lossless compression methods, Proceedings of the Symposium Educational
Technologies on Electronic Platforms in Engineering Higher Education (TEPE 2005), Technical University of Civil
Engineering of Bucharest, 27-28 May, pp. 211-218, CONSPRESS Publishing House.
Rdescu R., Harbatovschi C., 2006. Compression methods using prediction by partial matching, Proceedings of the 6th
International Conference Communications 2006 (COMM2006), pp. 65-68, Bucharest, Romania, 8-10 June.
Rdescu R., Blnescu C., 2006. Lossless text compression using the star (*) transform, Proceedings of the 6th International
Conference Communications 2006 (COMM2006), pp. 69-71, Bucharest, Romania, 8-10 June.
Rdescu R., C. Bonta, 2008. Design and implementation of a dictionary-based archiver, Scientific Bulletin, Electrical Engineering
Series C, University Politehnica of Bucharest, Vol. 70, Nr. 3, pp. 21-28.
Rdescu R., 2009. Transform methods used in lossless compression of text files, Romanian Journal of Information Science and
Technology (ROMJIST), Publishing House of the Romanian Academy, Bucharest, Vol. 12, Nr. 1, pp. 101-115.
Rao K.R., P.C. Yip (editors), 2001. The transform and data compression handbook, Boca Raton, CRC Press LLC.
Salomon D., 2007. Data Compression: The complete reference, 3
rd
Ed., Springer, Berlin-New York.
Salomon D., 2008. Concise Introduction to data compression, Springer.
Sayood K. (editor), 2005. Introduction to data compression. 3rd Edition, Morgan Kaufmann Series in Multimedia Information and
Systems.
Storer J.A., 1998. Data compression: Methods and theory, Computer Science Press.
van Lint J.H., 1992. Introduction to coding theory, Springer, Berlin-New York.
Wayner P., 1999. Data compression for real programmers, Elsevier.
www.rar.com, Grupo RAR (accessed December 2009)
Biographical notes
Radu Rdescu is an Associate Professor at the Faculty of Electronics, Telecommunications and Information Technology from the Polytechnic
University of Bucharest, Romania. In 1992, he registered as a member of the IEEE Information Theory Society. He became PhD in Electronics in
1998. He made traineeships at technical universities in Darmstadt, Germany (1997) and Lyon, France (1999-2000 and 2001-2002, specializing in
e-learning). He prepared 15 books and manuals, 14 guides for practical works, 7 new disciplines of study, and over 110 scientific papers. He
worked on 24 national and international research projects. He participated in cooperation programs with the Federal Polytechnic School in
Lausanne, Switzerland, Sdertrns University of Stockholm, Sweden, and Politecnico di Torino, Italy. In 2006, he was awarded the Creativity
Prize at the National Conference on Virtual Learning in Bucharest. His major scientific contributions are in the fields of information theory,
signal processing, e-learning systems, computer architecture, peripherals, and multimedia.
Received September 2009
Accepted October 2009
Final acceptance in revised form December 2009
International Journal of Engineering,
Science and Technology (IJEST)
2009 MultiCraft Limited. All rights reserved
No part of this journal may be reproduced by any means, nor transmitted, nor translated into any language without permission of
the publisher
Published in Nigeria by
MultiCraft Limited
14, Olajide Street, Iyana Ipaja, Lagos.