Vous êtes sur la page 1sur 79

B

No.119B
2017
Journal of Science & Technology 119 (2017)

CONTENTS

1. Experimental Studies to Verify the Effect of Chip Shrinkage Coefficient on Cutting Forces 1 16-107
and Surface Roughness in High Speed Milling of A6061 Aluminum Alloy
Pham Thi Hoa1,2, Mac Thi Bich1,2, Banh Tien Long1, Nguyen Duc Toan1*
1
Hanoi University of Science and Technology
2
Department of Mechanical Engineering

2. Dynamics and control of a four-bar mechanism with relative longitudinal vibration of the 6 16-158
coupler link
Nguyen Van Khang1, Nguyen Sy Nam2*
1
Hanoi University of Science and Technology
2
National University of Civil Engineering

3. Development of Plasma-Mig Hybrid Welding Process for Butt Joint Welding of Thick 11 16-115
Plate Steel
Lam Tran1*, Van Anh Nguyen2, Shinichi Tashiro2, Manabu Tanaka2,
Thuc Ha Nguyen1
1
Hanoi University of Science and Technology
2
Osaka University

4. The Influence of Electrical Parameters on the Penetration of Tungsten into the SKD61 16 17-009
Workpiece Surface in PMEDM using Tungsten Carbide Powder
Le Van Tao1,2,*, BanhTien Long1, Tran Xuan Thai 1, Nguyen Thi Hong Minh 1
1
Hanoi University of Science and Technology
2
Military Technical Academy

5. Modeling and Force Analysis of an Electrothermal Micro Gripper with Amplification 22 17-011
Compliant Mechanism
Dang Bao Lam1*, Nguyen Tuan Khoa1,2, Nguyen Dang Thuan3, Pham Hong Phuc1
1
Hanoi University of Science and Technology
2
Queensland Micro- and Nanotechnology Centre, Griffith University, Australia
3
Smart System Laboratory, Hanbat National University, Korea

6. Application of Sinusoidal Phase Modulation Technique for Infrared Spectrum 28 17-046


Measurement by Fourier Transform Method
Doan Giang1,2, Nguyen Van Vinh1 , Nguyen Thi Phuong Mai1, Vu Thanh Tung1*
1
Hanoi University of Science and Technology
2
Military Institute of Environmental Chemistry

7. A Method for Capturing Accuracy and Pose Repeatability of Articulated Industrial Robots 32 16-041
Duong Minh Tuan*, Le Duc Do - Hanoi University of Science and Technology OL

8. Optimal Parameters of Linear Dynamic Vibration Absorber for Reduction of Torsional 37 16-159
Vibration
Vu Xuan Truong1, 2, Khong Doan Dien2, Nguyen Duy Chinh2, Nguyen Duc Toan 3,*
1
Graduate University of Science and Technology, VAST
2
Hungyen University of Technology and Education
3
Hanoi University of Science and Technology
Journal of Science & Technology 119 (2017)

9. Effect of quenching media on distortion of C-Ring specimen made by 100Cr6 steel - 43 16-036
Simulation and Experiment OL
Tran Thi Xuan1*, Nguyen Van Tu1, Le Thi Chieu2, Vu Dinh Toai1
1
Hanoi University of Sciense and Technology
2
Vietnam Casting Metallurgy Society

10. Dynamic Analysis of Complex Composite Tubes by Continuous Element Method 48 16-026
Le Thi Bich Nam1, *, Nguyen Manh Cuong1, Tran Ich Thinh1,
Duong Pham Tuong Minh2, Le Quang Vinh3
1
Hanoi University of Science and Technology
2
Thai Nguyen University of Technology
3
Viet tri University of Industry

11. Fatigue Life Prediction under Multiaxial Variable Amplitude Loading Using a Stress 54 16-057
Invariant Based Criterion
Vu Quoc Huy*, Vu Dinh Quy, Le Thi Tuyet Nhung
- Hanoi University of Science and Technology

12. Performance Evaluation of a 2D Front Tracking Method a Direct Numerical Simulation 59 16-137
Method for Multiphase Flows
Vu Van Truong - Hanoi University of Science and Technology

13. Crystallization Kinetics of Mechanically Alloyed Al80Fe20 Amorphous Powder 66 16-121


Nguyen Thi Hoang Oanh, Tran Quoc Lap, Pham Ngoc Dieu Quynh,
Le Hong Thang, Nguyen Thi Anh Nguyet, Pham Ngoc Huyen, Nguyen Hoang Viet*
- Hanoi University of Science and Technology

14. Effects of Compaction Pressure on the Properties of In-Situ Hybrid Carbide-Reinforced 71 16-039
Copper-Based Composite Synthesized by Powder Metallurgy OL
Le Minh Hai - Hanoi University of Science and Technology
Journal of Science & Technology 119 (2017) 001-005

Experimental Studies to Verify the Effect of Chip Shrinkage Coefficient on


Cutting Forces and Surface Roughness in High Speed Milling of A6061
Aluminum Alloy
Pham Thi Hoa1,2, Mac Thi Bich1,2, Banh Tien Long1, Nguyen Duc Toan1*
1
Hanoi University of Science and Technology, No. 1, Dai Co Viet, Hai Ba Trung, Hanoi, Viet Nam
2
Department of Mechanical Engineering, Hungyen University of Technology and Education, Hungyen, Vietnam
Received: June 14, 2016; accepted: June 9, 2017

Abstract
This paper studied the relationship between the cutting force, surface roughness and chip shrinkage
coefficient through the affect of cutting parameters, i.e., cutting speed, feed rate and uncut chip thickness.
Experimental results of the chip shrinkage coefficient, cutting force and surface roughness at various cutting
parameter values for high-speed milling of A6061 aluminum alloy were presented in this study. The results
show that the cutting force and surface roughness can be derived based on the relationships with chip
shrinkage coefficient.
Keywords: High-speed milling, A6061 aluminum alloy, Chip shrinkage coefficient, Surface roughness,
Cutting force.

1. Introduction1 ensure workpiece surface roughness at a desired


quality, properly cutting parameters should be
High-speed milling (HSM) has become an
established. Several researchers studied the
innovative technique, which keeps improving
relationships of surface roughness and cutting
progressively. It is particular important to model of
parameters such as nose radius, clearance angle,
the HSM process in order to aid for prediction of the
cutting speed, feed rate, depth of cut, rake angle [7].
cutting process variables. Consequently, modeling of
Nowadays, research on surface roughness influenced
HSM process is essential for the design and
by cutting parameters is continuing for enhancing
optimization of the cutting conditions.
product quality at low cost.
Cutting force in a milling process is one of the
Another concern in modeling the HSM process
most important issues for the selection of machining
is estimation of chip geometric parameters, such as
parameters, such as feed rate and spindle speed [1].
chip thickness, chip length, etc. Since accuracy of the
Many researchers using both experiment and
model for chip parameter estimation directly affects
simulation approaches for prediction of cutting
the accuracy of the cutting force predictions, accurate
forces of HSM processes [2-4]. Compare to
chip modeling is always desired in the estimation of
experiment approach, simulation of a milling process
cutting forces, especially in micro milling [8]. Chip
using Finite Element Method (FEM) show a
shrinkage coefficient defined as the ratio of the uncut
beneficial of providing more detail information of
chip length by the actual chip length, can be a
cutting process variables, such as cutting forces, tool
prominent parameter for modeling of a cutting
stresses and temperatures [5]. Nowadays,
process. However, studies on the chip shrinkage
development of cutting force model for HSM is
coefficient have been limited in literature.
increasing interests toward higher precise cutting
force estimation and applicable for different cutting This study investigate the effect of cutting
conditions. parameters, i.e., cutting speed, feed rate and uncut
chip thickness on the chip shrinkage coefficient,
One of the major challenges of milling at high
cutting force, surface roughness. The relationship
speeds is that high-speed milling leads to the high
between this factor are found the chip shrinkage
temperature and stress growing at the interfaces of
coefficient and cutting force, surface roughness and
chip-tool or workpiece-tool resulting in unexpected
cutting parameters are found through experimental
roughness of workpiece surface finish [6]. In order to
measurement for high-speed milling of A6061
aluminum alloy.
*
Corresponding author: Tel: (+84) 988 693 047
Email: toan.nguyenduc@hust.edu.vn
1
Journal of Science and Technology

2. Experimental setup device.


2.1 Workpiece material
The workpiece used in this study is aluminum
alloy A6061, which has the hardness of 97HB. The
chemical composition of workpiece material is
represented in Table 1. Several workpieces used for
experiment are shown Figure 1. The workpiece
dimensions are 70x30x70 (mm).
Table 1. Chemical composition of the workpiece material
(%).
Si Fe Cu Mn Mg
0,4-0,8 0,3 0,05-0,3 0,10 0,8-1,2 Fig. 2. Surface roughness tester
Cr Zn Ti Al The Sartorius Volume Comparator (S224-1S)
0,05-0,30 0,25 0,15 remaining scale is used to determine the weigh of the chip after
cutting. The scale parameters are as follows: capactiy
of 220 gr, readability of 0.1 mg.
The chip shrinkage coefficient (K) can be
calculated by following formula [11]:
1000.Q
K (1)
.L.S.t
where Q is weight of the chip (gr), is material
density (g/cm3), l is chip length (mm), S and t are
feed rate (mm/rev) and uncut chip thickness (mm),
respectively.

Fig .1. The workpieces used in experiments 3. Design of experiment


The effects of cutting speed, feed rate, uncut
2.2 Milling experiment
chip thickness, on chip shrinkage coefficient, force
All the experiments are performed on a HS cutting, surface roughness are examined using a
Super MC500 high-speed milling machine three-factor/three-level full factorial design [10]. The
maximum feed rate of 30 m/min, maximum spindle range of each factor is set at three different levels as
speed of 30.000 rpm, travel distances of the shown in Table 2. Figure 3 shows the experimental
operating platform in the X, Y and Z directions of set-up of the milling process.
500 mm, 400 mm and 300 mm, respectively. Dry
milling condition with carbide insert cutting tool
(APMTT1604PDTR TC300) and diameter of 40 mm
is used for milling.
2.3 Measurement equipment
The cutting forces are measured by force
measurement device Kisler, which is equipped with a
force sensor (Kisler 9257B). The maximum load
capacities of the device in X, Y and Z directions are
1500N, 1500N and 5000N, respectively. The
sensitivity of sensor in X, Y and Z directions are 7.39
pC/N 7.39 pC/N and 3.72 pC/N, respectively. The
measured data is collected by an acquisition system
using DASYlab 10.0 software.
The surface roughness of the machined Fig. 3. Experimental set-up
workpiece is measured by a surface roughness tester
(Mitutoyo SJ400). The roughness values are in m.
Figure 2 shows the surface roughness measuring

2
Journal of Science and Technology

Table 2. Cutting parameters for the experiment F bV


1
b2 b3 b4
f t (3)
Level
No. Parameter Unit Level 1 Level 2
3 Ra c1V c2 f c3 t c4 (4)
V
1 (cutting m/min 1000 1130 1256 where ai, bi, ci (i = 1...4) are the constants to be
speed) determined. Using curve fitting tool in Minitab17,
f those constants can be determined as shown in Table
2 mm/min 800 1350 1800
(feed rate) 4.
t
3 (uncut chip mm 0,5 1,0 1,5 Table 4. Fitted constants obtained by surface fitting
thickness) method
Table 3. Experimental results i 1 2 3 4
a 0,889866 0,0090714 0,0339467 -0,1402635
V f F Ra 0,092838
No t (mm) K b 1264,012 -0,36893 0,047466
(m/ min) (m/min) (N) (m)
c 5820,781 -1,378843 0,04978739 0,10138765
1 1000 800 0,5 1,294 135,71 0,64 Figures 4-6 show K, F, and Ra as a function of
2 1256 800 0,5 1,304 126,98 0,39 the cutting parameters, i.e., V, f and t, respectively,
3 1000 1800 0,5 1,322 120,88 0,60 obtained using equations (2), (3) and (4) with the
4 1256 1800 0,5 1,370 111,73 0,31 constants in Table 4. Figure 4 shows that increasing
cutting speed leads to the increase of K. On the other
5 1000 800 1,5 1,144 146,36 0,53
hand, K is decreased with increasing depth of cut.
6 1256 800 1,5 1,102 128,42 0,48 This figure also shows that t has a great influence on
7 1000 1800 1,5 1,137 139,71 0,55 K, while the effect of f on K is minor. Besides,
8 1256 1800 1,5 1,133 117,38 0,35 cutting speed increases leading to the decrease in
9 1130 1350 1 1,236 125,46 0,44 contact area between the chip and the front of the
10 1130 1350 1 1,236 124,82 0,44 tool. Consequently, chip shrinkage coefficient is
1130 1350 increased [11]. In order to obtain the optimal cutting
11 1 1,236 126,72 0,44
parameter values for minimizing K, MAPLE
4. Results and discussions software is utilized based on NLPSolve command.
Optimal values for V, f and t are 1000 m/min, 800
4.1 Influences of V, f and t on the K, F and Ra mm/min, and 1.5 mm, respectively.
Table 3 shows the experiment results of K, F, Ra as a From Figure 5, increasing V, f or t all reduces F.
function of V, f and t. Using curve-fitting tool, the This is because at high-speed cutting, the generated
relationship between K, F, Ra dependence on V, f and heat can soften the materials thus decreasing cutting
t is established. That relationship is described by the forces [11]. Using NLPSolve command, the optimal
following equation (2), (3),(4). parameters of V, f and t for the objective function of
minimizing F are also found equal to 1256 m/min,
K a1V a2 f a3 t a4 (2)
1800 mm/min, and 0.5 mm, respectively.

Fig. 4. The relationship between K and cutting parameters V, t and f. a) Fixed V, b) Fixed f, c) Fixed t

3
Journal of Science and Technology

Fig. 5. The relationship between F and cutting parameters V, t and f. a) Fixed V, b) Fixed f, c) Fixed

Fig. 6. The relationship between Ra and cutting parameters V, t and f. a) Fixed V, b) Fixed f, c) Fixed t

Fig. 7. The relationship between F and K Fig. 8. The relationship between Ra and K
Figure 6 indicates that Ra increases with f and t on the relationship between F and K as well as
increasing f and t but reduces with increasing V. This Ra and K, the five-level full factorial design was
is because under high-speed cutting, the built up edge assigned for cutting speed (V) of 1000, 1064, 1128,
phenomenon would disappears leading the reduction 1192, 1256 m/min; feed rate (f) of 800, 1050, 1300,
of surface roughness [11]. Similar to K and F, the 1550, 1800 mm/min and uncut chip thickness (t) of
optimized values of V, f and t for minimizing Ra are 0.5, 0.75, 1.00, 1.25, 1.50 mm, respectively.
1256 m/min, 1800 mm/min and 0.5 mm, respectively.
It is seen that minimum values of F and Ra are
4.2 The relationship between F and K, Ra and K 117N and 0.403m, respectively, which are all
obtained at K = 1.312. When K is equal to 1.154,
This section analyzes the relationship between F
maximum values of F and Ra are obtained equal to
and K as well as, Ra and K based on Eqs. (2-4). By
eliminating V, f and t from Eqs. (2-4), the relationship 146N and 0.64m, respectively.
between F and K, Ra and K are found. As shown in The figures 7 and 8 also summarize the affected
Figures 7-8, In order to verify the effect of various V, trend of cutting parameters i.e., cutting speed, feed

4
Journal of Science and Technology

rate and uncut chip thickness on the cutting force (F) prediction of cutting forces in end milling with
and surface roughness (Ra) related with chip application to cor-nering cuts, International Journal of
shrinkage coefficient (K) as discussing in detail on Machine Tool Design and Research 22 (1982) 722.
section 4.1. From those figures, the optimal cutting [3] H.J. Fu, R.E. DeVor, S.G. Kapoor, A mechanistic
parameters can be obtained by minimizing cutting model for prediction of the force system in face
force (F), surface roughness (Ra) and chip shrinkage milling oper-ation, ASME Journal of Engineering for
coefficient (K). In order to minimize the F and Ra, the Industry 106 (1984) 8188.
maximum of cutting speed (V) and minimum of feed [4] W.P. Wang, Solid modelling for optimizing metal
rate (f) also uncut chip thickness (t) should be set. removal of three-dimensional end milling, Journal of
However, the decreasing of uncut chip thickness (t) Manufac-turing Systems 7 (1984) 5766.
will increase the chip shrinkage coefficient (K)
[5] T. O. Zel, T. Altan, Modeling of high speed
therefore (t) will be chosen based on the productivity machining processes for predicted tool forces stresses
of manufacturing process. and temperatures using FEM simulations, in:
5. Conclusions Proceedings of the CIRP International Workshop on
Modeling of Machining Oper-ations, Atlanta, GA,
This paper presents an experimental study on (1998) 225234.
relationship between cutting force, surface roughness [6] Tugrul Ozel, Taylan Altan, Process simulation using
and chip shrinkage coefficient when high speed finite element method - prediction of cutting forces,
milling of A6061 aluminum alloy. Some conclusions tool stresses and temperatures in high-speed flat end
are given as follows: milling, International Journal of Machine Tools &
Manufacture 40 (2000) 713738.
1. The relationship between the cutting force, surface
roughness and chip shrinkage coefficient through [7] Mehmet Alper, lhan ASLTRK, Effects of Cutting
cutting parameters e.g., cutting speed, feed rate, uncut Tool Parameters on Surface Roughness, International
chip thickness are explicitly described by Refereed Journal of Engineering and Science, 4(8)
(2015) 15-22.
mathematical functions.
[8] Ali Mamedo, Ismail Lazoglu, An evaluation of micro
2. The optimal cutting parameters for chip shrinkage milling chip thickness models for the process
coefficient, cutting force and surface roughness can mechanics, the International Journal of Advanced
be found by maximizing the cutting speed (V) and Manufacturing Technology, (2015) 1-7
minimizing the feed rate (f), which are useful for
[9] Swan MS. Incorporation of a general strain-to-failure
practical milling of A6061 aluminum alloy.
fracture criterion into a stress-based elasticity model
Acknowledgements: This research is funded by through a time-to-failure softening mechanism. M.Sc.
Vietnam National Foundation for Science and Thesis in Mechanical Engineering University of Utah,
Technology Development (NAFOSTED) under grant USA.(2012)
number 107.02-2016.01. [10] K. Venkata, M. Krishnam, G. R. Janardhana,
Optimization of Cutting Conditions for Surface
References Roughness in CNC End Milling, 12( 3) (2011) 383
[1] Wu Baohai, Yan Xue, Luo Ming, Gao Ge, Cutting 391.
force prediction for circular end milling process, [11] Banh Tien Long, Tran The Luc and Tran Sy Tuy.
Chinese Journal of Aeronautics, 26 (4) (2013) 1057 Metal Cutting Principles, 2nd Ed, Science and
1063. Technics Publishing House, (2013) (In Vietnamese)
[2] W.A. Kline, R.E. DeVor, J.R. Lindberg, The

5
Journal of Science & Technology 119 (2017) 006-010

Dynamics and Control of a Four-Bar Mechanism with Relative


Longitudinal Vibration of the Coupler Link
Nguyen Van Khang, Nguyen Sy Nam*
1
Hanoi University of Science and Technology No. 1, Dai Co Viet Str., Hai Ba Trung, Ha Noi, Viet Nam
2
National University of Civil Engineering , No. 55, Giai Phong Str., Hai Ba Trung, Ha Noi, Viet Nam
Received: October 12, 2016; accepted: June 9, 2017

Abstract
In the mechanisms and machines operating at high speeds, the elastic vibration of links is inevitable. In this
paper the dynamic modeling and controller design for a flexible four-bar mechanism are studied. The fully
coupled non-linear equations of motion are obtained by using the Lagranges equations with multipliers for
constrained multibody systems. The resulting differential-algebraic equations are solved using numerical
methods. A simple PD controller is designed to reduce the influence of the elastic link on the desired motion.
Keywords: Dynamic analysis, control, Four-bar mechanism, Ritz-Galerkin method, Vibration.

*
1. Introduction A fourbar mechanism OABC is shown in
Figure 1. The mechanism consists of the rigid crank
Traditionally, dynamic analysis and control of
OA of length l1, the flexible rod AB of length l 2 and
mechanisms have been based on the assumption that
the rigid rod BC of length l 3, the distance OC is l0,
the links behave as rigid bodies. The demand for high
speed lightweight machinery requires a redesign of is the external torque acting on the crank joint. e1(0)
the current mechanisms. Unfortunately, reducing the and e(20) are the unit vectors of the fixed coordinate
weight of four-bar mechanisms and/or increasing
system Ox0y0. e1 and e 2 are the unit vectors of the
their speed may lead to the onset of elastic
oscillations, which causes performance degradations reference coordinate system Axy which is rotated
such as misfeeding in the case of the card feeder with an angle 2 to the fixed one. Three
mechanism in Sandor et al [1]. Therefore, the variables 1, 2 and 3 are the angles between the x0-
dynamic analysis and control vibration of flexible axis and crank OA, the x0-axis and flexible link AB,
mechanisms are required. Although dynamic analysis the x0-axis and output link BC, respectively.
of flexible mechanisms has been the subject of
numerous investigations [1-6], the control of such u(l2,t)
systems has not received much attention [2-4]. Most l2 B x
of the work available in the literature which deals y
u(x,t)
with vibration control of flexible mechanisms employ x
an actuator which acts directly on the flexible link. y0 e2
A 2
The effect of the control forces and moments on the e1
overall motion is neglected. An alternative method
would be to control the vibrations through the motion e(20) 1 3
x0
of the input link. O C
( 0)
e 1
The current study deals with the control of a
Fig. 1. Diagram of the four-bar mechanism
four-bar mechanism with a flexible coupler link. An
actuator is assumed to be placed on the input link
which applies a control torque. A simple PD control It should be noted that in general the constraint
is designed which requires measurements of the equations depend on the elastic deformations. In this
position and angular velocity of the input link only. study, the transverse deformations are neglected.
Therefore, the constraint equations here depend on
2. Equations of motion of the four-bar mechanism
the longitudinal deformations:
f 1 l1 cos 1 l2 u( l2 ,t ) cos 2 l3 cos 3 l0 0
* f 2 l1 sin 1 l2 u( l2 ,t ) sin 2 l3 sin 3 0
Corresponding author: Tel.: (+84) 977.244.292
Email: nam.nguyensy@gmail.com (1)

6
Journal of Science & Technology 119 (2017) 006-010

Ritz-Galerkin method is applied to the coupler The strain energy according to Reddy [9] is given by
link. The axial deformation of the coupler link is l 2
u N N
2

EA dx EA kij qi q j
written as [7, 8] 1 1
(9)
N
2 0
x 2 i 1 j 1

u( x ,t ) X i ( x ).qi ( t ) (2) l2
i 1
where k ij X X dx
i j
where qi(t) are the modal coordinates and Xi(x) are 0

the mode shapes of the rod. The boundary conditions


The Lagrange's equations for constrained
are as follows:
holonomic systems are [8]:
u (l 2 , t )
u(0,t) = 0; EA 0 (3) d T T 2 f
x
dt j

j




k k
j Qj

(10)
j k 1
The mode shapes are given as [7]:
where j are the generalized coordinates which
2i 1 x include rigid body coordinates 1, 2, 3 as
X i ( x ) Bi sin
(4)
2 l2 well as elastic modal coordinates qj; fk are the
constraint equations, 1 and 2 are Lagrange
To simplify the equation (4) take Bi = 1. multipliers; and Qj are the generalized forces. By
substituting equations (1, 8, 9) into equation (10), we
The kinetic energy of the four-bar mechanism
obtained the equations of motion of the system as:
shown in Figure 1 is given by
l1l 22
1 1 1
l2 I
l12 l 2 1 2 cos( 1 2 )
I O12 I C 32 rM2 dx (5)
O
T TOA TBC TAB 2
2 2 20 N N
l12 cos 1 2 C q i i l1 sin1 2 C q i i
i 1 i 1
where IO and IC are the mass moments of inertia of
l l 2 N
the input and output links with respect to the joint 1 2
22 sin( 1 2 ) l1 22 sin1 2 Ci qi (11)
axes, respectively, is mass per unit length of the 2 i 1
N
2 l1 2 cos1 2 C q l
coupler link, and rM is the position vector of a point sin 1.1 l1 cos 1.2
i i 1
M on the coupler link given as i 1

rM rA x u e1 l1l 22 l23
N
(6) 1 cos1 2 l11 cos1 2 Ci qi 2
2 i 1 3
The coordinates of the point M in the fixed N N l l 2
N
coordinate system are given as:
2 2 Di qi mij qi q j 1 2 12 sin1 2

i 1 2
i 1 j 1
xM l1 cos 1 ( x u ) cos 2 N N N N
y M l1 sin 1 ( x u ) sin 2
(7) l112 sin1 2 Ci qi 2 2 Di q i

mij q i q j

i 1 i 1 i 1 j 1
By differentiating the equations (7) combined l 2 u B sin 2 .1 l 2 u B cos 2 .2
with equation (4), and then substituting into the (12)
equation (5), the kinetic energy is obtained as
I C3 l3 sin3 1 l3 cos3 2 0 (13)
l l 2
1
1

T I O l12 l2 12 I C 32 1 2 1 2 cos1 2 N
l11 sin 1 2 C i mij q j
2 2 2
l23 2 2 N N N
2 2 2 Di qi mij qi q j
j 1

6 2 i 1 N
l1 12 cos 1 2 C i 22 Di mij q j
i 1 j 1
(14)
N N N

2
mij qi q j l112 cos1 2 Ci qi
i 1 j 1 i 1 N
j 1

N EA k ij q j 1 cos 2 2 sin 2 i
l11 sin1 2 Ci qi (8) j 1
i 1
l2 l2 l2 where i = 1,2,..,N;
where Ci X i dx , Di xX i dx ; mij X i X j dx when i 2 k 1, k 1,2 ,...
1
0 0 0 i
1 when i 2 k , k 1,2 ,...

7
Journal of Science & Technology 119 (2017) 006-010

3. Dynamic analysis clearer, since the larger torque causes larger elastic
deformations.
A set of (3+N) differential equations (11-14) and
two algebraic constraint equations (1) are the motion Table 1. Parameters of four-bar mechanism
equations of mechanism with an elastic link.
Constant Description Value
Therefore, we have (5+N) differentialalgebraic
[unit]
equations with (5+N) variables given as
l0[m] Length of the ground link 0.4064
M( s ,t)s sT (s,t ) p1 (s, s, t ) (15) l1[m] Length of the input link 0.0635
f(s,t) = 0 (16) l2[m] Length of the coupler link 0.3048
l3[m] Length of the output link 0.3048
where s = [1 2 3 q1 qN 1 2]T ; is vector of
Lagrange multipliers; f is the vector of constraint IO[kgm2] Moment of inertia of the 7.466
equations. In this paper, the Lagrange multipliers input link x10-6
partition method is used to solve the system of IC[kgm2] Moment of inertia of the 2.002
differential-algebraic equations (15) and (16) output link x10-3
numerically. Some other algorithms can be found in [kg/m] Mass per unit length 0.2237
Khang [8]. E[N/m2] Modulus of elasticity 2.06x1011
For comparison purposes, another set of
A[m2] Cross-sectional area of 8.19
simulations is carried out by assuming that the
the coupler link x10-6
coupler link behaves as a rigid body (i.e., neglecting
m1[kg] Mass of the input link 0.0142
the longitudinal deformations). The values of the
parameters used in the simulations are given in Table m2[kg] Mass of the coupler link 0.0682
1. The torque on the input link is given by:
m3[kg] Mass of the output link 0.0682
sin( 2t / Tm ) t Tm
(t ) 0 (17)
0 t Tm

where 0 is the peak torque and T m is the duration of


the torque.
The initial conditions are selected as follows:
angle of input 10 = /2, angular velocity of input link
10 0 , elastic deformations qi0 = 0 and elastic
deformations velocity qi0 0 . By using the Newton -
Raphson method for solving constraint equations (1a,
b) with the initial conditions as above, we obtain the
initial positions of the other links as:
20 0.6752 rad , 30 2.1564 rad , 20 0 , 30 0 . Fig. 2. Crank angle, 0 = 0.03 Nm.
rigid, flexible.
Figures 2 and 3 compare responses of rigid and
flexible mechanism with a peak torque magnitude of
0 = 0.03 Nm and Tm = 1s. As mentioned in section 1,
because of the fully coupled nature of the equations,
the rigid body coordinates (e.g., input and output link
angular displacements) are affected by the elastic
deformation of the coupler link. However, this effect
is negligible when the peak of torque is small.
Another set of simulations is carried out with a
peak torque magnitude of 0 = 0.1 Nm, Tm = 1s and
the simulations are performed during the period from
0 to 3s. The responses for the flexible and rigid
models of the four-bar mechanism are shown in
Figures 4 through 6. The effect of flexibility is now Fig. 3. Output angle, 0 = 0.03 Nm.
rigid, flexible.

8
Journal of Science & Technology 119 (2017) 006-010

4. Control flexible four - bar mechanism


The governed equations are highly non-linear
differential equations. Most control designs require a
linear unconstrained model. In this case a
straightforward linearization is not possible since it
may be difficult to find an operating point. However,
the equations can be linearized around a rigid body
trajectory. Since the main goal of the controller is to
suppress vibrations, a rigid body trajectory can be
used as the nominal trajectory and the deviations
from this can be assumed small. The real trajectory of
flexible mechanism can be obtained as follows

1 1d y1

2 2 d y2
Fig. 4. Crank angle, 0 = 0.1 Nm (18)
rigid, flexible. 3 3 d y3
y3i qi
where 1d, 2d, 3d represent the rigid mechanism
trajectory; y1, y2, y3 are deviation of flexible trajectory
versus rigid trajectory. It is assumed that y1, y2, y3 ,
y3+i are small.
In order to investigate whether it is possible to
suppress vibrations of the flexible link by a control
torque applied to the input link, a simple control
strategy, in particular PD controller is used. The
target of the controller is that y1, y2, y3 , y3+i approach
zero when time approaches infinity (or large
enough). The control torque applied to the input link
is given by:
c K P y1 K d y 1 (19)

where Kp, Kd are the PD controller gains,


Fig. 5. Output angle, 0 = 0.1 Nm. y1 1 1d ; y 1 1 1d . Thus, the total torque
rigid, flexible. acting on the input link is + C.

Fig. 6. Longitudinal deformation of flexible coupler Fig. 7. Crank angle with PD controller, 0 = 0.1 Nm
link, 0 = 0.1 Nm. rigid, flexible.

9
Journal of Science & Technology 119 (2017) 006-010

numerically to simulate the system behavior. A


simple PD controller was designed which does not
require measurement of the elastic deformations. The
controller has been shown to be efficient in
suppressing the vibrations of the flexible link as well
as controlling the rigid body motion.
References
[1] Sandor, G. N. and Erdman, A. G. (1984) Advanced
Mechanism Design: Analysis and Synthesis.
Englewood Cliffs, NJ: Prentice-Hall.
[2] Sung, C. K. and Chen, Y. C (1991) Vibration control
of the elastodynamic response of high-speed flexible
linkage mechanisms. ASME Journal of Vibration and
Acoustics 113, 14-21.
Fig. 8. Output angle with PD controller. 0 = 0.1 Nm.
rigid, flexible. [3] Beale, D. G. and Lee, S. W. (1995) The applicability
of fuzzy control for flexible mechanisms. ASME
Design Engineering Technical Conferences DE-Vol.
84-1, 203-209.
[4] Liao, W. H. Chou, J. H. and Horng, I. R (1997).
Robust vibration control of flexible linkage
mechanisms using piezoelectric films. Smart Materials
and Structures 6, 457-463.
[5] Schwab, A. L. and Meijard, J. P. (1997) Small
vibrations superim - Posedon non-linear rigid body
motion. Proceedings of the 1997 ASME Design
Engineering Technical Conferences, Sacramento,
CA,1-7.
[6] Yigit, A. Scott, R.A. and Ulsoy, A.G (1988), Flexural
motion of aradially rotating beam attached to a rigid
body. Journal of Sound and Vibration 121, 201 - 210.
[7] Khang, N.V (2005), Engineering vibrations (in
Fig. 9. Longitudinal deformation of flexible coupler Vietnamese), Science and Technics Publishing House
link with PD controller, 0 = 0.1 Nm.
[8] Khang, N.V (2017), Dynamic of multibody system (in
rigid, flexible. Vietnamese, 2. Edition) Science and Technics
Now, we will suppress vibrations in case the Publishing House.
peak torque magnitude of 0 = 0.1 Nm and Tm = 1s [9] Reddy, J.N (2002), Energy Principles and Variational
(e.g., Figures 4 6). The parameters used in the Methods in Applied Mechanics, John Wiley and Son,
simulations are given in Table 1. The controller gains ISBN: 978-0-471-17985-6.
are chosen as Kp = 0.5, Kd = 0.2. The calculating
[10] Bajodah, A. H. Hodges, D. H. and Chen, Y. H (2003)
results are shown in Figs 7 9. These New form of Kanes equations for constrained
results show that controller is able to suppress the systems. Journal of Guidance, Control, and Dynamics
vibrations and control the link angular motions. In 26:7988.
Figure 9 the longitudinal deformation is suppressed
within 0.85 s.
5. Conclusions
In this study, dynamic modeling and control of a
four-bar mechanism with flexible coupler link has
been investigated and it was assumed that there is
only axial deformation. The non-linear equations of
motion are obtained through a constrained
Lagrangian approach and described by a set of
differential-algebraic equations. The governed
differential-algebraic equations are solved

10
Journal of Science & Technology 119 (2017) 011-015

Development of Plasma-Mig Hybrid Welding Process for Butt Joint


Welding of Thick Plate Steel
Lam Tran1*, Van Anh Nguyen2, Shinichi Tashiro2, Manabu Tanaka2, Thuc Ha Nguyen1
1
Hanoi University of Science and Technology - No. 1, Dai Co Viet Str., Hai Ba Trung, Ha Noi, Viet Nam
2
Osaka University, 11-1 Mihogaoka, Ibaraki, Osaka, 567-0047, Japan
Received: October 06, 2016; accepted: June 9, 2017

Abstract
Welding is present in all industrial sectors as a necessary technological process. One of the principal
directions for the progress of the welding is the development of hybrid welding processes. Plasma-MIG
hybrid welding was developed several decades ago. Nowadays, it becomes a bright technology in materials
processing. One of the versions of the PlasmaMIG processes is basically a combination of a keyhole
plasma arc with a MIG arc in order to deliver greater welding speeds, deeper weld penetration, and reduced
heat input. In this paper, plate to plate butt joint welds were conducted on mild steel plates and the aims of
this research is developed a Plasma-MIG hybrid welding process for single pass welding of thick steel
plates. As a result, it was found that the successful single-sided welding in one pass with complete
penetration and improve the weldability of welding joints in comparison with MIG welding process.
Keywords: Plasma keyhole, GMAW, Plasma-MIG hybrid welding, Hybrid arc.

1. Introduction This process called Super-MIG, which was recently


validated as a viable alternative to GTAW in tube
Welding* is present in all industrial sectors as a
joining application [2]. Jeff Palms has established the
necessary technological process. One of the principal
hybrid Plasma-GMAW (Super-MIG) weld process as
directions for the progress of the welding is the
a viable process for welding Titanium. Compared to
development of hybrid welding processes. Plasma-
traditional GMAW welding, this hybrid process
MIG hybrid welding was developed several decades
dramatically increases welding speed and penetration
ago and nowadays, it becomes a bright technology in
(both more than double) while producing weld joints
materials processing. The hybrid welding technology
with minimal spatter and with superior profile and
combines the deep penetration characteristics of
mechanical properties. Weld joints observed had a
Plasma Arc Welding (PAW) with the high weld
high degree of visual and metallurgical quality, free
deposition rates of MIG. The arc of PAW and the arc
of porosity, cracks and contamination [3]. Emel
of MIG are quite different welding heat sources but
Taban, Erdinc Kaluc and Alfred Dhooge have
both work under a gaseous shielding atmosphere at an
investigated the properties of hybrid (Plasma + Gas
ambient pressure, making it possible to combine these
tungsten arc) welded joints of modified 12% Cr
heat sources in a unique welding technique. The
stainless steel conforming to EN 1.4003 and UNS
combination of the two processes can deliver greater
S41003 steels using austenitic stainless steel type of
welding speeds under variable root opening
consumables such as 309 and 316 [4]. Welding
conditions, deeper weld penetration, and reduced heat
Solutions from the North American also reported that
input. In turn, lower heat input results in a narrower
Plasma-MIG hybrid welding process was able to
heat-affected zone (HAZ) and less distortion [1].
replace SAW with the hybrid process enabled the
Recent years, the interest has been increased in 100% joint penetration requirement to be met with
applying PAW process in industry due to the higher single sided welding in one pass on heavy thickness
welding speeds providing improved productivity and plate of mild steel [1].
producing welds with high penetration/width ratios.
However, the explanations given above and from
Since, Plasma Laser Technologies (Yokneam, Israel)
the literature on the Plasma-MIG hybrid welding are
developed a new hybrid Plasma-MIG welding
still indistinct and are little quantified, as well as
process that is helping meet the challenges of
mostly dating back to the 1970s and 1980s when the
increased demand, faster cycle times and more
technology available was unable to make the process
efficient manufacturing for tube joining application.
viable for the industry at that time. Therefore, there is
a need to develop new technologies and knowledge
* for this process. In this paper, plate to plate butt joint
Corresponding author: Tel: (+84)983.077.322
welds were conducted on mild steel plates and the
Email: lam.tran@hust.edu.vn
11
Journal of Science & Technology 119 (2017) 011-015

aims of this research is developed a Plasma-MIG constant current characteristics and base metal. The
hybrid welding system for single pass welding of configuration of the torches were set up based on the
thick steel plates. As a result, it was found that the distance and angle between the crossing positions of
successful single-sided welding in one pass with the electrodes-axis and surface on base metal shown
complete penetration and improve the weldability of as Fig.2, thus the leading Plasma and trailing MIG
welding joints in comparison with MIG welding were configured. Concerning the arc ignition steps,
process. In addition, the metal transfer from MIG Plasma arc was started firstly and weld pool was
wire was imaged by using high speed video camera formed on the surface of base metal, and then MIG
(HSVC) in order to know the interaction between the arc was started. In addition, metal transfer from MIG
plasma arc flow and the MIG arc promotes wire wire filler was imaged by using a high speed video
heating and current transfer at the anode spot (at the camera for evaluating the condition under which the
end of the MIG wire) where the molten weld metal metal transfer was stabilized as shown Fig.3.
droplets form and subsequently detach.
In order to develop a Plasma-MIG hybrid
2. Experimental procedure welding process for single pass welding of thick steel
plates, plate to plate butt joint welds were conducted
The classical representation of Plasma-MIG
on mild steel plates by varying experimental
hybrid welding process was shown in Fig.1. This
parameters such as the plate specifications including
Plasma electrode establishes an arc at the leading
the thick and initial position of base metal plate,
position of the welding process, and a keyhole is
plasma current, the energy input rate of MIG process,
created within the base material by the plasma arc.
the wire feed rate, welding speed. This paper also
GMAW follows and operates typically in the
presents an example of experimental results in which
conduction welding mode to fill the void created by
the weld has complete penetration, very good
the plasma arc. The interaction between the plasma
metallurgical, without porosity, cracks, and undercuts
arc flow and the GMAW arc promotes wire heating
in comparison with MIG welding process. The
and current transfer at the anode spot (at the end of
parameters of tests were shown in Table.1 and the
the GMAW welding wire) where the molten weld
initial plate positioning for hybrid welding of plate to-
metal droplets form and subsequently detach. The
plate butt joints was shown in Fig.2. The bead
resultant magnetic force F, shown in Fig.1, occurs as
appearance and the bead cross section of Plasma-
a result of the interaction of the electric currents
MIG hybrid welding and MIG welding was observed
passing through the two electrodes [1].
using the conventional techniques of macrograph on
The experimental apparatus consists of a Plasma cross-sections taken from the welded test plates (300
torch, a MIG torch, MIG power source with the mm bead). For each test plate, a cross-sections was
constant voltage characteristics, plasma source with cut approximately in the middle of the bead.

Table.1 Experimental parameters


Parameters Value Unit
Base metal Mild steel; Size: 300x50x9 (mm)
MIG welding wire JIS Z3312; Wire diameter: 1,2 mm
Groove angle 45 Degree
Plasma welding current 140 Ampere
MIG welding current 105 Ampere
MIG welding voltage 21 Voltage
Distance between the tip and base metal for MIG 20 mm
Arc length of Plasma 5 mm
Welding speed 12 cm/min
Wire feed rate 240 cm/min
Band pass filter 960 nm
Frame rate 3000 fps
Shielding gas for MIG and Plasma welding Argon
Gas flow rate 7.5 l/min
Back shielding gas flow rate 7.5 l/min
The distance between two torches 20 mm
The angle between two torches 10 Degree

12
Journal of Science & Technology 119 (2017) 011-015

Fig.1 Principle of Plasma MIG hybrid welding. 1Work-piece; 2Plasma jet; 3Plasma nozzle; 4Melting
metal; 5Plasma arc electrode axis; 6Wire axis; 7Angle between electrodes axes; 8-Tungsten electrode; 9
Consumable electrode (wire); 10MIG arc; 11Low temperature plasma; 12Wire current Iw direction; 13
Plasma current Ip direction; 14Magnetic forces F applied to plasma arc; 15Magnetic forces applied to MIG
arc

Unit: mm and dgree


Fig.2 The schematic illustration of the configuration of Fig.3 The photograph of the experimental setup
Plasma and MIG torch including Plasma torch, a MIG torch and the high
speed video camera (HSVC).
(a) (b)

Droplet Droplet

Fig.4 Observation of weld pool and droplet during welding. (a) The metal transfer of MIG welding
imaged by HSVC; (b) The metal transfer of Plasma-MIG hybrid imaged by HSVC.

Fig.5 The bead appearance of MIG welding (Welding current: 105 A; welding voltage: 21V; wire
feed rate: 240 cm/min; welding speed: 12 cm/min)

13
Journal of Science & Technology 119 (2017) 011-015

Fig.6 The bead appearance of Plasma-MIG hybrid welding

the torch axis during high-speed welding. The


resultant effect is a substantial increase in the plasma
arc rigidity and stability leading to a substantial
increase of penetration depth and welding speed.
Fig.5 presents the bead appearance observed by
using an optical microscope of MIG welding. The
bead with not completed weld bead at the top surface
and uncomplete joint penetration at the bottom
surface were found because of the heat input of MIG
welding process is not enough. Fig.6 presents the
bead appearance observed by using an optical
microscope of Plasma-MIG hybrid welding. The bead
Fig.7 The bead cross-sections of MIG welding with good quality at the top surface and complete
(Welding current: 105 A; welding voltage: 21V; joint penetration with a single pass at the bottom
wire feed rate: 240 cm/min; welding speed: 12 surface were found because of the heat input of
cm/min) Plasma-MIG hybrid welding process is enough.
The cross section of MIG weld illustrated in Fig.7
shows a poor metallurgical integrity of the weld with
undercuts at the top surface of the weld and lack of
fusion at the bottom surface of the weld. The weld
has uncomplete penetration. The cross section of
Plasma-MIG hybrid weld illustrated in Fig.8 shows
very good metallurgical integrity and consistency of
the weld without porosity, cracks, and undercuts. The
weld has improved wettability and complete
penetration [5].
Fig.8 The bead cross-section of Plasma-MIG
hybrid welding 4. Conclusions

3. Results and discussion The paper discussed the ability of Plasma-MIG


hybrid welding process for butt joint welding of thick
The metal transfer of both Plasma-MIG hybrid plate steel. The following conclusions are deduced
welding and MIG welding was imaged using HSVC from this study:
as shown in Fig.4. As seen in the figure the
interaction between the plasma arc flow and the MIG 1) The metal transfer of both Plasma-MIG
arc promotes wire heating and current transfer at the hybrid welding and MIG welding was imaged using
anode spot (at the end of the MIG welding wire) HSVC. As seen in the figure the interaction between
where the molten weld metal droplets form and the plasma arc flow and the MIG arc promotes wire
subsequently detach. The resultant magnetic force F, heating and current transfer at the anode spot (at the
shown in Fig.1 causes deflection of the plasma arc end of the MIG welding wire) where the molten weld
toward the front of the weld pool, thus compensating metal droplets form and subsequently detach. The
for the plasma arcs natural tendency to trail behind resultant effect is a substantial increase in the plasma

14
Journal of Science & Technology 119 (2017) 011-015

arc rigidity and stability leading to a substantial [2] Plasma Laser Technologies. Hybrid Plasma/MIG
increase of penetration depth and welding speed. Welding Reduces Welding Time. Power Engineering,
ProQuest Central. 112(8), (2008) 75-77.
2) The Plasma-MIG hybrid welding technology
is capable of achieving single-sided complete joint [3] Jeff Palms. Joining Methodologies for Titanium Alloys
- Hybrid Plasma and MIG combination. Advanced
penetration welds of the butt-joint welding of thick Industrial Technology, (2004).
plate steel with good weld shape, dimensions, and
metallurgical integrity in comparison with MIG [4] Emel Taban, Erdinc Kaluc and Alfred Dhooge. Hybrid
welding process. (plasma + gas tungsten arc) weldability of modified
12% Cr ferritic stainless steel. Materials and Design. 30
Acknowledgements (2009) 42364242.
The work was jointly supported by JICA Project [5] Shinichi Tashiro and Manabu Tanaka. Development of
(AUN/SEED-Net) with SRJP program and plasma MIG brazing process for dissimilar metal
TANAKAs Lab, JWRI, Osaka University, JAPAN. joining of aluminum to steel. International Symposium
on Interfacial Joining and Surface Technology (IJST),
References (2014).
[1] Madison Heights. Hybrid Welding: An Alternative to
SAW. Welding Solutions, Inc. (2007).

15
Journal of Science & Technology 119 (2017) 016-021

The Influence of Electrical Parameters on the Penetration of Tungsten into


the SKD61 Workpiece Surface in PMEDM using Tungsten Carbide Powder

Le Van Tao1,2,*, BanhTien Long1, Tran Xuan Thai 1, Nguyen Thi Hong Minh 1
1
Hanoi University of Science and Technology No. 1, Dai Co Viet Str., Hai Ba Trung, Ha Noi, Viet Nam
2
Military Technical Academy - No. 236, Hoang Quoc Viet, Cau Giay, Hanoi, Viet Nam
Received: January 10, 2016; accepted: June 9, 2017

Abstract
In the study, the authors studied the effect of the electrical current (Ip); the time-on discharge (Ton) on the
penetration of tungsten into the surface of SKD61 steel at different concentrations of tungsten carbide
powder used in powder mixed electrical discharge machining. The research has shown that the electrical
current and the time-on discharge can impact greatly on the penetration of tungsten and the carbidation of
the steel surface. In the mode of low electrical discharge, the penetration of tungsten into the surface and
the carbidation of the surface is better than that of the mode of high electrical discharge.
Keywords: EDM, PMEDM, Surface modification, Tungsten carbide powder.

1. Introduction1 separation of carbon element in the dielectric fluid


that enables the carbidation process on the surface of
The material SKD61 is a material with the
workpiece in PMEDM. Therefore, the surface quality
mechanical properties suitable for molds and machine
is improved considerably, especifically the micro
parts. Especially when heat-treated or chemically
hardness and the abrasion resistance. These are the
treated, SKD61 can achieve good mechanical
main reasons for many researchers to focus on
properties. Tungsten is a metallic element with the
PMEDM.
low thermal expansion and high strength, a high
melting temperature of about 3422C. In addition, the In 1980, Erden and Bilgin [3] detected the effect
resistance to oxidation, acid and alkali corrosion of of impurities in the dielectric liquid of EDM method,
tungsten is very good. These characteristics are also the authors have verified the experiments and theory
shown when tungsten is penetrated into SKD61 to determine the influence of impurities, which found
surfaces. The valuable surface characteristics created that adding the alloy powders impurities in the
by the penetration of tungsten is being useful in tools dielectric liquid that has improved the quality surface
and molds in the practice, particularly in the of workpiece after machining. Following the ensuing
mechanical engineering industry. years many authors have studied the effect of
impurities are mixed in the dielectric liquid of EDM,
In the method of processing, the electrical
specifically with some typical authors:
discharge machining (EDM) is widely used in
mechanical engineering. EDM is used for machining Wang et al [4] studied the impact of mixed the
high-strength material that are difficult with other alloy powders (Al and Cr) in the oil dielectric of
methods. In recent years, there have been some EDM process. Wang proved the electric parameters,
studies to improve the quality of surface workpiece the nature and concentration of the alloy powders in
by EDM. One of the methods to improve the the dielectric liquid are influential in technology
machining quality is to add powder into the dielectric properties and quality surface. Thus the coated metal
fluid. In the process of powder mixed EDM and the surface roughness are changed.
(PMEDM), the conductive particles mixed in the
Mohri et al [5,6,7] studied the effect of the
dielectric fluid reduces the insulation possibility of
silicon powders on the quality surface of workpiece
the dielectric fluid and therefore increases the
after machining. The results are surface resistant to
electrical discharges between the electrode and the
corrosion and the surface roughness (Ra) of less than
workpiece [1]. Besides, the spark discharges are more
2 m.
even and extended [2]. In addition, during spark
discharges, a thermal channel is formed which allows Uno et al [8] showed that the nickel powders
the melting of the metal particles and the chemical mixed in the dielectric liquid of EDM process, that
has transformed the surface of workpiece in bronze,
*
Corresponding author: Tel: 0912505036 aluminum. The nickel powders was used to coat a
Email: taoitd@yahoo.com surface layer on the workpiece to make high abrasion

16
Journal of Science & Technology 119 (2017) 016-021

resistance. Also authors pointed that coating Table 4. The Particle Size in Weight Percentage of
thickness increased with increasing concentrations of Tungsten Carbide Metal Powder
nickel powders.
5.5m 11m 16m 22m 31m
In the world, very few authors studied tungsten 5.23% 25.98% 59.74% 89.35% 98.93%
carbide powders mixed in the dielectric liquid of
EDM process to alter surface properties of workpiece. 2.1. Materials and Equipment
Within the framework of the paper, the authors
investigate the impact of electrical parameters on the The experiment used Daido Steel SKD61-
surface properties such as micro hardness of the steel Amistar (JIS- Japan) with the chemical composition
SKD61 as well as the study of carbidation process of as shown in Table 1. The dielectric fluid is Shell Oil
the surface after PMEDM. EDM Fluid 2. The technical properties are shown in
Table 2. The particle size and the chemical
2. Experiment composition in percent by weight of tungsten carbide
The total experiment plan is shown in Figure 1. metal powder is shown in Table 3 and Table 4.
The following sessions describe in details the
materials and methods of the experiments. 2.2. Experiment Method
An electrical discharge machine from the
Aristech Company, model CNC-460 EDM, was used
to remove the upper part of the SKD61 workpiece
to obtain dimensions as in Table 5. The copper
electrode polarity was negative. In this experiment,
tungsten carbide metal powder was mixed into the
dielectric fluid with the concentrations as in Table 5.
Table 5. Experimental Conditions
Deposition Condition Detail
Current (A) (Ip) 1A, 2A
Pulse on (s) (Ton) 16 s, 32 s, 50 s, 200 s
Pulse off (s) (Toff) 50s
The dielectric fluid Shell EDM Fluid 2
Polarity of Cu-electrode Negative ()
Fig. 1. Experimental flow
Current voltage (V) 80-120 V
Powder concentration (g/l) 20; 40; 60
Table 1. The Chemical Composition in Weight The parameters of the process are given in Table
Percentage of SKD61 5. The chemical composition was measured by
C% Si% Cr% Mo% V% Energy-dispersive X-ray spectroscopy (EDX) on a
0.38 1.0 5.0 1.25 1.0 scanning electron microscope JSM6610LA- JEOL
(JAPAN). Layer coating in surface of workpiece
was taken by a microscope AXIO-A2M. Micro
Table 2. The Technical Properties of Shell Oil EDM hardness of surface was determined by micro
Fluid 2 hardness tester DURAMIN- STRUERS
Properties Unit Value
3. Results and discussion
0
Velocity at 40 C cSt 2.25
Density at 150C kg/l 0.773 3.1. Analyzing the content of the element tungsten
0 penetration into the surface of workpiece
Freezing temperatures (max) C -27
Thermal conductivity W/m 0C 0.01 Using the EDX method to determine the
chemical composition of surface by region. As shown
Table 3. The Chemical Composition of Tungsten in Figure 2
Carbide Metal Powder As shown in Figure 3a;3b;3c, the chemical
composition of surface is determined at the mode I p =
C% Co% Fe% W% Other components 1 A, Ton = 16 s, 60g/l. The tungsten content of the
5.56 11.9 0.02 82.5 <0.01 surface is averaged among the three surface region on
the workpiece.

17
Journal of Science & Technology 119 (2017) 016-021

Fig.2. The region determine the chemical


composition of surface by EDX
001
2400

2100 W: 62.13%
Si
1800
W
1500 Fe
Cr
Counts

1200
V W
900 C Fe W Mo FeKesc
Fe
Cr Cr W
600 Mo W W
V V V Cr Fe Fig. 4. The tungsten content penetration into the
W
300 surface of workpiece at Ip = 1A
0
0.00 1.00 2.00 3.00 4.00 5.00 6.00 7.00 8.00 9.00 10.00

keV

Fig. 3a. The region I


002
2400

2100
Si
1800
W
W: 62.38%
1500
Counts

Cr
1200 Fe
V W
900
Fe FeKesc
C Fe
Cr W Mo Cr W W
600 V Mo W
V V Cr Fe W
300

0
0.00 1.00 2.00 3.00 4.00 5.00 6.00 7.00 8.00 9.00 10.00

keV

Fig. 3b. The region II


003
2400

2100
Si W: 62.71%
1800
Fe
W Fig. 5. The tungsten content penetration into the
1500
Cr surface of workpiece at Ip = 2A
Counts

1200 Cr
V W FeKesc
900 Fe The tungsten content penetrate into the surface layer
C Fe W Mo Cr W
W W of workpiece at Ip = 1A; Ip = 2A from figures 4 and 5:
600
V Mo
V V Cr Fe W
300 According to the graph in Figure 4, 5:
0
0.00 1.00 2.00 3.00 4.00 5.00 6.00 7.00 8.00 9.00 10.00
Considering at the pulse on time, or time-on
keV
(Ton), the element tungsten always penetrated into the
surface of SKD61 steel. The only at two modes of Ton
Fig. 3c. The region III = 200 s, Ip = 1A, the concentration of 20 g/l and 40

18
Journal of Science & Technology 119 (2017) 016-021

g/l (Figure 2), the tungsten does not penetrate into the dielectric or the based material. This explains the fact
surface of workpiece. that the micro hardness on the surface of workpiece
in PMEDM is higher than the micro hardness on the
The cause of this phenomenon can be explained
surface of workpiece in EDM. The improvement
as follows: The heat of the electrical discharges has
comparing between the micro hardness on the surface
molten the tungsten carbide powder which later on
of workpiece in PMEDM and the micro hardness on
penetrate into the surface of workpiece. The element
the surface of workpiece in EDM is lowest by 2.76%
tungsten does not penetrate into the surface of
at the mode Ton = 200 s; Ip = 1 A and concentration
workpiece at Ton = 200 s mode, Ip = 1A, the
20g/l. The highest improvement of micro hardness on
concentration of 20 g/l and 40 g/l, is due to the
the surface of workpiece between PMEDM and EDM
prolonged sparks time, the high bubble pressure of
is 129.17% at the mode T on = 16 s; Ip = 1 A and
the previous period reduced the concentration of
concentration 60g/l.
tungsten powder in the next period, leading to little or
no penetration of tungsten carbide into the workpiece
surface.
At the short time of eletrical discharges (Ton),
the tungsten content in the surface of workpiece are
more than that at the long time of eletrical discharges
(Ton). At the mode Ton = 16 s, the tungsten content
in the surface of workpiece is more than at the Ton =
32 s; 50 s; 200 s at Ip = 1 A and Ip = 2 A. Also,
according to the diagram in Figure 4, Ton = 16 s; Ip =
1 A and concentration of 60 g/l having the highest
tungsten content into the surface of workpiece, which
is 62.407%.
This phenomenon is due to the short time of
electrical discharges, which leads to the low bubble
pressure of the previous period, leading to a higher
concentration of tungsten powder in the next period. Fig. 6. Micro hardness on the surface of
Therefore, during the next electrical discharge period, workpiece (HV) at Ip = 1 A
the tungsten penetration into the surface of workpiece
in the discharge channel forming region is much
higher. Also, the time of electrical discharge and the
current of electrical discharges are at reasonable
levels to generate a heat channel to melt the tungsten
powder and then generates the energy imbalances in
the surface of workpiece, enabling of tungsten
penetration into the surface of workpiece.
3.2. Micro hardness on the surface of workpiece.
Using the micro hardness tester DURAMIN-
STRUERS to determine the micro hardness of
surface. The micro hardness of the surface is an
average value of three point on the surface of
workpiece.
According to the graph in Figure 6, 7, the micro
hardness on the surface of workpieces in PMEDM
method are higher than the micro hardness of the
surface of workpieces in the EDM method. Fig. 7. Micro hardness on the surface of
workpiece (HV) at Ip = 2 A
This phenomenon can be explained by the
appearance of the tungsten content in the surface of
workpiece after machining, according to Figure 8, 9.
The tungsten content are changed in to carbide on the
surface of workpiece under the thermal effect of the
EDM process. The carbon may come from the

19
Journal of Science & Technology 119 (2017) 016-021

This was partly due to the lower tungsten


content of the surface. Also, the high concentration of
the metal powder particles during forming the
discharge channel has resulted in an unstable spark
discharges condition [3]. This is also the main cause
of reduction in chemical carbidation process due to
the formation of heat channels.
Considering the micro hardness using the
tungsten powder, where Ip = 1 A; Ip = 2 A, the micro
hardness increases with the increasing concentration.
However, the micro hardness at Ip = 1 A; Ton = 50 s;
60g/l is lower than the micro hardness at 40 g/l.
Similarly, the micro hardness at Ip = 2 A; Ton = 32 s;
40 g/l is lower than the micro hardness at Ip = 2 A;
Ton = 32 s; 20 g/l, the micro hardness at Ip = 2 A; Ton
= 50 s; 60 g/l is lower than the micro hardness at Ip
= 2 A; Ton = 50 s; 20 g/l and 40 g/l.
4. Conclusions
The research investigating the tungsten
penetration into the surface of workpiece under the
influence of the electrical discharges curent (Ip), the
time electrical discharges (Ton) and the powder
concentration in PMEDM using tungsten carbide
alloy powder mixed into the oil dielectric has
achieved the following new results:
Fig. 8. The tungsten carbide phase of longitudinal section 1. The micro hardness on the workpiece surface
at Ip = 1 A, 500 times magnification in PMEDM is improved as compared to that of EDM.
2. The highest tungsten content at Ip = 1 A; Ton
= 16 s and concentration 60 g/l, where the
penetration into the surface of workpiece is 62.407%.
3. The improvement of the micro hardness of
the workpiece surface of PMEDM as compared to
that of EDM is lowest by 2.76% at the mode T on =
200 s; Ip = 1 A and concentration 20 g/l. The highest
change of micro hardness of the workpiece surface
between PMEDM and EDM of 129.17% was
obtained at the mode Ton = 16 s; Ip = 1 A and
concentration 60 g/l.
4. At Ip = 1 A; Ton = 16 s with all the different
concentrations, the tungsten penetration into the
surface of workpiece is highest and there is a
significant improvement the micro hardness of
PMEDM in comparison with EDM.
References
[1] K. Furutani, A. Saneto, H. Takezawa, N. Mohri, H.
Miyake, Accertation of titanium carbide by electrical
Fig. 9. The tungsten carbide phase of longitudinal discharge machining with powder suspended in
working fluid, Prec. Eng. 25 (2001) 138 144.
section at Ip = 2 A, 500 times magnification
[2] W.S. Zhao, Q.G. Meng, Z.L. Wang, The application
of research on powder mixed EDM in rough
machining, J. Mater. Process. Technol.129 (2002)
3033.

20
Journal of Science & Technology 119 (2017) 016-021

[3] A. Erden, S. Bilgin, Role of impurities in electric International Symposium on Machine Tool
discharge machining,in: Proceedings of 21st Designand Research, UK, 1985, pp. 329336.
International Machine Tool Design and Research
Conference, Macmillan, London, 1980, pp. 345350. [7] N. Mohri, J. Tsukamoto, M. Fujino, Surface
modification by EDMan innovation in EDM with
[4] C.H. Wang, Y.C. Lin, B.H. Yan, F.Y. Huang, Effect semi-conductive electrodes, in: Proceedings of Winter
of characteristics of added powder on electric Annual Meet ASME, vol. 34, 1988, pp. 2130.
discharge machining, J. Jpn. Inst. Light Met. 42 (12)
(2001) 25972604. [8] Y. Uno, A. Okada, Y. Hayashi, Y. Tabuchi, Surface
integrity in EDM of aluminum bronze with nickel
[5] N. Mohri, N. Saito, M.A. Higashi, A new process of powder mixed fluid, J. Jpn. Soc. Elec. Mach. Eng. 32
finish machining onfree surface by EDM methods, (70) (1998) 2431 (in Japanese).
Annals CIRP 40 (1) (1991) 207210.
[6] N. Mohri, J. Tsukamoto, M. Fujino, Mirror-like
finishing by EDM, in: Proceedings of the 25th

21
Journal of Science & Technology 119 (2017) 022-027

Modeling and Force Analysis of an Electrothermal Micro Gripper with


Amplification Compliant Mechanism

Dang Bao Lam1*, Nguyen Tuan Khoa1,2, Nguyen Dang Thuan3, Pham Hong Phuc1
1
Hanoi University of Science and Technology, No. 1, Dai Co Viet, Hai Ba Trung, Hanoi, Viet Nam
2
Queensland Micro- and Nanotechnology Centre, Griffith University, Australia
3
Smart System Laboratory, Hanbat National University, Korea
Received: January 13, 2017; accepted: June 9, 2017

Abstract
This work reports a novel design of a micro gripper, in which V-shaped electrothermal actuators are used to
create gripping force and a compliant mechanism is integrated to amplify displacements of the actuators.
The gripper is designed to handle micro samples of various sizes from 5 m to 50 m by applying
appropriate driving voltages. Those voltages are ranged from 5 V to 25 V, which are relatively low in
comparison with driving voltages of the electrostatic micro grippers. The compliant mechanism with
amplifying ratio 5.2, arranged between the actuators and the jaws, is aimed to compensate small strokes of
the actuators. Simulation by Finite-Element Analysis has also been carried out to confirm results of the
theoretical calculation and designing process. The micro gripper can be implemented in micro devices such
as micro robots or micro assembling systems, in which it can perform gripping and transporting tasks.
Keywords: Micro gripper, V-shaped actuator, Compliant mechanism.

1. Introduction* disadvantage, using of the amplifying compliant


micro mechanism is an effective solution.
Today acronym MEMS, which stands for Micro
Electro Mechanical System, has become more and Compliant mechanisms, which are jointless and
more familiar with all of us. MEMS technology has monolithic mechanical device, are very appropriate to
been widely researched and rapidly developed since use in the MEMS systems instead of the conventional
the last decades of twentieth century. Nowadays, mechanisms with classic revolute and prismatic joints
MEMS products have been applied into numerous [12-14]. Another reason for implementation of the
areas such as biomedical engineering, automobile compliant mechanism into micro gripper is the output
industry, military industry, aviation and space motion. All the electrostatic, electrothermal and
technology etc. as well as into human daily life. In piezoelectric micro actuators produce relatively small
those micro systems and devices, there is a demand strokes. The compliant mechanisms can help to
for micro grippers, which can manipulate tiny objects amplify those displacements and make them proper
with sizes ranged from few micrometers to hundreds for gripping task. In this work, the authors present the
of micrometers. Various types of micro actuators micro gripper driven by the V-shaped electrothermal
have been used for driving those grippers, including actuators and amplifying mechanism, which can grip
electrostatic [1-3], piezoelectric [4-6], shape-memory the micro samples sized up to 50 m and can be
alloys SMA [7-9] and electrothermal [10, 11]. In applied in micro robot or micro analysis systems.
comparison with other actuation methods, the micro
electrothermal grippers have the advantage of lower 2. Theoretical calculation
driving voltage (comparing with the electrostatic 2.1. Configuration and geometrical displacements
grippers), simple fabrication process (comparing with
the piezoelectric and SMA grippers) and large 2.1.1. Working principle
generated forces. On the other hand, those thermal Figure 1 shows the configuration of the micro
devices also have some disadvantages, such as gripper, which consists of three main parts: actuating
thermal dependence, lower working frequencies and unit with the V-shaped electrothermal actuators (1),
relatively small displacement. To overcome the last amplifying unit with the compliant mechanism (2)
and gripping jaws (3). The beams of the V-shaped
actuators are fixed at one end with the anchors (4),
the other end of the beams are connected to the main
*
Corresponding author: Tel.: (+84) 904.359.539 beam (5). When applying a driving voltage, the wings
Email: lam.dangbao@hust.edu.vn
22
Journal of Science & Technology 119 (2017) 022-027

of the V-shape beam will expand and push the main Because only geometrical displacements of the
beam (5) forward. The compliant mechanism (2) with compliant mechanism and the jaws are taken into
the flexural joints (6) is designed to convert consideration, we substituted the flexural joints with
displacement in y-direction into rotation movement of revolute joints and omit the elastic forces. The
the jaws (3), and also to amplify displacement of the equivalent model of the gripper is shown in figure 2.
actuators. Calculation is being carried out with only one wing
due to the symmetries of the gripper structure.
43
y Because three points F, B and C belong to one
rigid body, in order to find out displacement of the tip
x F of the jaw, position of the points B and C must be
36
2 determined. In figure 3 we can see under the input
displacement , the jaw i.e. BC will rotate an angle .
Using geometrical equations, the displacements of the
04 point B and C in x-direction and y-direction can be
calculated as:
65
) (1)
Dy DD' x OO' x AB 2 ( AU d )2 AB.sin( BAU

1 ) (2)
Cy CC' x OO' x AB 2 ( AU d )2 AB.sin( BAU
Fig. 1. Configuration of the micro gripper
Bx BBx' BC 2 CO x BC 2 CO 2
2
(3)
F

) (4)
By BB' y OO' AB2 ( AU d )2 AB.sin( BAU
Where B,C,D are the new position of B, C, D under
influence of the input displacement . From position
of BC and BC we can find the virtual center of
E
rotation I as below:

W
B 'C y' BH x BC y BH x'
I D
C H

B L L L B 'C y' BC y
L I (5)
60 L
90 B 'C y' BH y BCx BH x' BCx BH x BC y BH y
C
D D
B 'C y' BC y
A
60
And the rotation angle can be expressed as follows:
Fig. 2. Equivalent model of the micro gripper 2
L L
3L2 3
2
y
2
cos (6)
I 2L
C
H
B
H
C Where the length of BC, CD and BD at the initial
O
O
position is L. And finally, we have the lateral
B
D displacement of the jaw tip F as well as the
D amplifying ratio KA of the compliant mechanism:

d
cos cos .IF (7)
A U U x
K A 5.2 (8)

Fig. 3. Displacement of the links For example, there is input set of parameters
with AB = 200 m, BD = BC = 100 m, FH = 600
2.1.2. Geometrical displacements
m, and the angles are designed with the values as
shown in figure 2. We can calculate that for creating

23
Journal of Science & Technology 119 (2017) 022-027

25 m displacement for one gripper jaw, only 4.8 m beam (5) FS1 equals 100 N (see figure 6). Assuming
input displacement is needed. that deformation in the gripper structure is purely
elastic, meaning that relation between the input force
2.2. Displacements of the V-shaped actuators
and displacement of the gripper jaws is linear. With
Calculation for the V-shaped electrothermal the value 3.15 m of output displacement, we can
actuators has been presented in [15]. According to calculate the stiffness of the gripper structure: kS =
that paper, if we have the thermal actuator with FS1/3.15 = 31.75 N/m.
dimension as shown in figure 4, we can calculate
force F and displacement S as follows:

S l 2 sin 2 sin 2 2l.l l.sin (9)


l
F 2nEbh sin (10)
l
D

Fig. 6. Displacement of the gripper jaw


b Similarly, with simulating gripping force FG1 = 100
N located on the tip of the gripper jaws, the
displacement obtained by simulation is 2.48 m (see
h

figure 7). Therefore, the stiffness of the gripper jaw


can be calculated as: kG = FG1/2.48 = 40.32 N/m.
Fig. 4. V-shaped actuator

18 18

16 16

14 14
Displacement (um)

12 12
Force (mN)

10 10

8 8

6 6

4 4

Force
Fig. 7. Simulation of gripping force
2 2
Displacement

0 0 3.2. Displacements
0 5 10 15 20 25 30 35
Voltage (V) To drive the gripper, the V-shaped electro-
thermal actuator with following specification was
Fig. 5. Relation between voltage and displacement, chosen: number of beams n = 10, the length of each
force of the actuator beam l = 750 m, beam width b = 4 m, beam height
Where b, h and l are the width, height and length of h = 30 m, and the range of the driving voltages 030
V. Table 1 shows the relation between voltages and
the actuator beam. is the slope angle of the beam.
displacement S of the actuator. We have to take into
And n is the number of the beams in the actuator. In
consideration that simulation was carried out just
figure 5 is the graph showing the relation between
only for the thermal actuator itself.
driving voltages and forces, as well as displacements
of the V-shaped actuator. Table 1. Relation between displacements and driving
voltages of the actuator
3. Simulation
3.1. Stiffness calculation U(V) 5 10 15 20 25 30

Simulation was carried out with simulating S(m) 1.76 3.12 5.39 8.56 12.64 17.63
thermal-expansion force located at the tip of the main

24
Journal of Science & Technology 119 (2017) 022-027

To simulate whole system with AB = 200 m, Comparing S and , we can find the compressing
BD = BC = 100 m, FH = 600 m, we can find the ratio KC between displacements of the main beam of
displacements of the jaw tip of the gripper (see in the V-shaped actuator when it works alone and when
table 2). In section 2.1, we have already found the it is integrated into the gripper:
amplifying ratio of the gripper KA is 5.2. Therefore,
KC S
we can also find out the displacements of the 2.4 (11)
thermal V-shaped actuator while it is integrated into
the gripper. Deviation between displacements and S 3.3. Calculation of the minimum voltage for
can be explained by the fact that when the actuator is gripping micro objects
connected to the compliant mechanisms and the jaws,
it will operate as a member in the complex system
and can only be able to produce smaller
displacement.
In figure 8, we can see simulation result of the
displacement = 18.46 m of the jaw under the
driving voltage of 20 V. The calculated and simulated
displacements of the gripper are shown in figure 9.
The curve of the theoretically calculated
displacements and the curve expressing simulating Fig. 10. Jaws and micro object
results are almost identical. In figure 10, we can see the jaws and the micro
Table 2. Relation between driving voltages and sample with diameter D m. Supposed the initial gap
displacements and between the jaws is G m.

U(V) 5 10 15 20 25 The displacement of one jaw to approach the


object is calculated as follows:
(m) 3.80 6.74 11.62 18.46 27.26
(m) 0.73 1.29 2.23 3.55 5.24 x = = (G-D)/2 (12)
Using equation (8) and (11), the displacement S
of the V-shaped actuator can be expressed as:
S =KC. = KC./KA (13)
From (13) we can establish relation between the
minimum voltage for driving the micro gripper and
the sample diameter in form of the graph as shown in
figure 11.
3.4. Calculation of the gripping force with the micro
sample with diameter 30 m
Fig. 8. Simulation of complex gripping system
From (12), with the initial gap G=60 m, we can
calculate the displacement of one jaw to approach the
GD
40
object as: x 15 m
35
2
30 From the relation between the driving voltage
Displacement (um)

25
and displacement (Figure11 - table 2), if we consider
that this relation with U ranged from 15 V to 20 V is
20
approximately linear, corresponding with x = 15 m,
15
we can obtain Umin = 17.47 V.
10
Simulated The process of handling the micro object can be
5
Calculated split into 2 following phases. Firstly, the voltage
0
0 5 10 15 20 25 30
increases to the value of Umin, and the jaws are
Driving voltage (V) approaching and touching the object. And secondly,
the voltage continues to increase to U*, and generates
Fig. 9. Displacements of the gripper the gripping force Fk. In this second phase, the
gripping force has the value equaling the elastic force

25
Journal of Science & Technology 119 (2017) 022-027

generated when one jaw is compressed by x, and Fk systems, in which it can perform gripping and
can be expressed as follows: transporting tasks.
Fk kG . x => x= Fk/ kG Acknowledgments
This research is funded by Vietnam National
The stiffness kG = 40.32 N/m has been
Foundation for Science and Technology
obtained by simulation (Fig. 7). We can calculate the
Development (NAFOSTED) under grant number
voltage U for gripping the micro sample with
107.01-2015.18
diameter D with the gripping force Fk as follows:
U= Umin+ Ux (14) References
[1]. Heng-Chung Chang, Julius Ming-Lin Tsai, Hsin-
From (14), the gripping force generated while Chang Tsai, Weileun Fang - Design, fabrication, and
working with the sample with diameter 30 m can be testing of a 3-DOF HARM micromanipulator on (1 1
calculated as shown in the table 3. 1) silicon substrate, Sensors and Actuators A 125
(2006) 438445.
Table 3. Relation between driving voltages and
gripping force Fk [2]. A.Nikoobina, M.Hassani Niaki. - Deriving and
analyzing the effective parameters in microgrippers
Fk(N) x(m) Ux(V) U(V) performance, Scientia Iranica B (2012) 19 (6), 1554
100 2.48 3.26 20.73 1563.
200 4.96 6.97 24.44 [3]. Olivier Millet et al. - Electrostatic actuated micro
gripper using an amplification mechanism, Sensors
300 7.44 10.71 28.19 and Actuators A 114 (2004) 371378.
400 9.92 13.25 30.73
[4]. S.K. Nah, Z.W. Zhong - The A microgripper using
500 12.40 15.57 33.04 piezoelectric actuation for micro-object manipulation,
Sensors and Actuators A 133 (2007) 218224.
60
[5]. Chang-Seong Jeon, Joon-Shik Park, Sang-Yeol Lee,
50 Chan-Woo Moon - Fabrication and characteristics of
out-of-plane piezoelectric micro grippers using
40 MEMS processes, Thin Solid Films 515 (2007) 4901
Diamerer D (um)

4904.
30
[6]. Bong-Seok Kim, Joon-Shik Park, Byoung Hun Kang,
20 Chanwoo Moon - Fabrication and property analysis
of a MEMS micro-gripper for robotic micro-
10
manipulation, Robotics and Computer-Integrated
Manufacturing 28 (2012) 5056
0
5 10 15 20 25
Voltage (V)
[7]. S.B.Choi, Y.M.Han, J.H.Kim, C.C.Cheong Force
tracking control of a flexible gripper featuring shape
memory alloy actuators, Mechatronics 11 (2001) 677-
Fig. 11. Relation between sample diameter and 690.
driving voltage
[8]. Z.W. Zhong, C.K. Yeong - Development of a gripper
4. Conclusion using SMA wire, Sensors and Actuators A 126 (2006)
375381.
In this paper, the design, calculation and
simulation of the electrothermal micro gripper have [9]. M.Kohl, B.Krevet, E.Just - Shape memory alloy
been presented. The compliant mechanism with the device, Sensors and Actuators A 97-98 (2002) pp.
646-652.
amplifying ratio KA= 5.2 was used to magnify the
input displacement of the V-shaped actuator. The [10]. Timothy Moutlon, G.K. Ananthasuresh -
gripper is designed to work with the driving voltages Micomechanical devices with embedded electro-
ranged from 5 V to 25 V, and to grip the micro object themal-compliant actuation, Sensors and Actuators A,
vol. 90 (2001), pp. 38-48.
with diameter ranged from 5 m to 50 m. This
device can be fabricated by the bulk micromachining [11]. B.E. Volland et al. - Duo-action electro thermal micro
technologies using only one photomask on a SOI gripper, Sensors and Actuators A, vol. 97-98 (2007),
(silicon-on-insulator) wafer. Simulation has been pp. 646-652.
carried out to confirm the results of calculating work. [12]. Sridhar Kota et al. - Design of Compliant
The micro gripper can be implemented in micro Mechanisms: Applications to MEMS, Analog
devices such as micro robots or micro assembling

26
Journal of Science & Technology 119 (2017) 022-027

Integrated Circuits and Signal Processing (2001), [14]. Byoung Hun Kang, John T. Wen - Design of
29,pp. 715. Compliant MEMS Grippers for Micro-Assembly
Tasks, Proceedings of the 2006 IEEE/RSJ,
[13]. Xiantao Sun, Weihai Chen, Yanling Tian, Sergej DOI: 10.1109/IROS.2006.282626.
Fatikow,Rui Zhou, Jianbin Zhang, and Manuel
Mikczinski - A novel flexure-based microgripper with [15]. Nguyen Tuan Khoa, Dang Bao Lam, Dinh Khac
double amplification mechanisms for micro/nano Toan, Phm Hong Phuc - Design and fabrication of
manipulation, Review of Scientific Instruments 84, the micro bi-directional motor driven by electro-
085002 (2013). thermal actuators, Journal of Science and
Technology, Vol. 69(2013), pp. 62-68.

27
Journal of Science & Technology 119 (2017) 028-031

Application of Sinusoidal Phase Modulation Technique for Infrared


Spectrum Measurement by Fourier Transform Method
Doan Giang1,2, Nguyen Van Vinh1 , Nguyen Thi Phuong Mai1, Vu Thanh Tung1*
1
Hanoi University of Science and Technology No. 1, Dai Co Viet Str., Hai Ba Trung, Ha Noi, Viet Nam
2
Military Institute of Environmental Chemistry, Hanoi, Vietnam
Received: April 04, 2017; accepted: June 9, 2017

Abstract
The Fourier Transform Infrared (FTIR) spectrometer is widely utilized for the detection and identification of
gas in laboratories and open environments. The measurement sensitivity and range of the spectrometry are
limited due to the strength of the absorption or emission signals. This study proposes the use of the
sinusoidal phase modulation technique to improve the signal to noise ratio (SNR) of the detected signal of
an FTIR spectrometer. In this technique, a sinusoidal signal is applied to a voice coil to create movement of
a mirror. Hence, the intensity of the interference signal is a series of harmonics. A synchronous detection
(lock-in amplifier) is then utilized to detect and amplify only one suitable harmonic and removed all other
harmonics and noise. Therefore, the SNR of the harmonic is improved significantly. In this paper, a weak
infrared emission from a commercial heat-lamp is detected successfully using the proposed system.
Keyword: FT-TR spectroscopy, Frequency modulation, Phase modulation, Michelson interferometer.

1. Introduction* choppers and the synchronous detection (lock-in


amplifier technique). The main disadvantage of this
Fourier Transform Infrared (FT-IR)
method is that the beam of radiation is interrupted by
spectrometers are powerful instruments for
the chopper, the reduction in output is significant.
measurements of the intensity of infrared radiation as
Wavelength/frequency modulation technique has the
a function of frequency or wavelength [1, 2]. The
advantages over the amplitude modulation method
instruments are based on the idea of the interference
[7]. Both the reduction in output and the background
of radiation between two beams to generate an
noise are minimal [8]. However, this technique
interferogram. The intensity of the interference signal
requires a high-cost electro-optic modulator (EOM)
is a function of the optical path difference (OPD)
and it still has some residual amplitude modulation.
change between two beams. When Fourier Transform
algorithm on the signal is performed, the frequency In this paper, a simple phase modulation method
(wavelength) respond can be determined. Different to improve the signal to noise ratio of an FTIR
FT-IR spectrometers used different interferometers, spectrometer is proposed. In the proposed system, the
such as Michelson interferometer [3], Fabry-Perot OPD between two arms of the Michelson
interferometer [4], and grating interferometer [5]. interferometer is modulated by modulating the
Among these kinds of the spectrometer, the FT-IR oscillating of the mirror. Hence, the intensity of the
spectrometers using the Michelson interferometers interference signal is series of harmonics and each
are preferred. These have some advantageous features harmonic is a function of the OPD. Using the lock-in
over other techniques such as high precision and high amplifier technique [9], any harmonic of the
energy throughout. In this paper, the characteristics of interference signal can be detected accurately without
the FT-IR spectrometer based on the Michelson noise effect. The frequency/wavelength respond is
interferometer is first investigated. then determined using Fourier Transform method. In
the experiment, our proposed system is utilized to
Actually, the influence of environmental
detect an infrared radiation from a commercial heat-
background limits the measurement precision of the
lamp.
FT-IR spectrometer. To remove the background
effects, some modulation methods were employed. 2. Measurement principle
The earliest spectrometer used a chopper to modulate
2.1 Phase modulation Michelson interferometer
the intensity of the radiation sources [6]. The
background noise can be eliminated using mechanical The schematic diagram of the FT-IR
spectrometer based on the Michelson interferometer
is shown in figure 1. The radiation from an IR source
* that is placed at the focal point of a parabolic mirror
Corresponding author: Tel.: (+84) 976.516.396
Email: tung.vuthanh@hust.edu.vn (PM1) propagates to a beam splitter (BS). The beam

28
Journal of Science & Technology 119 (2017) 028-031

splitter is made of a special material that transmits


half of the radiation striking it and reflects the other
half. One beam passes through the beam splitter to a
fixed mirror and the second reflects off the beam
splitter to a moving mirror. The moving mirror is
driven by a voice coil actuator and it can move back Equation (3) shows that the interference signal of the
and forth precisely around a balanced point. The modulated interferometer is a series of harmonics.
fixed and movable mirrors reflect the radiation back Therefore, any harmonic from the signal can be
to the beam splitter. Another parabolic mirror (PM2) detected using the lock-in amplifier technique [8, 9].
directs the combined beam into an IR detector. When the signal enters a lock-in amplifier, it is first
Concurrently, a He-Ne laser propagates the same path multiplied by a reference value at a chosen frequency
with the IR radiation. The displacement of the and then passes through a low-pass filter (LPF).
moving mirror is determined accurately using the Therefore, the amplitude of any harmonic at a
interference signal of He-Ne laser that is collected significant modulation frequency can be accurately
using a photo-detector. The infrared spectrum is detected and all other higher order harmonics are
obtained by first collecting an interferogram using the removed. Hence, we can detect a pure harmonic
interferometer, and then performing a Fourier without noise. This signal is then amplified with a
Transform on the interferogram to obtain the suitable factor using an amplifier that is integrated
spectrum. into the lock-in amplifier. In this study, the first
harmonic is utilized. Using the lock-in amplifier, the
In this section, we propose a new method to intensity of the first harmonic is
improve SNR of an FT-IR spectrometer using the ph-
. (4)
When a polychromatic radiation source enters
the Michelson interferometer, but has a spectral
distribution given by I( ) as shown in Eq. (2), and the
light at different frequency is incoherent, then the
total intensity can be found be adding intensities for
different
. (5)
In the same way, as a monochromatic is used, using
the lock-in amplifier technique, we can determine the
Fig. 1. Schematic diagram of FT-IR spectrometer.
total intensity of the first harmonic at different
PM: Parabolic mirror; M: Mirror; BS: beam splitter;
FG: function generator; LIA: lock-in amplifier; DAQ: frequency
data acquisition; PC: personal computer. . (6)
-ase modulation technique. When the moving mirror
The right-hand side is nothing more than the
is modulated at a modulation frequency so that
the delay time between two arms of the Michelson sine form of the Fourier transform of I 1( ) so we
interferometer ( =OPD/c; c is the speed of light) have succeeded in writing an explicit form of the
varies with time as the following equation [10] relation ) = F{ }. When the Fourier
Transform algorithm is performed, both the
(1) amplitude and frequency of all components of the
where is the initial delay time caused by the radiation spectrum are determined.
unbalanced length between two arms and is the 2.2 Determination of radiation frequency of based
modulation excursion. on the zero path difference point
For a monochromatic radiation of frequency, the The zero path difference point (ZPD) is located
intensity of the interference signal is given as where the moving and fixed mirrors are the same
distance from the beam splitter. Therefore, all
(2) components of radiation with different frequencies
where is the average intensity. Using the are in-phase at the ZPD. Their contributions are all at
maximum and a very strong signal is produced by the
Bessel function, Eq. (2) is given by
IR-detector. When the OPD increases, different

29
Journal of Science & Technology 119 (2017) 028-031

frequencies produce interference peaks at different (a) (b)


positions of the movable mirror.
The laser He-Ne is utilized to measure the
displacement of the moving mirror from the ZPD.
The laser beam propagates the same path as the IR-
radiation in the interferometer and produces its own
interferogram at a photo-detector. This signal is used
as an extremely accurate measure of the OPD.
Therefore, when the moving mirror moves away from
the ZPD, the position of any peak in the
interferogram caused by different frequencies of the
IR- radiation is determined. A noteworthy is that, at Fig. 4. He-Ne laser interference signal. (a):
any peak position of the interferogram from the ZPD, interference fringes; (b): intensity of the interference
the displacement of the moving mirror is the signal.
wavelength of the radiation. The ZPD point can be detected by monitoring
3. Experiments the interference signal when all different frequencies
of radiation were in phase and they made a very
The experimental system is shown in figure 2. A strong signal as shown in figure 3. The ZPD point
commercial heat-lamp was used as an IR source. The was the biggest spike in the center of the burst. The
spectrum of the lamp was first measured using a interference signal of He-Ne laser is shown in figure
commercial radiometer (12-550 Mark III radiometer, 4. This signal was used to determine the displacement
Infrared Systems Development Corp.) and used as a of the movable mirror from the ZPD
reference. A modulation frequency of 20 Hz was
supplied for the voice coil actuator to modulate the The spectrum of the heat lamp measured using
OPD, hence the delay time between two arms of the our proposed system is shown in figure 5(a).
interferometer was modulated. A lock-in amplifier Concurrently, the spectrum of the lamp was measured
(PS1 Sciencetech Inc.) was used to detect the first using the commercial radiometer (12-550 Mark III
harmonic from the interference signal. The cutoff radiometer, Infrared Systems Development Corp.),
frequency of the lock-in amplifier was 1Hz. The figure 5(b). The measurement range of the radiometer
interference signal of radiation is collected using an covers the 1 to 16 um range. The experiment results
IR-detector (MCT-14-10-LN, Sciencetech Inc.) that using our proposed system and using the commercial
was cooled using Nitrogen liquid. radiometer show the same spectrum. The strongest
radiation was figured out at the wavelength of 3,2
m. It means that the spectrum of the heat lamp was
successfully determined using our proposed system.

(a)

Fig. 2. Experimental system.

(b)

Fig. 5. IR spectrum of the heat lamp, (a) the spectrum


obtained using the radiometer, (b) the spectrum
Fig. 3. Interferogram of the heat-lamp source obtained using our system (cutoff frequency of 1Hz).

30
Journal of Science & Technology 119 (2017) 028-031

4. Conclusion spectrometer with laser reference system, Sensors


and Actuators A: Physical Vol.149, (2009), 221-228.
The sinusoidal phase modulation FT-IR
spectrometer was performed. The advantageous [5] Lucey, Paul G., and Jason Akagi. "A Fabry-Perot
features of the sinusoidal phase interferometer and interferometer with a spatially variable resonance gap
employed as a Fourier transform spectrometer." SPIE
the lock-in amplifier detection were analysed. The Defense, Security, and Sensing, International Society
first harmonic of the modulated interference signal for Optics and Photonics, (2011).
was used for the spectrum measurement of broadband
radiation. The spectrum of a commercial heat-lamp [6] Bhattacharyya, et al., Wavelength modulation
was determined using our proposed method. This spectroscopy using novel mechanical light chopper
blade designs, Review of scientific instruments,
result opened a direction to develope FT-IR
Vol. 76, (2005), 083903.
spectrometer for a broadband radiation source such as
a blackbody or IR-lamps. [7] Lindsay, I. D., et al. "Mid-infrared wavelength-and
frequency-modulation spectroscopy with a pump-
References modulated singly-resonant optical parametric
oscillator." Optics Express, Vol. 14, (2006), 12341-
[1] Galina I. Dovbeshko, et al., FTIR spectroscopy
12346.
studies of nucleic acid damage, Talanta, Vol. 53,
(2000), 233-246. [8] Thanh-Tung Vu, et al., Accurate displacement-
measuring interferometer with wide range using an I2
[2] R. Harig, and G. Matz, Toxic cloud imaging by
frequency-stabilized laser diode based on sinusoidal
infrared spectrometry: A scanning FTIR system for
frequency modulation, Measurement Science and
identification and visualization, Field Analytical
Technology, Vol. 27, (2016), 105201.
Chemistry & Technology, Vol. 5, (2001), 7590.
[9] Thanh-Tung Vu, et al., Sinusoidal frequency
[3] L. Genzel and J. Kuhl, A new version of a Michelson
modulation on laser diode for frequency stabilization
interferometer for Fourier transform infrared
and displacement measurement, Measurement,
spectroscopy, Infrared Physics, Vol. 18, (1978), 113-
Vol. 94, (2016), 927-933.
120.
[10] Hariharan P, Optical Interferometry 2nd Edition,
[4] Lee, Feiwen, et al. "A MEMS-based resonant-
Academic Press, Elsevier, 2003.
scanning lamellar grating Fourier transform micro-

31
Journal of Science & Technology 119 (2017) 032-036

A Method for Capturing Accuracy and Pose Repeatability of Articulated


Industrial Robots
Duong Minh Tuan*, Le Duc Do
Hanoi University of Science and Technology No. 1, Dai Co Viet Str., Hai Ba Trung, Ha Noi, Viet Nam
Received: December 21, 2016; accepted: June 9, 2017

Abstract
This paper presents a low cost method for measuring accuracy and pose repeatability of industrial robots,
which can be used for robot home calibration, accuracy improvements, robot control compensation or
evaluating robot features. The method is proposed based on the kinematical relations of the robot joints and
the geometrical data from the robot controller with respect to several defined coordinate systems. The
experimental setup consists of three dial gauges, capturing position measurements of a KUKA KR6/2 robot
end-effector. The positions of the end-effector are employed to estimate robot pose accuracy and
repeatability. Moreover, experimental results demonstrate the efficiency of the method and can be used for
improving the robot accuracy as well as robot pose repeatability in industrial applications.
Keywords: Pose Repeatability, Accuracy, Industrial robot, KUKA KR 6/2

1. Introduction capture the robot poses for robot accuracy evaluation


* and calibrations. The measurement results show the
Recently, industrial robots (IRs) have been
efficiency of the method.
more and more used to realize continuous operations
such as prototyping, premachining of cast parts as 2. Method
well as end-machining of middle tolerance parts [1].
The method and the experimental setup
In addition, for repetitive tasks such as placing,
including a KUKA KR6/2 (6kg of payload) robot at
welding, and assembling, the repeatability of IRs
the laboratory of Machine-Tools and Tribology are
typically ranging from 0.03 to 0.1 mm and their
presented in this section. To derive the accuracy and
several millimeters accuracy are commonly used.
the repeatability from the experimental data, the
Thus, measurement approaches for accuracy and
coordinate systems must be defined for the proposed
repeatability have been used for enhancing actual
measurement method. All coordinate systems are
robot positioning accuracy. Lombard and Perrot [2]
depicted in Fig. 1. The world frame so-called w is
have implemented an automatic method to measure
fixed. The User frame can be defined at an arbitrary
robot accuracy. The most widely applied approach is
location in the robot cell (Fig. 2) with a designed
to use theodolite systems for calibration
point of a working table as the origin of the frame.
measurements [3]. Another high precision devices
The Tool frame has the origin at the Tool Center
like Coordinate Measuring Machines (CMMs) have
Point (TCP), see Fig.1.
been widely used for industrial dimension
measurement [4]. Laser Tracking Systems (LTS) [5] 2.1 Principle for Robot Pose Measurements
can combine the advantages of a large working space,
high accuracy and dynamic pose measurements. KUKA Robot language runing on VxWork - a
real time operating system (parallel with Windows),
However, they are high cost, and hence it is
controls the robot. This can provide the position and
prohibitive to calibrate and measure robot poses with
the orientation of the end-effector (Tool frame) via 6
low costs. A simple approach to measure an error
along a direction by a dial gauge has been discussed articulated joint angles. Thus, the robot pose can be
in [6]. However, this could only give accuracy for displayed on the screen of the KUKA Control Panel
(KCP). For example, when the robot moves to a
one direction. Using three dial gauge fixture for
industrial robots has been investigated in [7,8]. specified location described by pose k for which the
Herein, we employ a general low cost method with a pose variables are defined. That means that the pose
simple measurement fixture using three dial gauges to variables are determined by three position
components ( xk , yk and z k ) and three orientation
terms, Yaw ( Ak ), Pitch ( Bk ) and Roll (Ck ). As
* illustrated in Fig. 1, the robot changes pose 1 into
Corresponding author: Tel: 0947 036 686
Email: tuan.duongminh@hust.edu.vn
32
Journal of Science & Technology 119 (2017) 032-036

pose 2, it needs a transformation matrix called 1 T2 The real or measured transformation matrix from
and these poses are expressed in the World pose 1 to pose 2 can be computed as
coordinates defined by w T1 and w T2 , respectively.
F1m
1 u
u
F2m u F1m 1T2m and 1T2m u
F2m , (5)
Similarly, with respect to the User frame, one can
derive two transformation matrices for pose 1 and 1R m 1
p2m
pose 2 written as u F1 and u F2 , respectively. The
1
T2m 2 , (6)
0 1
controlled transformation matrix from the World
coordinate system to the Tool one describing pose k with 1 R2m and 1 p2m are the rotation matrix and
is evaluated as the position vector from pose 1 to pose 2,
w Rc w
pkc respectively.
w
Tkc wTk1 1Tk2 2Tk3 3Tk4 4Tk5 5Tk6 k , (1)
0 1 The robot absolute pose accuracy or the
positioning and orientation accuracy are defined as a
where wTkc is the transformation matrix from the difference between the measured transformation
w frame to Tool frame of pose k; i 1Tki presents the matrix 1T2m and the controlled counterpart 1T2c .
transformation matrix from joint frame i to joint Obviously, these matrices are independent on the
World and the User frames.
frame i-1, (i=1,2,,6); and w Rkc is the controlled
rotational matrix from the w frame to Tool frame of 2.2 Experimental Setup
pose k calculated as

cos Ck sin Ck 0 cos Bk 0 sin Bk


w
Rkc sin Ck cos Ck 0 0 1 0
0 0 1 sin Bk 0 cos Bk
1 0 0
0 cos A sin A
k k

0 sin Ak cos Ak
(2)
w c
with p is the controlled position vector
k

pkc xk yk zk .
w T
expressed in the w system reads as Fig. 1. Coordinate system definition
Moreover, we obtain the relation for controlled
transformation matrices of poses 1 and 2 with respect
to different frames as seen in Fig. 1

T2c wT1c
1 1
T2c wT1c 1T2c
w 1
T2c . (3)

Consequently, the controlled transformation


matrix from pose 1 to pose 2 can be determined as
follows

1R c 1
p2c
1
T2c 2 , (4)
0 1

where 1R2c and 1 p2c are the rotation matrix and


the position vector from pose 1 to pose 2 which are Fig. 2. Experimental setup for robots pose
used to evaluate the pose measurement data, measurements of KUKA KR6/2.
respectively. On one hand, the transformation matrix
from pose 1 to pose 2 is also calculated by the KRL Fig. 2 shows the experimental setup for robot
software using joint sensor measurements (in each pose measurements with a fixture supporting three
robot joint). On the other hand, the matrix is specified dial gauges with a specified space in which a tooling
by the commands in the robot program on the KCP. cube (A=50mm, made of aluminum) attached to the
robot end-effector can be easily placed. The tooling

33
Journal of Science & Technology 119 (2017) 032-036

cube is fixed on the end-effector at the origin of the indications of the dial gauges are recorded. The next
Tool frame and three orthonormal faces with respect step is to measure and record the coordinates of point
to x , y and z axes are well calibrated. The three P in Fig. 3. At last, the measures Px , Py and Pz are
dial gauges have resolution of 0.05 mm and their substituted into Eq. (7). Therefore, the actual position
stroke is of 18 mm. When the three orthonormal faces coordinates; u pkm ( xk , yk , zk ) of the robot end-effector
of the tooling cube touch the tips of the gauges, the
with respect to the User is achieved.
slide rods of the gauges are depressed in the x , y
and z directions. If the movement range of the cube 2.4 Orientation Accuracy
is within the stroke, the dial gauges can directly To determine the orientation of the robot end-
indicate the change of the tooling cube position. The effector, the dial gauges can be placed so that they are
setup is fixed on the table and referred to the not orthogonal together i.e. their axes do not intersect
reference plate (User frame). into a point. Consequently, three measuring points
2.3 Positioning Accuracy P1 ( x1 , y1 , z1 ) , P2 ( x2 , y2 , z2 ) and P3 ( x3 , y3 , z3 ) create
The Tool Center Point can be measured. We call a plane on which the orientation of the robot end-
a reference point on the gauge fixture is P with which effector is defined. By investing this plane for each
sample one can obtain the orientation of the robot
the TCP at pose k can be calculated by
tool frame.
xk Px (0.5 A M1 D1 ) Mathematically, the plane passing the three

yk Py (0.5 A M 2 D2 ), (7) points is constructed and the P1 (x1 , y1 , z1 ) lies in the
z P (0.5 A M D )
k z 3 3
plane yields the plane equation as

where xk , yk and zk are the coordinates of the a( x x1 ) b( y y1 ) c( z z1 ) 0 , (8)


TCP k with respect to the User frame. Px , Py and Pz where a, b and c are the direction numbers of the
are the coordinates of the point P on the gauge fixture plane, respectively. On the other hand, the P2 and P3
which can be determined from measurement lying in the plane yields
dimensions as shown in Fig. 3. D1 , D2 and D3 are
the maximum travel limits of gauges 1, 2 and 3, a( x2 x1 ) b( y2 y1 ) c( z2 z1 ) 0
. (9)
respectively. a( x3 x1 ) b( y3 y1 ) c( z3 z1 ) 0
D3 M3 With three variables and three equations (8) and
(9), the system of equations is solved to obtain the
P3 solution as
a ( y3 y1 )( z2 z1 ) ( y2 y1 )( z3 z1 ) ,
(10)
Z Y P2 ( x2 x1 )( y3 y1 ) ( x3 x1 )( y2 y1 )
M2
P1 b ( x3 x1 )( y2 y1 ) ( x2 x1 )( y3 y1 ) ,
O (11)
X P D2 ( x3 x1 )( z2 z1 ) ( x2 x1 )( z3 z1 )
c ( x3 x1 )( y2 y1 ) ( x2 x1 )( y3 y1 ) .
(12)
( x2 x1 )( y3 y1 ) ( x3 x1 )( y2 y1 )
1
M

P
1
D

Thereby, the normal vector of the measured


Fig. 3. Dimensions of the experimental setup. plane can be evaluated by its cosines

The measurement procedure of the TCP can be cos m


z a / q; cos zm b / q; cos zm c / q , (13)
addressed as follows:
At first, the robot moves to a designed pose k. where q a 2 b 2 c 2 . The direction vector
The tooling cube is installed on designed positions projected on z axis is
P1 , position. As the fixture moves so that the cube
z cos zm l cos zm m cos zm h , (14)
touches the tips of the dial gauges generating P1 , P2
and P3 points, then fixes the fixture in that place. The where l , m, and h are unit direction vectors of
the User frame. Similarly, the direction vector of the

34
Journal of Science & Technology 119 (2017) 032-036

x axis is x cos xm l cos xm m cos xm h when the Table 1. Positioning repeatability (mm)
x axis passing P1 and P2 Direction x Axis y Axis z Axis
Max errors 0.125 0.165 0.115
cos xm ( x2 x1 ) / r Min errors -0.145 -0.115 -0.170

cos x ( y2 y1 ) / r ,
m
(15) Mean -0.025 -0.036 0.045
STD 0.045 0.036 0.069
cos x ( z2 z1 ) / r
m

The mean and the standard of deviation (STD)


of experimental data were performed for 48 samples
where r ( x2 x1 ) ( y2 y1 ) ( z2 z1 ) .
2 2 2
or 48 times. Obviously, Fig. 4 describes that the
Therewith, the direction of y axis is computed as current robot position repeatability is mostly inside a
sphere (radius of 0.1 mm as the specified
y z x cos yml cos ym m cos ym h . repeatability of the robot by KUKA Company). Of
course there are some experimental measurements
The rotation matrix of the Tool coordinate outside the sphere, larger than 0.1 mm. However, the
system with respect to the User frame reads as measurements are inside the sphere with 95% of
confidence (as the Gauss distribution).
cos xm cos ym cos zm

Rk cos xm
u m
cos ym cos zm . (16)
cos xm cos xm cos zm

Consequently, we combine the position
measurement results and derive the actual or
measured transformation matrix from the User frame
k given by

u Rm u
pkm
u
Fkm k . (17)
0 1
u
Substituting Fkm into Eq. (5) the measured
transformation matrix 1F2m from pose 1 to pose 2 is
explicitly obtained in which both the robot position Fig. 4. Measurements of repeatability.
accuracy and orientation accuracy are evaluated.
Thus, we can clearly define the accuracy of the robot The orientation accuracy represented by the
formulated as rotational matrix in (16) can be derived from the
corresponding triangle. As shown in Fig. 5, the

R2 R2 R2
1 1 c 1 m
triangles connecting three measurement points
1 . (18)
( P1, , P2 , and P3 ) describe the orientation errors from
p2 p2 p2
1 c 1 m

which orientation repeatability 1 R2m can be derived.


If we reset all measurements of dial gauges of
the pose 1 to zero then the robot pose repeatability
can be derived properly. The pose 2 of the robot is
repeatedly measured with respect to the pose 1 as the
reference for repeatability study. Thus the 1 p2m is the
positioning repeatability and the 1 R2m describes the
orientation repeatability of the robot.
3. Results and Discussion
This section describes the measurements using
the proposed method on the KUKA robot.
Fig. 5. Orientation repeatability described as triangles
Experimental data obtained for 1 p2m shows the
( P1, , P2 , and P3 ).
efficiency of the method. The KR 6/2 robot has the
repeatability of 0.1mm. 1 p2m was calculated by With 48 samples for the robot pose we
invoking the Eq. (18) and shown in Table 1. computed the orientating repeatability as

35
Journal of Science & Technology 119 (2017) 032-036

0.6460.031 0.00080.002 0.2580.257 processed to compute the position repeatability and


orientation repeatability comprehensively. Therefore,
R2 0.00080.002
1 m
0.6460.031 0.4550.022 studying robot accuracy by this setup is convenient
0.000010.00001 0.000010.00001 0.6450.031
and flexible, particularly in large workshops.
. (19) Furthermore, with the measured accuracy and
repeatability solutions to improve the robots can be
These above values in (19) are Means and STDs found, e.g. improving robot control laws with error
of each entry of the rotational matrix from pose 1 to compensation. Thus the robot accuracy and pose
pose 2. It is obvious that the changes of the repeatability can be enhanced for continuous
orientation of the robot end-effector and positioning operations.
repeatability are acceptable if the robot can be used
for material manipulating. In terms of operations with Acknowledgments
continuous paths such as part machining the changes This research is funded by Hanoi University of
in the orientation might be large and can cause Science and Technology (HUST) under project
significant errors on machined parts. number T2016-PC-058.
The experimental results easily obtained from References
the measurement approach with very low cost dial
gauge fixture. This setup can be successfully used for [1] J. Bauer, M. Friedmann, T. Hemker, Analysis of
large working space of IRs in robot cells or in Industrial Robot Structure and Milling Process
Interaction for Path Manipulation, Process Mach.
industrial applications. The equipment is very flexible Interact. (2013) 245263.
and be placed at anywhere the robot can reach for
experiments. The calibration of the gauges and fixture [2] J. Lombard, J.C. Perrot, Automatic Measurement of
can be done with ease. Therewith the method is the Positioning Accuracy of Industrial Robots, Ann.
helpful for robot calibration at work. In addition to CIRP, 32 (1983) 297299.
that, the robot accuracy can be correctly estimated [3] D.E. Whitney, Industrial Robot Forward Calibration
with low cost. Method and Results, J. Dyn. Syst. Meas. Control 108
(1986) 18.
Moreover, the pose repeatability can be easily
obtained from the method. The method helps evaluate [4] B.W. Mooring and S.S. Padavala, The effect of
the robot accuracy and repeatability after long time kinematic model complexity on manipulator
accuracy, Proceedings, 1989 Int. Conf. Robot.
use. More importantly, based on the evaluation the
Autom. (1989) 593598.
robot accuracy can be enhanced in order to be used in
high accuracy process such as robot milling. [5] S. Decker, H. Gander, M. Vincze, J.P. Prenninger,
Therewith the robot errors must be compensated by Dynamic measurement of position and orientation of
improving the robot control model [9]. In addition, robots, IEEE Trans. Instrum. Meas. 41 (1992) 897
901.
the stiffness of the robot should be taken into account
since it strongly influences on the positioning [6] G.C. Smith, R.A. Smith, A non-contact method for
accuracy, in particular, for different poses and detecting on-line industrial robot position errors using
positions in the IRs volumes. The larger distance of a microwave doppler radar motion detector, Int. J.
the end-effector is, the larger deflection by the robot Adv. Manuf. Technol. 29 (2006) 605615.
end-effector becomes (lower accuracy) [10]. [7] D.D Le, Investigation of robot repeatability of a
However, compared with the accuracy the pose KUKA KR6/2, Thesis, Hanoi University of Science
repeatability is less influenced by target location. and Technology (2006).
Thus, the repeatability is an important robot [8] W. Xu, J.K. Mills, A New Approach to the Position
characteristic. Herein, the load was neglected due to and Orientation Calibration of Robots, Measurement,
the light tooling cube. Factors affecting the (1999).
repeatability will be discussed in the future work.
[9] B. Liu, F. Zhang, X. Qu, A method for improving the
4. Conclusion pose accuracy of a robot manipulator based on multi-
sensor combined measurement and data fusion,
The study presented the low cost measurement Sensors (Switzerland) 15 (2015) 79337952.
method in detail, which can be used for any type of
industrial robot topologies. Both pose accuracy and [10] M.T. Duong, X.T. Nguyen, Calculation method for
deflections of robot manipulators in loading
pose repeatability can be obtained from the method.
conditions, J. Sci. Technol. 64 (2008) 5156.
Interestingly, with very low cost experimental setup
we were able to demonstrate that this approach is
successfully realized on this equipment with the
KUKA KR6/2 robot. The experimental data were

36
Journal of Science & Technology 119 (2017) 037-042

Optimal Parameters of Linear Dynamic Vibration Absorber for Reduction


of Torsional Vibration

Vu Xuan Truong1, 2, Khong Doan Dien2, Nguyen Duy Chinh2, Nguyen Duc Toan 3,*
1
Graduate University of Science and Technology, Vietnam Academy of Science and Technology, Hanoi, Vietnam
2
Faculty of Mechanical Engineering, Hungyen University of Technology and Education, Hungyen, Vietnam
3
School of Mechanical Engineering, Hanoi University of Science and Technology, Hanoi, Vietnam
Received: October 24, 2016; accepted: June 9, 2017

Abstract

This paper presents three analytical methods to determine optimal parameters of the passive mass-spring-
disc dynamic vibration absorber (DVA), such as the ratio between natural frequency of DVA and shaft, damping
ratio of DVA. The original model presented by Den Hartog, Luft and Warburton are solved and has shown in
good agreement. Three analytical methods is then adopted for torsional shaft model. The simulation results
indicate that the effectiveness in torsional vibration could be reduced. Finally, the optimal parameters of DVA
were applied to decrease the shaft torsional vibration considering the vibration duration and stability criterion.
Keywords: Dynamic vibration absorber, Torsional vibration, Fixed-points theory.

1. Introduction where is the ratio of the absorbers mass to the


*
The dynamic vibration absorber (DVA) or tuned- primary structures mass.
mass damper (TMD) is a widely used passive vibration Since then, the xed-points theory and DVA
control device. When a mass-spring system, referred to structures have become one of the design laws used in
as primary system, is subjected to a harmonic excitation optimizing design of the damped and undamped
at a constant frequency, its steady-state response can be primary system [6-8].
suppressed by attaching a secondary mass-spring Luft proposed methodology MEVR (maximum of
system or DVA. This idea was pioneered by Watts in equivalent viscosity resistance) for the original model
1883 and Frahm in 1909. However, a DVA consisting [2]. Later, Warburton used minimum of quadratic
of only a mass and spring has a narrow operation region torque method (MQT) and found that the damping in
the neutralizer can also be optimized [3]. The results
and its performance deteriorates significantly when the
was given by
exciting frequency varies. The performance robustness
can be improved by using a damped DVA that consists 1 / 2 (1 3 / 4)
opt ; opt (3)
of a mass, spring, and damper. The key design 1 4(1 / 2)(1 )
parameters of a damped DVA are its tuning parameter
and damping ratio. This paper presents three analytical methods such
as FPM, MQT and MEVR to determine optimal
The optimization technique for original model that parameters of the dynamic vibration absorber (DVA)
is described in detail by Den Hartog [1]. The optimum for new shaft model such as the ratio between natural
tuning ratio of the neutralizer was found as a function of frequency of DVA and shaft, the damping ratio of
the neutralizers mass given by absorber. Since then we compare and evaluate optimal
effectiveness of those methods. Based on the main idea
1
opt (1) is to build a program that calculates to prove optimal
1 analytic solution of the original model, which applies to
and the damping ratio of the absorber torsion shaft model. Optimal parameters are presented
3 as very neat analysis. The simulation results indicate
opt (2) that the effectiveness in torsional vibration reduction.
8 1

*
Corresponding author: Tel.: (+84) 988.693.047
Email: toan.nguyenduc@hust.edu.vn
37
Journal of Science and Technology

2. Shaft modeling and equations of vibration J a J aa nca e22a nka e12a 0 (5)

The natural frequency of the DVA and the shaft,


respectively

ka ks
a s (6)
ma Jr

Introducing the dimensionless parameters


DVA
ma e e
; a ; 1 ; 2 (7)
r a mr r r r
0 ks ka
a ca
; ; (8)
M(t) ca s s maa
Jr Ja
where is the frequency of excitation torque
Fig.1. Modeling of the shaft system with DVA. Therefore, Eqs.(4) and (5) can be expressed by
In this study, the shaft model shown in Figure 1 is
considered. The shaft is modeled as a torsion spring 1
2 2
a 2s
M
Jr
(9)
which has stiffness ks and a disc which has moment of
inertia is Jr and rotating at the constant angular 2 2a n s 2a
velocity 0 is disturbed by harmonic torque M(t). The (10)
n 2 2s 2a 0
passive mass-spring-disc dynamic vibration absorber
(DVA) is attached on the shaft to minimize the The matrix form of Eqs.(9,10) are expressed as
torsional vibration of the shaft.
Mq + Cq + Kq = F (11)
hub
rotor passive disk
where q a
T

e1 e2
The mass matrix, viscous matrix, stiffness matrix
r a and excitation force vector can be derived as
r ka
ca
M(t)
1 2 2 0 0
M 2
C 2 (12)

2
0 n s

M (t )
2 0
Fig. 2. Modeling of the DVA K s 2
F Jr (13)
0 n s
2 2
0
Figure 2 shows the model of the DVA used in this
study. The DVA contains a passive disk and springs-
dampers system. The radius and moment of inertia of 3. Determine optimal parameters of the DVA
the passive disk are R, Ja, respectively. The shaft and 3.1. Fixed-points theory for optimal design
the passive disc are linked together by springs and
dampers system. The stiffness of each spring is ka. The The forced vibration of this system will be of the form
viscous coefficient of each damper is ca. n is the
number of springs-dampers. The angular displacement M (t ) M eIt (14)
of the rotor is r and the torsional vibration of shaft
Thus, the stationary response of this system which can
can be written as (t ) r 0t . be written as
The relative angular displacement between the rotor
(t ) eIt , a (t ) a eIt (15)
and passive disk as a .

The system equations of motion can be expressed by where and a are complex amplitude vibration of

J r J a J aa ks M (t ) (4)
the primary system and DVA, respectively.

38
Journal of Science and Technology

Substituting Eqs.(14-15) into Eq.(11), this becomes some of damping ratio. For c = 0 or c becomes infinite
so the amplifier fuction curve becomes infinite. That
means some where in between there must be a value of
2 1 2 2 damping ratio for which the peak becomes a minimum.

2 Two other curves are draw in Fig. 3, for = 0.1 and 0.4.
2


2i 0 0 1 M The first step of this method is to specify two


0 n 2 0 (16) fixed points. Suppose that two points (S and T) with
a ks
horizontal coordinates as a 1, 2. The conditions for A
s

2s 0 does not depend on the damping ratio is expressed as


2
0 n
2 2
s
follows
A
0 (20)
Hence the stationary response of the primary
system is expressed as
Substituting Eq.(19) into Eq.(20), this becomes
A iA2 M
1 (17)
A3 iA4 k s ( A12 A42 A22 A32 )
0, (21)
A12 A22 2
where A
2
4
2
A 3
2 2

A32 A42 2
A1 2 2 n 2 2 ; A2 n 2 ;
A12 A42 A22 A32 0 (22)
A3 n n n
2 2 2 2 2 2 2 4 2 2 2 2 2
From Eq.(22), we have
A4 3 2 2 n 3 2 n 2 n
A1 A2
1
1
(23)
After short caculation the Eq.(17) we obtained the A3 A4
real amplitude of the vibration response, which can be
written as
A1 A2
(24)
A12 A22 2 M M A3 2
A4 2
(t ) A (18)
A3 A4 k s
2 2 2
ks
We obtain the value of A at two points (S, T) these
where A is called the amplifier function that is are expressed as follows
defined by
A2
AS (25)
A A
2 2 2
A4 1
A 1 2
(19)
A A 2
3
2 2
4

A2
AT 2
(26)
A4

1 Den Hartog [1] reported that the graph of amplifier


function does not change in between the two peaks (S,
0
T) when the vertical coordinates of the S and T must be
equal. In this condition, we have
0 0.4 AS AT (27)

0.1 The optimal parameter of and are specifed by


S T
solving Eqs.(23-27) which can be written as


* (28)
n 2 1
Fig. 3 The graph of amplifier function
1
1,2
2
1,2
* 2
(29)
Figure 3 shows a plot of the amplifier fuction with 2 2 2 1

39
Journal of Science and Technology

Then, the optimum absorber damping can be 3.2. Minimum of quadratic torque (MQT) for optimal
identified as follows design
A The state equations of Eqs.(9,10) are expressed as [3,9]:
0 (30)
y (t ) By (t ) H f M (t ) (38a)
Eq. (19) gives where


T
A2 A32 A42 2 A12 A22 2 (31) y a a (38b)

The system matrix B is derived in [4,9] and has the


Taking derivative of Eq.(31) with respect to , this
form
becomes
0 E
B -1
A A A -M -1C
(38c)
A A3 3 AA32
2
A1 1 -M K

2 (32)
A4 2 A A2 where E is matrix unit, E 2
A A4
2
AA4 A2

In this study, the B matrix can be obtained as
Substituting Eq.(30) into Eq.(32) we obtain
0 0 1 0
0 0 0 1
A3 A
A2 A3 A1 1 B s
2
n s 2
2 2
0 n s 2

2
A A4
A4
A2
A2
(33)
2 1 n 2 2
2s 2 1 n s
2 2 2 2

s 0
2


2 2
(39)
Substituting Eqs.(28-29) into Eq.(33), this becomes
Matrix of excitation force is obtained as [4,9]
A1 A
A1 A2 A3 3 Hf
1 T
0 M 1F 0 0 J r1
T
J r1 (40)
M (t )
12 (34)
A2 A
A2 A2 A4 4
1
The quadratic torque matrix P is solution of the
Lyapunov equation [3]

and BP + PBT + S f Hf HTf = 0 (41)

where Sf is the white noise spectrum of the excitation


A1 A
A1 A2 A3 3 torque. The quadratic torque for vibration of shaft is

22 (35) determined by solving the Eq.(41)
A2 A
A2 A2 A4 4 n 2 4 4 1 2 2
2
n 2 4 (1 2 )
Brock [5] reported that the optimal value of n
damping as follows 1 2 2 2 (2 2 ) 4
P11 S f (42)
2 3s n 2 mr2 r2 4
12 22
opt * (36)
2 Minimum condition are expressed as
Substituting Eqs.(34-35) into Eq. (36) we obtain the
optimal value of damping ratio as following P11 P11
0; 0 (43)
* *
2
3
opt * (37)
2 2n(1 2 ) The optimal parameters of the DVA for design that was
determined by solving the Eqs.(42,43)

40
Journal of Science and Technology

2 n(2 )
2 ctd c td
opt (44) 0; 0 (54)
2 n (1 2 ) * *

The optimal parameters of the DVA are determined by


2 2 (4 3 2 )
opt (45) solving the equation Eqs.(53-54)
2 2 n(1 2 )(2 2 )

3.3.Maximum of equivalent viscous resistance opt * (55)
(MEVR) for optimal design
n(1 2 )

The first step of this method is to specify these


2
quadratic torques. By solving the Eq.(41) these opt * (56)
quadratic torques for vibration of shaft were obtained as 2 n

Sf 4. Numerical simulations and discussions


P32 (46)
2 2 r4 mr2 2s In this paper, we survey the shaft with the
parameters in Table 2. The shaft rotating is disturbed by
the harmonic torque M(t) of amplitude 5 Nm and
4 2 4 n2 4 4 n2
Sf 2 4 2 2 frequency 18.849 rad/s.
4
n 2 n
2 2 2

P33 (47) Table 1. Value of optimal parameters


2s n 2 mr2 r4 4
Parameters FPM MQT MEVR
S f ( 2 2 n 2 ) 0.6670 0.6703 0.6737
P34 (48)
2ns 2 r4 mr2 2 0.0656 0.0537 0.0541

After short caculation the Eqs.(4,5) we obtained Table 2. The input parameters for simulation

mr r2 ks nka e12a nca e22a M (49) Parameters Value Units


mr 5.0 kg
Hence the equivalent resistance torque on the primary ma 0.1 kg
structure which was obtained as 0.1 m
M eqv nka e12a nca e22a (50) 0.1 m
e1 0.06 m
Substituting Eqs.(7,8) into Eq.(50), this becomes e2 0.08 m
n 6 -
M eqv nma 22s 2 r2a n mas 2 r2a (51)

Thus the equivalent resistant coefficient of the DVA on


the primary structure was obtained as

n ma s 2 r2 a

nma 2 2s 2 r2 a
ctd (52)
2

If the primary system is excited by random moment


with a white noise spectrum Sf, then the average value of
Eq.(52) are the components of the matrix P in Eq.(41),
Lyapunov equation, this means

n mas 2 r2 P34
Fig. 4. Torsional vibration with optimized DVA
nma s r P32
2 2 2 2

ctd (53) Table 1 describes the optimum value that


P33
corresponding to the input data of Table 2. Simulation
Maximum condition are expressed as results with optimal parameters described in Fig. 4.

41
Journal of Science and Technology

These results show that torsional vibration of shaft (NAFOSTED) under grant number 107.02-2016.01.
without DVA has a harmonic form amplitude of about
References
0.02 rad.
[1]. Den Hartog J.P., Mechanical Vibrations, 4th Edition,
Figure 4 shows that in the first 0.4s, the amplitude McGraw-Hill, NY, (1982).
of the torsional vibration reduces rapidly. Effectiveness
of the optimal DVA using FPM is highest in [2]. Luft, R.W., Optimal Tuned Mass Damper for building,
comparison to the two other methods, however, the J. Struct. Div., ASCE, 105(12) (1979) 2766-2772.
difference between the viration response curves are [3]. Warburton G.B., Optimum absorber parameters for
negligible, especially vibration responses of the system various combinations of response and excitation
with MQT and MEVR are nearly the same. This shows parameters, earthquake Engineering and Structural
the strength of the fixed-points theory compared to Dynamics, (1982) 381-401.
other analytical methods. After the above period, the [4]. Khang N.V, Engineering Mechanics, Viet Nam
torsional vibration of the shaft shifts to the steady state Education Publishing House, (2009).
with a very small amplitude of about 1.20E-03 rad. At
[5]. Brock J.E, A Note on the Damped Vibration Absober,
this stage, the vibration responses with optimal DVA J. Appl. Mech., 13(4). A-284, 1946
determined by all methods are almost identical.
[6]. Nishihara O, and Asami T, Close-form solutions to the
5. Conclusions exact optimizations of dynamic vibration absorber
(minimizations of the maximum amplitude
In this paper, three analytical methods have been
manification factors), Journal of Vibration and
developed and examined for new shaft model. The Acoustics, 124, 576-582, 2002.
same procedure as in the conventional analytical theory
has been used to derive the optimum tuning and [7]. Liu K, and Liu J, The damped dynamic vibration
damping ratios of the device. Research results are absorber: revisited and new result, Journal of Sound
and Vibration 284, 1181-1189, 2005.
verified by numerical simulation with high reliability.
The optimal parameters were determined in analytical [8]. Anh N.D, and Nguyen N.X, Design of TMD for
form and furthermore leads to the simple explicit damped linear structures using the dual criterion of
formulas. The results presented in this paper may offer equivalent linearization method, International Journal
new ways of using the device over the conventional one. of Mechanical Sciences 77, 164-170, 2013.

Acknowledgements [9]. Nagashima I., Optimal displacement feedback control


law for active tuned mass damper, Earthquake
This research is funded by Vietnam National engineering and structural dynamic, 30: 1221-1242,
Foundation for Science and Technology Development (2001).

42
Journal of Science & Technology 119 (2017) 043-047

Effect of Quenching Media on Distortion of C-Ring Specimen Made by


100Cr6 Steel - Simulation and Experiment
Tran Thi Xuan1*, Nguyen Van Tu1, Le Thi Chieu2, Vu Dinh Toai1
1
Hanoi University of Sciense and Technology - No. 1, Dai Co Viet Str., Hai Ba Trung, Ha Noi, Viet Nam
2
Vietnam Casting - Metallurgy Society, 4th floor, No.91 Lang Ha Str., Dong Da, Hanoi
Received: December 18, 2016; accepted: June 9, 2017

Abstract
Polymer quenchants based on polyalkylene glycols (PAG) are currently the most widely used type of
aqueous quenchants. In this paper, the numerical simulation method by Sysweld software was used to
predict the distortions of C-ring model using 100Cr6 steel when quenched in the difference media. The
simulation results are in good agreement with the experimental values, and they also showed that the top of
C-ring was opened after deformation. The gap opening of C-ring specimens when quenched in PAG 20%
solution is the smallest in comparison to the specimens quenched in water, oil and PAG 10% solutions. On
the other hand, the research results also show that the distortion at the top of C-ring is the largest and they
are decreased from the top downward the bottom of C-ring.
Keywords: Distortion, quenching, numerical simulation, C-ring, 100Cr6 steel, poly(alkylene glycol)

1. Introduction* quench some kind of materials. Subsequently,


polymer quenchants were developed and are used to
Heat treatment is used to improve some of the
control distortion and to prevent the crack in steel
mechanical properties of steel components, and
parts during the quenching process. Currently, a
commonly involves a quenching step which may
number of polymers have been used as the
cause undesired geometrical distortions in the
quenchants for heat treating applications. These
processed parts. The dimensional accuracy of these
include: poly(vinyl alcohol), poly(alkylene glycol),
parts is affected and leads to production and
Poly(acrylamide), polyvinylpyrolidine, poly(sodium
economic losses. An example of this situation is the
acrylate), poly(ethyl oxazoline). Of these, the most
production of rolled and heat treated rings with large
commonly used are poly(alkylene glycol) PAG,
diameters and small thickness, quenching causes out-
poly(vinyl pyrrolidone) PVP, and poly(ethyl
of-roundness of the rings.
oxazoline) PEOX. However, the most common
In quenching process, the quenchant (namely quenchants encountered for induction heat treating
the heat transfer coefficient of quenchant) decided to applications worldwide are PAG type [1].
microstructure, hardness, residual stresses and
The simulation of metal heat treatment was
distortion of quenched components. Requirements set
studied by some authors, but the simulation of fully
that the quenchants must have the high enough of the
mechanical physical metallurgical behavior is
cooling rate to transform austenite into martensite
researched by the very little author, because of the
with the very high hardness, at the same time they
limit of the simulation software and computer
have the small enough of the cooling rate to limit the
capacities. Among the commercial simulation
distortion of quenched details. Therefore, the research
software today, Sysweld software is considered as the
in order to limit the distortion of the steel parts when
strongest and most complete application for the heat
quenching is very important.
treatment of metals. For deformation problems,
The oil quenchants are often used, but they Sysweld not only calculates the distortion due to
present both fire and smoke hazards. For many years, thermal expansion as other software, it also calculates
water-in-oil emulsions were used to obtain better fire- the distortion caused by the phases transformation as
resistance but these fluids were particularly shown in Fig. 1. [2].
susceptible to biological attack. Water quenchants
Based on above advantages, the Sysweld
although harmless and cheaper, but the very high
software was selected to simulate the distortion of C-
cooling rate of its can cause the crack in quenched
ring model made by 100Cr6 steel when quenched in
components and therefore it is not appropriate to
various media, they are including water, oil and PAG
liquid solution of 10% and 20%. After simulation, the
*
Corresponding author: Tel.: (+84) 986.985.861 results are verified by experiment.
Email: xuan.tranthi@hust.edu.vn
43
Journal of Science and Technology

Fig. 3. FEM model of the C-ring specimen.


Fig. 1. The strain (distortion) caused by phases PAG 10%
transformation in steel [2]
PAG 20%

Oil

Fig. 4. HTC of some quenching media

Fig. 2. Geometry and dimensions of C-ring (all


dimensions in mm)
Water
2. Methodology
2.1 Modeling of the quenching process
The testing of the Navy C-ring type was
described by Narazaki and Totten (2006) as a
procedure to evaluate the propensity for quenching
distortion in several materials. M. Manivannan et al
(2014) used this test to study and predict the
distortion in heat treatment components [3]. Until
now, the C-ring has been one of the most common Fig. 5. HTC of water quenchant
types of specimen used for observing the dimensional Combine the two features mentioned above, the
changes (distortion) after heat treatment. In this author uses the type of element which has 5 faces and
research, the shape and dimensions of the C-ring 15 nodes with straight and curved edges combination
specimen are given in Fig. 2. The material employed for meshing the C-ring model. To achieve the
for the C-ring specimen was a 100Cr6 (DIN 17230) necessary precision, the size of the elements must be
with the nominal chemical composition (certified by small enough [4] and in this study, the model has
the supplier) of 1.04% C, 0.26% Si, 0.33% Mn, meshed with the biggest element edge is 2 mm.
0.31% Ni, 1.53% Cr, 0.01% Mo, 0.01% Mo.
To build the simulation project, the model must
Using Sysweld software to discrete the C-ring also be assigned a full set of parameters on the
model in Fig. 2, we get the finite element model as mechanical - physical - metallurgical properties of the
shown in Fig. 3. Here, by the cross section of the material and the other simulation conditions. For the
model have the curve, so to ensure the accuracy when problem of quenching, the heat transfer coefficient
meshing the model, the authors use element type has (HTC) at the surface of the model which is the
curved edges. On the other hand, in the direction of boundary condition of simulation. The value of an
thickness, the model has a uniform cross-section HTC function at the surface of the model is
shape, so the element type with the straight edge will determined by experiment and will be incorporated
fit in the direction of thickness. into the Sysweld software to simulate (Fig. 4, 5):

44
Journal of Science and Technology

2.2 Experimental Procedure 3. Results and Discussion


The C-ring specimens were heated up to 850 C Using Sysweld software to perform the
in Nabertherm (Germany) furnace and held at this simulation of quenching process for the model in Fig.
temperature for 20 minutes. Then, the samples are 3 with the full material parameters of 100Cr6 steel
quenched in the water, oil and PAG quenchants of and boundary conditions as shown in the Fig.s 4 and
10% and 20%. 5, we have determined the distortions of the model
after quenched 500 seconds corresponding to the
The measurement of the C-ring dimensions
quenchants as follows:
before and after the heat treatment was performed on
the coordinate measuring machine, model YM21 3.1. Distortion of the C-ring specimen when
(Fig. 6) with an accuracy of 0.001 mm. The quenched in PAG 10%
dimensional change was analyzed: gap opening G
Fig. 8 displays the simulated displacement in the
based on Fig. 7, the dimensional change may be
x direction of the points in the cross-section of the C-
expressed as: G = n n
ring which is quenched in PAG 10% solution. The
results show that in the right half of C-ring model, the
nodes are displaced in the positive direction of the Ox
axis and in the left half of C-ring, the nodes are
displaced in the negative direction of the Ox axis.
Thus, the distortion on the specimen tends to open up
the C ring. The node 480 is displaced along the
positive direction of the Ox axis with a distance of
approximately 0.066 mm, while node 402 is
displaced along the negative direction of the Ox axis
with the same distance. It means that the gap opening
of the C-ring is about 0.132mm. The simulated results
in Fig. 8 also show that the distortion of C-ring
specimen decreases from the top (node 480) to the
Fig. 6. Universal Microscopy YM21 bottom (node 432). At the node 801, the displacement
is about 0.02 mm and the displacement of node 432 is
nearly zero.
The comparison between the experimental and
simulated values for the distortion of C-ring when
quenched in PAG 10% solution as shown in Table 1.
The compare shows that the relative difference for
gap opening is smaller than 4%.
Table 1. Comparison of experimental and simulated
distortions of the C-ring after quenched in PAG 10%
Quenchant Gap opening (G) (mm) Difference
Fig. 7. C-ring specimen in before (gap opening n) and (%)
after (gap opening n) quenching. Experimental Simulation
PAG 10% 0.128 0.132 3.03

3.2. Distortion of the C-ring specimen when


quenched in PAG 20%
The computation results of displacements in the
x-direction of the points in the cross section of the C-
ring model when quenching in PAG 20% as shown in
Fig. 9. In this case, the computation results show that
the biggest displacement in the x direction (Ux) is
0.043 mm, achieved at node 480 and displacement at
the node 402 is -0.043 mm. It means the gap opening
of the C-ring is about 0.086 mm. The computation
results in Fig. 9 also show that the Ux distortion at the
Fig. 8. Simulated displacements in the x direction of
the C-ring, after quenched in PAG 10% solution

45
Journal of Science and Technology

bottom zone of the C-ring is the smallest, namely the Table 2 presents a comparison between
displacement of the node 432 is zero. the experimental and computation values for the
distortion of the specimen when quenched in PAG
Comparing these results with the sample results
20% solution. In this case, the gap opening
are quenched in PAG 10% solution we can see that in
computation result of the C-ring is only of 0.086 mm
this case, the samples have the smaller deformation.
and the measuring result on the coordinate measuring
This means that when quenching in polymer
machine is 0.08 mm. Thus, the corresponding error,
quenchants if the concentration of polymer is higher
in this case, is 7%.
then the distortion of the sample is smaller. This can
be explained by the cooling rate of the specimen to be 3.3. Distortion of the C-ring specimen when
quenched in PAG 20% is lower than it has been quenched in water
quenched in PAG 10% [5].
With such samples, in water heat treatment as we
Table 2. Comparison of experimental and simulated obtain the results as shown in Fig. 10. The simulation
distortions of C-ring after quenched in PAG 20% results in this case also reflect a trend of deformation
of the sample is similar to the heat treatment in the
Quenchant Gap opening (G) (mm) Difference polymer quenchants. That is, after quenching, the C-
Experimental Simulation (%) ring samples are also deformed with trend opens and
PAG 20% 0.08 0.086 7 the deformation of the sample is also distributed
symmetrically through the center. The nodes in the
Table 3. Comparison of experimental and simulated right half are moved in the positive direction of the
distortions of C-ring after quenched in water Ox axis, while the nodes in the left half are moved in
Quenchant Gap opening (G) (mm) Difference the opposite direction, making the sample was opened
Experimental Simulation (%) - This is like observing in experimental samples. The
Water 0.176 0.18 2.2 simulated deformation in Ox direction at node 480, in
this case, is 0.09 mm and -0.09 mm at nodes 402.
That is the gap opening of the C-ring is 0.18 mm.
Comparing these results with the results of the
specimen to be quenched in PAG 10% and PAG 20%
solution, we see that the samples, in this case, have
the greater deformation. It is caused by the cooling
rate of the sample when quenched in the water is
higher than PAG solution [6].
Compare the measuring gap opening of the C-
ring with the simulated values when quenched in
water, we received the result as shown in Table 3. In
this case, the relative difference for gap opening is
smaller than 3%.

Fig. 9. Simulated displacements in the x direction of 3.4. Distortion of the C-ring specimen when
the C-ring, after quenched in PAG 20% solution. quenched in the oil
When the simulation of the quenching process of
the research model in traditional quenchant is oil, we
received the displacement in an x-direction of the C-
ring model as shown in Fig. 11. Like the simulated
results of the quenched model in water, PAG 10%
and 20%, when the specimen was quenched in oil, the
displacement in the x-direction at node 480 is also
largest (about 0.059mm). In this case, the gap
opening of the C-ring model is about 0.118 mm.
Thus, the distortion of the C-ring model, when
quenched in oil, is smaller in compare with the model
which quenched in PAG 10% and water, but larger
than the quenched model in PAG 20%.

Fig. 10. Simulated displacements in the x direction of The comparison between the experimental and
the C-ring, after quenched in water. simulated distortions of the C-ring when quenched in

46
Journal of Science and Technology

oil as shown in Table 4. It is seen that the relative also reflect the properties of the quenching media as
difference for gap opening is smaller than 5%. shown in the Fig.s 4 and 5: the HTC of the specimen
to be quenched in PAG 20% is smallest, it will have
The simulated and measured distortions of the
minimal distortion.
C-ring specimens, when they were quenched in the
different quenchants, are shown in Fig. 12. We see Acknowledgments
that if the specimen is quenched in water then the gap
This research is funded by Hanoi University of
opening of C-ring is largest; whereas if that was
Science and Technology (HUST) under project
quenched in PAG 20% solution, it will have minimal
number T2016-PC-150.
distortion.
4. Conclusions
Table 4. Comparison of experimental and simulated
distortions of C-ring after quenched in oil. Based on the simulation and actually measuring
results of C-ring deformation above, the following
Quenchant Gap opening (G) (mm) Difference conclusions can be drawn:
Experimental Simulation (%)
Oil 0.113 0.118 4.2 The simulation results based on finite element
method are in good agreement with the experimental
values. The maximum difference between the
experimental and the simulated values for the C-ring
gap opening was only about 7%. I.e. above
simulation method is accurate and reliable.
The gap opening of the C-ring specimen that
quenched in PAG 10% solution is smaller than the
specimen quenched in water but larger than quenched
specimen in oil.
The quenching distortion of C-ring specimen is
the largest at the top (including the nodes 480, 402,
474 and 396) and decreases to zero at the bottom of
C-ring (node 432).
With PAG 20% aqueous solution, the gap
Fig. 11. Simulated displacements in the x direction of
opening of the quenched C-ring specimen is the
the C-ring, after quenched in oil.
smallest. i.e. the PAG 20% quenchant is most suitable
to quench the 100Cr6 steel when compared with oil,
water or PAG 10% quenchants.
References
[1] George E. Totten, Polymer quenchants for introduction
heat treating applications: The basic, Union Carbide
Corporation 771 Old Saw Mill River Road Tarrytown,
NY 10591.
[2] ESI Group, Sysweld 2010 Reference manual, January
2010.
Fig. 12. Comparison of distortion when quenched in
different quenchants: simulation and experiment. [3] M. Manivannan, D. O. Northwood & V. StoilovNavy,
Use of Navy C-rings to study and predict distortion in
The simulated and measured results also shown heat treated components: experimental measurements
that the distortion of the specimen, when quenched in (2014), University of Windsor, Ont., Canada.
PAG 10%, is nearly equal to the distortion of the [4] Bathe, K. J., Finite Element Procedures in Engineering
sample when quenched in the oil. In the process of Analysis (1982) Prentice-Hall, Englewood Cliffs.
measuring the deformation, due to the device's
accuracy and the operator's subjective factors, the [5] Mr. P.K.Deval, New Generation Polymer Quenchant for
heat transfer coefficient is additionally calculated Heavy Forgings, Dy. General Manager Hardcastle
Petrofer Pvt. Ltd
based on the simulation software through the cooling
speed value provided by experimental measurements. [6] M. Eshraghi-Kakhki, M.A. Golozar, A. Kermanpur,
These factors can also lead to some minor differences Application of polymeric quenchant in heat treatment
between experiment and calculation. These results of crack-sensitive steel mechanical parts: Modeling and
experiments (2010), Elsevier Ltd.

47
Journal of Science and Technology

48
Journal of Science & Technology 119 (2017) 048-053

Dynamic Analysis of Complex Composite Tubes


by Continuous Element Method
Le Thi Bich Nam1, *, Nguyen Manh Cuong1, Tran Ich Thinh1,
Duong Pham Tuong Minh2, Le Quang Vinh3
1
School of Mechanical Engineering, Hanoi University of Science and Technology
2
Faculty of Mechanical Engineering, Thai Nguyen University of Technology
3
Department of Mechanical Engineering, Viet tri University of Industry
Received: December 08, 2016; accepted: June 9, 2017
Abtract
This paper presents a new continuous element model for studying the dynamic behavior of complex tubes
composed by cylindrical shell, annular plate and conical shell which are widely used in automobile exhausts,
sewer pipes and pipelines. Based on the analytical solutions of the system of differential equations for
cylinders, cones and annular plates, the dynamic stiffness matrix for complex composite tubes has been
established and assembled. Natural frequencies and harmonic responses for composite tubes have been
calculated and validated by comparing with the literature and with Finite Element Method. The assembly
procedure of Continuous Element Method demonstrated remarkable advantages in terms of precision,
volume of data storage, calculating time and larger range of studied frequencies.
Keywords: Complex shell, tube, Continuous Element Method, composite shell

1. Introduction* Naghdis thin shell theory in conjunction with a


multilevel partition technique. Kang [6] investigated
The complex composite tubes have many
the free vibration of joined thick conicalcylindrical
applications in various branches of engineering such
shells with variable thickness using a three-
as automobile, aeronautical, marine, civil and power
dimensional Ritz method.
industry. Hence, the comprehension of dynamic
behaviors of such structures is of great important in The main draw-back of traditional methods like
order to design and fabric safer and stronger FEM is the discretization of the domain which causes
composite shell structures. Different methods for errors in dynamic analysis, especially in medium and
analyzing free vibrations of isotropic and composite high frequencies. The Continuous Element Method
joined conical-cyclindrical shells have been applied. (CEM) or Dynamic Stiffness Method (DSM) based
Liang et al. [1] have investigated the dynamic on the closed form solution of the system of
characteristics of a symmetric cross-ply laminated differential equations of the structure is developed to
conical shell with an annular plate at the top end overcome these difficulties. The CE models for
using the transfer matrix method. The free vibration composite cylindrical shell presented in works of
of joined isotropic conical-cyclindrical shells has Tran Ich Thinh and Nguyen Manh Cuong [7] imposes
been solved by Irie et al. [2] using the transfer matrix a considerable advancement of the study on CEM for
approach. Caresta and Kessissoglou [3] have composite structures. Recently, the new research for
analyzed the free vibrations of joined truncated thick laminated composite joined cyclindrical-conical
conical-cyclindrical shells using a power series shells by Nguyen Manh Cuong et al. [8] has
solution and a wave solution method. Kouchakazadeh emphasized the strong capacity of DSM in
and Shakouri [4] presented a study dealing with assemblying complex structure. Le Thi Bich Nam et
vibrational behavior of two joined cross-ply al. [9] have proposed a new continuous element
laminated conical shells, joined cylindrical-conical formulation for cross-ply annular plates.
shells using thin-walled shallow shell in which the
However, there is a lack of researches on
expressions among stress resultants and deformations
combined cylinder-annular plate-cone structures. The
are extracted as continuity condition at the joining
purpose of this paper is presenting a new continuous
section of the cones. Recently, Qu et al. [5] proposed
element for thick combined cross-ply laminated tubes
free vibration characteristics of conicalcylindrical
composed by cylinder, annular plate and conical
spherical shell combinations with ring stiffeners by
shells taking into account the effect of shear
using a modified variational method, Reissner
deformations and rotational inertia as well as the
* continuity conditions at joined sections. Obtained
Corresponding author: Tel.: (=84) 24.3868.0103 results using this model have been compared with
Email: nam.lethibich@hust.edu.vn

48
Journal of Science & Technology 119 (2017) 048-053

those of FEM and with available results in other with R=R1+s.sin (3)
investigations and good agreement was obtained and
advantages of CE model have been confirmed. 2.1.3. Force resultantsdisplacement relationships

2. Formulation of thick cross-ply composite shells Qs


of revolution
Q
2.1. Theory of composite conical shells
2.1.1. Constitutive relations
The plane stress-reduced stiffnesses of a
laminate composite composed by N orthotropic layers
are: Fig. 2. Force and moments resultants on shells
E1 E E2
Q11 , Q12 12 2 , Q22 , Fig. 2 illustrates the force and moment
1 1221 1 1221 1 1221 resultants on a shell. The forces-displacements
Q66 G12 , Q44 G23 , Q55 G13 (1) expressions for laminated composite conical shell are
N
1 N k 4 written as follows:
Aij Q ( z k 1 z k ) , Aij Qij ( z k 1 z k ) (i, j 4,5),
k
ij
4

k 1 4 k 1 u0 A12 v B
N S A11 u0 sin 0 w0 cos B11 S 12 S sin
1 N 1 N s R s R
Bij Qijk ( z k21 z k2 ), Dij Qijk ( z k31 z k3 ) (i, j 1,2,6)
2 k 1 3 k 1 w0 cos 1 w0
QS kA55 S , Q kA44 0
where Ei,Gij, 12, 21: elastic constants of the kth layer, s R R (4)
Aij, Bij, Dij: laminate stiffness coefficients and zk-1 and u0 A22 v B
N A12 u0 sin 0 w0 cos B12 S 22 S sin
zk are the boundaries of the kth layer. s R s R
2.1.2. Strains, stress and forces resultant v 1 u 0 sin 1 S sin
N S A66 0 v0 B66
s R R R s R
Following the First Order Shear Deformation u0 B12 v0 w0 cos S D12
Therory (FSDT), the displacement components are M S B11 u0 sin D11 S sin
s R R s R
written as (see Fig. 1):
u0 B22 v D
M B12 u0 sin 0 w0 cos D12 S 22 S sin
us, , z, t u0 s, , t zs s, , t , s R s R
v0 u 0 sin 1 S sin
vs, , z, t v0 s, , t z s, , t M S B66 0 D66
s R R R s R
ws, , z, t w0 s, , t (2) where k is the shear correction factor (k=5/6), (Ns, N,
Ns) are force resultants, (Ms, M, Ms) are moment
with u0,v0,w0: displacements of the point Mo at the
resultants and (Qs,Q) are shear force resultants at s
median radius of the shell and S, : rotations of a and directions, respectively
transverse normal about the -axis and s-axis.
2.1.4. Equations of motion
The equations of motion using the FSDT for
laminated composite conical shell are [8]:
N S sin N S
N S N 1 I 0u 0 I 1 S
s R R
N S 2 sin 1 N cos
N S Q I 0 v0 I 1
s R R R
M S sin
Fig. 1. Geometries of a shell edge by FSDT M S M 1 M S QS I 1 u0 I 2S
s R R (5)
2.1.2. The strain-displacement relations of conical
M S 2 sin 1 M
shell M S Q I 1 v0 I 2
s R R
u 0 1 S sin QS 1 Q sin cos
S k S kS s QS N I 0 w 0
s R s R s s R R R
N zk 1
1 v 0 cos 1 w0
u 0 sin w0 cos , Z 0 with I i z dz (i 0,1,2)
(k ) i

R R R k 1 zk

1 v0 1 u 0 sin
where (k): material mass density of the kth layer.
k S sin S v0 ,
R s R R

49
Journal of Science & Technology 119 (2017) 048-053

The equations for cylindrical shells and annular 4.1. Continuity conditions
plates can be derived from equations (4), (5) by using
The continuity conditions for assembling tubes
= 0o and = 90o respectively.
composed by cylinder, cone and annular plate are
3. Continuous element for thick composite shells of
revolution expressed as follows [3]:
3.1. State vector u1 = u2 cos w2 sin, v1 = v2, s1 = s2 , (9)
The state-vector for the investigated conical w1 = u2 sin + w2 cos, Qs1 = Ns2 sin + Qs2 cos,
shells is yT = {u0, v0, w0, S, , NS, NS, QS, MS,
Ns1 = Ns2 cos - Qs2 sin, Ms1 = Ms2, Ms1 = Ms2
MS}T. Using the Fourier series expansion, state
variables for the circumferential wave m are written 4.2. Assembly procedure
as:
The performant assembly procedure of
uo (s, , t ), wo (s, , t ), s (s, , t ), N S (s, , t ), QS (s, , t ), M S (s, , t )T Continuous Elements presented in our previous
u

research [8] is used to construct the Dynamic
( s), wm ( s), s m ( s), N S m ( s), QS m ( s), M S m ( s) cos me it
T
m
m 1
Stiffness Matrix for combined cylindrical-annular
vo ( s, , t ), ( s, , t ), N ( s, , t ), Q ( s, , t ), M ( s, , t )T plate-conical shells. First, the dynamic stiffness

matrix for cylinder, annular plate and cone must be
v ( s ), m ( s ), N m ( s), Q m ( s ), M m ( s ) sin me it (6)
T
m
evaluated separately. Then the dynamic stiffness
m 1 matrix for the tube can be constructed by employing
Substituting (6) in (4) and (5), a system of the assembly procedure similar to those of FEM for
differential equations in the s-coordinate can be assembling stiffness matrix illustrated in Fig. 3.
expressed in the matrix form as: dym / ds Am (s, )y m
Kcylinder
with Am is a 10x10 matrix (7)
Kannular
3.2. Dynamic stiffness matrix K() K() = plate
Kcone
Then, the dynamic stiffness matrix K()m for
conical shell is determined by [8], [9]:
T121 T11 T121
K ( ) m 1 Fig. 2. Construction of dynamic stiffness matrix for
T21 T22 T12 T11 T22 T121 m composite tubes
L

A( s , ) ds T T 5. Results and discussions


with Tm e 0 11 12 (8)
T21 T22 5.1. Modal analysis
4. Continuous element method for composite tubes A Matlab program based on presented
formulations is developed for analyzing the vibration
Consider a joined cylindrical shell-annular
of a joined cylinder-annular plate-cone tube. First, a
plate- conical shell in Fig. 3 with R1: radius of the
composite conical shell with an annular plate at the
cylinder, R2 and R3: small and large radius of the
top end subjected to the free-clamped boundary
cone. L1 and L2: lengths of the cylinder and cone
condition is taken for example. Geometric properties
along its generator, h: thickness of shells, : half cone are R = 200 mm, R1 = 180 mm, R2 = 100 mm, h = 2
angle. mm, and L2 = 100 mm where R and R1: radius of
large and small edges of the cone, R2: inner radius of
the annular plate, material parameters are E1 = 135
GPa, E2 = 8.8 GPa, G12 = G13 = G23 = 4.47 GPa and
12 = 0.33, = 1600 kg/m3 with layer configuration
[90/0/0/90] (Material 1). Obtained dimensionless
frequencies =R(h2/D11)1/4 calculated by the
present formulation and by Liang et al. [1] using both
FEM and Transfer matrix method are compared in
Table 1. Here m represents the number of
circumferential waves and n the number of axial half
Fig. 3. Geometry of a tube composed by cylinder, annular
plate and cone waves. It is easy to remark that the present method
gives high precise results which are very closed to
those from Liang et al. [1], especially to the transfer

50
Journal of Science & Technology 119 (2017) 048-053

matrix method solution. Tiny differences (0.07% to model is valid for analyzing the considered tube with
3.49%) between our formulation and those in [1] different modes and boundary conditions.
confirm that the present solution is exact and the
5.2. Harmonic responses
present formulations can be applied to study the
dynamic behavior of tubes. This section confirms important advantages of CE
models with respect to other methods for complex
There is no comparative data from literature for
structures. Fig. 3 shows the comparison of FE and CE
complex composite tubes. Therefore, obtained natural
harmonic responses for the studied tube subjected to
frequencies of our CE program will be validated with
the C-C (left) and F-C boundary condition (right).
respect to those of FEM. For FE models, the Ansys
SHELL181 elements with the 180x20 mesh assure It is seen that using a raw mesh (10x80) the
good results after a convergence test. The parameters discrepancies between FE and CE curves are noticed
of the considered tube are: h = 0.002m, R1 = 0.1m, R2 from 1495.5 Hz for the 1st test and at 1660 Hz for the
= 0.18m, R3 = 0.2 m, L1 = L2 cos = 0.1 m, Material second one. More precise FE solutions are obtained
1. The comparison of CE and FE results for clamped- by using a finer meshing (20x120) because the FE
clamped (C-C), free-clamped (F-C) and supported- curves coincide with CE one within a larger
clamped (S-C) composite tubes for various vibration frequency range. However, important differences
modes is illustrated in Table 2. It is seen from this between two solutions still occur from 2556 Hz for
table that good agreements are noticed between CE the 1st test and from 2413,5 Hz for the second one. It
and FE solutions. CE model give excellent solutions is clear to remark that FE models converge towards
which are close to FE results with small errors those of CE when reducing the element sizes.
varying from 0% to 3.75%. The precision of our
Table 1. Comparison of R 4 h 2 / D11 for a cross-ply combined conical shell-annular plate with layer scheme
[90o/0o]s, Material 1 and R1 = 200mm, h1 = h2 = 2 mm, R2 = 180mm, 1 = -11.77o

Method m=1 Errors (%) m=2 Errors (%)


n=0 FEM [1] 5.4672 0.18 11.745 0.35
Transfer matrix [1] 5.4575 0 11.786 0
CEM 5.3911 1.23 11.7452 3.49
n=1 FEM [1] 5.9070 0.88 11.876 0.65
Transfer matrix [1] 5.8554 0 11.799 0
CEM 5.8660 0.18 11.7904 0.07
n=2 FEM [1] 7.1143 0.79 12.048 0.39
Transfer matrix [1] 7.0588 0 12.001 0
CEM 7.1306 1.00 12.117 0.95

Frequency (Hz) Frequency (Hz)


-40 -60
2656.5
1.5
178.5
355.5
532.5
709.5
886.5
1063.5
1240.5
1417.5
1594.5
1771.5
1948.5
2125.5
2302.5
2479.5

2833.5
709.5

1594.5

2479.5
1.5
178.5
355.5
532.5

886.5
1063.5
1240.5
1417.5

1771.5
1948.5
2125.5
2302.5

2656.5
2833.5

-60 -80
-80 -100
20 log10|w|

20 log10|w|

-100 -120
-120
-140
-140
-160
-160
-180
-180 FEM 80x10 FEM 80x10
-200
-200 FEM 120x20 FEM 120x20
-220 CEM 3 -220 CEM

Fig. 3. Comparison of harmonic responses computed by FEM and by CEM for a C-C cylinder-annular plate-cone tube
(Left) and a F-C cylinder-annular plate-cone tube (Right)

51
Journal of Science & Technology 119 (2017) 048-053

Table 2. Comparison of natural frequencies for a composite tube with C-C, F-C and S-C boundary conditions (h = 0.002m,
R1 = 0.1m, R2 = 0.18m, R3 = 0.2 m, L1 = L2 cos = 0.1 m. Material 1)

Boundary Mode Frequency (Hz)


Condition n m FEM (180x20) CEM Error (%)
1 1 1036.1 1035.7 0.04
1 2 1047.4 1047.7 0.03
1 3 1102.6 1104.2 0.14
1 4 1235.6 1238.2 0.21
C-C 1 5 1463.2 1464.9 0.12
1 6 1665.8 1660.2 0.34
1 7 1755.2 1745.7 0.54
2 5 1825.5 1820.7 0.26
1 1 184.51 187.8 1.75
1 2 268.54 279.0 3.75
1 3 619.83 626.3 1.03
2 2 1063.5 1065.3 0.17
F-C 2 1 1072.2 1072.2 0.00
2 0 1120.4 1121.5 0.10
2 3 1122.2 1124.3 0.19
1 4 1144.7 1147.9 0.28
1 1 1035.9 1035.5 0.04
1 0 1043.0 1042.6 0.04
1 2 1047.2 1048.9 0.16
1 3 1101.7 1105.9 0.38
S-C 1 4 1236.1 1241.8 0.46
1 5 1466.8 1470.1 0.22
1 6 1674.7 1662.1 0.76
2 1 1725.4 1714.6 0.63

Using a minimum meshing, CE model The exactness and validity of continuous model
demonstrates considerable advantages when dealing for different properties of geometry, vibration mode
with complex structures in terms of calculating time and boundary conditions have been confirmed by the
and the saving of data storage volume. For example, excellent agreements between obtained results with
the required time for plotting FE harmonic response those published by other researches and by Finite
curves of the 1st test are 143 minutes (10x80 FE Element Method. In conclusion, the proposed
mesh), 289 minutes (20x120 FE mesh) but is only 81 Continuous Element model consists an interesting
minutes for CE model with only 3 elements. These approach to calculate the natural frequencies of thick
CE advantages are verified for different boundary combined cross-ply laminated tube with high
conditions and for all frequency range. accuracy, especially for medium and high frequencies
where other current methods give unreliable
6. Conclusions
solutions. With a minimum meshing for complex
This research has presented a new Dynamic structures, our model accelerates the calculating
Stiffness Matrix for thick combined cross-ply speed and saves the volume of data storage.
laminated tube composed by cylinder, annular plate
The introduced Continuous Element can be
and conical shells taking into account the effect of
developed to resolve the problem of shells on elastic
shear deformations and rotational inertia as well as
foundation and containing flowing fluid, shells on
the continuity conditions at joined sections.
non-homogenous elastic foundations, shells with
damping or shells with stiffeners.

52
Journal of Science & Technology 119 (2017) 048-053

References [6] Jae-Hoon Kang (2012). Three-dimensional vibration


analysis of joined thick conical-cylindrical shells of
[1] Sen Liang, H.L. Chen, Tianning Chen, Michael Yu revolution with variable thickness, J. of Sound and
Wang (2007). The natural vibration of a symmetric Vibration 331(2012) 41874198.
cross-ply laminated composite conical-plate shell,
Comp. Struct. 80 (2007) 265278. [7] Tran Ich Thinh, Nguyen Manh Cuong (2013).
Dynamic stiffness matrix of continuous element for
[2] T. Irie, G. Yamada and Y. Muramoto (1984). Free vibration of thick cross-ply laminated composite
vibration of joined conical-cylindrical shells, Journal cylindrical shells. Compos Struct; 98:93102.
of Sound and Vibration (1984) 95(l), 31-39.
[8] Nguyen Manh Cuong, Tran Ich Thinh and Vu Quoc
[3] Mauro Caresta, Nicole J. Kessissoglou (2010). Free Hien (2014). Vibration analysis of cross-ply
vibrational characteristics of isotropic coupled composite joined conical-cylindrical shells by
cylindricalconical shells, Journal of Sound and Continuous Element Method, Proceedings of the
Vibration 329 (2010) 733751. International Conference on Engineering Mechanics
[4] M.A.Kouchakzadeh, M.Shakouri (2014). Free vibration and Automation-ICEMA3, Pages 401-408.
analysis of joined cross-ply laminated conical shells, [9] Le Thi Bich Nam, Nguyen Manh Cuong, Tran Ich
International Journal of Mechanical Sciences Thinh (2014). Continuous Element formulation for
78(2014)118125. vibration of thick composite annular plates and rings,
[5] Yegao Qu, Shihao Wu, Yong Chen, Hongxing Hua Tuyn tp cng trnh Hi ngh C hc k thut ton
(2013). Vibration analysis of ring-stiffened conical quc K nim 35 nm thnh lp Vin C hc,
cylindricalspherical shells based on a modified 10/4/1979-10/4/2014, (2), Pages 319-324
variational approach, International Journal of
Mechanical Sciences 69 (2013) 7284.

53
Journal of Science & Technology 119 (2017) 054-058

Fatigue Life Prediction under Multiaxial Variable Amplitude Loading


Using a Stress Invariant Based Criterion
Vu Quoc Huy*, Vu Dinh Quy, Le Thi Tuyet Nhung
Hanoi University of Science and Technology, No. 1, Dai Co Viet Str., Hai Ba Trung, Ha Noi, Viet Nam
Received: May 06, 2016; accepted: May 05, 2017

Abstract
Predicting fatigue damage for structural components subjected to variable amplitude loadings is a complex
issue. In order to estimate fatigue life under those loading conditions, a multiaxial fatigue criterion must
gather with a fatigue damage accumulation rule that allows capturing different damage mechanisms when
they are activated. In this paper, combinations of a stress invariant based fatigue criterion with some
damage accumulation rules are carried out to deal with variable amplitude loadings. An approach combining
three tools, including multiaxial criterion, multiaxial S-N curves and cumulative damage rule are used in this
study. Results show good correlations of fatigue life between experimental and predicted results for 1045
steel.
Keywords: Variable amplitude loading, multiaxial fatigue criterion, Damage accumulation.

1. Introduction* S-N curves and cumulative damage rule) are used in


this study. The approach includes following steps:
Damage evaluation of variable amplitude
loadings is a challenge in multiaxial fatigue Step 1: choose a reference S-N curve (under
characterization. Under the variable amplitude fully reversed torsion, for instance) and identify the
loadings, different damage mechanisms can be function of the reference S-N curve (power law
activated. To predict fatigue life, a multiaxial fatigue function) from experimental data:
criterion must gather with a fatigue damage
t1
accumulation rule that allows capturing the different xya (1)
damage mechanisms when they are activated. In 1 N
general, four categories of fatigue criteria can be
where and are material parameters, xya is shear
distinguished: critical plane approaches, integral
approaches, approaches based on the stress invariants stress amplitude and N is number of cycles.
and energy approaches [1]. Among multiaxial fatigue Step 2: predict the fatigue limit values of
criteria, the approaches based on stress invariants different constant amplitude multiaxial loading cases
have an obvious advantage of computation time by a multiaxial fatigue criterion.
compared to the critical plane and the integral
approaches. A recent stress based criterion proposed Step 3: build the S-N curve of each constant
by Vu et al. [1] shows a very good prediction quality amplitude multiaxial loading case from its fatigue
for a wide range of experimental data conducted on limit value and the function of reference S-N curve
various steels. Regarding damage accumulation rules, [3].
a comprehensive review of many approaches can be Step 4: estimate the cumulative damage and the
found elsewhere [2]. In this paper, combinations of
fatigue life of the multiaxial variable amplitude
the Vu et al. criterion and some damage accumulation
loading (many blocks) by using a cumulative damage
rules are carried out to deal with variable amplitude
rule.
loading conditions.
In this study, a fatigue criterion based on stress
2. Fatigue life prediction methodology invariants is used in step 2 and different cumulative
2.1. Prediction approach rules (Miner, DCA) will be applied in step 4. More
details about step 1 and step 3 may be found in [3].
In order to predict the fatigue life under
multiaxial variable amplitude loading, an approach 2.2. A fatigue criterion based on stress invariants
combining three tools (multiaxial criterion, multiaxial
Vu et al. [1] proposed a multiaxial criterion for
high cycle fatigue based on invariants of macroscopic
* stress tensor (denoted as Vu criterion). By
Corresponding author: Tel.: (+84) 904.169.355
Email: huy.vuquoc@hust.edu.vn
54
Journal of Science & Technology 119 (2017) 054-058

introducing quantity J2,mean, the multiaxial fatigue The presence of J2,mean quantity allows capturing
endurance criterion is established as follows: accurately effects of phase shift and frequency on
fatigue limit of material under multiaxial loading. The
f 1 J '2 (t ) 2 2 J 2,2 mean 3 I f I1, a , I1, m prediction capacity of the criterion is tested on 119
iso-frequency axialtorsion experiments and some
(2)
other more complex loadings: biaxial loading and
where 1, 2, 3 and are material parameters; J '2 (t ) asynchronous loading and the results show that the
and J 2,mean capture shear stress effect and phase shift criterion is in good accordance with the experimental
data [1]. Under proportional loading, the assessment
effect. J '2 (t ) is the second invariant of stress of the criterion can be carried out from an analytical
amplitude part defined from the deviator of the solution. Under others loading cases, the numerical
a
amplitude of the stress tensor S (t ) and J 2,mean is the implementation of the criterion is very simple and can
be easily integrated in a damage model or in a finite
mean value of J '2 (t ) during period T: element code.

1 a 2.3. Cumulative damage rules


J '2 (t )
a
S (t ) : S (t ) (3)
2 Damage accumulation models aim to account
T
accumulated damage in material based on damage
1
T 0
J 2, mean J '2 (t ) dt (4) parameters. Basically, damage parameters are used to
estimate fatigue strength under certain stress level
and loading path. Then, a relation between the
I f I1, a , I1, m is a function of I1,a and I1,m reflecting estimated fatigue life and the loading cycles is
respectively the effects of amplitude and mean value performed. Among the cumulative damage rules, the
simplest and well-known is the linear accumulation
of the hydrostatic stress. The values of I1,a and I1,m
rule proposed by Miner [4] as follows:
are defined from I1 (t ) tr ( ) that is the first
# blocks # cycles ni
invariant of the stress tensor: D (10)
i 1 N i , f j

j 1
1
I1, a max I1 (t ) min I1 (t ) (5)
2 tT tT
where D is the accumulated damage, ni is the applied
number of cycles at the life level Ni,f. Miners rule
I1, m
1

max I1 (t ) min I1 (t )
2 tT tT
(6) predicts that the failure occurs when D 1.
Damage Curve Approach rule (DCA) is a
Vu criterion is identified from two fatigue nonlinear damage accumulation [5]. The DCA
limits, under fully reversed torsion ( t1 ) and under concept is that damage accumulation proceeds along
fully reversed tension ( f 1 ), for instance. The value the curve associated with the life level at which a
of 1, 2, 3 and are determined from the two fatigue cycle ratio is applied. In general case when K block
limits: loadings are applied before failure occurs, the
equation for DCA becomes:
t1 (7)
( N1 / N 2 )0.4 ( N 2 / N3 )0.4
t f /3
2 2
{[(n1 / N1 ) n2 / N 2 ] ...
3 1 1
(8) (11)
f 1 )0.4
nK 1 / N K 1} nK / N K 1
( N K 1 / N K

4
1 2 1 (9) Note that the subscripts 1, 2, K-1, K are the
2
sequence numbers of the loadings as they occurs. A
In order to capture the effects of phase shift and mean
particular interest in (11) is the exponent 0.4. It is
stress, Vu et al. [1] proposed to distinguish two
shown that this value is reasonable for many
categories of metals based on ultimate strength Rm:
materials.
Low-strength metals (Rm < 750 MPa):
3. Application to multiaxial variable amplitude
1 = 0.65 and 2 = 0.8636; I f I1, a , I1, m I1, a I1, m loading
High-strength metals (Rm > 750 MPa): 3.1. Multiaxial S-N curves
1 = 0.65 and 2 = 0.8636;
This paragraph introduces the application of the
I f I1, a , I1, m I1, a
f 1
I1, m mentioned above approach for 1045 carbon steel
t1
under multiaxial variable amplitude loading. The

55
Journal of Science & Technology 119 (2017) 054-058

main mechanical characteristics of this steel are:


Youngs modulus E = 205 (GPa), yield stress Rp0.2 =
280 (MPa) and ultimate strength Rm = 580 (MPa).
The fatigue limits in fully reversed torsion ( t1 ) and
fully reversed tension ( f 1 ) have been reported in [6]
as: t1 169 (MPa), f 1 240 (MPa). The parameters
and in equation (1) are identified using fully
reversed torsion test data and the results are:
1206.39 , 0.7805 . With these values of
, and the fatigue limits of different multiaxial
constant amplitude loading cases, the functions of S-
N curves are as follows:
Under fully reversed torsion: Fig. 2. S-N curve under reversed tension
1.2812
xya
N 1206.391.2812
169
(12)
xya
Under fully reversed tension:
1.2812
xa
N 1206.39 1.2812
(13)
xa 240
where xa is normal stress amplitude.

Under in phase tension torsion (stress ratio


k = xya / xa = 0.5):

xa
1.2812 Fig. 3. S-N curve under in phase tension torsion
N 1206.39 1.2812
(14) (k=0.5)
xa 190
Under in phase tension torsion (stress ratio
k = 1):
1.2812
xa
N 1206.391.2812 (15)
xa 132

Fig. 4. S-N curve under in phase tension torsion


(k=1)
Fig. 1 to Fig. 4 show the correlation between
experimental data [6] and predicted values for S-N
curves under fully reversed torsion (Fig. 1), fully
reversed tension (Fig. 2), in phase tension torsion
k=0.5 (Fig. 3) and in phase tension torsion k=1
(Fig. 4). It can be seen that the predicted S-N curves
Fig. 1. S-N curve under reversed torsion fit well with the experimental points both in values
and in trends for all studied loading cases. This

56
Journal of Science & Technology 119 (2017) 054-058

reveals that the Vu criterion is suitable for the Miner or DCA damage rule is then used to estimate
prediction of multiaxial S-N curves. the cumulative damage and the coresponding fatigue
life of each load case. Table 2 shows the experimental
3.2. Experiment data
and the predicted fatigue life of all studied loading
A testing campaign of variable amplitude cases.
loading is carried out by Flacelire in [6]. The
Table 2. Fatigue life of 1045 steel
variable amplitude loading is created by multiple
blocks of constant loading. These blocks have Speci Test Miner DCA
different loading natures, including fully reversed Case -men Nf (x105) Nf (x105) Nf (x105)
torsion (denoted as To), fully reversed tension (cycles) (cycles) (cycles)
(denoted as Ta), in phase tension-torsion (denoted as
TaTo) (k = 0.5) and in phase tension-torsion (TaTo) 1 6,27
1 6,02 5,94
(k = 1). Many loading profiles are conducted based 2 6,21
on two principal element blocks (denoted as Block A
1 4,84
and Block B). All the load cases are resumed in Table
1. One or several specimens are tested for each load 2 2 5,59 6,03 6,10
case. For more details, with load case 1, the two 3 9,54
element blocks are fully reversed tension in 105
cycles (Ta) and fully reversed torsion in 105 cycles 3 1 5,65 5,66 5,78
(To). The loading profile of load case 1 includes 4 1 3.93 5,84 5,99
repeatedly one block A followed by one block B until
5 1 5.23 6,24 6,36
the failure of specimen. The same loading profiles are
applied for load cases 2, 6 and 7. For the loading 1 2,58 2,02 1,71
profile of load cases 3, 4 and 5, the block A is 6 2 1,27 2,64 2,54
respectively repeated 1, 2 and 3 times before the
block B is applied until the failure. 3 4,18 2,75 2,65

Table 1. Variable amplitude tests on 1045 steel [6] 1 4,89 5,4 5,32
7 2 6,37
Load Block A Block B 3,55 3,62
Loading profile 3 4,12
case (cycles) (cycles)
1 Ta (105) To (105) (A/B)/(A/B)/... For better comparison of the results, fatigue life
5 5 predictions are illustrated in Fig. 5. These graphs
2 To (10 ) Ta (10 ) (A/B)/(A/B)/...
display the predicted life versus the experimental life
5 5
3 To (10 ) Ta (10 ) (A)/(B/B/...) on log-log coordinates. The solid line represents
4 5
To (10 ) 5
Ta (10 ) (A/A)/(B/B/...) perfect correlation between experimental and
5 5
predicted values. The outer bound (dash lines) is
5 To (10 ) Ta (10 ) (A/A/A)/(B/A/B...) related with a factor of two of fatigue life. Data points
6 TaTo TaTo that fall below the solid line represent conservative
k=1 k=0.5 (A/B)/(A/B)/... estimations and points above represent non-
(105) (105) conservative prediction. The Fig. 5a shows the
correlation between the estimations using Miner rule
7 TaTo TaTo and the experimental data while the Fig. 5b shows the
k=0.5 k=1 (A/B)/(A/B)/... results by using DCA rule. As shown in Fig. 5a and
(105) (105) Fig. 5b, there is not a significant difference in the
predictions using Miner rule and using DCA rule.
3.3. Results and discussion This reveals that the effect of load sequence is
For the experimental tests shown in Table 1, relatively small for the considered loading cases. The
fatigue life predictions were carried out based on the predicted results of both Miner rule and DCA rule are
methodology mentioned in section 2.1. For each quite satisfactory. All the data points are in the limit
block, the life level Ni,f is determined from S-N curve zone. It can be concluded that the fatigue life
functions (12, 13, 14, 15) depending on the nature of prediction method gives good estimation for 1045
loading (Ta, To, TaTo k = 0.5 or TaTo k =1). The steel.

57
Journal of Science & Technology 119 (2017) 054-058

1,E+06 1,E+06

Non-Conservative Non-Conservative
Estimated fatigue life (cycles)

Estimated fatigue life (cycles)


Miner rule DCA rule
1,E+05 1,E+05

Conservative Conservative

1,E+04 1,E+04
1,E+04 1,E+05 1,E+06 1,E+04 1,E+05 1,E+06
Experimental fatigue life (cycles) Experimental fatigue life (cycles)

Fig. 5. (a) Miner rule and (b) DCA rule fatigue life correlation of 1045 steel under variable amplitude loading

4. Conclusion Development (NAFOSTED) under grant number


107.02-2014.25.
In this paper, the criterion based on stress
invariants developed by the present authors, was used References
to account fatigue damage accumulation under
[1] Q.H. Vu, D. Halm, Y. Nadot, Multiaxial fatigue
variable amplitude loading conditions. The criterion for complex loading based on stress
concluding remarks are as follows: invariants, International Journal of Fatigue. 32 (2010)
10041014.
The Vu criterion is suitable for the prediction
of multiaxial S-N curves. [2] A. Fatemi, L. Yang, Cumulative fatigue damage and
life prediction theories: a survey of the state of the art
Under variable amplitude loading, the for homogeneous materials, Int. J Fatigue. 20(1)
prediction approach combining three tools (1998) 9-34.
(multiaxial criterion, multiaxial S-N curves,
[3] V. Papadopoulos, Long life fatigue under multiaxial
damage accumulation rule) gives good loading, International Journal of Fatigue. 23 (2001)
estimation of fatigue life. 839849.
Both Miner and DCA damage rules are [4] M. A. Miner, Cumulative damage in fatigue, Journal
appropriate for prediction of fatigue life of of Applied Mechanics. 67 (1945) A159-A164.
1045 steel.
[5] S.S. Manson, G.R. Halford, Practical implementation
The prediction methodology used in this study of the double linear damage rule and damage curve
could be modified to deal with more complex loading approach for treating cumulative fatigue damage, Int.
such as random loading in service. Cycle counting Journ. of Fracture. 17 (1981) 169-192.
techniques and/or overload effects need to be taken [6] L. Flacelire, Contribution la modlisation du
into account in further study. dommage en fatigue multiaxiale dun acier C36
Confrontation lexprience. PhD Thesis, University
Acknowledgments of Poitiers, France (2004).
This research is funded by Vietnam National
Foundation for Science and Technology

58
Journal of Science & Technology 119 (2017) 059-065

Performance Evaluation of a 2D Front Tracking Method a Direct


Numerical Simulation Method for Multiphase Flows
Vu Van Truong
Hanoi University of Science and Technology - No. 1, Dai Co Viet Str., Hai Ba Trung, Ha Noi, Viet Nam
Received: August 24, 2016; accepted: June 9, 2017

Abstract
This paper evaluates the performance of a direct numerical simulation (DNS) method called "front-tracking"
for multiphase flows. The interface separating two fluids or two phases is represented by connected
elements that move on a fixed rectangular grid used for solving the Navier-Stokes equations. The phases
values of material properties are specified by an indicator function that is reconstructed from the interface
point location. The interface points are updated by the velocities, which are interpolated from the velocities
on the fixed grid. The method is evaluated through a thorough investigation of the performance using a
variety of verification and validation test cases including advection of the interface, computations of the
surface tension, and interplay of the viscous and interfacial tension terms. The method is then used to
simulate the evolution of the RayleighTaylor instability. Good agreement in comparison of the present
method with the previous literature proved the accuracy and capability of the method.
Keywords: DNS, Front-tracking, Performance evaluation, Multiphase flow, Rayleigh-Taylor instability

1. Introduction* cubic interpolated pseudo-particle (CIP) method. For


further discussions of these methods, the reader is
Multiphase flows play an important role in the
referred to Sussman et al. [3] for the LS method, [4]
workings of nature and engineering problems. In
for the phase field method, and Takewaki et al. [5] for
terms of mathematics, multiphase problems are very
the CIP method. Instead of advecting the marker
difficult. Therefore, exact analytical solutions are
function directly, the interface between the different
available only for the simplest problems. In addition,
fluids can be tracked using marker points, and then
experimental studies of multiphase flows are not easy
the marker function is reconstructed from the
to carried out. Accordingly, computational fluid
information of the interface points. These methods
dynamics, including direct numerical simulations
are referred to as "front-tracking" (FT) methods. One
(DNS), becomes a standard tool in multiphase flow
of the most popular FT method is that introduced by
research.
Unverdi and Tryggvason [6]. Detailed description of
For DNS, it is necessary to solve the full Navier- the method and its applications can be found in [7]. In
Stokes equations, and a number of different the above-mentioned approaches, only one single
approaches have been developed and applied. One of field governing equations are used, and the boundary
the pioneering works calls back to Harlow and Welch conditions, e.g., the surface tension force, at the
[1], in which the authors distributed marker particles interface are introduced to the equations as the source
throughout the fluid region. They solved the terms. Accordingly, these methods are called "one-
governing equations on a regular grid that covers the fluid" approaches. A long with development of the
fluid-filled and the empty part of the domain. one-fluid formulation methods, other techniques were
Accordingly, the method is called "marker-and-cell" also explored, such as the boundary-fitted lagrangian
(MAC) method. The next generation of methods for method
multiphase flows was developed gradually from the
The front-tracking method introduced in [6] has
MAC method. One of the most known methods is the
been widely used in multifluid and multiphase
volume of fluid (VOF) method that was introduced
problems [7], and recently in our works [1012].
and discussed by Hirt and Nichols [2]. In the VOF
Despite the wide use of the front-tracking method, its
method, the different fluids are identified by a marker
evaluation has been conned to validation problems
function that takes different values in the different
specic to the particular applications of interest to the
fluids. Other maker function methods include the
respective authors. This calls to question the
level set (LS) method, the phase field method, and the
performance of the method for the various multiphase
ow problems. Accordingly, this paper presents a
Corresponding author: Tel.: (+84) 915.058.146 detailed analysis of the behavior of the method over a
Email: truong.vuvan1@hust.edu.vn wide range of verication and validation problems

59
Journal of Science & Technology 119 (2017) 059-065

including advection of the interface, computations of xnf 1 xnf Vt (3)


the interfacial tension, and interplay of the viscous
and interfacial tension terms. These problems are where V is the velocity at the front point, which is
commonly used to verify and validate the accuracy interpolated from the fixed grid points using a smooth
and capability of the DNS methods for multiphase weighting function [13]
flows. We focus here only 2D problems.[8], and the
lattice-Boltzmann method [9]. V wij uij (4)

wij d x f ih d y f jh (5)

1 4h 1 cos r 2h , r 2h
d r (6)
0 , r 2h

where h is the grid spacing (Fig. 1), and i and j is the


indices of the fixed grid point. The information of the
front points is used to reconstruct an indicator that
has a value of one in one fluid and zero in the other:

I x x f n f dS (7)
f

Fig. 1. An interface represented by connected Then the density and viscosity fields are updated:
elements on a fixed grid. Information is passed
between the front points and the fixed grid 1 I 1 I 2 ; 1 I 1 I 2 (8)
2. Numerical method To calculate the interfacial tension force, we first
The fluids are assumed incompressible, calculate the net force on each front element:
immiscible and Newtonian. All phases are treated as
one fluid with variable density and viscosity . In Fl nds t s ds t 2 t1 (9)
s s
terms of the one-fluid formulation, the governing
equations include: where t is the tangents of the end points of each
element. After that, this force is transferred to the
u t uu p u uT fixed grid (so that it is included in the solution of the
Navier-Stokes equations) using the same smooth
weighting function, i.e. Eq. (6),
g n f (x x f )dS (1)
Fij Fl wij sl h2
f
(10)
u 0 (2) l

Here, u is the velocity vector, p is the pressure, g is Here sl is the length of the element. In the following,
the gravitational acceleration, and t is time. The we briefly describe the solution procedures.
superscript T denotes the transpose. The last term in
Suppose n time steps have been completed, to
Eq. (1) accounts for the interfacial tension force at the
calculate the solution at time level n+1 carry out the
interface. At the interface, denoted by f, is the following steps:
interfacial tension coefficient, is twice the mean
curvature, and nf is the unit normal vector to the 1. Update the position of the interface points
interface. The Dirac delta function (xxf) is zero [Eq. (3)]
everywhere except a unit impulse at the interfaces xf. 2. Reconstruct the indicator function, update the
The above equations are discretized using a second- material properties, and calculate the interfacial
order centered difference approximation for the tension force
spatial derivatives and an explicit predictor-corrector
method for time integration. The discretized 3. Calculate an intermediate velocity field:

u* tA n n u n n 1
equations are solved on a fixed, staggered grid using
the MAC method [1]. (11)

The interface separating two fluids is where the advection, the diffusion, the gravitational
represented by connected points on a fixed grid (Fig. body force and the interfacial tension force in Eq. (1)
1). The movement of the interface points is given as are denoted by A

60
Journal of Science & Technology 119 (2017) 059-065

4. Find the pressure field by solving the Poisson u, v 50 y 314, x 50 314 (14)
equation:
The grid resolution is 100100. Fig. 2 indicates
un1 u* t 1
n 1
p (12) that after one revolution, the interface of Zalesak's
disk is almost identical to the initial shape.
where un1 0 3.2. A circular drop in a vortical flow field
5. Compute the divergence-free time level n+1 This accuracy test introduced by Bell et al. [16]
fluid velocity field is to test how well a method to resolve thin laments
on the scale of the mesh which can occur in stretching
u n 1 u* t p n 1 (13) and tearing ows. A circular drop with a radius of
0.15 is initially located at (0.5, 0.75) in a box of
This solution procedure for time integration is first
order, to produce a second-order scheme, the 1.01.0 (Fig. 3a). The velocity field is given as:
technique described by Esmaeeli and Tryggvason u 2 cos t T sin 2 x sin y cos y
[14] is used. More detailed description of the method (15)
v 2 cos t T sin y sin x cos x
2
can be found in [7].
3. Performance tests where T is the period. This velocity field first
The following, we present the verication and stretches the drop into a thinner lament that is
validation of the method. The congurations to be wrapped around the center of the box, then slowly
adopted in these tests follow directly from the reverses and pulls the lament back into the initial
respective references, including dimensions and circular shape, i.e. at the end of the period T, the
velocity elds. shape should be the same as the initial one (Fig. 3a).
The grid resolution is 3232 with T = 8. Fig. 3b
3.1. A notched disc in rotating ow shows the drop shape at time t = T computed by the
Solid body rotation of a notched disc introduced present front-tracking method (FT) the dash line in
by Zalesak [15] (Fig. 2) is a test commonly used for Fig. 3b in comparison with the initial shape.
evaluating the accuracy of a method in maintaining a Obviously, the difference is barely visible. In contrast,
sharp corners. The initial data is a slotted circle the level set method (LS) [17] produced remarkable
centered at (50,75) with a radius of 15, a slot width of difference (the dash-dot line in Fig. 3b) even though a
5, and a slot length of 25. The domain is 100100. much finer grid 256256 was used.
3.3. Stationary drop
This accuracy test is to test how well a method
to predict the pressure difference between the inside
and outside of the drop. This pressure difference is
induced by the surface tension force acting on the
interface, as given by Laplace's law. For a circular
droplet in equilibrium the velocity should be exactly
zero. However, because of numerical errors, the
velocity field is not zero, and is referred to "spurious
currents" (Fig. 4). A good method should produce the
accurate pressure difference with spurious currents as
small as possible. There are three dimensionless
numbers that characterize the flow: the Laplace
number La d D d2 where D is the drop diameter,
and density and viscosity ratios d c and
d c . The subscripts d and c respectively represent
the fluids inside and outside the drop. The maximum
nondimensional velocity, i.e., the capillary number
Fig. 2. Comparison of the Zalesak disk interface
Ca, is defined as Ca dU max .
before (top) and after one rotation (bottom)
The velocity field is given as:

61
Journal of Science & Technology 119 (2017) 059-065

We consider a circular drop with a radius of 0.5 different techniques for computing surface tension,
placed at the center of a box of 22 with all other e.g., the continuous surface force (CSF) and the
properties set to unity except for (Fig. 4a). The grid continuous surface stress (CSS) [18], the magnitude
resolution is 6464. Accordingly, the value of is of the spurious currents produced by the front-
equal to the value of La. For instance, La = 0.12 tracking is much smaller as shown in Table 1.
yields = 0.12, and according to the YoungLaplace 3.4. A damped surface wave
equation the pressure difference is pexact = /R =
0.24. Fig. 4 indicates that the method predicts the To verify the interaction of the viscous term
pressure rise reasonable well with Ca = 2.6510-4. In with the surface tension term, we perform a
comparison with the VOF method implemented with simulation a damped surface wave between two
superposed immiscible uids, as shown in Fig. 5, and
compare the computational results with the initial
value theory of Prosperetti [19]. In a [0, 2][0, 2]
domain, two uids are initially separated by an
interface dened by

Fig. 3. Accuracy test for the advection of the interface


points. (a) The initial shape (t = 0) and the shape at
half period (t = T/2). (b) The shape of the drop at the
end of the period (t = T ) computed by the FT method,
i.e. the dash-line, compared with that computed by the Fig. 4. (a) The spurious currents generated by a
LS method [17], i.e. the dash-dot line (the solid line circular drop with La = 0.12 and (b) the distribution of
representing the initial shape). The LS method was the resulting pressure field along the line y = 1.0. The
used with a 256256 grid resolution grid resolution is 6464 with a domain of 22

62
Journal of Science & Technology 119 (2017) 059-065

Table 1. Measurements of spurious currents with three surface tension methods

Number of grid points/D Ca


La d c d c
VOF FT CSS (VOF) CSF (VOF) FT

0.120 1 1 30 32 610-3 1.410-2 2.610-4


0.357 1 1 32 32 310-3 1.210-2 2.710-4

y A0 cos 2 x (16) oscillation frequency 0 = 1 1 . Fig. 5b


where the wavelength is set to 2, and the initial shows the temporal evolution of the wave amplitude
amplitude A0 is set to 0.01. The boundary conditions normalized by in comparison with the theory of
are shown in Fig. 5a. Three grid resolutions are used: Prosperetti [19]. The figure shows that while 3232
3232, 6464 and 128128. The surface tension predicts an incorrect frequency the finner grid
coefficient is set to 2. The densities of two fluids 128128 produces a satisfactory result. This conrms
are identical and set to 1 = 2 = 1. The kinematic that the method is capable of accurately predicting
this ow.
viscosity of both uids is set to 0.064720863. The
time is non-dimensionalized by the inviscid 3.5. RayleighTaylor instability
We consider the growth of a two-dimensional
RayleighTaylor instability that has been studied by
numerous methods to characterize the quality of
interface transport methods, see, for example [20], as
shown in Fig. 6a. Two immiscible fluids with the
denser one at the top are placed in a box of 14. The
interface separating two fluids is defined as:
y 2 A0 cos 2 x (17)

where A0 is taken to be 0.05. The top fluid has a


density 1 = 1.225, while the bottom fluid has a
density of 2 = 0.1694. Both fluids have the same

Fig. 5. Damped surface wave: (a) computational


domain, and (b) evolution of the amplitude of the
wave versus nondimensional time = 0t Fig. 6. RayleighTaylor instability problem

63
Journal of Science & Technology 119 (2017) 059-065

Fig. 7. Temporal evolution of the interface computed by the front-tracking method (left) in comparison with the
results predicted by Herrmann [20] (right)
viscosity 1 = 2 = 0.00313. The gravity acceleration some other direct numerical simulation methods, e.g.,
is set to 9.81. The boundary conditions are periodic at level set method and volume of fluid method for the
the left and right sides, and no-slip at the top and problems investigated in this study, with the similar
bottom. A grid resolution of 128512 is used. As grid resolutions the front-tracking method yielded the
time progresses, the heavy fluid (top) falls into the results which are more accurate, especially, in the
lighter fluid (bottom) due to gravity, and rolls up into case of calculating surface tension forces. In addition,
two counter-rotating vortices as shown in Fig. 6b. the present 2D method is quite simple and very easy
to be implemented. This facilitates computations of
Fig. 7 shows the evolution of the interface shape many multiphase problems.
at different times in comparison with the results
predicted by another numerical method of Herrmann However, the results presented in this paper are
[20] in which the author used a much finer grid of limited to the 2D cases. Therefore, in future research,
5122048. In each frame of Fig. 7, the left is the we will investigate some 3D problems to evaluate the
present result while the right is Herrmann's result. performance of the 3D front-tracking method.
Excellent agreement has been archieved. This Acknowledgments
confirms that front-tracking method produces the
accurate results for this multiphase problem. This research was supported by Hanoi
University of Science and Technology (HUST) under
5. Conclusion grant number T2016-PC-028.
We have presented the results of a number of References
verification and validation problems for the front-
tracking method, which has been widely used for [1] F.H. Harlow, J.E. Welch, Numerical calculation of
multiphase problems. The interface separating two time-dependent viscous incompressible flow of fluid
phases or fluids is represented connected elements with free surface, Phys. Fluids. 8 (1965) 21822189.
that are used to calculate the interfacial tension force. [2] C.. Hirt, B.. Nichols, Volume of fluid (VOF) method
The discretized governing equations are solved by the for the dynamics of free boundaries, J. Comput. Phys.
second-order predictor-corrector method. The various 39 (1981) 201225.
problems have been solved: a notched disk in rotating [3] M. Sussman, E. Fatemi, P. Smereka, S. Osher, An
ow, a circular drop in a vortical flow field, a improved level set method for incompressible two-
stationary drop, and a damped surface wave. The phase flows, Comput. Fluids. 27 (1998) 663680.
method is then used to simulate the Rayleigh-Taylor
[4] D. Jacqmin, Calculation of two-phase NavierStokes
problem. The numerical results produced by the
flows using phase-field modeling, J. Comput. Phys.
method are reasonably accurate and satisfactory. This 155 (1999) 96127.
confirms and supports the accuracy of the method for
numerous multiphase problems. In comparison with [5] H. Takewaki, A. Nishiguchi, T. Yabe, Cubic
interpolated pseudo-particle method (CIP) for solving

64
Journal of Science & Technology 119 (2017) 059-065

hyperbolic-type equations, J. Comput. Phys. 61 (1985) [12] T.V. Vu, G. Tryggvason, S. Homma, J.C. Wells,
261268. Numerical investigations of drop solidification on a
cold plate in the presence of volume change, Int. J.
[6] S.O. Unverdi, G. Tryggvason, A front-tracking Multiphase Flow. 76 (2015) 7385.
method for viscous, incompressible, multi-fluid flows,
J. Comput. Phys. 100 (1992) 2537. [13] C.S. Peskin, Numerical analysis of blood flow in the
heart, J. Comput. Phys. 25 (1977) 220252.
[7] G. Tryggvason, B. Bunner, A. Esmaeeli, D. Juric, N.
Al-Rawahi, W. Tauber, J. Han, S. Nas, Y.-J. Jan, A [14] A. Esmaeeli, G. Tryggvason, Computations of film
front-tracking method for the computations of boiling. Part I: numerical method, Int. J. Heat Mass
multiphase flow, J. Comput. Phys. 169 (2001) 708 Transfer. 47 (2004) 54515461.
759.
[15] S.T. Zalesak, Fully multidimensional flux-corrected
[8] C.W. Hirt, J.L. Cook, T.D. Butler, A Lagrangian transport algorithms for fluids, J. Comput. Phys. 31
method for calculating the dynamics of an (1979) 335362.
incompressible fluid with free surface, J. Comput.
Phys. 5 (1970) 103124. [16] J.B. Bell, P. Colella, H.M. Glaz, A second-order
projection method for the incompressible Navier-
[9] X. Shan, H. Chen, Lattice Boltzmann model for Stokes equations, J. Comput. Phys. 85 (1989) 257
simulating flows with multiple phases and 283.
components, Phys. Rev. E. 47 (1993) 1815.
[17] D. Enright, R. Fedkiw, J. Ferziger, I. Mitchell, A
[10] T.V. Vu, S. Homma, G. Tryggvason, J.C. Wells, H. hybrid particle level set method for improved interface
Takakura, Computations of breakup modes in laminar capturing, J. Comput. Phys. 183 (2002) 83116.
compound liquid jets in a coflowing fluid, Int. J.
Multiphase Flow. 49 (2013) 5869. [18] G. Tryggvason, R. Scardovelli, S. Zaleski, Direct
numerical simulations of gas-liquid multiphase flows,
[11] T.V. Vu, G. Tryggvason, S. Homma, J.C. Wells, H. Cambridge University Press, Cambridge; New York,
Takakura, A front-tracking method for three-phase 2011.
computations of solidification with volume change, J.
Chem. Eng. Jpn. 46 (2013) 726731. [19] A. Prosperetti, Motion of two superposed viscous
fluids, Phys. Fluids. 24 (1981) 12171223.
[20] M. Herrmann, A balanced force refined level set grid
method for two-phase flows on unstructured flow
solver grids, J. Comput. Phys. 227 (2008) 26742706.

65
Journal of Science & Technology 119 (2017) 066-070

Crystallization Kinetics of Mechanically Alloyed Al80Fe20


Amorphous Powder

Nguyen Thi Hoang Oanh, Tran Quoc Lap, Pham Ngoc Dieu Quynh, Le Hong Thang,
Nguyen Thi Anh Nguyet, Pham Ngoc Huyen, Nguyen Hoang Viet*
Hanoi University of Science and Technology No. 1, Dai Co Viet Str., Hai Ba Trung, Ha Noi, Viet Nam
Received: June 15, 2016; accepted: June 9, 2017

Abstract
The crystallization kinetics of an Al80Fe20 amorphous powder alloy were investigated by thermal analysis.
Crystallization of amorphous Al80Fe20 during continuous heating undergoes four stages. The first-stage
crystallization leads to the formation of fcc-Al from amorphous matrix. The next stages are the
decomposition of the residual amorphous phase into several intermetallic compounds. The activation
energies of the alloy were calculated from differential scanning calorimetry data using the Kissinger, Ozawa
and AugisBennett models. The non-isothermal crystallization kinetics are analyzed by Johnson-Mehl-
Avrami equation. The value of the Avrami index indicated that the crystallization is interface - controlled
growth.
Keywords: amorphous alloys, mechanical alloying, crystallization kinetics, Avrami exponent

1. Introduction* DSC has also led to the study of the crystallization


kinetics by so-called non-isothermal methods.
Al-rich metallic glasses have generated
Several reports on the successful formation of
considerable research interest because of the excellent
an amorphous phase through MA have been
mechanical and chemical properties. Tensile strength
published for Al80Fe20 amorphous alloy [2, 8-10]. But
of Al-based amorphous alloys is 2-5 times higher
there is a lack of studies regarding the crystallization
than their conventional crystalline counterparts [1-3].
kinetics of Al80Fe20 amorphous alloy.
Their high tensile strength can be further enhanced if
In this study, the thermal stability as well as the
fcc-Al nano-particles are homogeneously dispersed
crystallization kinetics of the mechanically alloyed
within a certain size and fraction range through
Al80Fe20 amorphous powder has been investigated
primary crystallization [4, 5]. One of the critical
using DSC in non-isothermal modes. The value of the
aspects of their applications is thermal stability, as the
Avrami index is calculated by Johnson-Mehl-Avrami
amorphous state is a non-equilibrium phase which
equation to determine crystallization mechanism of
irreversibly crystallizes upon heating. The
Al80Fe20 amorphous powder.
crystallization kinetics are very important for the
development of amorphous alloys and nanocrystalline 2. Experimental
materials, the properties of which are strongly
Al80Fe20 amorphous alloy powder was prepared
affected by the crystallization process. Therefore, the
via mechanical alloying process after 60h of milling
crystallization kinetics of amorphous alloys have
(more details in [11]). The structure of the as-
been studied extensively. Controlling the
received samples was confirmed by XRD
microstructure development from the glassy
measurements using RIGAKU RINT-2000 with
precursors requires detailed understanding of the
CuK (=1.5405) radiation. Morphology of the
specific mechanisms influencing structural
amorphous powder samples was observed by a field
transformations. Moreover, crystallization studies are
emission scanning electron microscope (FE-SEM).
essential for the proper choice of the consolidation
The crystallization kinetic of the powders was
parameters in order to maximize densification and, at
evaluated by non-isothermal DSC under a continuous
the same time, retaining the desired microstructure [6,
flow of Ar gas (70 mL/min) at heating rates of 5, 10,
7].
20 and 40 K/min using NETZSCH STA 409C, where
Differential scanning calorimetry (DSC)
platinum cups were used as containers.
technique allows a rapid and precise determination of
crystallization temperatures of amorphous materials. 3. Results and disscution
Fig. 1 shows the XRD pattern of Al80Fe20
* powder mixture presented a fully amorphous
Corresponding author: Tel.: (+84) 904.777.570
structure after 60 hours of milling.
Email: viet.nguyenhoang@hust.edu.vn
66
Journal of Science & Technology 119 (2017) 066-070

crystallization peaks shift to higher temperatures. The


peak temperature (Tp) values at different heating rates
are summarized in Table 1.
Table 1. Characteristic temperature at crystallization
peaks of Al80Fe20 powder at different heating rates
Heating rate, Tp1, Tp2, Tp3, Tp4,
K/min C C C C
5 360.9 412.0 486.0 576,.9
10 366.1 424.0 496.5 587.6
20 371.7 438.2 506.8 596.4
Fig. 1. X-ray diffraction patterns of Al80Fe20 40 373.6 445.9 512.5 601.3
amorphous powder.
Similar observation for the temperature peak for
the first crystallization peak of those amorphous
samples were made by F. Zhou [8] with Tp1 about 400
o
C. These amorphous alloys have crystallization
temperature range from 300 oC to 640 oC by F. Zhou
and from 350 oC to 630 oC in this study.
The activation energy of the crystallization
process gives important information regarding the
thermal stability of the sample. It can be evaluated
5m from constant-rate heating DSC curves taken at
different heating rates using the Kissinger Ozawa and
Fig. 2. FE-SEM image of Al80Fe20 amorphous Augis-Bennett equations, as given by equation (1),
powder after 60h of milling. (2), (3), respectively: [12]

E
ln 2 a const (1)
T RTp
p
Ea
ln( ) const (2)
RTp

E
ln
T T a const (3)
p o RTp

where is the heating rate, T p is the temperature


at the exothermal peak, R is the gas constant and Ea is
the activation energy of crystallization. Figure 4-6
show that Kissinger plot ln(/T p2) versus 1000/Tp,
Fig. 3. DSC curves of Al80Fe20 amorphous powder at Ozawa plot ln() versus 1000/Tp, Augis-Bennett plot
various heating rates. ln(/Tp-To) versus 1000/Tp, which yields straight lines
with a good fit, respectively. Table 2 presents results
Fig. 2 illustrates the SEM/EDS observation for of the activation energy calculated through three
as-received Al80Fe20 amorphous powder. It can be methods.
seen that fine powder particles, the particle size
mostly below 15 m, were agglomerated to form Table 2. Activation energy (Ea [kJ/mol]) of Al80Fe20
larger particles amorphous powder for the crystallization stages
determined via three methods
Fig. 3 presents the DSC diagram for the Al80Fe20
amorphous powder as a function of temperature taken Active Energy, kJ/mol
Methods
at different heating rates. As can be seen, this powder Peak 1 Peak 2 Peak 3 Peak 4
has four crystallization peaks, which means that Kissinger 510.1 230.2 362.6 493.2
powder undergoes four crystallization stages.
Ozawa 520.7 241.8 375.4 507.6
Moreover, increasing the heating rate from 5 to 40
o
C/min caused all position of the exothermic Augis-Bennett 515.4 236.0 369.0 500.4

67
Journal of Science & Technology 119 (2017) 066-070

crystalline grains during the phase transition, which


can be obtained by Johnson-Mehl-Avrami (JMA)
equation: [12]

x(t ) 1 e k
n

(4)
where x is the crystallization volume fraction at
time t, n is the Avrami exponent and k is the reaction
rate constant related to absolute temperature
described by Arrhenius equation:
Ea

k ko e RT

(5)
Fig. 4. Kissinger plots of the Al80Fe20 amorphous
powder. where is a constant, is the activation
energy, R is the gas constant and T is the absolute
temperature.
There are 2 methods to determine the Avrami
parameter. The first method was proposed by Ozawa.
We have:
d ln( ln(1 x))
n
d ln T
(6)
The value of x at any selected T is calculated
from the ratio of the partial area of the crystallization
peak at the selected temperature T to the total area of
the exothermic peak. Fig. 7 shows diagram of
crystallized volume fraction for Al80Fe20 amorphous
Fig. 5. Ozawa plots of the Al80Fe20 amorphous powder.
powder.

Fig. 7. Crystallized volume fraction x for Al80Fe20


Fig. 6. Augis-Bennett plots of the Al80Fe20 powder at different heating rates.
amorphous powder.
Combining equation (6) and plot (7), at any
It can be seen, the values of the activation fixed temperature, we can consider the Avrami
energies calculated from three models are parameter to be 0.91 in the first crystallization event.
approximate. Therefore, we can use one of the three
methods to calculate the activation energy. The second method to calculate Avrami
parameter is through the activation energy calculated
The Avrami index (n) gives detailed information by Kissinger method, as following
on the nucleation and growth mechanism of new

68
Journal of Science & Technology 119 (2017) 066-070

R ln( ln(1 x)) amorphous alloy were made by F. Zhou et al. [8],
n( x ) (7) and M. Krasnowski [2].
1
Ex
T
The crystallized volume fraction is also
determined by measuring the corresponding partial
area of the exothermic peak. Plotting ln[-ln(1-x)]
versus ln(1/T) with x between the range of 15% to
85% of transformed fractions, the JMA plots at
different heating rates are obtained as in Fig. 8.

Fig. 9. XRD patterns from amorphous Al80Fe20 alloy


after heat treatment at temperatures at (a) 413, (b)
468, (c) 535 and (d) 670 C.
4. Conclusion
Crystallization kinetics of mechanically alloyed
Al80Fe20 amorphous powder have been investigated
Fig. 8. JMA plots for 1st crystallization peaks of using DSC in non-isothermal modes. The
Al80Fe20 amorphous alloys at different heating rates. crystallization behavior of amorphous powder occurs
The Avrami index was obtained by the slopes of in four stages in the temperature range of 350 and 630
o
these plots. The Avrami index (n) is 0.80 in the first C. The primary phase of fcc Al together with
crystallization process. According to calculated maintaining amorphous phase in the first
Avrami index calculated by 2 methods is approximate crystallization event followed by formation of
to 1. The Avrami index usually between 1 and 4 if the Al13Fe4, Al3Fe and Al6Fe intermetallic phases in the
growth of the crystal is diffusion controlled. With n second crystallization event. At the higher
less than 1, the crystal growth has been shown to be crystallization temperature in the third crystallization
interface controlled [13]. A low value of n has also stage, intermetallic phases of Al13Fe4 and Al6Fe
been reported by other investigators in the primary occurred. In the final exothermic event, phases of fcc-
crystallization of amorphous alloys. This value Al, Al13Fe4 and AlFe3 can be realized. The values of
suggesting that the transformation in this stage is activation energy calculated from three methods
interface-controlled growth [14]. Kissinger, Ozawa and Augis-Bennett are almost
same. The Avrami exponent is less than 1 for the first
In order to determine the products of crystallization peak, suggesting that the
crystallization, milled powders were annealed in the transformation was interface - controlled growth.
DSC by heating at 20 C/min to temperature in the
range of 413 and 670 C, coressponding to the end Acknowledgments
temperatures of four crystallization reactions. Fig. 9 This research is funded by Vietnam National
shows XRD spectra from the amorphous Al80Fe20 Foundation for Science and Technology
alloy after heat treatment at different temperatures. Development (NAFOSTED) under grant number
After heating to 413 C, the amorphous alloy began 103.02-2012.19.
to crystallize into fcc-Al phase and remain
amorphous phase. After increase heating temperature References
to 468 C intermetallic phases of Al13Fe4, Al3Fe and [1]. John H. Perepezko and Rainer J. Hebert,
Al6Fe can be detected from XRD pattern in Fig. 8 (b). Amorphous Aluminum AlloysSynthesis and
At higher temperature of 535 C cleary diffraction Stability. JOM, 54 (2002) 34-39.
peaks of Al13Fe4 and Al6Fe phases can be seen Fig. 8
(c). At the final heating temperature of 670 C, no [2]. M. Krasnowski and T. Kulik, Nanocrystalline
amorphous phase can be retained, phases of fcc-Al and amorphous Al-Fe alloys containing 60-85%
and Al13Fe4 can be obtained. Similar observation of Al synthesised by mechanical alloying and
regarding products of structural changes for the phase transformations induced by heating of

69
Journal of Science & Technology 119 (2017) 066-070

milling products. Materials Chemistry and Al80Fe20 alloy powders prepared by ball milling.
Physics, 116 (2009) 631-637. Journal of Non-Crystalline Solids, 250252,
Part 2 (1999) 704-708.
[3]. Akihisa Inoue, Amorphous,
nanoquasicrystalline and nanocrystalline alloys [9]. J. Noetzel, D.C. Meyer, A. Tselev, A. Mcklich,
in Al-based systems. Progress in Materials P. Paufler, F. Prokert, E. Wieser, and W. Mller,
Science, 43 (1998) 365-520. Amorphization of Fe/Al: bulk and thin-film
effects. Applied Physics A, 71 (2000) 47-54.
[4]. Sergio Scudino, Kumar B. Surreddi, Hoang V.
Nguyen, Gang Liu, Thomas Gemming, Mira [10]. Wang Genmiao, Zhang Daoyuan, Chen Huiyu,
Sakaliyska, Ji S. Kim, Jens Vierke, Markus Lin Bixia, Wang Weihua, and Dong Yuanda,
Wollgarten, and Jurgen Eckert, High-strength Formation and properties of Fe20Al80 amorphous
Al87Ni8La5 bulk alloy produced by spark plasma powder. Physics Letters A, 155 (1991) 57-61.
sintering of gas atomized powders. Journal of
[11]. Nguyen Hoang Viet, Nguyen Thi Hoang Oanh,
Materials Research, 24 (2009) 2909-2916.
Pham Ngoc Dieu Quynh, Tran Quoc Lap, and
[5]. Akihisa Inoue and Hisamichi Kimura, High- Kim Ji Soon. Bulk Amorphous Al80Fe20
strength Al-based nanostructure alloys. Current Produced by Mechanical Alloying and Spark-
Opinion in Solid State and Materials Science, 2 Plasma Sintering. in The 2nd International
(1997) 305-310. Conference on Advanced Materials and
Nanotechnology 2014. Hanoi: Bach Khoa
[6]. P. P. Choi, J. S. Kim, O. T. H. Nguyen, D. H.
Publishing house.
Kwon, Y. S. Kwon, and J. C. Kim, Al-La-Ni-Fe
bulk metallic glasses produced by mechanical [12]. Miray elikbilek, Ali Erin Ersundu, and
alloying and spark-plasma sintering. Materials Sheyla Aydn, Chapter 6 - Crystallization
Science and Engineering: A, 449-451 (2007) Kinetics of Amorphous Materials, in Advances
1119-1122. in Crystallization Processes, Y. Mastai, Editor.
2012, InTech. p. 127-158.
[7]. K. B. Surreddi, S. Scudino, M. Sakaliyska, K.
G. Prashanth, D. J. Sordelet, and J. Eckert, [13]. S. W. Du and R. V. Ramanujan, Crystallization
Crystallization behavior and consolidation of and magnetic properties of Fe40Ni38B18Mo4
gas-atomized Al84Gd6Ni7Co3 glassy powder. amorphous alloy. Journal of Non-Crystalline
Journal of Alloys and Compounds, 491 (2010) Solids, 351 (2005) 3105-3113.
137-142.
[14]. J.W. Christian, The Theory of Transformations
[8]. F. Zhou, R. Lck, M. Scheffer, D. Lang, and K. in Metals and Alloys. 1975, Netherlands:
Lu, The crystallization process of amorphous Pergamon, Oxford.

70
Journal of Science & Technology 119 (2017) 071-075

Effects of Compaction Pressure on the Properties of In-Situ Hybrid


Carbide-Reinforced Copper-Based Composite Synthesized
by Powder Metallurgy
Le Minh Hai
Hanoi University of Science and Technology - No. 1, Dai Co Viet, Hai Ba Trung, Hanoi, Viet Nam
Received: December 21, 2016; accepted: June 9, 2017

Abstract
An in-situ Cu-(Nb,Ti)C composite was prepared through mechanical alloying of Cu-Nb-Ti-C mixture, followed
by pressing and sintering. The obtained results showed that compaction pressure has a great influence on the
properties of the bulk composite. Density, microhardness and electrical conductivity of the bulk composite
increased with increasing compaction pressure as increasing pressure provides better packing and lower
porosity. The microhardness increased from 369 to 443 HV and the electrical conductivity of the Cu-(Nb,Ti)C
composite increased, going from 7.46 to 8.93 (m-1x108) as the compaction pressure was increased into the
range of 500 - 1000 MPa. A good combination of microhardness of 425 HV and electrical conductivity of
8.62 x 10-8 (m)-1 of the Cu-15%vol. (Nb,Ti)C composite was obtained at compaction pressure of 800MPa.
Keywords: Cu, (Nb,Ti)C, mechanical alloying, sintering, compaction pressure.

1. Introduction* The bulk composite was synthesized from the


achieved in-situ Cu-(Nb,Ti)C powder via powder
Mechanical alloying (MA) is a potential powder
metallurgical routes including compaction by uniaxial
metallurgical route for preparing in-situ ceramic
cold pressing and sintering in an inert atmosphere.
particle-reinforced metal matrix composites (MMCs)
Consequently, powder consolidation step has a
from a mixture of starting elemental powders by high
significant influence in determining the final
energy ball-milling [1-3]. The greatest advantage of
properties of the in-situ copper matrix composite.
this route is it is able to develop nanostructured
Compaction pressure is expected to play an important
materials such as nanocomposites [4]. For example,
role in developing a process to produce a bulk in-situ
Krasnowski et al. [5] have produced FeAl30%TiC
Cu-(Nb,Ti)C composite with a good combination of
nanocomposite by MA and hot-pressing consolidation
strength and electrical conductivity. In present work,
from pure elemental Fe, Al, Ti, and C powders.
we studied the influence of the compaction pressure on
Nanocomposite materials provide the possibility for
both mechanical and electrical conductivity of the bulk
enhanced functionality and multifunctional properties
Cu-15% vol. (Nb,Ti)C composite.
in contrast with their more limited single component
counterparts [6]. 2. Experimental procedure
In our previous work [7], the in-situ Cu-15% vol. The starting powders used were pure elemental
(Nb,Ti)C nanocomposite powder has been Cu, Nb, Ti and C powders, purchased from Sigma-
successfully synthesized via MA at room temperature Aldrich, with mixture composition corresponding to
without subsequent heat treatment. After 10 hours of Cu-15vol.% (Nb,Ti)C. The copper powder was of
milling the peaks observed for the starting Nb and Ti 99.8% purity with an average particle size of 32.9 m;
powders are no longer observable, while the peaks of the niobium powder was of 99.9% purity with an
niobium-titanium carbide (Nb,Ti)C appear implying average particle size of 4.7 m; the titanium powder
that a significant volume fraction of carbide had was of 98% purity with a particle size of 30.2 m; and
already formed. Maximum volume fraction of the the graphite powder was 99.99% pure with an average
carbide was reached after 20 hours of milling. The particle size of 4.08 m.
formed in-situ carbide (Nb,Ti)C particles exhibited an
The mixture of powders was milled using a
excellent interfacial bonding with the copper matrix
Fritsch Pulveristte 6 planetary ball mill in an argon
and a good dispersion with particle size ranging from
atmosphere with a rotation speed of 400 rpm. The ball-
20 to 80 nm.

*
Corresponding author: Tel.: (+84) 912.098.484
Email: hai.leminh@hust.edu.vn
71
Journal of Science & Technology 119 (2017) 071-075

to-powder ratio was 10:1 with 10mm-diameter Increasing pressure provided better packing and led to
stainless steel balls. a decrease in porosity with the formation of new
particle contacts [9]. High pressure caused localized
After 20 hours of milling, the as-milled powders
deformation at the contacts allowing new contacts to
were extracted for further characterization by X-ray
form as the gaps between particles collapsed [10-11].
diffraction (XRD) analysis, scanning electron
Consequently, further gains in green density require
microscopy (SEM) and energy dispersive X-ray
greater expenditures of energy from the external
(EDX) analysis. The as-milled powders were
pressure source.
compacted in a cylindrical steel mould and then the
compacted composites were sintered by vacuum As can be observed in the green density curve in
sintering. In order to investigate the influences of Figure 2, the response of the as-milled composite
compaction pressure, the sintering temperature was powders during uni-axial pressing can be divided into
fixed at 900oC and the compaction pressure was varied two stages. During the first stage (below 800MPa)
in a range of 500-1000MPa. The density of the sintered densification increased abruptly until the start of the
in-situ Cu-(Nb,Ti)C composite was determined based second stage (above 800 MPa), where the densification
on Archimedes principle using a Satorious electronic rate lowered and remained constant, reflecting the
analytical balance and reported as relative density. particle work hardening. The sintered densities of the
Micro-hardness and electrical conductivity of the composites followed an increasing trend similar to that
composite were measured using a Shimadzu Vickers of green density. Figure 2 also shows clearly that the
micro-hardness tester at a load of 100 g and a four- sintered densities are higher than the green densities.
point probe Changmin Tech CMT-SR2000N, This phenomenon can be explained by a reduction in
respectively. surface tension, dimension shrinkage of the composite
and a reduction of internal voids during sintering [12].
3. Results and discussion
Because the compaction pressure plays a
In order to study the effects of compaction
significant role in the densification of the specimen, it
pressure on the properties of the composite, the Cu-15
is expected that varying the pressure of compaction
vol.% (Nb,Ti)C as-milled powder after 20 hours of
should drastically alter the microhardness. The
milling was compacted at different pressures in the
average values of five runs of Vickers microhardness
range of 500-1000 MPa. SEM images of the polished
for the sintered composites with different compaction
cross-sections of the sintered composites produced
pressures are given in Figure 3. The microhardness
with different compaction pressures are shown in
values increased from 369 to 443 HV as the
Figure 1 (a-f). Increasing the compaction pressure
compaction pressure increased from 500 to 1000 MPa.
causes considerable change in the compact
This increase in microhardness is the result of work
microstructure as the size and fraction of pores
hardening due to severe plastic deformation during
decrease remarkably. Furthermore, the pore shape
uni-axial compaction. On the other hand, it is well
varies as well. According to Gessinger et al. [8], the
known that the mechanical properties of powder
shape of pores shows more acute angles when
metallurgical materials are degraded by the presence
consolidating at higher pressure. Therefore, the driving
of pores which reduce the effective cross-sectional
force for shrinkage increases with the compacting
area and negatively affect strength. When the pressure
pressure since the curvature of the core increases.
of compaction is kept low, the interparticle contacts
As applied pressure increases, the density of the are randomly distributed with the presence of voids
powder compact increases while porosity decreases. between particles. When the pressure of compaction is
Determination of the densification behaviour of a increased, particle contacts get rearranged to fill these
powder is usually based on the measurement of the voids and pores, resulting in greater densification of
density of the powder compact as a function of the the specimen and increased microhardness.
compaction. Figure 2 shows the density of the sintered Furthermore, higher compaction pressure also
composite as a function of compaction pressure. An increased the dislocation population, resulting in an
application of higher loads increases both green and initially faster sintering rate [11]. Higher compaction
sintered densities of the composites. The green density pressure, therefore, contributes to the increase of
increases from 5.55 g/cm3 [corresponding to 73.3% hardness of the sintered composite.
theoretical density (TD)] to 5.91 g/cm3 (78.9% TD),
while the sintered density increases from 6.09 g/cm3
(82.1% TD) to 6.79 g/cm3 (91.8% TD) as compaction
pressure increases from 500 MPa to 1000 MPa.

72
Journal of Science & Technology 119 (2017) 071-075

(a) (b)

(c) (d)

(e) (f)

Fig. 1. SEM images showing morphology of Cu-15% vol. (Nb,Ti)C sintered pellets with different compactions
pressure (a) 500 MPa (b) 600 MPa (b) 700 MPa (c) 800 MPa (d) 900 MPa (e) 1000 MPa
the composite to obtain good conductivity. High
porosity compaction led to a high resistance to the
The dependence of electrical conductivity on
conduction of electron because the pores reduced the
compaction pressure is plotted in Figure 4. Like
effective cross-sectional area for electron movement
microhardnes, conductivity of the Cu-(Nb,Ti)C
[13].
composite also increased, going from 7.46 to 8.93
(m-1x108) as the compaction pressure was increased Figure 5 indicates that a good combination of
into the range of 500 - 1000 MPa. Electrical microhardness and electrical conductivity of the
conductivity of the sintered composite was found to be composite was obtained at 800 MPa of compaction
a strong function of compaction pressure. Higher pressure, where the microhardness and conductivity
applied compaction pressure increased considerably were 430.5 HV and 8.77 x 108 (m)-1, respectively,
the electrical conductivity of the sintered composite. suggesting that to produce Cu-(Nb,Ti)C composite
The different electrical conductivity might be with high microhardness and high electrical
attributed mainly to the existence of residual pores. conductivity, the optimum compaction pressure is
Presence of closed porosities in the compact restricts 800MPa.

73
Journal of Science & Technology 119 (2017) 071-075

1.0

Electrical conductivity (m-1x107)


7.4 Green density
Sintered density 0.9
7.0
Density (g/cm3)

6.6
0.8
6.2
0.7
5.8

5.4 0.6

5.0
400 500 600 700 800 900 1000 1100 0.5
Compaction pressure P (MPa) 400 500 600 700 800 900 1000 1100
Compaction pressure P (MPa)
Fig. 2. Effect of compaction pressure on the green and
sintered density at 900oC of Cu-15%vol. (Nb,Ti)C Fig. 4. Effect of compaction pressure on the electrical
composites conductivity of Cu-(Nb,Ti)C composite

450 10.0

Electrical conductivity (m-1x108)


900 MPa
430
Micro-hardness (HV)

9.0
800 MPa
410
600 MPa 1000
8.0
390 700 MPa MPa

370 500 MPa


7.0

350
400 500 600 700 800 900 1000 1100 6.0
Compaction pressure P (MPa) 350 380 410 440
Microhardness (HV)
Fig. 3. Effect of compaction pressure on the
microhardness of of Cu-15%vol. (Nb,Ti)C composites Fig. 5. Variations in microhardness and electrical
sintered at 900oC conductivity of Cu-15%vol. (Nb,Ti)C composites with
compaction pressure
4. Conclusion
Compaction pressure played a significant role in
the density, consequently on the microhardness and References
electrical conductivity of the bulk sintered Cu-
[1] M.S El-Eskandarany, Mechanical alloying for
(Nb,Ti)C. As applied pressure increases, the density of
fabrication of advanced engineering materials, Noyes
the powder compact increases while porosity Publications, New York, USA (2001) 16-18
decreases. The microhardness values increased from
369 to 443 HV and the electrical conductivity of the [2] C. Suryanarayana, Mechanical Alloying and Milling,
Cu-(Nb,Ti)C composite increased, going from 7.46 to Marcek Dekker, New York, USA (2004) 59-66, 83-87
8.93 (m-1x108) as the compaction pressure was [3] Lu L., Lai M. O. (1998), Mechanical Alloying, Kluwer
increased into the range of 500 - 1000 MPa. The Academic, USA, p.48.
obtained results suggest that to produce Cu-(Nb,Ti)C [4] L. Froyen, J.D. Wilde, Materials Science Forum Vols
composite with high microhardness and high electrical 437-438 (2003) 141-144.
conductivity, the suitable compaction pressure is
800MPa. [5] M. Krasnowski, T. Kulik, Journal of Alloys and
Compounds 448 (2008) 227-233.

74
Journal of Science & Technology 119 (2017) 071-075

[6] K.M. Hanif, R.W. Meulenberg, F.F. Strouse, Journal [10] W.F. Wang, Powder Metallurgy 40(4), (1997) 295-
of the American Chemical Society 124 (2002) 11495- 300.
11502.
[11] R.M. German, Powder Metallurgy Science, Second
[7] L.M. Hai and T.D. Huy, Journal of Science and edition, Metal Powder Industries Federation, New
Technology 112 (2016) 80-84. Jersey (1994) 173-175.
[8] Gessinger G.H., Powder metallurgy 3(1) (1971) 29-32 [12] M.P. Groover, Fundamentals of modern
manufacturing: Materials, Process & Systems, John
[9] I.M. Moon, J.S. Choi, Powder Metallurgy 28 (1) Wiley & Sons, New York (2002) 334-348.
(1985) 21-28.
[13] L. Li, Y.S. Wong, J.Y.H. Fuh, L. Lu, Materials and
Design 113 (2001) 563567.

75