Vous êtes sur la page 1sur 14

DISCUSSIONS AND CLOSURES

agreement with the authors data, as shown in Figs. 1ac. How-


Discussion of Hydraulics of Tangential ever, at the downstream end vicinity, for Q = 8 and 10 l / s, the
Vortex Intake for Urban Drainage by D. Yu sharp increase observed in the flow depth is not predicted by Eq.
and J. H. W. Lee 2. Runs for Q = 12, 14, and 16 l / s are further considered in Figs.
March 2009, Vol. 135, No. 3, pp. 164174.
DOI: 10.1061/ASCE0733-94292009135:3164

Oscar Castro-Orgaz1
1
Research Engineer, Instituto de Agricultura Sostenible, Consejo Superior
de Investigaciones Cientificas, Finca Alameda del Obispo, E-14080,
Cordoba, Spain. E-mail: oscarcastro@ita.csic.es

The discusser has read with interest this work on tangential vortex
intakes, where a systematic 1D approach was proposed. The au-
thors proved that flow features such as the head-discharge rela-
tionship or the vortex flow zone may be approximated
theoretically. However, other important results including the ex-
perimental free surface profiles in the nonprismatic inlet were
presented but not further analyzed. The authors argued that due to
the energy loss within a hydraulic jump and zones with rapidly
varied flow, the 1D approach may not be used. However, as dem-
onstrated by Castro-Orgaz et al. 2008, nonprismatic channel
flow may be approximated by a relatively simple approach based
on the gradually varied flow theory. The present discussion aims
at providing a 1D analysis of the data presented by the authors
using this approach, a point so far overlooked in the paper. Limi-
tations for the 1D analysis are highlighted, and questions on fea-
tures of general interest are asked of the authors.
The energy head H for sloping channel flow with streamlines
nearly parallel to the bed is given by Montes 1994; Hager 1999

Q2
H=z+h+ 1 + S2o 1
2gA2
where z = channel bed elevation; h = flow depth measured verti-
cally; Q = discharge; b = channel width; A = cross-sectional area
= b h; and So = bed slope. Differentiation of Eq. 1 assuming po-
tential flow, that is, dH / dx = 0, results in

Q2 A
So + 1 + S2o
dh gA3 x
= 2
dx Q2 A
1 1 + S2o 3
gA h
Eq. 2 was numerically solved using a standard fourth-order
Runge-Kutta method, with the critical point as the boundary con-
dition at the corresponding extreme of the nonprismatic channel.
A singular point analysis Castro-Orgaz et al. 2008 for this study
results in spirals i.e., there are no control sections inside the
nonprismatic channel reach. For a nonprismatic channel the criti-
cal depth varies with x, such that different control points results
for the up- and downstream extreme sections. Fig. 1 shows the
solution of Eq. 2 for Test 1 in dimensionless form using the
upstream critical depth for a horizontal channel hc = Q2 / gB21/3 Fig. 1. Comparison between computed Eq. 2 and observed free
as the scaling. surface profiles for nonprismatic tapering channel, Test 1, and Q
For discharges of Q = 4, 8, and 10 l / s the critical point was = a 4 l / s, b 8 l / s, c 10 l / s, d 12 l / s, e 14 l / s, f 16 l / s, g
taken at the upstream channel section, resulting in excellent 17 l / s, and h 20 l / s

544 / JOURNAL OF HYDRAULIC ENGINEERING ASCE / AUGUST 2010


high dh / dx value near the critical depth. The good agreement of
computations with observation for Test 1 indicates that a 1D mod-
els does not necessarily involve a failure in the solution due to
energy dissipation within a hydraulic jump, as stated by the au-
thors. It simply means that two different backwater curves for
potential flow may be computed starting from the extremes of the
tapering channel if a hydraulic jump appears, and that the energy
loss is concentrated to a local hydraulic jump.
Fig. 2 shows the solution of Eq. 2 for Test 3. Runs for Q
= 2, 4, 6, and 7 l / s involved a supercritical flow profile i.e., an
upstream critical depth control. The agreement between compu-
tations and observations is again excellent, as shown in Figs.
2ad. Runs for Q = 7.5, 9 and 11 l / s corresponding to subcritical
flow with a downstream critical depth control are considered in
Figs. 2eg, resulting in good agreement. As previously ob-
served, the major discrepancies also appear here at the down-
stream end vicinity, because nonhydrostatic pressure effects are
not accounted for by Eq. 2.
In conclusion, the 1D model is a reasonable approach for free
surface profiles in nonprismatic channels, despite the inherent
limitations of this theory. Discrepancies result mainly at the
downstream end of the vicinity, because streamline curvature may
be important there. The discusser would like to ask the authors:
1. Are there pressure readings, or ADV measurements, avail-
able from their experimental program? In the photos of the
paper it appears that no pressure taps were installed. These
data may be useful to highlight the influence of streamline
curvature on the flow features.
2. Did the authors record the 2D shock-wave pattern in the
nonprismatic channel reach i.e., their shape and position? It
may have influenced several flow features, particularly the
flow depth at the junction vicinity.

References

Castro-Orgaz, O., Giraldez, J. V., and Ayuso, J. L. 2008. Transcritical


flow due to channel contraction. J. Hydraul. Eng., 1344, 492496.
Hager, W. H. 1999. Wastewater hydraulics: Theory and practice,
Springer, Berlin.
Montes, J. S. 1994. Potential flow solution to the 2D transition form
mild to steep slope. J. Hydraul. Eng., 1205, 601621.

Fig. 2. Comparison between computed Eq. 2 and observed free


surface profiles for non-prismatic tapering channel, Test 3, and Q Closure to Hydraulics of Tangential Vortex
= a 2 l / s, b 4 l / s, c 6 l / s, d 7 l / s, e 7.5 l / s, f 9 l / s, and g Intake for Urban Drainage by D. Yu and J.
11 l / s H. W. Lee
March 2009, Vol. 135, No. 3, pp. 164174.
DOI: 10.1061/ASCE0733-94292009135:3164
1df using the upstream critical point, yet the experimental data
suggest the simultaneous existence of a downstream critical point,
with a hydraulic jump forming within the channel. Therefore, a
Daeyoung Yu1 and Joseph H. W. Lee, F.ASCE2
1
Deceased; formerly Research Asst. Prof., Dept. of Civil Engineering,
backwater equation starting at the downstream control was also The Univ. of Hong Kong, Pokfulam Rd., Hong Kong SAR, China.
considered. The entire experimental free surface profile is seen to 2
Prof., Dept. of Civil Engineering, The Univ. of Hong Kong, Pokfulam
be well predicted by a combination of the two computed backwa- Rd., Hong Kong SAR, China. E-mail: hreclhw@hkucc.hku.hk
ter profiles.
Finally, runs for Q = 17 and 20 l / s Figs. 1g and h support
the existence of a downstream control from which free surface The writers would like to thank the discusser for his interest in
profiles were determined, resulting in an excellent agreement with our work on tangential vortex intakes. The comments and sugges-
observations. At the downstream end the 1D approach predicts tions are well appreciated. Essentially, the discussion revolves
flow depths below the experimental data, with the corresponding around the use of a one-dimensional 1D approach to analyze the

JOURNAL OF HYDRAULIC ENGINEERING ASCE / AUGUST 2010 / 545


free surface profile in the tapering and sloping inlet; based on the 385
comparison of predicted and observed depths for model no. 1 and
model no. 3 Figs. 1 and 2 of discussion, it is claimed that the
gradually varied flow GVF analysis is a success. We would like
to clarify the nature of the complex flow in the tapering inlet, and Tangential intake
elaborate on the role of 1D analysis in the context of the hydraulic
design of tangential vortex intakes.

20
Supercritical and Transitional Flow in Vortex Inlet
10

Our proposed theory for the hydraulic design of tangential intakes 280
is based on heuristic concepts founded on the 1D theory; it rep-
20
resents an advance over previous designs based mainly on expe-
rience or physical model studies Jain and Ettema 1987. 10
However, the limitations of the 1D theory are fully recognized, as 240
there are 3D flow features in the tapering and downward sloping
inlet section, including the junction. Consider first a small dis- 20
charge that drains freely into the vortex dropshaft. The flow will 10
be supercritical in the inlet. Fig. 1 shows the measured transverse 200
free surface profile for model no. 14 cf. Table 1 of paper for
Q = 1.5 L / s. The shock wave in the supercritical flow can readily 20
be seen. The sharp reduction of channel width results in a non- 10
uniform transverse depth distribution; the water depth at the ta- 160
pering side rises substantially higher, nearly twice the depth on
the opposite side. As the shock wave propagates downstream, the 20
transverse difference in depth gradually decreases. This uneven 10
transverse water surface profile is found in the supercritical flow
120
for both smooth stable and hydraulic jump cases. In the experi-
ments, the longitudinal depth variation is measured along the cen- 20
terline of the channel. The transverse free surface profile was
10
measured for selected runs. No independent pressure measure-
ments were made. 80
For larger discharges, when the free draining capacity is ex-
ceeded Q Q f , the vortex flow in the dropshaft would exert an
upstream influence, with the formation of a hydraulic jump with
significant energy loss. In addition, the 3D swirling flow at the
downstream junction with the dropshaft and the inlet flow near

the junction are nonhydrostatic see Fig. 8 of the paper. It is
Flow direction
indeed gratifying that useful results for this complex flow can be
obtained using the 1D concepts. 0
An appreciation of the 3D flow featuresthe transverse shock 0 50 100
wave in the tapering inlet, the complex 3D flow in the dropshaft,
Fig. 1. Measured transverse water depth profile model no. 14, Q
and inlet flow near the junctioncan be gained by viewing the
= 1.5 L / s; all lengths in mm
video clips of both stable and unstable designs at http://
www.aoe-water.hku.hk/vortex.

Prediction of Free Surface Profile in the Inlet


Channel 100

The proposed theory is aimed at the development of a generic 50


approach to the design of tangential intakes. It is important that
the intake can handle the whole range of flows in a smooth and
y (mm)

0
stable manner, without any abrupt change in depth and conse-
quent overflows. The design is based on the determination of the
50
free drainage discharge Q f and the control shift discharge Qc, 2 L/s
which does not explicitly require the prediction of the free surface 4 L/s
100 6 L/s
profile in the tapering inlet channel.
For stable designs Fig. 9 of the paper, energy conservation is 50 0 50 100 150 200 250 300
a valid approximation; the approach channel depth is given by the x (mm)
head-discharge relation, and the depth at the junction is given by
the critical depth. It is well-known that a 1D analysis can be used Fig. 2. Water surface profile in tapering inlet channel model no. 3
to compute the GVF in a nonprismatic channel. For example, the by GVF computation

546 / JOURNAL OF HYDRAULIC ENGINEERING ASCE / AUGUST 2010


standard-step method accounting for bottom friction and References
expansion/contraction loss can be readily adopted to compute the
GVF using the energy equation Henderson 1966; Subramanya Henderson, F. M. 1966. Open channel flow, Macmillan, New York.
1982. Fig. 2 shows an example of such a GVF computation of Jain, S. C., and Ettema, R. 1987. Vortex-flow intakes. Swirling flow
the free surface in the tapering channel for model no. 3 adopting problems at intakes, IAHR Hydraulic Structures Design Manual, Vol.
a Mannings n = 0.009 for perspex for the supercritical flows 1, J. Knauss, ed., Balkema, Rotterdam, The Netherlands, 125137.
before Qc is reached Q = 2 , 4 , 6 L / s. The upward sloping free Mehrotra, S. C. 1974. Circular jumps. J. Hydr. Div., 1008, 1133
surface toward the junction due to the reduction of cross-section 1140.
area can be clearly seen. The results would be practically the Sadler, C., and Higgens, M. 1963. Radial free surface flow. M.S.
same by neglecting friction. Although this calculation tacitly ne- thesis, Massachusetts Institute of Technology, Cambridge, Mass.
glects any transverse depth variation, the computed profile in the Subramanya, K. 1982. Flow in open channels, Tata McGraw Hill, New
supercritical reach will be comparable to the measurements simi- Delhi.
lar to those shown in Fig. 1a and Fig. 1b, or Fig. 2ac of the Watson, E. J. 1964. The radial spread of a liquid jet over a horizontal
discussion. plane. J. Fluid Mech., 20, 481499.
Wong, C. K. C., Yu, D. Y., and Lee, J. H. W. 2010. Numerical simu-
On the other hand, for an unstable design such as model no.
1, for larger flows Q Q f there is downstream influence leading lation of aerated flow in a supercritical flow diversion intake. Proc.,
to a hydraulic jump induced by the vortex flow in the dropshaft. 17th IAHR-APD Congress, B. Melville, ed., The University of Auck-
land, Auckland, New Zealand.
For these cases, the downstream control is not precisely known,
and there is energy dissipation across the jump. It is clear the
GVF profiles presented in Fig. 1cf of the discussion deviate
greatly from the observations along most of the channel. This is
expected in view of the inability of Eq. 2 in discussion to Discussion of Analysis of PVC Pipe-Wall
account for the energy loss across the jump, and the violation of
the hydrostatic approximation near the downstream boundary
Viscoelasticity during Water Hammer by
junction. A. K. Soares, D. I. C. Covas, and L. F. R.
When a hydraulic jump is involved, a standard GVF analysis Reis
proceeds by: 1 solving the supercritical GVF profile from an September 2008, Vol. 134, No. 9, pp. 13891394.
upstream control section; 2 solving the subcritical GVF profile DOI: 10.1061/ASCE0733-94292008134:91389
from a downstream control section; and 3 determining the loca-
tion of the jump by locating the section when the momentum Huan-Feng Duan1
balance is satisfied i.e., the depth from the upstream and down- 1
Dept. of Civil and Environmental Engineering, Hong Kong Univ. of
stream profiles are conjugate jump heights. Science and Technology, Clear Water Bay, Kowloon, Hong Kong.
For typical GVF calculations of open channel flows, the jump E-mail: ceduan@ust.hk
length is negligible compared with the region of interest. For the
present problem, however, the length of the jump would account
for a significant portion of the channel, and the upstream and The authors contributed a timely analysis of a PVC pipe system
downstream limits of the jump are not easily defined. This is during water hammer events. As stated in the original paper, the
further complicated by an expanding jump with changing dis- contribution of PVC pipe-wall viscoelasticity to the attenuation,
charge per unit width channel width is greater downstream and dispersion, and shape of the transient pressure head was analyzed
the ratio of jump length to jump height is unknown Sadler and based on the hydraulic-viscoelastic transient solver HVTS de-
Higgins 1963; Watson 1964; Mehrotra 1974. veloped by Covas and her coauthors Covas et al. 2004, 2005.
Our experimental investigation was supplemented by exten- This discussion mainly focuses on the provided data and model
sive GVF calculations of the free surface profile in the tapering results which present some inconsistencies on further inspections.
inlet channel. In view of the aforementioned uncertainties, the The tested system was a simplified pipeline system of length
heuristic theory was finally adopted for the hydraulic design. The about 85.40 m from tank to location P07 18.10 m from tank to
writers agree a more detailed 1D analysis of the free surface pump and 67.30 m from pump to Location P07. The model pa-
profile in the short steep channel would be beneficial. However, rameters were specified as: wave speed a = 460 m / s; diameter D
the transverse depth variation, the unknown energy losses due to = 75 mm; wall thickness e = 5.2 mm; E0 = 3.069 GP1.
the shock wave and across the hydraulic jump, and the unknown From their Fig. 2a, in the earlier stage, the total time interval
jump length all pose significant challenges to the formulation of a for three periods 1 period= 4L / a is about 1.0 s i.e., t = 1.0 s.
general robust model. Second, the downstream boundary condi- Therefore, this result seems to conflict with the provided informa-
tion at the junction is known only for stable designs for which Qc tion, as 12 L / a is much greater than t. Is there a mistake in the
can be reasonably predicted. The determination of this stable con- data provided?
dition must take into account the interaction of the tapering chan- Further, from many experiments Grellmann and Seidler
nel flow with the vortex flow in the dropshaft. Finally, we remark 2001, the PVC material is very different from some others, such
that 3D numerical model calculations using the volume-of-fluid as polyethylene PE used in Covass study, polyamide PA, and
VoF method have given reasonable predictions for this type of polyimide PI, and so on. That is, the PVC pipe holds an obvious
transcritical flows Wong et al. 2010. nonlinear deformation under the condition of external pressure
loading. And it will again be dependent on the frequency of the
Acknowledgments loaded pressure. In water hammer systems, that frequency is de-
pendent on the different system of interest. It shows that it may be
This work was supported by the Hong Kong Research Grants problematic when the linear approximated formulas from Covas
Council RGC HKU7143/06E. et al. 2004, 2005 is applied for PVC pipes. Therefore, it is also

JOURNAL OF HYDRAULIC ENGINEERING ASCE / AUGUST 2010 / 547


hard to demonstrate the authors final results that the creep effect derstanding of the pressure wave reflections, Fig. 1 depicts the
in PVC pipe is much smaller than that in HDPE pipes. measured pressure signal at the downstream end of the pipeline
Location P07.
After ball valve closure at t 0.7 s, an overpressure is gen-
References erated at the downstream end of the pipeline that propagates up-
stream and is affected by different features of the pipeline: 1 the
Covas, D., Stoianov, I., Mano, J., Ramos, H., Graham, N., and Maksimo- first gradual decrease of pressure is due to five pipe branches
vic, C. 2004. The dynamic effect of pipe-wall viscoelasticity in located along the pipeline; water inside the branches is com-
hydraulic transients. Part 1: Experimental analysis and creep charac- pressed relieving the maximum overpressure; 2 the second pres-
terization. J. Hydraul. Res., 425, 516530. sure decrease is caused by a cross-sectional increase along the
Covas, D., Stoianov, I., Mano, J., Ramos, H., Graham, N., and Maksimo- electromagnetic flow meter site; the inner pipe diameter changes
vic, C. 2005. The dynamic effect of pipe-wall viscoelasticity in from 75 to 101 mm and downstream the flow meter, it reduces
hydraulic transients. Part 2: Model development, calibration and veri- again to 75 mm; 3 another cross-sectional area reduction leads
fication. J. Hydraul. Res., 431, 5670. to a pressure increase; the pipe inner diameter between pump and
Grellmann, W., and Seidler, S., eds. 2001. Deformation and fracture check valve decreases from 75 to 15 mm; 4 after the diameter
behaviour of Polymers, Springer-Verlag, Berlin. change, the pressure wave reaches the centrifugal pump and a
large pressure decrease is observed as the flow passes through the
pump impeller. The flow in the pipeline rapidly reduces to zero,
and then reverses through the pump while it is rotating in the
Closure to Analysis of PVC Pipe-Wall normal direction. The check valve located downstream the pump
did not close instantaneously and its complete closure occurred
Viscoelasticity during Water Hammer by some time after the back flow was established. This caused
A. K. Soares, D. I. C. Covas, and L. F. R. an instantaneous stoppage of the reverse flow with the corre-
Reis sponding pressure rise 5. The check valve was described by an
September 2008, Vol. 134, No. 9, pp. 13891394. in-line valve with a steady-state orifice equation, and the closure
DOI: 10.1061/ASCE0733-94292008134:91389 time and the reverse velocity were calibrated using observed data
Fig. 2.
Alexandre Kepler Soares1; Ddia I. C. Covas2; and As the check valve closes, a positive pressure wave propagates
downstream the pipeline Fig. 1, feature 5; the pressure drop
Luisa Fernanda R. Reis3
1
Asst. Prof., Dept. of Sanitary and Environmental Engineering, Federal feature 6, observed immediately after it, is caused by the in-
Univ. of Mato Grosso, Ave. Fernando Correa da Costa, 78060-900 crease of the pipe cross section at the flow meter location.
Cuiab, MT, Brazil. E-mail: aksoares@ufmt.br The upstream check valve closure associated with the down-
2
Asst. Prof., Dept. of Civil Engineering, Instituto Superior Tcnico, stream ball valve closure created two dead ends between which
Technical Univ. of Lisbon TULisbon, Ave. Rovisco Pais, 1049-001 the transient pressure wave propagated. Fig. 3a presents dimen-
Lisbon, Portugal. E-mail: didia.covas@civil.ist.utl.pt sionless pressure variation H-Ho / HJ at Locations P06 down-
3
Prof., Dept. of Hydraulic and Sanitary Engineering, So Carlos School stream the check valve and P07 immediately upstream the ball
of Engineering, University of So Paulo, Ave. Trabalhador so- valve, in which H0 = steady state pressure head and HJ
carlense, 13566-590 So Carlos, SP, Brazil. E-mail: fernanda@
= theoretical Joukowsky overpressure HJ = a0Q0 / gA, where a0
sc.usp.br
= elastic wave speed; Q0 = initial steady state flow rate; g
= gravity acceleration; and A = pipe cross-sectional area. When
The writers would like to thank the discusser for the interest the pressure reaches its maximum value at P06, pressure at P07
shown and for the opportunity to discuss further the application of has its minimum value, and vice versa.
the viscoelastic model for hydraulic transient analyses in PVC After this detailed analysis, the wave speed was estimated
pipes presented in the paper. In response to the discussion, two based on the traveling time of the first pressure wave between two
main issues in the paper deserve special attention and are dealt consecutive pressure transducers, t* = L / a, where L is the distance
with in the following paragraphs. between the transducers. According to pressure data collected at
Locations P06 and P07, wave speed was estimated as approxi-
mately 440 m / s Fig. 3b, t* = 0.136 s, L = 59.80 m. As the wave
speed in plastic pipes is a time-dependent function rather than a
Experimental Data Analysis and Wave Speed constant parameter as it relates to linear elastic materials due to
Estimation unsteady friction and pipe-wall viscoelasticity Covas et al. 2004,
2005, the elastic wave speed was estimated as 460 m / s and the
Prior to inverse calculations, a detailed analysis of the observed PVC creep function was calibrated by using only one Kelvin
transient pressure signal was carried out to define appropriate Voigt element. This led to the wave speed variation shown in
boundary conditions i.e., determination of downstream ball valve Fig. 4.
closure, the check valve closure and back flow, etc. as well as to Given the complexity of the system, the elastic wave speed
characterize the wave speed. Observed transient pressures showed could not be estimated by the time interval for a period 1
particular wave reflections associated with different features of period= 4L / a as proposed by the discusser.
the pipeline, such as lateral pipe branches along the pipeline;
changes in main pipe cross-sectional area at the electromagnetic
flow meter and at the smaller diameter pipe between the pump PVC Rheological Behavior
and the check valve; and reverse flow due to ball valve closure.
These pressure wave reflections were extensively explored by The writers agree with the discusser concerning the PVC rheo-
Soares et al. 2008, and are summarized here. For a better un- logy. Indeed the PVC material is very different from high-density

548 / JOURNAL OF HYDRAULIC ENGINEERING ASCE / AUGUST 2010


volume fraction. Whereas viscoelasticity is the result of the dif-
fusion of molecules inside an amorphous material, PVC shows
little change in molecular structure, and its creep deformation is
very low compared with other plastics due to limited molecular
motion at room temperature, in contrast to polyethylene PE,
which has greater molecular motion in its amorphous part Aklo-
nis and MacKnight 1983; Ferry 1970; Ward and Hadley 1993.
This also validates the use of only one KelvinVoigt element for
description of PVC creep function while three elements are nec-
essary for HDPE pipes Covas et al. 2005; Gally et al. 1979.
With regard to the stressstrain behavior, at low values of
stress the time-varying creep strain is linearly related to the stress.
The creep compliance function is then independent of the magni-
tude of stress and linear viscoelasticity with the Boltzmann su-
Fig. 1. Observed transient pressure signal at the downstream end of perposition principle is applicable. At higher values of stress, the
the pipeline Location P07 linearity is no longer valid and the creep compliance function
becomes stress-dependent nonlinear viscoelasticity Dean et al.
1995. This usually happens when the deformations are large or if
the material changes its properties under deformations. For ex-
polyethylene HDPE used in Covass study. Each of these ma- ample, HDPE presents a nonlinear viscoelastic behavior at high
terials has some crystalline volume fraction, and a remaining strain levels strains 5% Zhang and Moore 1997. In Covass
amorphous volume fraction. In HDPE, the crystalline part is ap- study, the maximum strains of transient events generated were
proximately 90% and it is fixed while the amorphous part is flex-
0.3%, and the linear viscoelastic model could be used. The maxi-
ible; the PVC has an amorphous structure and little crystalline
mum strains generated in Soaress study were 0.04%. In accor-
dance with Dean et al. 1995, creep compliances for PVC also
depend on the magnitude of the applied stress when it exceeds
about 4 MPa. In Soaress study, the stress never exceeds 0.6 MPa.
Based on this analysis, the authors believe the linear viscoelas-
tic model could also be applied for PVC pipes, and the experi-
mental results showed that the PVC pipe-wall viscoelasticity is
much smaller than that in HDPE pipes.

Final Remarks

In conclusion, the authors believe that creep functions obtained


by tensile tests are important for the characterization of the vis-
coelastic behavior of PVC as a pipe material, providing a good
prior estimation. However, as stated by Covas et al. 2004, these
cannot correspond to the exact creep function of the PVC, when
integrated in a pipe system, particularly for buried pipes. This is
because mechanical tests cannot account for: 1 the variability of
Fig. 2. Observed transient pressure signal at the upstream end of the material properties; 2 a slight anisotropy of the pipe; 3 the
pipeline Location P06 and simulated flow rate at check valve increase in pipe stiffness due to axial constraints; 4 loading fre-
quency and pipe relaxation; and 5 uncertainties in unsteady fric-

Fig. 3. a Dimensionless observed pressure variation at the upstream end P06 and the downstream end of the pipeline P07; b observed
overpressure due to ball valve closure

JOURNAL OF HYDRAULIC ENGINEERING ASCE / AUGUST 2010 / 549


Discussion of Unifying Criterion for the
Velocity Reversal Hypothesis in Gravel-Bed
Rivers by D. Caamao, P. Goodwin, J. M.
Buffington, J. C. P. Liou, and S.
Daley-Laursen
January 2009, Vol. 135, No. 1, pp. 6670.
DOI: 10.1061/ASCE0733-94292009135:166

Bruce J. MacVicar1; Colin D. Rennie2; and Andr G. Roy3


1
Asst. Prof., Dept. of Civil and Environmental Engineering, Univ. of
Waterloo, 200 University Ave. West, Waterloo, Ontario, Canada N2L
3G1. Email: macvicab@yahoo.ca
2
Assoc. Prof., Dept. of Civil Engineering, Univ. of Ottawa, 161 Louis
Fig. 4. Calibrated creep function and wave speed
Pasteur St., Ottawa, Ontario, Canada K1N 6N5.
3
Prof., Canada Research Chair in Fluvial Dynamics, Dept. of Geography,
Universit de Montral, C.P. 6128, Succursale Centre-Ville, Montreal,
Quebec, Canada H3C 3J7.
tion losses due to different inflow rates in the referred system,
unsteady friction effects are negligible when compared to pipe- The authors present a criterion to distinguish pool-riffle sequences
wall viscoelasticity, and were thus assumed to be described by the in which a velocity reversal occurs from those in which it does
creep function. not. This would explain the variability that has been found among
The calibration of transient hydraulic models involves not only various studies on the hydraulics of pools and riffles and allow an
the determination of parameters related to viscoelasticity but also assessment as to whether the velocity reversal phenomenon is
to unsteady friction if it is not included in the creep function. It significant for the formation and maintenance of pools in gravel-
further involves several analyses of boundary and internal condi- bed rivers. The criterion is based on the mass and energy conser-
tions, pressure wave reflections, wave speed estimation, steady vation equations calculated at two cross sectionsone in the pool
state friction, etc., as well as the identification of other dynamic and the other in the riffle. The application of the criterion to
effects such as cavitation and fluid-structure interaction. In view several data sets appears to distinguish the presence/absence of
of these complexities, the analysis of pressure transients in plastic velocity reversal based on the section-averaged velocity.
pipes still remains a daunting task. In spite of its apparent usefulness, the approach proposed by
the authors may suffer from important limitations. The selection
of the two cross sections included in the analysis is of critical
References importance. There is much morphological variability in pools and
riffles and guidelines should be specified for the identification of
Aklonis, J. J., and MacKnight, W. J. 1983. Introduction to polymer representative sections. More importantly, however, partial flow
viscoelasticity, 2nd Ed., Wiley, New York. velocity reversal may occur without a reversal of bulk velocity.
Covas, D. I. C., Stoianov, I., Mano, J. F., Ramos, H. M., Graham, N., and Keller 1971 originally proposed that a reversal occurs for flow
Maksimovic, C. 2004. The dynamic effect of pipe-wall viscoelas- velocity vectors near the streambed, where they are likely to have
ticity in hydraulic transients. Part I: Experimental analysis and creep a significant effect on sediment transport. The authors speculate
characterization. J. Hydraul. Res., 425, 516530. that if the bulk velocity reversal is linked to a commensurate
Covas, D. I. C., Stoianov, I., Mano, J. F., Ramos, H. M., Graham, N., and reversal of shear stress and transport capacity then the pool will
Maksimovic, C. 2005. The dynamic effect of pipe-wall viscoelas- scour relative to the riffle p. 68. Considering the lateral vari-
ticity in hydraulic transients. Part II: Model development, calibration ability of flow and shear stress in riffles and pools as discussed by
and verification. J. Hydraul. Res., 431, 5670. Booker et al. 2001 and MacWilliams et al. 2006 and the im-
Dean, G. D., Tomlins, P. E., and Read, B. E. 1995. A model for non- plications of flow depth variations on near-bed hydraulics as pre-
linear creep and physical aging in polyvinyl chloride. Polym. Eng. sented by MacVicar and Roy 2007, the link between mean flow
Sci., 3516, 12821289.
velocity at a cross section herein called bulk velocity and shear
Ferry, J. D. 1970. Viscoelastic properties of polymers, 2nd Ed., Wiley,
stress is not straightforward. Although the authors acknowledge
New York. that their one-dimensional analysis neglects the variability of flow
Gally, M., Guney, M., and Rieutord, E. 1979. An investigation of
velocity in space and time, partial or momentary reversal of ve-
pressure transients in viscoelastic pipes. J. Fluids Eng., 101, 495
locity may be common in pool and riffle sequences thus limiting
499.
Soares, A. K., Covas, D. I. C., and Reis, L. F. R. 2008. Experimental
the applicability of the proposed criterion. As will be shown in
and numerical analysis of hydraulic transients in PVC pipes. Proc.,
this discussion, a reversal of near-bed velocity can occur without
a corresponding reversal of bulk velocity. Furthermore, we show
10th International Conference on Pressure Surges: Surge Analysis
System Design, Simulation, Monitoring and Control, S. Hunt, ed.,
that flow turbulence should be considered in addition to the ve-
BHR Group Limited, Cranfield, U.K. 317332. locity for the estimation of shear stress in pools and riffles.
Ward, I. M., and Hadley, D. W. 1993. An introduction to the mechanical Field data for this discussion were obtained from the study of
properties of solid polymers, Wiley, New York. MacVicar and Roy 2007. The field site is Moras Creek, a
Zhang, C., and Moore, I. D. 1997. Nonlinear mechanical response of 6.0-m-wide and 0.70-m-deep gravel-bed river with a median bed
high-density polyethylene. Part I: Experimental investigation and particle size of 60 mm located in Quebec. The pool is forced, as
model evaluation. Polym. Eng. Sci., 372, 404413. its location is determined by a large tree and root wad that con-

550 / JOURNAL OF HYDRAULIC ENGINEERING ASCE / AUGUST 2010


Fig. 1. Mean downstream velocity U normalized by the mean profile velocity Umean and turbulent kinetic energy ke profiles for a forced-pool
field site a and b and a straight pool in a flume experiment c and d

stricts the width of the channel. The pool length is approximately


16 m, or 2.7 times the width of the channel. The entry slope is
between 4.5 and 7. Velocity measurements were made at seven
discharges up to and including the bank-full discharge. Additional
flow measurements were made in a straight 0.6 m deep,
1.5-m-wide flume with a median bed particle size of 10 mm. A
straight pool with no width constriction was constructed in the
flume. The pool had a total length of 7.29 m or 4.9 times the
width of the channel, a residual depth of 0.25 m and entry and
exit slopes of 5.
Measurements at three sections are presented for both the
flume and field examples. The normal flow section was measured
close to the upstream limit of the pool over what is considered the
riffle. The pool head section was located in the upstream portion
of the pool where the water depth is increasing. The pool tail
section was located in the downstream portion of the pool where
the water depth is decreasing. The tree that forces the pool in the
field example was located between the pool head and pool tail
sections. Results from the middle of the pool are not critical to
our discussion but are available MacVicar and Roy 2007. Mea-
surements were made at the bank-full discharge in the thalweg for Fig. 2. Bulk velocity and the near-bed velocity in the channel thal-
the field case, and at a normal flow depth of 0.24 m in the channel weg as it varies with discharge up to the bank-full discharge
centerline for the flume case. 4.9 m3 / s for three sections in a forced-riffle pool

JOURNAL OF HYDRAULIC ENGINEERING ASCE / AUGUST 2010 / 551


Estimates of mean downstream velocity U and turbulent ki- curs in the pool tail. As a result of the change in shape of the
netic energy ke were calculated from two-minute time series velocity profile, the near-bed velocity is also greater than the bulk
measured at 20 Hz for the field case and 50 Hz for the flume case. velocity for this section.
Downstream velocity U was calculated as the time-averaged Riffles and pools are defined by changes in bed elevation. We
value. Turbulent kinetic energy ke is defined as present clear evidence that bulk flow velocity cannot be assumed
to be related to shear stress in pools and riffles due to the vari-
ke = 0.5u2i + v2i + w2i 1 ability in channel bed morphology and to the presence of nonuni-
form flow. A criterion based on the bulk velocity that is driven by
where u, v, and w are the downstream, lateral, and vertical com- changes in channel width, while attractive for its simplicity, is
ponents of the flow velocity; the subscript i denotes the instanta- insufficient to determine if a near-bed velocity reversal occurs and
neous velocity fluctuation relative to the mean; and the overbar may mislead our understanding of the dynamics of pools and
indicates that it is the mean value. For the field case, only the riffles. Instead, we argue that the sensitivity of near-bed values of
downstream and vertical components of the flow were measured. both downstream velocity and turbulent kinetic energy to changes
Based on the common observation that u2i v2i w2i Nezu and in depth holds considerable promise for a critical test of the ve-
Nakagawa 1993, it was assumed that locity reversal hypothesis and an eventual explanation of pool and
riffle formation.
v2i = 0.5u2i + w2i 2

This assumption was not required for the flume case. References
Where the channel boundary is nonuniform, the shapes of U
and ke profiles differ from what is observed in normal flow Fig. Booker, D. J., Sear, D. A., and Payne, A. J. 2001. Modelling three-
1. Relative to the normal flow section, downstream velocities are dimensional flow structures and patterns of boundary shear stress in a
smaller in the pool head and greater in the pool tail in the lower natural pool-riffle sequence. Earth Surf. Processes Landforms,
third of the profile Fig. 1a and c. In the upper third of the 265, 553576.
Keller, E. A. 1971. Areal sorting of bed material: The hypothesis of
profile, the inverse effects are observed and the highest velocities
velocity reversal. Geol. Soc. Am. Bull., 823, 753756.
occur in the pool head. These effects are more pronounced in the Kim, S. C., Friedrichs, C. T., Maa, J. P. Y., and Wright, L. D. 2000.
field case than in the flume case, perhaps as a result of the rela- Estimating bottom stress in tidal boundary layer from Acoustic Dop-
tively shorter pool or the wood in the field case. The shape of the pler Velocimeter data. J. Hydraul. Eng., 1266, 399406.
ke profiles differs between the field and flume cases near the water MacVicar, B. J., and Roy, A. G. 2007. Hydrodynamics of a forced
surface in the normal flow section Fig. 1b and d. High turbu- riffle-pool in a gravel-bed river: 1. Mean velocity and turbulence in-
lence near to the water surface in the field case may be the result tensity. Water Resour. Res., 43, W12401.
of water surface waves and may indicate that true normal flow did MacWilliams, M. L. J., Wheaton, J. M., Pasternack, G. B., Street, R. L.,
not occur in the field case. Nevertheless, the shape of profiles in and Kitanidis, P. K. 2006. Flow convergence routing hypothesis for
the pool head and the pool tail is similar in both cases. The high- pool-riffle maintenance in alluvial rivers. Water Resour. Res., 42,
W10427.
est levels of turbulent kinetic energy occur in the lower third of
Nezu, I., and Nakagawa, H. 1993. Turbulence in open-channel flows,
the profile in the pool head in both the flume and field cases.
A.A. Balkema, Rotterdam, The Netherlands.
These results are significant for a discussion of the distribution
of shear stress in pools and riffles because both the near-bed
downstream velocity and turbulent kinetic energy can be used to
estimate shear stress on the bed Kim et al. 2000. They show that
a peak in shear stress occurs in the pool-head as a result of high Discussion of Sediment Transport
turbulence, while the change in the shape of the velocity profile
results in high shear stress near to the bed in the pool tail.
Modeling ReviewCurrent and Future
To demonstrate that a near-bed velocity reversal can occur in Developments by A. N. Papanicolaou,
the absence of a bulk velocity reversal, we calculated both veloc- M. Elhakeem, G. Krallis, S. Prakash, and
ity values from field measurements at seven different discharges J. Edinger
Fig. 2. The near-bed velocity was calculated as the mean of all January 2008, Vol. 134, No. 1, pp. 114.
measurements made in the thalweg in the lower 20% of the pro- DOI: 10.1061/ASCE0733-94292008134:11
file. The bulk velocity was calculated as the cross section average
from 5 to 7 velocity profiles spaced at 1 m intervals across the
Chih Ted Yang, F.ASCE1
width of the channel. As a result of lateral flow concentration see 1
Borland Prof. of Water Resources and Director of Hydroscience and
MacVicar and Roy 2007 for discussion, near-bed velocities in the Training Center, Colorado State Univ., Fort Collins, CO 80523-1372.
thalweg are greater than the bulk velocities at relatively low dis- E-mail: ctyang@engr.colostate.edu
charges in the normal flow section and the pool head. Near-bed
velocities in these sections peak at discharges less than half of the
bank-full value and no velocity reversal occurs between the nor- The authors of the article provide a comprehensive review of
mal flow section and the pool head. In the pool tail, and similar to sediment transport models available for solving engineering prob-
results in the pool head, no reversal of bulk velocity occurs. The lems. This type of forum article should be beneficial to students,
cross-sectional area of this section is always greater than the researchers, and practicing engineers on the type of models avail-
cross-sectional area of the normal flow section. However, a ve- able and how to select a proper model for solving a particular
locity reversal does occur in the near-bed velocity between these problem. There is no universally acceptable model that can most
two sections. Above the half bank-full discharge 2.5 m3 / s, cost-effectively give the best results for all conditions. The suc-
the highest near-bed velocity of the three measured sections oc- cess of the application of a computer model depends, to a large

552 / JOURNAL OF HYDRAULIC ENGINEERING ASCE / AUGUST 2010


Fig. 1. Comparison between the measured and simulated longitudi-
nal profiles of the delta in the Tarbela Reservoir Yang and Simes
2002
Fig. 2. Comparison between the measured and calculated scour
depths along the Tigris River Othman and Wang 2004
degree, on the modelers understanding of the basic theories in
fluvial hydraulics and field conditions. If more than one model
can be used for solving a particular situation in the field, the diction of channel geometry change along an unlined emergency
discusser prefers a model based on sound simple theories with spillway of Lake Mescalero, New Mexico; Tarbala Reservoir
reliable and stable numerical solutions. sedimentation processes in Pakistan; scour depth and grain-size
The article classified computer models as one, two, and three distributions below a dam on the Tigris River in Iraq; and channel
dimensional, with their strengths and weaknesses well explained. cross section and pattern developments along the unlined emer-
Generally speaking, 1D models are suitable for long-term and gency spillway of the Willow Creek Reservoir in Montana.
large-scale simulations while 3D models are suitable for detailed Fig. 1 shows the comparison between the surveyed and com-
studies of local phenomena if 3D data are available for calibration puted bed profiles using GSTARS3 after 22 years of operation of
and verification. Under certain situations, it may be desirable to the Tarbela Reservoir in Pakistan. Fig. 2 shows the comparisons
have a model that can give quasi-2D or quasi-3D results based on between the measured and calculated scour depths using
1D numerical solutions. GSTARS 2.1 at three stations of the Tigris River below the Mosul
The article briefly mentioned the GSTARS Molinas and Yang Dam in Iraq. Fig. 3 shows comparisons between measured and
1986 computer model developed by the U.S. Bureau of Recla- calculated bed material size distributions along the Tigris River
mation USBR as a 1D model. The original GSTARS using For- below the Mosul Dam using GSTARS 2.1. All these comparisons
tran IV for mainframe computers is no longer current. Subsequent and those cited in the references show that GSTARS models can
series of models have been developed and used by USBR and accurately simulate and predict the variation of reservoir and river
others in different countries. They are GSTARS 2.0 Yang et al. morphologic changes based on the stream tube concept and mini-
1998, GSTARS 2.1 Yang and Simes 2000, and GSTARS3 mum stream power theory without using any 2D or 3D models.
Yang and Simes 2002 using Fortran 77/90/95 for PC operation. The Willow Creek Reservoir located in Montana with an un-
Most of the original and older versions of GSTARS programs lined and unprotected emergency spillway that has a concrete
have been re-written, improved, and expanded in the more recent cutoff wall at the reservoir. GSTARS Molinas and Yang 1986
models. Only GSTARS 2.1 and GSTARS3 users manuals and was used to simulate and predict the variation of channel geom-
executable code are currently available for public use. Interested etry and pattern along the unlined emergency spillway of the Wil-
readers can contact the discusser or go to his Web site, http:// low Creek Reservoir. The simulation was based on 50% of the
www.engr.colostate.edu/ce, then click Academic Faculty/ probable maximum flood. A total of three stream tubes and
Hydraulic Engineering/Chih Ted Yang/Personal Website/ Yangs 1973 sediment transport formula were used in all simu-
GSTARS, and download these manuals free of charge. lations. The simulated and predicted results shown in Fig. 4 are
An important and unique feature of all GSTARS models is the
conjunctive use of stream tube concept and the theory of mini-
mum stream power. This feature allows the simulation and pre-
diction of qusai-2D hydraulic conditions along and across a
channel or a reservoir, and qusai-3D variations of channel geom-
etry and bed profile of a river or a reservoir. Interested readers
may consult the examples in the users manual of GSTARS2.1
and GSTARS3. More detailed explanations of the theoretical
basis and examples of applications of GSTARS2.0/2.1/3 are given
by Yang and Simes 2008, Simes and Yang 2008, Yang
2008, and in the Erosion and Sedimentation Manual U.S. Bu-
reau of Reclamation 2006. Laboratory and field data to test
GSTARS2.0/2.1/3 capabilities for simulating and predicting hy-
draulic and sediment conditions include, but are not limited to,
bed sorting and armoring resulting from bed degradation; knick
point movement; bed profile evaluation; reservoir delta develop-
ment and movement; local scour at the Mississippi River Lock Fig. 3. Calculated and observed grain-size distributions along the
and Dam No. 26 Replacement site near St. Louis, Missouri; pre- Tigris River Othman and Wang 2004

JOURNAL OF HYDRAULIC ENGINEERING ASCE / AUGUST 2010 / 553


Fig. 4. a Longitudinal bed and water surface profiles along the Willow Creek Reservoir emergency spillway; b narrow scour hole at station
B along the spillway; c change of channel pattern from straight to meandering; d formation of natural levee; e formation of central island and
divided flows; and f channel width change due to bedrock control

qualitative in nature because no field data are available for veri- sion processes. Fig. 4d shows that the eroded channel gradually
fication. formed a deep center channel with a natural levee on both sides.
Fig. 4a shows the longitudinal bed profile and the water sur- Fig. 4e shows that a central island is formed and the channel
face profile at time step 15 along the emergency spillway of the will have divided flows at low flows. Fig. 4f is a bedrock-
Willow Creek Reservoir. It is apparent that GSTARS can handle controlled section at Station O. The initial predicted erosion at
water surface profile computations for critical flow at stations A, time step 15 was in the vertical direction. However, the channel
F, and O; supercritical flow between A and B, F and G, and O and adjustment cannot move in the vertical direction due to bedrock
P; hydraulic jump between B and C, and H and I; and subcritical control, and erosion occurs in the lateral direction. The theory of
flow between C and E, and I and N. Fig. 4b predicted the nar- minimum stream power was used to guide GSTARS in the deter-
row scour hole at Station B immediately below the spillway crest. mination of optimum channel geometry adjustment. Fig. 4 dem-
Fig. 4c shows that a fairly uniform original cross section of a onstrates that GSTARS has the ability to make realistic
straight channel at Station C changed to an uneven cross section predictions of possible types of channel geometry and pattern
typically found at the bend of a meandering river during the ero- adjustments in accordance with the stream tube concept and the

554 / JOURNAL OF HYDRAULIC ENGINEERING ASCE / AUGUST 2010


minimum stream power theory regardless of the flow regimes and dm = 1.75q2/3S7/9 1
local constraints.
The aforementioned examples demonstrate that users are not where dm = median size of stone in the stone layer m; q = unit
limited to only 1D, 2D, or 3D models. If a 1D numerical solution discharge m3 / s / m; and S = slope of the channel m/m. In this
along stream tubes can provide reliable and realistic qusai-2D and development, it has been assumed that the value of Shields pa-
qusai-3D solutions, it may present a practical engineering tool for rameter for the critical shear stress is equal to 0.060 and that the
long-term large-scale simulation in the future. specific gravity of the stone material is equal to 2.65. Mannings
equation was used as the flow equation, with the following
Strickler-type relationship used to quantify Mannings n

References n = 0.049d1/6
m 2
For values other than 0.060, 2.65, and 0.049 as referred to above,
Molinas, A., and Yang, C. T. 1986. Computer program users manual the coefficient in Eq. 1 will be somewhat different than 1.75. In
for GSTARS (Generalized Stream Tube model for Alluvial River Simu-
fact, for the authors 76-mm sandstone particles having a specific
lation), U.S. Bureau of Reclamation Engineering and Research Cen-
gravity of 2.29, the coefficient given in Eq. 1 would be equal to
ter, Denver, Colorado.
Othman, K. I., and Wang, D. 2004. Application of GSTARS 2.1 model 2.30, which in part is indicative of the sensitivity of stone stability
for degradation in alluvial channels. Proc., 9th International Sympo- to specific gravity as also noted by Peirson and Cameron 2006.
sium on River Sedimentation, World Association of Sedimentation and A similar expression for the required stone size can also be
Erosion Research, Beijing, Vol. III, 15321537. derived from an equation developed by Stephenson 1979, as
Simes, F. J. M., and Yang, C. T. 2008. GSTARS computer models and shown by Smith and Kells 1995
their applications. Part II: Applications. Int. J. Sediment Res., 234,
dm = 1.62q2/3S7/9 3
299315.
U.S. Bureau of Reclamation Technical Service Center. 2006. Erosion which is remarkably similar to Eq. 1. The coefficient in Eq. 3
and sedimentation manual, U.S. Bureau of Reclamation, Denver. was derived on the basis of Stephensons C coefficient having a
Yang, C. T. 1973. Incipient motion and sediment transport. J. Hydr. value of 0.245, which is the midpoint of the values reported for
Div., 9910, 16791704. smooth pebbles 0.22 and crushed granite 0.27, a stone-specific
Yang, C. T. 2008. GSTARS computer models and sediment control in gravity of 2.65, a stone mass porosity of 40%, and an angle of
surface water systems. Key Note Address, 3rd Int. Conf. on Water repose of the stone material of 35. Using the mean angle of
Resources and Arid Environment, Riyadh, Saudi Arabia. repose or friction angle for the stone material reported by the
Yang, C. T., and Simes, F. J. M. 2000. Users manual for GSTARS 2.1,
authors of 48.8, a stone-specific gravity of 2.29 for their 76-mm
U.S. Bureau of Reclamation, Technical Service Center, Denver.
Yang, C. T., and Simes, F. J. M. 2002. Users manual for GSTARS3,
sandstone material, and a mean porosity of 46% for the randomly
U.S. Bureau of Reclamation, Technical Service Center, Denver. placed stone material as determined from the authors data, the
Yang, C. T., and Simes, F. J. M. 2008. GSTARS computer models and coefficient in Eq. 3 would be equal to 1.41.
their applications. Part I: Theoretical development. Int. J. Sediment Abt et al. 2008 introduce the coefficient of uniformity, Cu,
Res., 233, 197211. into the stability relationship, which they express as SI units
Yang, C. T., Trevio, M. A., and Simes, F. J. M. 1998. Users manual
for GSTARS 2.0, U.S. Bureau of Reclamation Technical Service Cen-
d50 = 97.82C0.70 0.68 0.70
u qf S 4
ter, Denver. where d50 = median stone size cm, which is the same as dm when
both are defined in terms of stone weight; Cu is the coefficient of
uniformity of the stone material dimensionless; and q f = design
unit discharge m3 / s / m. For a uniformly graded stone material
with, for example, Cu = 2, Eq. 4 is very similar to that given
Discussion of Placed Rock as Protection by Smith and Kells 1995 in Eq. 1 and Stephenson 1979 in
against Erosion by Flow down Steep Eq. 3.
Slopes by W. L. Peirson, J. Figlus, The relationship presented in Eq. 1 can be expressed in more
S. E. Pells, and R. J. Cox general terms as
September 2008, Vol. 134, No. 9, pp. 13701375. dm = Kq2/3S7/9 5
DOI: 10.1061/ASCE0733-94292008134:91370
where K is a coefficient which, as shown above, reflects the inte-
J. A. Kells, Ph.D., P.Eng., M.ASCE 1 gration of such parameters as stone-specific gravity, stone mass
1
Prof., Dept. of Civil and Geological Engineering, Univ. of porosity, angle of internal friction of the stone material, particle
Saskatchewan, 57 Campus Dr., Saskatoon, SK, Canada S7N 5A9. roughness, and the coefficient in the Strickler-type relationship
for Mannings n i.e., Eq. 2. In the work reported by Smith and
Kells 1995, which draws considerably on earlier work described
The discusser found the authors paper to be an interesting con- in Smith and Murray 1975, it was found that the value of K is
tribution to the Journal and to the advancement of the use of 1.8 for initial stone movement, 1.5 for initial failure of the stone
stone in the design of steeply sloping channels. The discusser and mass which thereafter heals itself provided that a sufficient
various associates have also carried out some studies on the sta- amount of stone material has been placed at the crest of the
bility of stone on steep slopes, which is the primary focus of this slope, and 1.2 for ultimate failure of the stone-paved slope. Here,
discussion. initial stone movement refers to the point at which a single stone
Smith and Kells 1995 present the development of a semi- is removed from the stone layer and transported to the bottom of
empirical relationship for the median size of stone required for the slope, initial failure is the point at which temporary exposure
stability in a two-dimensional open channel flow of the underlying slope material occurs e.g., exposure of the filter

JOURNAL OF HYDRAULIC ENGINEERING ASCE / AUGUST 2010 / 555


Table 1. Calculation of K Values from Eq. 5 Using the Authors Data the stability of the individual stones comprising the layer.
dm SG qtot,fail S On the assumption that the authors values of qtot,fail shown in
mm m3 / s / m m/m K their Table 2 reflect the unit discharge on the slope at the failure
condition not fully defined in the paper for randomly placed
76 2.29 0.118 0.20 1.10
stone, the discusser has determined the corresponding K values
76 0.084 0.30 1.01 from Eq. 5 using the authors data as shown in Table 1. As
76 0.054 0.40 1.08 indicated, the K values for the authors data are substantially less
109 2.37 0.233 0.20 1.01 than those given in Smith and Kells 1995, regardless of whether
109 0.156 0.30 0.96 one compares to Smith and Kellss initial K = 1.5 or ultimate
109 0.156 0.40 0.77 K = 1.2 failure conditions. This finding suggests that the findings
94 2.64 0.194 0.20 0.98 of Smith and Kells are more conservative than those given by the
94 0.183 0.30 0.74 authors. Of course, the differences in stone-specific gravity must
94 0.161 0.40 0.65 be considered, but this difference does not explain the difference
between the findings based on the authors data and those of
Smith and Kells. Moreover, the slope dependency that is evident
layer, and ultimate failure is when permanent exposure of some on the basis of the authors data was not observed in the work of
or all of the underlying slope material or filter layer takes place. Smith and Kells.
The discharge used in the analysis was the total discharge over Interestingly, for large slope values e.g., 0.40 m / m, the find-
the crest. The stone material was uniformly graded for all but one ings of Smith and Kells 1995 more or less agree with those of
test and no difference was detected in this regard and was clas- Peirson and Cameron 2006 for the ultimate failure condition cf.
sified as semirounded in shape. Fig. 4 in Peirson and Cameron 2006, but are less conservative for
Smith and Kells 1995 indicate that the discharge required for lower slopes for the same failure condition. Inclusive of some
ultimate failure was 2550% or more than that required to pro- margin of safety, Smith and Kells 1995 suggest that K = 1.8 in
duce initial failure. The larger increase in discharge was required Eq. 5 is a reasonably appropriate value for design purposes.
for the flat gradient tests i.e., S = 4% to 7% of Smith and Murray
1975. That a smaller increase in the discharge beyond the initial
failure condition was required to produce ultimate failure for the References
larger slope tested by Kells and Smith i.e., S = 20% is in accor-
dance with the authors findings about the reduced ductility of Abt, S. R., Thornton, C. I., Gallegos, H. A., and Ullman, C. M. 2008.
the armor instability processes with increasing armor slope. Round-shaped riprap stabilization in overtopping flow. J. Hydraul.
Although the work of Smith and Murray 1975 was based on Eng., 1348, 10351041.
Fathalla, A. M., and Kells, J. A. 1999. Stability of rock linings in steep
flat-gradient slopes ranging in value between 4% and 7%, Smith
channels. Proc., 14th Hydrotechnical Engineering Specialty Conf.,
and Kells 1995 subsequently showed that the same stability re-
CSCE, Regina, SK, Canada, Vol. II, 185194.
lationship could be applied on much steeper slopes of up to 20%.
Peirson, W. L., and Cameron, S. 2006. Design of rock riprap protection
They concluded in an inferential way that the applicability of the to prevent erosion by water flows down steep slopes. J. Hydraul.
same stability relationship, which was essentially developed on Eng., 13210, 11101114.
the basis of a flat sloping channel as implied by the use of Shields Smith, C. D., and Kells, J. A. 1995. Stability of riprap channel linings
stability parameter, was due to the increased interparticle forces on steep gradients. Proc., 12th Canadian Hydrotechnical Conf.,
that occur between the stones placed on a steep slope. Fathalla CSCE, Regina, SK, Canada, Vol. 1, 317326.
and Kells 1999 studied slopes as high as 35%, although they Smith, C. D., and Murray, D. G. 1975. Cobble lined drop structures.
developed a somewhat different relationship from that given in Can. J. Civ. Eng., 24, 437446.
Eq. 1. Among others, however, they concluded that it is impor- Stephenson, D. 1979. Rockfill in hydraulic engineering, Elsevier, New
tant to consider the stability of the stone layer rather than simply York.

556 / JOURNAL OF HYDRAULIC ENGINEERING ASCE / AUGUST 2010


Copyright of Journal of Hydraulic Engineering is the property of American Society of Civil Engineers and its
content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's
express written permission. However, users may print, download, or email articles for individual use.

Vous aimerez peut-être aussi