You are on page 1of 10

Engineering Applications of Computational Fluid Mechanics Vol. 6, No.1, pp.

100109 (2012)

SIMULATING FLOW OVER CIRCULAR SPILLWAYS BY USING


DIFFERENT TURBULENCE MODELS
H. Rahimzadeh *, R. Maghsoodi **, H. Sarkardeh # and S. Tavakkol
* Department of Mechanical Engineering, Amirkabir University of Technology, Tehran, Iran
E-Mail: rahimzad@aut.ac.ir (Corresponding Author)
** Department of Civil Engineering, Shahrood Branch, Islamic Azad University, Shahrood, Iran
#
Hydraulic Structures Division, Water Research Institute (WRI), Tehran, Iran

Department of Civil and Environmental Engineering, Amirkabir University of Technology, Tehran, Iran

ABSTRACT: Fluent software has been used to simulate flow over a circular spillway and results were compared
with experimental data. As the flow over a circular spillway is turbulent and has a free surface, its characteristics are
complex and often difficult to be predicted. This study assesses the performance of some turbulence models to
predict the hydraulic condition of flow over circular spillways. The Volume of Fluid (VOF) method is applied to
obtain the free surface in each case. Such cases include highly swirling flows, stress-driven secondary flows and
flows over circular spillways. Finally it is concluded that the results of RSM, RNG k-, Realizable k-, SST k-
turbulence models agree well with experimental data.
Keywords:

circular spillway, numerical simulation, fluent, turbulence models, VOF

developed the Bernoulli theorem along a


streamline to flow in open channels. Tadayon
(2009) analysed mean characteristics of
curvilinear flows by using Computational Fluid
Dynamics (CFD). Pettersson and Rizzi (2009)
used Fluent Software to compare two different
turbulence models accuracy in computing local
boundary layer properties with wind tunnel
measurements. Bagheri and Heidarpour (2010)
simulated flow over a circular-crested spillway
with an irrotational vortex to determine the
spillway discharge coefficient and velocity values
over the crest. Yazdi et al. (2010) simulated flow
around a spur dike with free-surface flow by
using fully three-dimensional, Reynolds-averaged
NavierStokes equation. They also applied the
volume of fluid method with geometric
reconstruction scheme to model the free-surface
flow. Rahimzadeh et al. (2010) also simulated
flow over stepped spillways by using Fluent
software. Unal and Goren (2011) presented a
comparative study based on the 3D computational
simulations of the flow around a circular cylinder.
They used three different two-equation turbulence
models in their simulations.

1. INTRODUCTION
Circular spillways are typically used for
measuring discharges and as a control device in
water systems. A circular spillway is an overflow
structure with a circular crest. In fact, they
provide a unique relationship between the
upstream head and the discharge. The
characteristics of the flow over circular spillways
have been a subject of interest to many
researchers. Vo (1992) experimentally found the
discharge coefficient of circular spillways as a
function of the dimensionless total head of the
approaching flow. Ramamurthy and Vo (1993a)
applied these equations to predict the velocity
distribution over a cylindrical spillway.
Ramamurthy and Vo (1993b) compared results of
their experiments with others. Chanson and
Montes (1997) described experiments of circular
spillways, with eight cylinder sizes, for several
spillway heights and for five types of inflow
conditions including partially-developed inflow,
fully developed inflow, upstream ramp and
upstream hydraulic jump. Heidarpour and
Chamani (2006) developed a method to predict
the velocity distribution based on the potential
flow past a cylindrical spillway. Hargreaves et al.
(2007) described the validation of CFD for
modelling free surface flows past a broad-crested
weir. Castro-Orgazet et al. (2008) presented a
generalized one-dimensional model with the
assumption of critical flow in a curvilinear
domain. Castro-Orgazand and Chanson (2009)

2. EXPERIMENTAL DATA
Experimental tests used in this numerical study
were conducted in a smooth channel by Vo
(1992). Six fixed-bed tests were conducted in a
flume with La = 1.800 m, R = 0.152 m and w =
1.164 m (Fig. 1). Other parameters are given in
Table 1.

Received: 14 May. 2011; Revised: 18 Aug. 2011; Accepted: 6 Oct. 2011

100

Engineering Applications of Computational Fluid Mechanics Vol.6, No. 1 (2012)

u
u
u i u j t i j
x
j x i

2
u
k t i ij (4)
x i
3

The Boussinesq hypothesis is used in the k-, and


k- models.

Fig. 1

3.1 Standard k- model (Launder and


Spalding, 1972)

Schematic view of circular spillway.

The effective viscosity is modelled as follows:

t C

Table 1 Selected experimental parameters.


(deg)

(deg)

H1 (m)

Q (L/s/m)

Test 1

90

90

0.1237

85.39

Test 2

90

90

0.1762

152.17

Test 3

90

90

0.2093

203.43

Test 4

60

45

0.0796

41.34

Test 5

60

45

0.1185

80.04

Test 6

60

45

0.1482

115.71

t k

k x j

Gk Gb YM

x j

p
ui u j 2 ul

ij

xi x j x j xi 3 xl

ui u j
x j

(6)

u i
t
x i

where u i and u i are the mean and fluctuating


velocity components, respectively. Substituting
expressions of this form for the flow variables
into the instantaneous continuity and momentum
equations and simplifying (and dropping the
overbar on the mean velocity, u ), we have

k is the

k ku i
t
x i

(1)

u i
u i u j
t
x j

(5)

turbulence kinetic energy, and is the turbulence


dissipation rate. The following equations are the
transport equations for the standard k- model.

In this section different turbulence models which


are used in the present research are briefly
described. In Reynolds averaging:

u i 0
t x i

where C is a constant (Table 2),

3. TURBULENCE MODELS

u i u i u i

k2

x j
C1

x j

Gk C3 Gb C2

(7)

2
k

Standard constants of the k- model are listed in


Table 2, and were used in the model.

(2)

3.2 Renormalization-group (RNG) k- model


(Yakhot and Orszag, 1986)
Effective viscosity and transport equations for the
RNG k- model are as follows:

2k
d

(3)

eff

d
1.72 3
1 100

(8)

eff

k ku i
t
x i

where equations 2 and 3 are called Reynoldsaveraged Navier-Stokes (RANS) equation,


u iu j is called Reynolds stresses, and should

x j

be modelled. Hinze (1975) related the Reynolds


stresses to the mean velocity gradients:

101

k
k eff
xj

G k G b Y M

(9)

Engineering Applications of Computational Fluid Mechanics Vol. 6, No. 1, (2012)

Table 2 Standard turbulence models constants.


C1

C2

C1

C2

Standard k-

1.44

1.92

0.09

1.0

1.3

RNG k-

1.42

1.68

Realizable k-

1.44

1.9

1.0

1.2

Standard k-

0.555

0.83

0.09

1
A 0 4.04, A s 6 , cos 1
3
S ij S jk S ki

W
, S S ij S ij ,
S
1 u
u
S ij j i
2 x i x j

u i
t
x i


eff
C1 Gk C3 Gb
k
x j

3
2 C 1 0 2
C2

1 3
k
k

x j

(10)

0.6321

2.3929
0 2.3929

0.3679

where 0 1.0 and

Sk

mol
eff

k ku j
t
x i

Realizable k- model (Shih et al., 1995)

A0 As

C1

(14)

C3 Gb


k
, S
C 1 max 0.43,

Standard constants of realizable k- model are


listed in Table 2, and were used in the model.

, U * Sij Sij ij ij

ij ij 2 ijk k , ij ij ijk k
kU


C1S

x j

(13)

where

(12)

The difference between the realizable k- model


and the standard and RNG k- models is that C
is no longer constant. It is computed from

t


x j

C2

The effective viscosity is modelled as follows:

k2

x j
YM


u j
t
x j

, 0 4.38, 0.012

t C

t

k

Gk Gb

xi

(11)

Standard constants of RNG k- model are listed


in Table 2, and were used in the model.
3.3

6W ,

Transport equations of the realizable k- model


are as follows:

where k and are computed using the


following formula:

1.3929
0 1.3929

3.4

The standard k- models (Wilcox, 1998)

The effective viscosity and transport equations of


this model are as follows:

where ij is the mean rate-of-rotation tensor

t *

viewed in a rotating reference frame with the


angular velocity k and
102

(15)

Engineering Applications of Computational Fluid Mechanics Vol.6, No. 1 (2012)

where * damps the turbulent viscosity causing a


low-Reynolds-number correction.

In k- and k-models, G k , G b , G , D and

(16)

Y M (Sarkar and Balakrishnan, 1990) are the


generation of turbulence kinetic energy due to the
mean velocity gradients, the generation of
turbulence kinetic energy due to buoyancy, the
generation of , the cross-diffusion term, and the
contribution of the fluctuating dilatation in
compressible turbulence to the overall dissipation
rate, respectively.

(17)

3.6 Reynolds stress model (RSM) (Launder


et al., 1975; Gibson and Launder, 1978;
Launder, 1989a and b)

k ku i
t
x i

x j

k
t
k x j

G k Y k

x j


t
x j

G Y

u i
t
x i

Standard constants of k- model are listed in


Table 2, and were used in the model.

k
t
k x j

G k Y k

G Y D

x j

u j

u i
u iu k
u j u k
g i u j g j u i
x
x

k
k

G ij Buoyancy Pr oduction
Pij Stress Pr oduction

u u
u u j
p i j 2 i
x
xi
xk xk
j

(19)

ij Pr essure Strain

ij Dissipation

2 k u j um ikm uium jkm

Fij Pr oduction by System Rotation

(20)

(21)

Of the various terms in the above transport


equation, C ij , D L ,ij , Pij , and Fij do not require
any modeling. However, DT ,ij (Lien and
Leschziner, 1994), G ij , ij (Gibson and Launder,
1978;, Fu et al., 1987;, Launder, 1989a and b),
and ij need to be modeled to close the equations.
To simulate the pressure strain (ij), Linear
pressure-strain is used.

where


uiuj

xk xk

DL ,ij Molecular Diffusion

(18)

u i
t
x i

x j

C ij Convection

DT ,ij Turbulent Diffusion

k ku i
t
x i

x j


uiuj uk p kj ui ik uj

xk

The effective viscosity and transport equations are


modelled in SST k- model, by equations 18, 19
and 20.

1
2ij ij F2

max * ,
1

u iu j
u k u iu j
t x k

Local Time Derivative

3.5 Shear-stress transport (SST) k- model


(Menter, 1994)

k
t

The turbulence stress components are,

1
,
F1 k ,1 1 F1 k ,2

1
F1 ,1 1 F1 ,2

k ,1 1.176, w ,1 2.0, k ,2 1.0, w ,2 1.168

ij is the mean rate-of-rotation tensor, is

4. FREE SURFACE MODELLING

The volume of fluid (VOF) method appears to be


a powerful computational tool for the analysis of
free-surface flows (Hirt and Nichols, 1981). The
tracking of the interface(s) between the phases is

damps the turbulent viscosity causing a lowReynolds-number correction and F1 , F2 are the
blending functions.
103

Engineering Applications of Computational Fluid Mechanics Vol.6, No. 1 (2012)

accomplished by the solution of a continuity


equation for the volume fraction of one (or more)
of the phases. For the qth phase, this equation has
the following form:

q
t

. .q 0

(22)

where q is the volume fraction of qth phase. In

(a)

each control volume, the volume fractions of all


phases sum up to unity. The following three
conditions are possible for each cell:

q 0 : the cell is empty (of the qth phase).

q 1 : the cell is full (of the qth phase).

0 q 1 : the cell contains the interface

(b)

between the qth phase and one or more other


phases.
It can be assumed the free surface is on the
volume fraction of 0.5 (Fluent Manual 2005,
Dargahi 2006). In each cell the average properties
are computed according to the volume fraction of
each phase. For example, density and dynamic
viscosity in each cell of two phases are:
1 1 1 1 2 , 1 1 1 1 2 ,
respectively. The phases are represented by the
subscripts 1 and 2.
In this research, the geometric reconstruction
method of Young (1982) is employed. This
method represents the interface between fluids
using a piecewise-linear approach. It assumes that
the interface between two fluids has a linear slope
within each cell, and uses this linear shape for
calculation of the advection of fluid through the
cell faces.

(c)
Fig. 2 Computational grid in the vicinity of spillway:
(a) 3D view, (b) horizontal view, (c) near the
crest.

5. BOUNDARY CONDITIONS AND MESH


GENERATION
Fig. 3

Both structured and unstructured meshes have


been used. Denser mesh was used around the
spillway, close to free surface and in the boundary
layer in order to provide a higher accuracy and
considering the viscous flow. The first grid
surface of the solid boundaries was at
y 0.0005 , which ensures that the first grid
surface off the wall is located almost everywhere
at y 1.0 y u * y and that at least two

Solution domain and boundaries for modelled


circular spillway.

Boundary conditions which were employed in this


investigation are (Fig. 3): Two different inlets
were needed to define the water flow (Inlet I) and
air flow (Inlet II). These inlets were defined as
stream-wise velocity inlets that require the values
of velocity. To estimate the effect of walls on the
flow, empirical wall functions known as standard
wall functions (Launder and Spalding, 1974) were
used. The upper boundary above the air phase
was specified as a symmetry condition, which
enforces a zero normal velocity and a zero shear
stress.

grid surfaces are located within the laminar sublayer (y+< 5.0), where y is the distance of the
first grid from the solid wall, u* is wall shear
stress, and is kinetic viscosity (Fig. 2).

104

Engineering Applications of Computational Fluid Mechanics Vol.6, No. 1 (2012)

This figure shows good agreement between


measured and computed velocities by different
turbulence models except the standard k-, the
standard k-, and the k- RNG models.
Fig. 6 compares the pressure profile at crest,
obtained from simulation and experimental results
in different tests.
Fig. 5 and 6 show that the results from standard
models do not have a good agreement with
experimental data. It could be concluded that in
standard turbulence models (k- and k-), both
models under curvature conditions do not have
good performance in predicting hydraulic
characteristics.
In Fig. 7 and Fig. 8, numerical velocity vector
field and the corresponding flow patterns
represented by the streamlines for the flow
upstream of the spillway are given for two
extreme available cases.

To complete the description of the CFD


simulation, the PRESTO pressure discretization
scheme was applied as it showed the best
convergence in this simulation. The first order
upwind momentum and turbulent kinetic energy
discretization scheme was applied (Karki and
Patankar, 1989). The PISO pressure-velocity
coupling algorithm (Issa, 1986) was used purely
because it is designed specifically for transient
simulations.
The
unsteady,
free-surface
calculations required fine grid spacing and small
initial time steps. The grid spacing used was
adequate for solution convergence and has shown
good agreement with the experimental results. A
time step equal to 0.001 was selected. During the
3D simulation runs, solution convergence and the
water-surface
profiles
were
monitored.
Convergence was reached when the normalized
residual of each variable was in the order of 0.001.
The free surface was defined by a value of VOF =
0.5, which is a common practice for volume
fraction results (Fluent Manual, 2005 and Dargahi,
2006). After the convergence of the numerical
solution, in order to obtain more accurate results,
the mesh was refined according to gradients of
two phases and velocities, and the model was run
subsequently. The final number of mesh in
various conditions changed in the range of
187026202881 cells. A sensitivity analysis was
used and the number of mesh increased to over
two times, which has shown the results of the
model were valid.

7. CONCLUSIONS

In the present numerical study, flow over circular


spillways was simulated by using a three
dimensional code (Fluent Software). Different
turbulence models with the VOF method for free
surface modelling were employed to simulate
fully 3D flow. The numerical simulation solved
the NavierStokes equations within the flow
domain upstream and downstream from a
spillway. In the present research, focus was on
performance of different turbulence models to
simulate the flow over the circular spillways. By
comparing the 3D simulation results with the
flume data obtained by other researchers, the
simulation was found to produce flow over a
circular spillway with sufficient accuracy by all
turbulence models except the standard k- and the
standard k-. The RSM turbulence model had the
best agreement among all turbulence models with
experimental data.

6. VERIFICATION

It was necessary to ensure the accuracy of the


model before employing the numerical model.
For this purpose, experimental cases which were
mentioned in previous sections were employed.
To evaluate the free surface, the first case was
selected regarding the available flume data.
Existing experimental results used for validation
include water surface profiles, pressure
distributions and streamwise velocity distributions
for flow over the circular spillways. Fig. 4
compares experimental results of Tests 1, 2 and 3
(considering Table 1 parameters) with computed
water surface profiles over the crest.
Fig. 4 shows that different turbulence models
have no meaningful effect on predicting the free
surface in this simulation. Therefore all numerical
results are on one line. Results of velocity
simulation were compared with all experimental
tests (Table 1) and have been shown in Fig.5.

Fig. 4

105

Water surface profiles for flow over circular


spillway.

Engineering Applications of Computational Fluid Mechanics Vol. 6, No. 1 (2012)

Fig.5

(a) Test 1

(b) Test 2

(c) Test 3

(d) Test 4

(e) Test 5
(f) Test 6
Comparison between computed and measured horizontal velocity profiles at the crest of circular spillway by
employing different turbulence models.

ij
k

Gk

NOMENCLATURE

ui

velocity component
mean velocity component

ui
ui

fluctuating velocity component

u*

wall shear stress

t
eff

Reynolds stresses

uiu j

Gb
G

density of fluid
pressure
viscosity of fluid
turbulent viscosity
effective viscosity

YM
Yk
106

Kronecker delta
turbulence kinetic energy
turbulence dissipation rate
generation of turbulence kinetic energy
due to the mean velocity gradients
generation of turbulence kinetic energy
due to buoyancy
generation of turbulence kinetic energy
due to
contribution of the fluctuating dilatation
in compressible turbulence to the overall
dissipation rate
dissipation of k due to turbulence

Engineering Applications of Computational Fluid Mechanics Vol. 6, No. 1 (2012)

Fig.6

Y
k

ij
S

ij
F 1 & F2
*

(a) Test 1

(b) Test 2

(c) Test 3

(d) Test 4

(e) Test 5
(f) Test 6
Comparison between computed and measured pressure head distributions at crest of spillway by employing
different turbulence models.

dissipation of due to turbulence


turbulent Prandtl number for k
turbulent Prandtl number for
turbulent Prandtl number for
the mean rate-of-rotation tensor viewed
in a rotating reference frame with the
angular velocity k
modulus of the mean rate-of-strain
tensor
specific dissipation rate
mean rate of rotation tensor
blending functions
Parameter used to damp the turbulent

Cij
D
DT,ij
DL,ij
Pij
Gij
ij
ij
Fij
q
y
107

viscosity causing a low-Reynoldsnumber correction term


convection term
cross diffusion term
turbulent diffusive transport
molecular diffusion
stress production
effects of buoyancy on turbulence
pressure strain
dissipation rate
production by system rotation
volume fraction of qth phase
distance of first grid from solid wall

Engineering Applications of Computational Fluid Mechanics Vol.6, No. 1 (2012)

3. Castro-Orgaz O, Girldez JV, Ayuso JL


(2008) Critical flow over circular crested
weirs. ASCE J. Hydraulic Engineering
134(11):1661-1664.
4. Chanson H, Montes JS (1997). Overflow
characteristics of cylindrical weirs. Research
Report CE154. Dept. of Civil Engineering,
University of Queensland, Australia, 96 pages
(ISBN 0 867767219).
5. Dargahi B (2006). Experimental study and 3D
numerical simulations for a free-overflow
spillway. ASCE J. Hydraulic Engineering,
132(9):899, (10):1061/07339429.
6. Fluent Manual (2005). Manual and User
Guide of Fluent Software, Fluent Inc.
7. Fu S, Launder BE, Leschziner MA (1987).
Modeling strongly swirling recirculating jet
flow with Reynolds-stress transport closures.
In Sixth Symposium on Turbulent Shear
Flows, Toulouse, France.
8. Gibson MM, Launder BE (1978). Ground
effects on pressure fluctuations in the
atmospheric boundary layer. J. Fluid Mech.
86:491-511.
9. Hargreaves DM, Morvan HP, Wright NG
(2007). Validation of the volume of fluid
method for free surface calculation: the
broad-crested weir. Engineering Applications
of Computational Fluid Mechanics 1(2):136146.
10. Heidarpour M, Chamani MR (2006). Velocity
distribution over cylindrical weirs. J.
Hydraulic Research 44(5):708711.
11. Hinze JO (1975). Turbulence. McGraw-Hill
Publishing Co., New York.
12. Hirt CW, Nichols BD (1981). Volume of
fluid methods for the dynamics of free
boundaries. Journal of Computational
Physics 39:201225.
13. Issa RI (1986). Solution of implicitly
discretized fluid flow equations by operator
splitting. J. Comput. Phys. 62:40-65.
14. Karki K., Patankar SV (1989). Pressure-based
calculation procedure for viscous flows at all
speeds in arbitrary configurations. AIAA
Journal 27:1167-1174.
15. Launder BE (1989a). Second-moment closure
and its use in modelling turbulent industrial
flows. International Journal for Numerical
Methods in Fluids 9:963-985.
16. Launder BE (1989b). Second-moment closure:
present... and future? Inter. J. Heat Fluid
Flow 10(4):282-300.
17. Launder BE, Spalding DB (1972). Lectures in
Mathematical Models of Turbulence.
Academic Press, London, England.

Case 1

Case 6
Fig.7

Numerical field velocity vectors.

Case 1

Case 6
Fig.8

Numerical field velocity streamlines.

REFERENCES

1. Bagheri S, Heidarpour M (2010). Overflow


characteristics of circular-crested weirs. J.
Hydraulic Research 48:515520.
2. Castro-Orgaz O, Chanson H (2009). Bernoulli
theorem, minimum specific energy, and water
wave celerity in open-channel flow. ASCE J.
Irrigation and Drainage Engineering
135(6):773-778.

108

Engineering Applications of Computational Fluid Mechanics Vol.6, No. 1 (2012)

32. Yakhot
V,
Orszag
SA
(1986).
Renormalization group analysis of turbulence:
I. Basic theory. Journal of Scientific
Computing 1(1):1-51.
33. Yazdi J, Sarkardeh H, Azamathulla HMD,
AB Ghani A (2010). 3D simulation of flow
around a single spur dike with free-surface
flow. Intl. J. River Basin Management
8(1):5562.
34. Young DL (1982). Time-dependent multimaterial flow with large fluid distortion. In
Numerical methods for Fluid Dynamics. Ed.
Morton KW, Baines MJ. Academic Press,
New York, 273285.

18. Launder BE, Spalding DB (1974). The


numerical computation of turbulent flows.
Computer Methods in Applied Mechanics and
Engineering 3:269-289.
19. Launder BE, Reece GJ, Rodi W (1975).
Progress in the development of a Reynoldsstress turbulence closure. J. Fluid Mech.
68(3):537-566.
20. Lien FS, Leschziner MA (1994). Assessment
of turbulent transport models including nonlinear RNG eddy-viscosity formulation and
second-moment closure. Computers and
Fluids 23(8):983-1004.
21. Menter FR (1994). Two-equation eddyviscosity turbulence models for engineering
applications. AIAA Journal 32(8):1598-1605.
22. Pettersson K, Rizzi A (2009). Comparing
different CFD methods accuracy in
computing local boundary layer properties.
Engineering Applications of Computational
Fluid Mechanics 3(1).
23. Rahimzadeh H, Yazdi J, Sarkardeh H (2010).
Validation of the volume of fluid method for
free surface flow over a stepped spillway.
Journal of Dam Engineering 21(2):171-183.
24. Ramamurthy AS, Vo ND (1993a).
Application of dressler theory of weir flow. J.
Appl. Mech. 60(1):163166.
25. Ramamurthy AS, Vo ND (1993b).
Characteristicsofcircular-crested weir. J.
Hydraulic Engineering 119(9):1055-1062.
26. Sarkar S, Balakrishnan L (1990). Application
of a Reynolds-stress turbulence model to the
compressible shear layer. ICASE Report 9018, NASA CR 182002.
27. Shih TH, Liou WW, Shabbir A, Zhu J (1995).
A new k- eddy-viscosity model for high
Reynolds number turbulent flows -model
development and validation. Computers
Fluids 24(3):227-238.
28. Tadayon R (2009). Modelling Curvilinear
Flows In Hydraulic Structures. Ph.D. thesis.
Concordia Univ., Montreal, Canada.
29. Unal UO, Goren O (2011). Effect of vortex
generators on the flow around a circular
cylinder: computational investigation with
two-equation turbulence models. Engineering
Applications
of
Computational
Fluid
Mechanics 5(1):99-116.
30. Vo ND (1992). Characteristics of Curvilinear
Flow Past Circular-Crested Weirs. Ph.D.
thesis. Concordia Univ., Montreal, Canada.
31. Wilcox DC (1998). Turbulence Modeling for
CFD. DCW Industries Inc., La Canada,
California.

109

You might also like