b009 PDF

CFD Simulation Of Flow in a Triangular Driven Cavity
Vilas kalamkar 1
Gopal R. Shevare 3
Department Of Aerospace Engineering, IIT Bombay, Mumbai-400076
Ashish Pande 2
Department of Mechanical Engineering, VNIT, Nagpur-440011
Abstract: The problem of steady incompressible viscous flow within a driven cavity is of primary
importance in computational fluid dynamics (CFD). Most probably the attention to the driven cavity
flow is due to the simplicity of the geometry and also the boundary conditions. Despite its simplicity,
the driven cavity flow shows interesting flow features as the Reynolds number increases. This paper
explains the numerical simulations of various triangular shapes driven cavities for different
Reynolds numbers and comparison of the results with numerical solutions found in the literature.
Finite volume based explicit solver is used to carry out the steady state numerical simulations. The
numerical scheme employed for the simulation can handle both constant and variable density flows.
The results of the computations for different triangular geometry and Reynolds numbers were
presented in the form of plots of the streamline traces in the flow field and located the eddies
formation.
Keywords: CFD, viscous-incompressible flow, Finite volume method, Reynolds number, triangular
driven cavity
Introduction
The study of steady incompressible viscous flow within a driven cavity is very important in
that it provides a good bench mark case for the validation of CFD codes. Square cavity has
attracted a lot of CFD researchers and numerous publications are available which provide a
lot of insight on the nature of flow. This is due to the fact that square cavity geometry is
simpler to simulate, as it does not have any sharp corners. It has hence provided for a good
test case for numerical scientists to improve on existing algorithms by optimizing them for
the square cavity geometry. However, in reality, we rarely come across a square geometry.
The flow past cavities with sharp corners is ubiquitous in nature and is encountered in many
engineering applications. The presence of a sharp corner increases the degree of difficulty of
solving and fewer publications are available in literature.
--------------------------------------------------------------------------------------------------
1 Research scholar (email:[email protected])
2 Postgraduate student,(email:[email protected],[email protected],[email protected])
3 Professor (email:[email protected])
The 2-D steady incompressible flow inside a triangular cavity has been studied numerically
by McQuain et al.[1] ,Jyotsna and Vanka [2], Li and Tang [3] and Gaskell et al. [4]. Apart
from numerical studies, Moffat [5] have studied the triangular cavity flow analytically in the
Stokes regime. Moffat [5] have shown that the intensities of eddy and distance of the eddy
from the corner follows a geometric sequence. Jyotsna and Vankas [2] and Gaskell et al. [4]
have compared their numerical results with the analytical solutions. McQuain et al [1] have
applied the Batchelors mean square law to triangular cavities flow and analytically obtained
the inviscid core vorticity for infinite Reynolds number.
The main objective of the present paper is to analyze the results obtained by numerical
simulations of flow through a lid driven 2-D triangular cavity. The results of the several
triangular cavities are generated and compared with existing analytical and numerical results
obtained from literature. Simulations have been done for cases not addressed in previous
work and CFD predictions with regard to the flow conditions are presented.
The size and type of cavity were chosen to be similar to that available in the literature so as
to provide a direct comparison for the present results. The Grid generation module (GridZ) of
CFDExpert is used for creating the grids. The grid is clustered at the surface and also at the
sharp corners.
Mathematical Modeling
The scheme employed for the simulation can handle both constant and variable density
flows. Since at low speed, the changes in density are very small, the conservation equations
are rewritten by transforming the dependent variables from conserved quantities to primitive
variables. Thus, pressure and momentums replace density and respective individual velocity
components. The equations get conditioned in a manner that system changes into a fully
compressible solver at high speeds and into an incompressible solver at low speeds [6].
The Navier-Stokes equations are written as:
∂
Γ
∂t ∫∫∫ QdV + ∫∫ [F − G ].dA = 0
where
[
Q = p, vx , v y , vz ,T ]T
F and G are the convective and viscous fluxes respectively.

p, v x , v y , v z , T denote pressure, velocity components and temperature respectively.
The preconditioning matrix is given by,

ρT
Θ 0 0 0 ρT vx
ρ
Θvx ρ
0 0 ρT v y
Θvy 0
Γ= Θvz ρ
0
ρT vz
0 0
ρv y ρ T H + ρC p
ΘH −1 ρv x ρv z
where
⎛ 1 ⎞
∂ρ Θ = ⎜ 2 − ρT ⎟
ρT = ⎜U ρC p ⎟⎠
∂T ⎝ r
Here ρ and U r are density and a reference velocity for an ideal gas defined as follows:
⎧ ε c , if v < ε c ⎫
⎪ ⎪
U r = ⎨ v , if ε c < v < c ⎬
⎪ ⎪
⎩ c , if v > c ⎭
where ε is a small number ( 10 −5 ).
The convective fluxes were evaluated by modified Roe flux difference splitting. The
diffusive part of the scheme is derived from the preconditioned system in which the acoustic
speed was replaced by a pseudo acoustic speed. This preconditioning based pseudo
incompressible solver enhances the convergence history of the density-based solvers at all
speeds. However, the time accuracy of the solution is lost in this process.
Results and Discussions

All the results presented in the paper have at least a 3-degree fall in the mass residue and
convergences of the solutions have been ascertained before presenting the results
1 Equilateral triangular cavity
Simulations were done for Reynolds number of 100, 350, 500, and 1000. Fig. 1 (a-d) shows
the qualitative comparison of the streamline contours for different Reynolds number. The
results indicate very good agreement of the present results with those reported in the
literature at different Reynolds numbers with certain deviation. The presence of a major re-
circulation zone, which splits in two or more zones for increasing Reynolds numbers, is
captured very well though only for primary and secondary eddy core at the left top corner
which may be due to the coarse grid used comparatively.
Table 1 shows the location of the center of the primary eddy with that of literature. The
location of the primary eddy as obtained in the present study compares very well with around
10 % deviation that obtained by Li and Tang for Re = 100 and 350. However for Re = 500,
there is 1% deviation in x-direction and around 15% deviation in y-direction compared to
McQuain and Ribbens.In general for Re = 500 there is considerable scatter in the locations
between that predicted in literature [1-5] and the present study, which may be due to number
of grid points and their arrangement. Results for Re = 1000 are in good agreement with the
only result available of Li and Tang with a deviation in y coordinate.
(a) Re = 100 (b) Re = 350 (c) Re = 500
(d) Re = 1000
Fig 1: Streamline contours for Equilateral triangular cavity
Reynolds Number Present Study Gaskell McQuain Li and Tang

Grid points(51x51) Grid points(6605) Grid points(200x200) Grid points(80x80)
100 (0.351,0.3668) (0.333,0.356) (0.329,0.355) (0.368,0.363)
350 (0.1625,0.221) (0.149,0.234) (0.173,0.265) (0.152,0.213)
500 (0.1174,0.2271) (0.128,0.217) (0.173,0.265) (0.108,0.213)
1000 (0.1068,0.153) - - (0.108,0.138)
Table 1 : Comparison of primary eddy centers for equilateral triangle
2 Right angled triangular cavity

This is the new case considered for left aligned right angle triangle. Simulations were done
for different Reynolds numbers of 500,1000,2500 and 7500. The results are shown in Fig.2 .
From figures it has been observed that tertiary eddy core at left hand side upper corner is
observed from Re = 2500 on wards though its development can be seen at around Re = 1000.
With the increase of Reynolds number from 500 on wards up to around 2500 it is noted that
center of primary eddy moves away from perpendicular side i.e. towards left side in negative
x-direction and downward side in negative y-direction (i.e. in 3rd quadrant) till the
development of tertiary eddy has not started. As the development of tertiary core progresses
from around Re = 2500 onwards, center of primary eddy core moves gradually more in
positive x-direction compared to positive y-direction. The primary eddy locations for these
cases is shown in Table 2.
(a) Re = 500 (b) Re = 1000 (c) Re = 2500 (d) Re = 7500
Fig 2: Streamline contours for Right angle triangular driven cavity
Reynolds Number Present Study

Grid Points(81x81)
500 (0.7107,0.7577)
1000 (0.7073,0.7493)
2500 (0.7023,0.7306)
5000 (0.7057,0.7306)
7500 (0.7090,0.7308)
Table 2: Primary eddy centers for left aligned Right angle triangle
3 Isosceles triangular cavity

Finally the isosceles triangle driven cavity for which Reynolds number is kept constant is
considered and simulation were carried out for different corner angles viz. 28,40,80 and 90
degrees for the number of grid points (81x81). The results are shown in Fig. 3. Our result for
corner angle 28 degrees is in good agreement with Jyotsna and Vanka[2] with hardly 1%
deviation in x coordinates and y coordinate almost the same, refer Table 3. But, there is quite
high deviation in result with respect to Gaskell[1] which may be due to the variation in grids
considered. It is observed that for sharp corner i.e. for small corner angle, number of eddies
formation increases as suggested by Moffat[5] and can be seen for corner angles 28 and 40
degrees; which are having primary and secondary eddy core compared to 80 and 90 degrees
corner angle with only primary eddy. It can be seen that position of secondary eddy almost
remains the same but, as we go from 28 degree to 40-degree, center of primary eddy moves
toward right corner away from y-axis. After that with increase in corner angle the primary
center moves backward.
(a) Angle = 28 (b) Angle = 40 (c) Angle = 80 (d) Angle = 90
Fig 3: Streamline contours for isosceles triangular cavity
Angle(Degree) Present Study Gaskell Vanka and Jyotsna

Grid Points(81x81)
28 (0.1520,-0.5453) (0.124,-0.568) (0.153,-0.545)
40 (0.4854,-0.2961) - -
80 (0.2481,-0.2626) - -
90 (0.2450,-0.2334) - -
Table 3: Location of primary eddy centers for Isosceles triangle
Conclusion
In the numerical analysis of triangular driven cavity, accurate solutions of various cases are
presented. The governing equations are solved up to very low residuals at various Reynolds
numbers. The computed solutions are compared with numerical solutions of standard
available results and it is found that the results are in good agreement for wide range of
Reynolds numbers. The triangular driven cavity flow can be a very good benchmark test
case for CFD studies to test performances of numerical methods. In future flow analysis can
be observed for further higher numbers of Reynolds numbers taking more care of corner
effect for 2D cases and their flow analysis can be further Studied for 3D cases.
References
[1]McQuain W. D., Ribbens C. J., Wang C. Y., Watson L. T.,1994, {`` Steady viscous flow in a
trapezoidal cavity''}, Computers and Fluids, 23, 613-626
[2]Jyotsna R., S. P. Vanka,1995, {``Multigrid calculation of steady, viscous flow in a triangular
cavity ''}, Journal of Computational Physics, 22, 107-117.
[3]Li M., Tang T.,1996, {``Steady viscous flow in a triangular cavity by efficient numerical
techniques ''}, Computers and Mathematics with Applications, 31,55-65 .
[4]Gaskell P. H., Thompson H. M., Savage M. D.,1999, {``A finite element analysis of steady
viscous flow in triangular cavities''}, Proceedings of the Institution of Mechanical Engineers Part C-
Journal of Mechanical Engineering Science, 213, 263-276.
[5] Moffatt H. K.,1963,{``Viscous and resistive eddies near sharp corner ''}, Journal of Fluid
Mechanics, 18, 1-18.
[6] Jonathan M. Weiss and Wayne A. Smith,1995, {''Preconditioning Applied to Variable and
Constant Density Flows''}, {AIAA Journal}, 33(11), 2050-2057.

b009 PDF

Uploaded by

Copyright:

Available Formats

b009 PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

b009 PDF

Uploaded by

Copyright:

Available Formats

CFD Simulation Of Flow in a Triangular Driven Cavity

The Navier-Stokes equations are written as:

F and G are the convective and viscous fluxes respectively.

The preconditioning matrix is given by,

Results and Discussions

(a) Re = 100 (b) Re = 350 (c) Re = 500

Fig 1: Streamline contours for Equilateral triangular cavity

Reynolds Number Present Study Gaskell McQuain Li and Tang

Table 1 : Comparison of primary eddy centers for equilateral triangle

2 Right angled triangular cavity

Fig 2: Streamline contours for Right angle triangular driven cavity

Reynolds Number Present Study

3 Isosceles triangular cavity

Fig 3: Streamline contours for isosceles triangular cavity

Angle(Degree) Present Study Gaskell Vanka and Jyotsna

Table 3: Location of primary eddy centers for Isosceles triangle

You might also like