See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/277475335

Numerical Simulation and Analysis of Supersonic flow over a flat plate.

Conference Paper · December 2014

CITATIONS READS
2 3,286

2 authors:

Sazid Zamal Hoque Paragmoni Kalita

Indian Institute of Technology Madras Tezpur University
19 PUBLICATIONS 104 CITATIONS 34 PUBLICATIONS 158 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Paragmoni Kalita on 01 September 2015.

The user has requested enhancement of the downloaded file.

 Numerical Simulation of Supersonic Viscous Flow over a Flat Plate

S. Z. Hoque1, P. Kalita2*

1
Department of Mechanical Engineering, Tezpur University, 784028,
Email: [email protected]
2* Department of Mechanical Engineering, Tezpur University, 784028,

Email: [email protected]

Abstract
In this paper, a numerical simulation and analysis of supersonic flow over a flat plate using FLUENT packages
as well as an in-house solver is presented. In the ANSYS FLUENT platform, Liou and Steffen’s Advection
Upstream Splitting Method (AUSM) is used for computing the inviscid fluxes. The in-house solver computes
the inviscid fluxes by using the AUSM as well as van Leer’s Flux Vector Splitting (FVS) schemes. Steady state
solutions of the two dimensional Navier-Stokes equations neglecting body forces are obtained by first order
Euler explicit time-marching technique. The inviscid fluxes in both ANSYS Fluent as well as the in-house
solver are computed by using a first order discretization scheme. The gradient terms in the viscous fluxes are
computed by Green’s theorem. Normalized temperature, velocity, entropy generation and Mach number are
calculated at the trailing edge of the flat plate for constant wall and adiabatic wall temperature case respectively.
The results obtained by ANSYS Fluent and the in-house solver for both the wall conditions are found to be in
good agreement.
Keywords: Supersonic, Fluent, Boundary layer

1 Introduction M = 3 using the FLUENT packages. The flexibility

and efficiency of FLUENT packages in employment
Computational fluid dynamics (CFD) is a of non-structured finite difference grids were
powerful tool for understanding the motion of fluids discussed. The accuracy of predictions with triangular
and gases. From food, agricultural industry to and polyhedral grids was analyzed. Isaev and Lysenko
aeronautics and automotive technology, (2009) further studied the unsteady supersonic flow in
computational fluid dynamics has been widely used step-shaped channel with the use of the CFX and
due to its flexibility and efficiency in reducing the FLUENT packages. Roe’s Flux Difference Splitting
effective design cost. Fluid dynamics, Mathematics (FDS) and the advection upwind splitting method
and high level computer programming language (AUSM) with second and third order scheme were
comprises the branch of computational fluid used to solve convective fluxes with triangle and
dynamics. Now a days, commercial CFD softwares polyhedral grids. It was found that existence of a right
(FLUENT, CFX, CFD-Post etc.) are commonly used corner in computational region caused a large phase
in many industries as well as academic institutes to change in the evolution of the flow. Tu et al. (2006)
understand and visualize the flow behavior including illustrated the ANSYS CFX solution of supersonic
supersonic and hypersonic flow of gases. flow over flat plate. Normalized velocity, pressure and
Supersonic flow over a flat plate with zero temperature profiles are plotted against normalized y-
incidence is a classical problem in computational fluid distance for isothermal and adiabatic conditions of the
dynamics. Koirala et al. (2013) investigated the flat plate. On the other hand, entropy generation due
supersonic flow over flat plate numerically. The to viscous dissipation is also an important parameter
variation of flow parameters at Mach numbers (M) to understand the complex nature of the flow. Budair
2.2, 3.2, 4.2, 5.2 and 10.2 and at plate temperatures (2001) presented an entropy analysis of flow
288.16 K and 373 K was computed. The normalized generated due to the motion of flat plate. The entropy
surface pressure along the flat plate was also generation in the space between the plates is more
computed. At the leading edge, the pressure change considerable at initial time than at later times. The
was found to be higher also pressure increased with solution of supersonic flow over flat plate can be used
increasing M. The leading edge shock as well as directly to any curved surface with large radius of
boundary layer thickness was distinguished better at convergence. Therefore, results obtained using CFD
M 10.2. Despande et al. (2010) studied the supersonic techniques have immense importance in designing the
flow over a moving rectangular protrusion on a flat large airfoils of aircraft wings.
plate. With the protrusion motion, the flow field was This paper is organized into 7 sections.
found to depend strongly on the velocity of Computational domain and methodology used is
protrusion. Isaev and Lysenko (2004) presented the discussed briefly in section 2. The governing
simulation of supersonic flow in a step channel with equations of fluid flow are described in section 3
along with the property relations of the fluid. energy and the equations of state. Neglecting the
Boundary conditions used in FLUENT as well as presence of body forces and volumetric heating value
in-house solver are discussed in section 4. Section 5 the two dimensional equations are:
consists of a brief description of simulation of the Continuity equation:
rectangular flow domain in FLUENT. Flow parameter ¶ρ ¶ ( ρu ) ¶ ( ρv )
results are discussed in section 6. Finally, in section 7, + + =0 (1)
conclusive remarks are drawn for the results obtained ¶t ¶x ¶y
using in-house solver as well as FLUENT packages. where, ρ is the density of the fluid, u- and v- are the
2 Methodology velocity components in x- and y- directions
respectively and t is the time.
Air is taken as a perfect gas with constant Prandtl
number and the viscosity is computed by using the x-momentum equation:
¶ ( ρu ) ¶ ( p + ρu ) ¶ ( ρuv ) ¶τ xx ¶τ yx
2
Sutherland’s law. The flow behaviour over flat plate
for Mach number M = 3 is investigated using an in- + + = + (2)
house solver as well as in ANSYS Fluent. In the ¶t ¶x ¶y ¶x ¶y
ANSYS Fluent platform Liou and Steffen’s AUSM y-momentum equation:
scheme (1993) is used to solve the inviscid fluxes.
¶ ( ρv ) ¶ ( ρuv ) ¶ ( p + ρv ) ¶τ xy ¶τ yy
2
The in-house solver uses both AUSM and van Leer’s + + = + (3)
FVS (Anderson et al., 1986) schemes. For the ¶t ¶x ¶y ¶x ¶y
purpose of comparison of different inviscid flux Energy equation,
¶E ¶ {( p + E ) u} ¶ {( p + E ) v}
schemes the order of accuracy is kept unity to have a
common benchmark. For laminar flow condition the + +
length of the flat plate is chosen to be extremely ¶t ¶x ¶y
small. ¶ ( uτ xx + vτ xy ) ¶ ( uτ yx +vτ yy ) ¶ æ ¶T ö (4)
= + + çk ÷
¶x ¶y ¶x è ¶x ø
¶ æ ¶T ö
+ çk ÷
¶y è ¶y ø
where, E is given by,
ì 1 ü
E = ρ íe + ( u 2 + v 2 ) ý (5)
î 2 þ
where, e is the specific internal energy. T is the
absolute temperature, k is the thermal conductivity.
The normal viscous stress components are t xx and
Figure 1 Two dimensional computational flow
domain t yy . t xy = t yx is the tangential viscous stress
Based upon the freestream parameters of air, the component. To solve the above system of equations,
Reynolds number at the trailing edge is computed as we need some more equations to reduce the number
2000. Thus the flow regime is laminar. For Fluent of unknown variables. For perfect gas, the density can
simulation, the computational domain is created in be calculated as,
ANSYS workbench. A uniform structure mesh of ρ= pRT (6)
rectangular elements is created using quadrilateral Also considering air as calorically perfect gas,
meshing in ANSYS mechanical. Generation of e= C v T (7)
appropriate mesh has a significant impact in where, Cv is the specific heat at constant volume.
converging the solution in Fluent. The schematic
Assuming the Prandtl number, Pr as constant, the
diagram of the flow domain is shown in the Figure 1.
thermal conductivity can be expressed as,
The length of domain is equal to the plate length and
the height is taken to be 6 times the boundary layer mC p
thickness (Tu et al., 2006). k= (8)
Pr
For the in-house solver, a computer C-code is
used to define the computational domain taking Here, Cp is the specific heat at constant pressure.
x-co-ordinate along the length of the plate and 4 Boundary Conditions
y-co-ordinate along the normal to the plate.
Boundary conditions play a vital role in
3 The equations of motion converging the solution of any numerical simulation.
The governing equations for fluid flow are based In general, the boundary-conditions are given
on the laws of conservation of mass and momentum, according to the sign of the eigenvalues of the
convective flux Jacobian matrix (Anderson John D. pressure in the dummy cell is set the same as that in
Jr., 1995). The boundary conditions implemented in the interior cell. The density is calculated from
the ANSYS Fluent solver are illustrated in Figure 2. pressure and temperature using the equation of state.
The free stream atmospheric pressure is taken as The u- and v-velocities in the dummy cell for both the
operating pressure for the flow problem. At inlet, wall conditions are obtained by using the no-slip
boundary conditions on velocity, pressure, boundary conditions.
temperature are given with respect to the free stream
condition (Dirichlet type). In Fluent, the pressure inlet 5 Fluent Simulations
condition is considered for the inlet boundary A mesh of 80X80 rectangular grids are created
condition. The static gauge pressure with respect to over the computational domain. The 2D domain is
the operating atmospheric pressure is taken to be zero. then loaded in ANSYS Fluent. Explicit form of the
Inflow, top boundary- Dirichlet density based solver is used to solve the governing
y
conditions based on freestream values equations for the steady flow case. The first order
Euler explicit scheme with a CFL number of 0.2 is
Computational domain
used for the time-marching of the Navier-Stokes
Outflow boundary condition, equation in Fluent. The viscosity for the air is
Flow all properties are extrapolated computed from the Sutherland’s viscosity law with
direction from upstream three co-efficient. A Fluent user defined function
M=3.0 u-
Flat plate, no-slip condition for (UDF) is written using the DEFINE_PROPERTY
component of velocity, adiabatic or
isothermal wall condition
macros to find the value of thermal conductivity (Eqn.
8). Now, the boundary conditions are established to
x the flow domain in Fluent as discussed in the above
section. The green cell-based method under the
Figure 2 Boundary condition for two dimensional gradient panel in Fluent is used. Solution initialization
laminar supersonic flow has been done with the free stream conditions of the
fluid flow. The solution convergence criterion is given
The total pressure is calculated as using the relation, as 1x10-7 for continuity, x-velocity, y-velocity and
g energy equation respectively. Continuity equation
p'0 + pop æ γ - 1 2 ö g -1 takes maximum iteration to converge.
'
= ç 1+ M ÷ (9)
ps + pop è 2 ø
6 Results and Discussion
where, p0' is the total pressure at inlet, ps' is the static
The Fluent solution is loaded in ANSYS CFD-
pressure just adjacent to the cell, pop is the operating
post for obtaining the numerical results. A variable for
pressure.
the non-dimensional y-distance in CFD-post, as
For the outflow boundary condition, the pressure
suggested by Van Driest (1952) is created. This
outlet boundary condition is considered at outlet with normalized y-distance variable is used in plotting the
zero back flow temperature. No-slip condition for profiles of various normalized flow field parameters.
velocity and density is considered for the stationary All the free stream values along with the
wall surface. The heat flux for adiabatic case is taken thermodynamic properties of the laminar flow over
to be zero. For the constant wall temperature case the the boundary layer are listed in Table 1.
temperature of the wall is taken to be the free stream Figure 3 shows the Fluent result of normalized
value of the temperature. We have used pressure far
velocity profiles for both adiabatic and constant wall
field boundary conditions with Mach number value as
conditions at the trailing edge. The boundary layer is
3 for the top (far field). Gauge pressure is taken in
thicker in adiabatic wall temperature case then
pressure panel and free stream value of temperature is
constant wall temperature case. Also, the formation of
taken for the temperature panel of the far field shock is more distinguishable in case of adiabatic wall
boundary condition. temperature case.
For the in-house solver, the inlet boundary The velocity profiles using van Leer’s FVS
conditions are specified by fixing the Mach number, scheme and AUSM scheme for adiabatic and
static pressure, static temperature and v-velocity of the isothermal wall conditions obtained using the
freestream. At the outlet all the variables are in-house solver are shown in Figure 4 and Figure 5
extrapolated from within the computational domain. respectively. van Leer’s FVS scheme provides a
At the top boundary the freestream conditions are thicker boundary layer than AUSM scheme in both
specified. For adiabatic wall condition, the the cases. But the leading edge shock effect is better
temperature gradient at the wall is made zero by visualized using AUSM scheme as the normalized
specifying the temperature at the dummy cell just
velocity profile changes abruptly as shown in Figure.
below the wall the same as that in the interior cell. For
isothermal wall the temperature in the dummy cell is
set the same as the specified wall temperature. The
Table 1 Data used for laminar flow case isothermal wall condition. Also the leading edge
shock wave is recognizable easily in both the cases.
Parameter Value Near the shock wave, temperature increases sharply
Plate length L, m 2.85X10-5 and after that it attains free stream value of
Speed Uinf , m/s 340.28 temperature.
Pressure Pinf, N/m2 101325
Temperature Tinf, K 288.16
Density, kg/m3 1.225
Mach number, M 3.0
Dynamic viscosity, µ, kg/m.s 1.7894X10-5
Specific gas constant R,
287
J/kg.K
Specific heat Cp, J/kg.K 1005.4
Prandtl number, Pr 0.71

Figure 5 Normalized u-velocity profiles for

isothermal wall temperature condition at trailing
edge obtained using the in-house solver

Figure 3 Normalized u-velocity profiles for the

adiabatic and constant wall temperature
conditions at trailing edge obtained using ANSYS
Fluent

Figure 6 Normalized temperature profiles for the

adiabatic and constant wall temperature
conditions at trailing edge obtained using ANSYS
Fluent
The Mach number profile at the trailing edge
obtained using Fluent is shown in Figure 9. Clearly it
can be seen that a stronger shock wave is formed for
adiabatic wall temperature condition. Also, the
Figure 4 Normalized u-velocity profiles for boundary layer is thicker for adiabatic case.
adiabatic wall temperature condition at trailing Figure 10 and 11 illustrate the Mach number
edge obtained using the in-house solver profile at the trailing edge using van Leer’s FVS and
AUSM scheme for adiabatic and constant wall
The non-dimensionalized temperature profile at temperature case respectively. A thicker boundary
the trailing edge for the adiabatic and isothermal wall layer is observed using van Leer’s FVS scheme. Near
conditions obtained using ANSYS-Fluent is shown in the leading edge of the shock, the Mach number
Figure 6. Figures 7 and 8 show the normalized decreases and after that it attains free stream value.
temperature profile using van Leer’s FVS and AUSM Entropy profile obtained using Fluent is shown in
schemes for adiabatic and isothermal wall temperature Figure 12. Entropy generation in adiabatic wall
cases respectively. It can be seen that results obtained temperature case is much higher than that of constant
using in-house solver are in good agreement with that wall temperature case at trailing edge due to higher
of Fluent results. A thicker boundary layer is observed value of temperature in adiabatic case.
using van Leer’s FVS scheme for both adiabatic and
 obtained using in-house solver is slightly higher than
the result obtained using FLUENT solver. A higher
rate of entropy generation can be seen using van
Leer’s FVS scheme in adiabatic wall temperature
case.

Figure 7 Normalized temperature profiles for

adiabatic wall temperature condition at trailing
edge obtained using in-house solver

Figure 10 Mach number profiles for adiabatic wall

temperature condition at trailing edge obtained
using in-house solver

Figure 8 Normalized temperature profiles for

isothermal wall temperature condition at trailing
edge obtained using in-house solver

Figure 11 Mach number profiles for constant wall

temperature condition at trailing edge obtained
using in-house solver

Figure 9 Mach number profiles for the adiabatic

and constant wall temperature conditions at
trailing edge obtained using ANSYS FLUENT
The entropy attains almost the datum value at
free stream as all the flow parameters attain free
stream values. The in-house solver results for entropy
at trailing edge are shown in Figure 13 and 14 for Figure 12 Entropy profiles for the adiabatic and
constant wall and adiabatic wall temperature case constant wall temperature conditions at trailing
respectively. The entropy generation for adiabatic case edge using ANSYS Fluent
 References
Anderson John D. Jr. (1995), Computational Fluid
dynamics, McGraw-Hill, Inc.

Anderson, W.K., Thomas, J.L. and van Leer, B.

(1986), Comparison of finite volume flux vector
splitting for the Euler equations, AIAA Journal, Vol.
24, No. 9, pp. 1435-1460.

Budair M. O. (2001), Entropy analysis of unsteady

flow on a flat plate, International Journal of Energy
Research, Vol. 25, pp. 519-524.

Deshpande V., Eshpuniyani B. and Sanghi S. (2010),

Computational study of supersonic flow over a flat
Figure 13 Entropy profiles for constant wall plate with moving protrusion, Proceedings of the 37th
temperature condition at trailing edge obtained and 4th national conference on Fluid Mechanics and
using in-house solver Fluid Power, IIT Madras, December 16-18.

Isaev S. A. and Lysenko D. A. (2004), Testing of the

fluent package in calculation of supersonic flow in
step channel, Journal of Engineering Physics and
Thermophysics, Vol. 77, No. 4, pp. 457-460.

Isaev S. A. and Lysenko D. A. (2009), Testing of

Numerical Methods, Convective Schemes, Algorithms
for Approximation of flows, and grid structures by the
example of a supersonic flow in a step-shaped channel
with the use of the CFX and Fluent packages, Journal
of Engineering Physics and Thermo-physics, Vol. 82,
pp. 321-326.

Koirala R. P., Kumar K. V. S., Adhikari S. and

Figure 14 Entropy profiles for adiabatic wall Praveen D. (2013), Numerical Simulation and
temperature condition at trailing edge obtained analysis of Supersonic flow over a flat plate, Journal
using in-house solver of Mechanical and Civil Engineering, Vol. 4, Issue 6,
pp. 36-41.
7 Conclusion
Supersonic flow over a flat plate is analyzed Liou M. S. and Steffen C. J. Jr. (1993), A New Flux
numerically with the use of FLUENT packages as Splitting Scheme, Journal of Computational Physics,
well as an in-house solver. A first order upwind Vol. 107, pp. 23-29
scheme is used to solve the convective fluxes in both
the solvers. The variation of temperature, velocity, Tu J., Yeoh G. H. and Liu. C. (2006), Computational
Mach number and entropy at the trailing edge of the Fluid Dynamics – A practical Approach , Elsevier, pp.
plate are analyzed for both constant wall and adiabatic 334-344.
wall temperature. From the analysis it is found that
the leading edge shock has a deep impact on the flow Van Driest, E. R. (1952), Investigation of Laminar
parameters. The u-velocity, temperature and Mach Layer in Compressible Fluids Using the Crocco
number change abruptly near the leading edge shock. Method, NACA Tech. Note 2579.
The AUSM scheme allows a better detection of shock
compared with van Leer’s FVS scheme. The
u-velocity decreases while the temperature and Mach
number increase near the shock wave from its free
stream values. Again, adiabatic wall temperature case
results stronger shock than constant wall temperature
case. The results obtained using the in-house solver
are in good agreement with the Fluent results.

View publication stats