Performance of some Limiters on the Higher Order Computation of

Inviscid Hypersonic Flows

Paragmoni Kalita
Abhijit Gogoi
 Contents
• Introduction
• The Governing Equations
• The numerical scheme
• The MUSCL approach and the limiter functions
• Results and Discussion
• Concluding Remarks
 Introduction

• Inviscid hypersonic flow

• Calculation of pressure co-efficient and wave drag co-efficient
• Mach independence
• Numerical stability of flux scheme

• Two gas models

• Perfect gas model
• Chemically reacting equilibrium gas model
• Polynomial correlations of Tannehill and Mugge
• AUSM scheme
 Introduction

• Limiters
• To suppress numerical oscillations in the regions of high gradients (e.g.
Shock).

• Two limiter functions

• Van Albada
• Hemker Koren
 The Governing Equations

• Eular Equations

U F G
  0
t x y

   u   v 
 u      vu 
 p  u 2
 G 
U   F
 v    uv   p  v2 
     
  em   p   em  u   p   em  v 
 The numerical scheme

• AUSM (Advection Upwind Splitting Method)

• Split the flux vector into convective part and acoustic part

(c) ( p)
FF F

  0 
 u   p
where, F
(c )
u   ,F  
( p)
 u   p
   
  h0  0 
 The numerical scheme
• Convective part
• Based upon the Mach number at the cell interface

 a 
  ua 
F1/2  M 1/ 2  
(c)

  va 
 

 0 L/ R
h a
   L , if M 1/ 2  0
where    L / R  
   R , if M 1/ 2  0

• Here the subscript ‘1/2’ refers to the interface between the upstream and
downstream sides ‘L’ and ‘R’ respectively.
 The numerical scheme

• Mach number at the cell interface

 
M 1/ 2  M L  M R

 1  M  M  , if M  1

 2
where, M 
 1  M  12 , otherwise
 4
 The numerical scheme

• Pressure at the interface

 
p1/ 2  pL  pR

 p  M  M  , if M  1

  2 M 
where, p 
 p  M  12  2 M  , otherwise
 4
 The numerical scheme

• The MUSCL approach and the limiter functions

Figure 1: The left and right states for a cell-face in a cell-centred finite volume method.
 The numerical scheme
• The MUSCL approach and the limiter functions

1
1 where,  L / R  1  ˆ  rL / R  1  ˆ   L / R
U R  U I 1   R  U I  2  U I  1  2
2
U I 1  U I U I 1  U I
1 such that, rR  and rL 
U L  U I   L  U I 1  U I  U I  2  U I+1 U I  U I-1
and   r  is a slope limiter with the symmetric property,
2

  r    1 / r 
̂

The numerical scheme

• Dependence of type of reconstruction on the value of 𝜅

𝜅 Type of reconstruction
1 Centered
1/3 Upwind-biased
0 Upwind-biased
-1 Upwind
 The numerical scheme

• Van Albada limiter

2r 𝑟2 + 1
𝜅=0  r   𝜓 𝑟 =
1 + 𝑟2
r 1
2

• The right and left states of the cell interface

U R  U I 1 
1
R
 

a b  b a 
2
  2

2
a  b  2
2 2

1
UL  UI  L so that,
2
aR  U I+2  U I+1 , bR  U I+1  U I
aL  U I+1  U I , bL  U I  U I-1
 The numerical scheme
• Hemker-Koren limiter
3𝑟
𝜅=1 3 𝛷 𝑟 = 2
2𝑟 − 𝑟 + 2
• The right and left states of the cell interface

U R  U I 1 
1
R
 
 2
 
2a   b  b  2 a
2

2
2a  2b  ab  3
2 2

1
UL  UI  L so that,
2
aR  U I+2  U I+1 , bR  U I+1  U I
aL  U I+1  U I , bL  U I  U I-1
 Results and Discussion
• Flow over semi-cylinder at Mach number 15.0
• The cylinder diameter is 10 m.

Figure 2: A typical coarse grid for the computation of hypersonic flow over a semi-cylinder

• For the actual computations a 201X201 grid along the r-θ is taken
Results and Discussion

Figure 3: Steady state density contour plots.

Results and Discussion

Figure 4: Steady state pressure contour plots.

 Results and Discussion

(a) (b)
Figure 5: Variation of density along the stagnation line (a) Perfect gas model (b) Equilibrium air
Model of Tannehill and Mugge.
 Results and Discussion

(a) (b)
Figure 6: Variation of Mach number along the stagnation line (a) Perfect gas model (b) Equilibrium air
model of Tannehill and Mugge.
 Concluding Remarks
• The first order scheme is found to produce more numerical oscillations
in the vicinity of the strong shock as compared with the second order
schemes with the limiter functions.

• Performance of both the limiter functions found identical. However on

a much resolved scale the Van Abada limiter is found to suppress the
numerical oscillation more than the Hemker Koren limiter.

• Some higher order reconstruction techniques with other limiter

functions may be tried for such strong shock problems and a more
detailed error analysis for all these schemes may be carried out in
future.