TRANSONIC SOLUTIONS FOR THE MACH REFLECTION OF WEAK SHOCKS

J. K. Hunter , A. M. Tesdall
Abstract

We present numerical solutions of the steady and unsteady transonic small disturbance equations that describe the Mach reection of weak shock waves. The solutions contain a complex structure consisting of a sequence of triple points and tiny supersonic patches directly behind the leading triple point, formed by the reection of weak shocks and expansion waves between the sonic line and the Mach shock. The presence of an expansion fan at each triple point resolves the von Neumann paradox. The numerical results and theoretical considerations suggest that there may be an innite sequence of triple points in an inviscid weak shock Mach reection. von Neumann paradox, weak shock Mach reection, self-similar solutions

Keywords:

1.

Introduction

For sufciently weak shocks, the von Neumann theory of shock reection shows that a standard triple point conguration, in which three shocks meet at a point, is impossible. However, experimental observations of weak shock reections off a wedge show a pattern that resembles a single Mach reection with a triple point. This apparent disagreement between theory and observation became known as the triple point, or von Neumann, paradox. Guderley [1] proposed that there is a supersonic region behind the triple point of a steady weak shock Mach reection and an additional expansion wave centered at the triple point, which provides a resolution of the paradox. This theory was not widely accepted for weak shock Mach reection off a wedge because no evidence of a supersonic region or an expansion wave was found in experiments or numerical solutions. However, recent numerical solutions [2, 4, 5] have shown that Guderleys explanation is correct; the supersonic region was not detected previously because it is extremely small. The solutions in
Department of Mathematics, University of California at Davis, Davis, California 95616, U.S.A. ). Department of Mathematics, University of California at Davis, Davis, California 95616, U.S.A. ( ).

 ( & % # !   5 ! &   2 ''0)'$"43!

 ( & % # !      ''0)'$" 1

2
[2] were for a shock reection problem for the unsteady transonic small disturbance (UTSD) equations that provides an asymptotic description of weak shock reection, while the solutions in [4, 5] were for shock reection problems for the Euler equations. These solutions show a supersonic region and indications of an expansion fan at the triple point, but none of them are sufciently resolved to show the detailed structure of the ow in the supersonic region. In [3], high-resolution numerical solutions of the UTSD shock reection problem were obtained for a range of parameter values corresponding to Mach reection. All of the solutions contain a supersonic region behind the triple point. This region consists of a sequence of supersonic patches formed by the reection of shock and expansion waves between the sonic line and the Mach shock. Each reected shock intersects the Mach shock, resulting in a sequence of triple points, instead of a single triple point. The presence of an expansion fan at each triple point resolves the von Neumann triple point paradox. In this paper, we present solutions of the steady TSD equations for a problem that describes the stationary Mach reection of weak shocks. The solutions contain a sequence of supersonic patches and triple points similar to, but more resolved than, those obtained in [3]. These results support the conjecture, suggested by theoretical considerations [2, 3], that there is an innite sequence of triple points in an inviscid weak shock Mach reection. In addition, we present two new solutions of the UTSD shock reection problem at parameter values that are closer to the detachment point than the ones in [3]. The supersonic regions in these solutions are remarkably small.

2.

The steady shock reection problem

We write the steady TSD equations in normalized form as

where , are proportional to the , uid velocity components, respectively. Equation (1) is hyperbolic in , elliptic in , and changes type across the sonic line

The steady shock reection problem we consider here consists of (1) in the subject to boundary conditions (cf. Fig. 1) region

"r ) B Pd)

if

ruth310(  6 " )' " 6 2' e E gY9fd)

if

if

DC E

"#& %$# "#! DC B 2 )  G  #&F G  " )' #&`6 108 "g)Pd!) s B ) B r q p`h 3108 i  6 " )' &c6b3108 " a`Y319X 2" )' $  6 2" )' S  B ) B S $ H 2 H WDVPUTR" QPI 65" 2" )' 65" 2" )' 73A@98 734310(

(1)

(2)

(3) (4) (5)

Transonic solutions for the Mach reection of weak shocks

C B

Figure 1. A diagram of the computational domain, showing global -contours for the solution corresponds to and to . The point lies in Fig. 2(b). The boundary at . is the incident shock generated at the wedge corner between and , and (the ow is from right to left), and is the triple point.

Here, and are constants, and is the location of the incident and Mach shocks. Physically, this problem corresponds to the reection of a shock generated by a supersonic jet incident on a wedge located at , off a rigid wall located at .

3.

The numerical method

To solve (1), (3)(5) numerically, we evolve a solution of the UTSD equation forward in time until it converges to a steady state, using line relaxation. We write the unsteady TSD equation as

where is the velocity potential with , . We dene a non-uniform grid in the direction and in the direction, where and . We also dene as , the neighborhood of the point , with length where . Similar denitions apply for the non-uniform grid . We denote an approximate solution of (6) by , is a xed time step, and discretize (6) in time using where (7)

where the ux function is dened by . We solve (7) by sweeping from right to left in , using the spatial discretization (8)

0 A2 5 Q @3 s 7s 3 5 s Q $ 0 @3 4I( %  5 5 C@4( ) b$ @3 0 EI( % s xw y FF! 8q  xw r $ FF! vq  s @3 I ! q tihe D q piu he q tihe D ! sq piu hfcd y g g g r g e )  6 ' Y s f80XY 5 5 " #& Ia 6 % ' ` 13 I %  Y s   I $ s I X% 13 X% 5 5 5 675 WV" 92A3T9' R Q 5 U 0 "( ) 0 92 % S0 EP% I( 7s 3 6 ( p 1H( 9' s  C1H( ) ) s 3 ) s 3 7s D  ( ( 2) 6 7C@4( ) 5G 8E( ) F' s X)) 5 5 C1@0 2 B A2  @@92 8@4( ) 6)) 7s ) 7 D ) 7s 3 5 0 7s 3 5  ( 6 C@34( 3" CsE( 0' s 30  @4(2) s 3 2 0 12 ) ( ))  % d %   65" 2" ) 73319' % 

s B ) &) d) B r

(6)

 

Qs 3 7s 3 7s D ' 7 3 7s 3 ' 5 Q G hI 6 Q0 8@4( 3" Q0 8E( 0F$ I 6 Q0 '4( 3" Q0 8@4( 0F5 G 0 EI( %  0 V1HI( %

 

"#u  %   %  

6 2' Y9fe  )

 2

# # $ 

& ' %

!" 

qi

& 2

4
Here, is a numerical ux function, and . We used a second-order ux limiter scheme, with a Lax-Wendroff ux as the higher order ux, and an Engquist-Osher ux as the lower order ux. We compute solutions on the nite computational domain illustrated in Fig. 1. We use a grid that is exponentially stretched away from the triple point toward , and (4) on . the boundaries. We impose the no-ow condition (5) on In our computations we take equal to the value of behind the incident shock . On the boundary we impose the Dirichlet data (3). Since the ow is supersonic on , no boundary condition is required there.

4.

Numerical results

Fig. 2(a)(c) shows -contour plots for solutions of the steady shock reec, and . The dashed line is tion problem (1), (3)(5) with equal to , the sonic line (2), showing that all of the solutions contain a supersonic region directly behind the leading triple point, the size of which increases rapidly with . The slight thickening of the incident shock visible in Fig. 2(a) and in Fig. 1 is caused by the use of a stretched grid. The nonuniform grids are stretched by amounts between and , and the number of points in our largest grid . is approximately In our most rened solution, with , a sequence of triple points formed by the reection of weak shock and expansion waves between the sonic line and the Mach shock is clearly visible. Fig. 2(d) is a plot of the sonic line alone, which shows the sequence of supersonic patches behind the leading triple point. The number of triple points and supersonic patches in the numerical solution increases with increasing resolution. As in the transonic controversy for shock-free ows over an airfoil, the smooth termination of a supersonic patch appears unlikely because of the overdetermination of boundary value problems for hyperbolic PDEs. This argument and the numerical results suggest that there may be an innite sequence of triple points in the inviscid solution. In Fig. 3 we present two new solutions of the UTSD shock reection problem to augment those in [3]. There, it was shown that the size of the supersonic region decreased rapidly with increasing , where is a parameter that measures the inverse shock slope, and the largest value of used was . In Fig. 3 plots and . The supersonic region for of -contours are shown for equal to is smaller than the region for given in [3] by a linear factor of approximately . The supersonic regions in Fig. 3 are so small that the relative resolutions of the solutions are too low to see the detailed structure of the ows inside the supersonic region.

7s D xw FF! y tg u  Q0 8E( iq ! tg e D  q pg e y i

!# G

 G #

 G#q

 G G # #

 G  # Dq i

#  q G  G !#

r 8

 G

qi

   #G

G   # Vq

qi

Transonic solutions for the Mach reection of weak shocks

0.6544

0.533

0.6543

0.532

0.6542

0.6541

y
0.531 0.53 -0.403

-0.5167

-0.5166

-0.5165

-0.5164

-0.402

-0.401

-0.4

0.468 0.467 0.466

0.467 0.466 0.465

0.465

0.464

0.464 0.463 0.462 0.461 0.46 -0.3475 -0.345 -0.3425 -0.34

0.463 0.462 0.461 0.46 -0.3475 -0.345 -0.3425 -0.34

Figure 2. Contour plots of near the triple point for increasing values of in (a)(c), and in (d) a plot of the sonic line behind the incident and Mach shocks, for the solution in (c). The contour spacing is 0.01 in (a)(c), and the dotted line is the sonic line. The regions shown contain ; (b) the rened uniform grids, which have the following numbers of grid points: (a) ; (c),(d) .

A

 G #! q i

    

 G #u q i

(c)

(d)

 G #u q i
x

Yu q i G

(a)

(b)

A 9

0.15368

0.1845

y/t

0.18445
0.15366

1.3777

1.37775

y/t
0.15367

1.44601

1.44602

1.44603

Figure 3. Contour plots of near the triple point for two new values of , for the unsteady TSD shock reection problem in [3]. The -contour spacing is 0.01 in (a)(b). The dotted line is the sonic line. The regions shown contain the rened uniform grids, which have the following ; (b) . numbers of grid points: (a)

References
Oxford, 1962. [2] [3] [4]

, Weak shock reection, J. Fluid. Mech., 410 (2000), pp. 235261. , Self-similar solutions for weak shock reection, Siam. J. Appl. Math., 63 (2002), pp. 4261. , Numerical simulation of weak shock diffraction over a wedge under the von Neumann paradox conditions, Computational Mathematics and Mathematical Physics, 39 (1999), pp. 13351345. , The von Neumann paradox in weak shock reection, J. Fluid. Mech., 422 (2000), pp. 193205

v v  t  ) 2 1   )  & # 1 9  6 3 1 ) 2  p f 2 i g IuQDIsr0 QUbY8 XDRq8 2 QRhBA

[5]

  )  & #  ) 2 1    G  E 3 0 (CRQPI!2 HFCDCBA

9 f8 2  d A  ) 2 1 a ` 8 G 2 W T @IeFB(BcXb !YHCXVUFS

9  6 3  ) 2 1   )  & @8 7$54'0 ('%$#

"      ! 

[1]

, The Theory of Transonic Flow, Pergamon Press,

 A

 A 

 G !#u q

(a)

(b)

G  #u q

x/t

x/t