Sei sulla pagina 1di 9

High Temperature, Vol.38, No. 3, 2000, pp. 444--452. Translatedfrom Teplofizika Vysokikh Temperatur, Vol.38, No. 3, 2000, pp.

468--476. Original Russian Text Copyright 9 2000 by Borodin. Peigin.

HEAT AND MASS TRANSFER AND P H Y S I C A L G A S D Y N A M I C S

Numerical Investigation of Supersonic Flow past Blunt Bodies of Intricate Shape at an Angle of Attack and Slip Angle
A. I. Borodin and S. V. Peigin
Research Institute of Applied Mathematics and Mechanics, Tomsk, Russia

ReceivedApril 15, 1999 Abstract--A supersonic flow of viscous homogeneous gas past blunt bodies of intricate shape at an angle of attack and slip angle is investigated numerically within the model of complete three-dimensional viscous shock layer using the time relaxation method. The main regularities are studied of the general structure of flow and of the distribution of pressure and heat flux along the surface. An analysis is performed of their dependence on the shape of the body, angle of attack and slip angle, Mach and Reynolds numbers, and on other determining parameters of the problem. The accuracy and range of validity of a number of approximate approaches to the solution of the problem are estimated. INTRODUCTION Numerical investigations of three-dimensional supersonic flows of viscous homogeneous gas in the vicinity of blunt bodies using different gasdynamic models of flow were previously performed in [1-9]. The solution of equations of three-dimensional laminar boundary layer under conditions of flow past blunt bodies with two symmetry planes at zero angle of attack was derived in [1]; at an angle of attack, in [2, 3]; and for the general case of flow at an angle of attack and slip angle, in [4]. An investigation within the model of three-dimensional hypersonic (thin) viscous shock layer using a number of simplifying assumptions of the distribution of longitudinal pressure gradients was performed in [5, 6]. Without simplifying assumptions for the general case of the absence of symmetry planes in the flow, this problem was treated in [7] and, based on the solution of equations of three-dimensional parabolized viscous shock layer, in [8]. A solution of simplified Navier-Stokes equations for flows with two symmetry planes with preassigned distribution of longitudinal pressure gradients was derived in [9]. A fairly detailed review is found in [10]. FORMULATION OF THE PROBLEM We will treat a three-dimensional supersonic flow of viscous homogeneous gas in the vicinity of a smooth blunt body subjected to flow at an angle of attack and slip angle. We will investigate this flow in curvilinear coordinates [7] {x/}, in which the coordinate x 3 is reckoned on a normal to the body surface, and the coordinates {xl, x 2} are selected on the body surface and are polar with the origin at the stagnation point; in so doing, we will assume, for definiteness sake, that x 1 is a marching coordinate, and x 2 is an angular coordinate. We will use the model of complete viscous shock layer [11] as the input gasdynamic model. As is demonstrated by the results of the analysis made in [12], this model may be used in a wide range of flow geometry and Mach and Reynolds numbers and enables one to perform calculations of flow past bodies at moderate supersonic velocity of incident flow and for the cases when the shock layer is not thin in the entire windward part of the body subjected to flow [ 10]. Dimensionless unsteady-state equations for a threedimensional complete viscous shock layer in the above-identified curvilinear coordinates {x i} will be written as [10]

p ( D u ~ + Ajku u ) = -4g~a~)g

Ro p ( D u 3 + Ajku 3 Su k ) = - OP

'

(1)

~ X 3'

0018-151X/00/3803-0444525.00 9 2000 MAIK"Nauka/Interperiodica"

NUMERICAL INVESTIGATION OF SUPERSONIC FLOW

445

p= ?-lpT,
a D-~+ u-s a + u3~3, --s

it=

T O) ,

stagnation point, the boundary conditions on the body surface and on the stagnation line have the form

Re =

p=V..R ~t(v~/r

x 3 = 0:

Us

= 0,

pu 3

G(x l,x2),

T = T(o(x l, x2).
x l = 0: au3lax I = aTlOx 1 = aplax 1 = O, U =U
I 2

(3)

Here, all of the linear dimensions are related to the characteristic linear dimension R; V~. and u i are physical components of the velocity vector; p=V~P, p.@,

(4)

=0.

V~ T/cp~, It( V2~/Cp~)l.t,o, and ~/, respectively, denote the


pressure, density, temperature, viscosity, Prandtl number, and the adiabatic exponent; The Latin superscripts take the values of 1,2, 3, and the Greek ones, the values of 1, 2. Summation is performed for a pair of repeating superscripts, and no summation is performed for the superscripts in parentheses. Here and below, the subscripts w, ~, and s indicate the values on the body surface, in the incident flow, and behind the shock wave, respectively. The covariant components of metrical tensor are gsl~ = a~13- 2bs[~x3 + a~'Xb~b~'f~(x3) 2, g3s = g s 3 = O, g33 = 1 , and g = det]lgu[I = gllg22 - glEg21; the components of symmetrical tensors asl~ and bsl3, respectively, denote the first and second quadratic forms of the surface; and the c o e f f i c i e n t s A~k are known functions of the metrical tensor components and their derivatives [ 10]. The set of equations (1) is solved with the boundary conditions on the shock wave, on the body surface, on the stagnation line, and on the outlet boundary x 1 = x , = const. The boundary conditions on the shock wave are provided by the unsteady-state generalized Rankine-Hugoniot conditions, which may be most simply formulated in coordinates which are naturally related to the shock wave surface and have the form
p(V3-O-~) = V 3 - c ~ ,
1

The upper boundary X l - - X1 was selected far downstream of the flow, so that the perturbations due to this boundary would not have a significant effect on the flow in the region being treated. The so-called soft boundary conditions were preassigned on this boundary, for which the second derivatives with respect to the marching coordinate x I of all sought functions on this boundary were zero.

NUMERICAL SOLUTION OF THE PROBLEM In view of the selected coordinates and for resolving the singularities in the input equations and boundary conditions on the stagnation line, the numerical solution of the initially boundary problem (1) involved the transition to new dependent and independent variables of the Dorodnitsyn type, which further enabled one to obtain such profiles of the sought functions in the cross section of the layer that would be close to self-similar ones and weakly dependent on the marching coordinate,

!Xrp&.x 3

9 :,,

r:x,
P,4t-gdx

( v 3 - ~ ) ( v s - v~) = rt av ~
Re~. an '

A:(- G(2)

l:Ra
X 3

l-tP 2,

P+p(V3-C~) 2 =

Poo+(V3oo-~) 2,

pu3/-g =

a~, au

aT =
oRe~ an

(V3_O~)[T_ To.+ (V37 ~ ) 2

ax'

ax 2-N;p'l-gdx3
o

( v ~ - ~ ) ~ ~so(~_vD(v~____ - vb]
puS/ g q g(ss)

avi aw a~'
ays aX 3 '

~t(~x~, ;)'
,,
s
- -

(5)
&-~-~c~ ~fs

_s
U

Here, ~) is the dimensionless velocity of the shock wave front on a normal to its surface, and n is the coordinate reckoned along this normal. On the assumption that the stagnation line is straight and coincides with the normal to the body surface at the HIGH TEMPERATURE Vol. 38 No. 3 2000

fs =

Ws3-s' j(~,)

bs - g 4 ~ ,

446

BORODIN, PEIGIN

A = o 0~ ,
01nA.

~lOf~ + ( 2

+ ~ - ~ r ) S l + 0~
Ox3 aa = OI'~ a.

,101nA'~

Of 2

The boundary conditions on the shock wave at 4 = 1 have the form

+--~-J2+.~ 0--7" ,

r01nA

lA gbtm (O:fa Of~Olngtm~

The continuity equation in this case will be satisfied identically, and the remaining equations (1) in new variables may be transformed to

:<u.

O Flg 02fa pgal~,~-_gOx~] = + a ,J


1 O (b ~ Of~)

g~fl

OP

(~')2-=pO~a

IA g OT = (u| -3
~3 (u .-~)

F T-T** + A 2 g ( Ox'~2
-o, -- fi~) 1 g~(u-fi~)(fi~

(7)

2b(m/~(l~) yM 2

J'

+(~r
A O(~

~~l~ 0~ a ((~1)2 - Ubcm'~-~) 0fc~ a 3 l~(~-I~b(13)Ofl~

p ~ m1

bT.)~t,"("ca() + 2AallU (~ '


br
0

b,,) 04

~gg= B0x~ A~IA,


-u-t-&
where

+ A~s(~l )2 + ~- fl- sbca)bcs)Of~Ofs

04 04

gOtl5 ,q/~0X 3

c,

_ a2g12-sig221w _ 90xa'~

~'

t o~ ~ J '

C2 -0[-! g 0T Y-1ATlnp]

algl2--a2gll(OX 3 _ 9 OX3] =
g

to~ A~j,

OT ,.2-aOfaOT
= ~ +~
-A(1-

lOP ~2-aOfaOP
pO~ p o4 o~~'

(6)

~5-~

the tilde corresponds to the parameters calculated in the coordinates related to the shock wave surface, and M.. is the Mach number. On the body surface at 4 = 0, the boundary conditions have the form 0fa = 0,

Y-Yl(l-lnp))OT~y-1y_T~OAlnp
02f aO2f ~, -lg(~')2-a-~ga~ 042 /)4 2

04

f a = 0,

T = Tw(~ 1,~2).

(8)

~[Au3_ ~.@glp1 = 30A y - 1 T 0,fg


A~ J u ~; v ~ 0x' ~+,4 04 04'
0u' (~I)2-a0fa0U 3 AotBb(a) 3 b(B)Of aOf B

+~+

.0X 3 U3 =-a-~(+
P : ~t~)

(~l)2-a0fa0x 3 0X3 ~'(~-'~+ 0X' y- 1

A~I(0x3")-1

' P:

Y pr.

Therefore, the initial problem reduced to finding the solution of the set of differential nonlinear equations relative to the functions of current fa, temperature T, and transverse coordinate x 3, the equation for which is a corollary of the equation of momentum in projection on a normal to the body surface, and of the equations of continuity and state. A steady-state solution of this problem was found numerically using the time relaxation method on the basis of modification of an implicit (relative to the coordinate 4) scheme [13] with the order of approximation 0 ( 5 4 ) 4 + O ( ~ 2 ) 2 + O ( ~ l) + O(~'~). In the convection operator, the derivatives with respect to time x and marching coordinate ~1 were approximated by finite unilateral backward differences, and on the circumferential coordinate g2, by central differences calculated on the basis of previous global iteration on the running circumference ~ 1 + ~i~2 = const. HIGH TEMPERATURE Vol.38 No. 3 2000

NUMERICAL INVESTIGATION OF SUPERSONIC FLOW The derivatives of pressure and coordinate x 3 along the marching and circumferential directions (required to determine the so-called "elliptic" terms in the set of equations (6)) were calculated on the basis of central differences with the accuracy of o ( ~ l ) 2 + O ( ~ 2 ) 2 and stored for the entire flow field. In so doing, in order to increase the stability of the calculation results, the thus obtained distributions were smoothed in the neighborhood of the stagnation point with due regard for the behavior of these functions in the vicinity of the stagnation line. In view of the nonlinearity of the initial problem, iterations were employed, namely, local iterations in deriving a solution between the body surface and the shock wave at the running point ~l = constl and ~2 = const 2 and global iterations in deriving a solution in the shock layer on the running circumference E1 = consh. After completion of each global iteration, the value of A(~l, ~2) was determined on the running circumference using the cyclic sweep algorithm [14], and the transition was made to the next global iteration on this circumference. In order to improve the accuracy of calculations, use was made of a nonuniform (on the coordinate ~) difference net which enabled one to identify in the solution the boundary layer at the body surface. An iteration algorithm analogous to that of [15] was used to construct this net. The solution was assumed to be converging if the maximum difference between all sought functions in all nodes of the difference net in the running and previous time steps did not exceed 10-2~'L An important feature of the suggested numerical method is that it does not require for its realization the presence of symmetry planes in the flow (and, in view of this, enables one to perform calculations for the most general case of flow past bodies at an attack angle and slip angle) and that, in calculating the coefficients of friction and heat transfer on the body surface, there is no need to numerically differentiate the obtained profiles of velocity and temperature in the cross section of the shock layer. Along with the case when the body surface was preassigned in an analytical form, a case of practical importance was treated in which the body surface was preassigned in a tabular form. The following approach was used to simulate the latter case. Assume that the values of the radius vector of the body surface are known only at the points of the difference net rg(XI , x~ ). Because, in reality, the values of rg(Xl, x~ ) are known with some random error, we can write

447

S is a random number from the interval [0, 1], and A is the characteristic value of the error amplitude. In order to calculate the coefficients in the equations, one needs to know the value of the function r and its partial derivatives Om+nr/Omxl~nX 2 up to the third order inclusive (m + n < 3). Let F(i, j) = Om+nrJOmx~OnX2(X~, X~). Then, for smoothing these functions, the following two-step iteration procedure is used, based on cubic B-splines [16] (k is the iteration number).

Step 1
The periodicity of the functions being smoothed on the circumferential coordinate ~2 (subscript j) (F(i, J + 1) = F(i, 1)) for all 1 < i < I and 1 < j < J was used to smooth F(i,j) on the coordinate x 2 as follows:

Fk+l(i,j) = Fk(i,j) + ~(Fk(t, j - 1) - 2El(i, j) + F~(i, j + 1)).


The iterations were terminated when the number of changes of sign of the second derivative of the function F on the coordinate x 2 ceased to vary.
1 .

Step 2
The boundary conditions at the stagnation point F(1,j) = 0 (form=0 F ( 1 , j ) = 0, and n # 0 ) ,

F(O,j) = -F(2, j) ( f o r m = 1,3),

F(0, j) = F(2, j) (for m = 0, 2) and the condition of smoothness at the right-hand end of F(I + 1, j) = 2F(L j) - F(I - 1, j) for all 1 < i < I and 1 < j < J were used to smooth the function F(i,j) on the marching coordinate as follows: Fk+ l(i, j)

= Fk(i, j) + ~(Fk(i - 1, j) - 2Fk(i, j) + Fk(i + 1, j)).


Similarly to Step 1, the iterations were terminated when the number of changes of sign of the second derivative of the function F on the coordinate x I ceased to vary. This procedure was first applied to the function

rg(Xl, x2), after which the smoothed function rg (x l, x 2)


was determined, and central differences were used to determine the approximate values of Org/Oxa(xl, x~ ) which were smoothed by the foregoing procedure. The second- and third-order derivatives were smoothed analogously.
$

rg(xl, x~) = ra(Xl, X~) + A ( 2 S - 1),


where rg(XJ, x~ ) is the exact value of the radius vector

r(x l, x2) of the body surface at the point of net (x~, x~ ),


HIGH TEMPERATURE Vol. 38 No. 3 2000

448

BORODIN, PEIGIN well as the absolute, Pw and Cq, and relative, pO and C~q, distributions of pressure and heat flux along the surface of the body subjected to flow (related to their values at the stagnation point). The numerical algorithm was tested by comparing the calculation results with the data of [17] on the shape of the shock wave and on the distribution of pressure along the surface of the body and shock wave. Some results of comparison are given in Fig. 1, where curves 1, 2, and 3 indicate the dependence on the central angle (the spherical coordinates were defined in [17]) of the distribution of pressure along the shock wave and along the body surface, as well as the shock wave departure. Here, a = l , b = c = 0 . 5 , R e = 1 0 S , M**=6, Tw=0.1, T = 1.4, and t~ = 0; the curves indicate the results of our calculations, and the points indicate the data of [ 17]. In addition, a comparison was made between the results obtained in the sequence of nested nets HM N r for different steps with respect to time (M, N, and K denote the number of points on the coordinates ~l, ~2, and 4, respectively). Some comparison results are given in Table 1, which contains the values of the heat-transfer coefficient Cq, of the pressure on the surface P~, and of the shock wave departure x~ at the stagnation point and at one of the points of ellipsoid midsection at R e = 104, o~ = O, b = O.7, c = O.3, and Tw=O.15. An analysis of the results reveals that an H20 32 s net is sufficient to derive a solution with an accuracy of up to 1%, and K = 5 is sufficient for moderate values of Re. The number of steps with respect to time required to derive a converging solution at Re = l0 s was 130-180 on the average, which is somewhat worse than the respective results obtained using a scheme of a high order of accuracy for two-dimensional flows [15], but much less than the number of iterations in time that is required to derive a steady-state solution using numerical methods of the second order of approximation on the transverse coordinate [ 18, 19]. We will dwell on some analytical results. Note, first of all, that the absolute values of the coefficients of fric-

p0, po, x
1.0

0.9:
0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 10 20 30 40 50 2 60 70 80 90 9, deg

Fig. 1.

DISCUSSION OF THE RESULTS We treated a flow past a triaxial ellipsoid at an angle of attack tt and slip angle ~, whose equation in Cartesian coordinates {z'} had the form (zl/a - 1)2 + (z2/b) 2 + (z3/c) 2 = 1. The determining parameters of the problem assumed the following values: to = 0.5, 0 = 0.71, a = 1, T = 1.2-1.4,

Tw = 0.05-0.3, Re = 102-106 , 0~

M** = 4-15, 0.3<c<3, ~.

0.3<b<3, ~, 0~

The process of solution involved determining the components of the velocity vector and the pressure and temperature in the cross section of the shock layer, as Table 1

Variants

Quantities
Cq(O, O) P~(O, O) x~ (0, O) Cq(1, O)
Pro(l, 0) x~ (1, 0)

H20x32x8
= 0.1 0.04592 0.9602 0.03366 0.002220 0.03164 0.3132

~r

32x8
0.05

~20_x32x8
0.01 0.04593 0.9601 0.03366 0.002084 0.02898 0.3162

~.40 32x8
_ _ _ _ _ 0.05 x 0.04630 0.9602 0.03339 0.002380 0.03327 0.3087

H40x64xS 8r = 0.025 0.04634 0.9602 0.03339 0.002361 0.03284 0.3112

H40x32x 16
8r = 0.05 0.04632 0.9602 0.03339 0.002493 0.03301 0.3087 No. 3 2000

0.04592 0.9602 0.03366 0.002172 0.03085 0.31477

HIGH TEMPERATURE Vol.38

NUMERICAL INVESTIGATION OF SUPERSONIC FLOW

449

(a)

(d)

(b)

(e)

,.... ~

&
J

Fig. 2.

tion and heat transfer on the body surface are very sensitive to the variation of almost all of the determining parameters of the problem. Therefore, our main attention was given to the analysis of the dependence of relative distribution of pressure and heat flux along the surface on the main determining parameters of the problem. Characteristic examples of distribution of isolines of relative heat flux along the surface of an ellipsoid for different values of angles ct and [3 are given in Fig. 2 (b = 0.7, c = 0 . 3 , Re = 103, Moo = 10, Tw = 0.1,)'= 1.2). Here, the angles o~ and 13 assumed the following values: (a) o~ = 0, [3 = 0; (b) o~ = 30 ~ 13= 0; (c) o~ = 60 ~ 13= 0; (d) tz = 30 ~ 13= 7.5~ (e) cz = 30 ~ 13= 15~ (f) cz = 45 ~ 13 = 15 ~ The correlation between Cartesian coordinates, whose axes are shown in Fig. 2, with the curviHIGH TEMPERATURE Voi. 38 No. 3 2000

linear coordinates ~a on the surface has the form x = ~lcos(~2), y = ~lsin(~2), 0<x,y_<l,

where x is the horizontal axis, and y is the vertical axis; the origin of coordinates (intersection of axes) coincides with the stagnation point. One can see that the heat flux maximum located, for the treated shape of the body at zero angle of attack, at the stagnation point of the body shifts from the stagnation .point as the angle of attack increases; the heat flux maxnmum shifts (while remaining in the symmetry plane of the flow) toward a point of maximum mean curvature of the body surface. The absolute magnitude of this maximum is 5-7%, and it increases with the angle of attack o~ and Reynolds number.

450 1.0~ 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1
I I | I I I I | |

BORODIN, PEIGIN similarity of flow in this region is the ratio of the radii of the principal curvatures of the body surface at the stagnation point. In order to check the accuracy of this approximation, define the range of its validity, and estimate the degree to which the shape of the body affects the solution upstream of the flow, we have performed calculations of flow, at zero angle of attack and zero slip angle, past a family of ellipsoids of different shapes preassigned by the condition b = Cl , ~ and c = C2 4~. One can readily see that, because the radii of the principal curvatures of the ellipsoid surface at the stagnation point, Rl and RE, are determined by the formulas R 1 = b2/c and R2 - c2/a, all ellipsoids of this family have the same values of R l = C~ and R2 = C22 at the stagnation point. An analysis of numerical solutions has revealed that, for all treated ranges of variation of the Reynolds and Mach numbers, this approximation is characterized by a fairy good accuracy. One can assume that the flow in the neighborhood of bluntness of the body is indeed locally self-similar. For all treated ellipsoids of the given family (the longitudinal axis a varied from 0.25 to 4.0), the extent of departure of the shock wave and the profiles of velocity and temperature of the components in the cross section of the layer depended weakly on the parameter a. As is seen in Table 2, the variation of the absolute , values , of the heat-transfer coefficient . .C. i/ at the stagnation pomt for the above-descnbed famdy of surfaces was likewise insignificant (Re = 103, M~. = lO, Tw=O.l,~l= 1.2, o~= 0). We further made a comparison of the results of numerical calculations for different methods of preassigning the body surface, namely, in analytical and tabular forms. Figures 3 and 4 illustrate examples of such comparison for the case of flow past an ellipsoid with axes b = 0.7 and c = 0.3 with r = 30 ~ and ~ = 15~ (the distribution of cOo along the marching coordinate ~l (0 < ~l < 1), Fig. 3, and that along the circumferential coordinate ~2 (0 < ~2 < 2~), Fig. 4). Here, the curves indicate the case of analytical preassigning of the body surface, and the points indicate the case of tabular preassigning of the surface with the random error amplitude A = 0.05; in Fig. 3, lines 1 and 2 indicate ~2 = 0, 3.14; in Fig. 4, lines I and 2 indicate ~l = 0.2, 0.55. An analysis of these results has demonstrated that the smoothing procedure suggested by us is fairly efficient and enables one to derive a numerical solution to the problem with an accuracy sufficient for practical applications in a wide range of variation of the amplitude of

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Fig. 3.

It was previously observed [7, 8] that, under conditions of hypersonic flow of viscous homogeneous gas past smooth blunt bodies within the framework of more approximate models of thin and parabolized viscous shock layer (TVSL and PVSL models), the relative distribution of heat flux C~q in the neighborhood of bluntness of the body is largely conservative with respect to the Reynolds number and the quantity Tw (for the cooled surface of the body). The results of calculations performed by us using the model of complete viscous shock layer have demonstrated that the quantity C~q is indeed rather weakly dependent on the above-identified parameters and, therefore, the inference about the conservatism of C~q (including that with respect to the variation of the Mach number) is valid within the model of viscous shock layer as well, this being true of the entire frontal surface of the body up to its midsection. The results of asymptotic analysis of the NavierStokes equations [20, 21] indicate that within the TVSL and PVSL models, because of the parabolic behavior of the respective sets of equations, the solution of the problem in the neighborhood of the stagnation line may be found irrespective of the solution in the neighboring regions. In so doing, the only geometric parameter of
Table 2

Ellipsoid axes Cq

a=4, b= 1.4, c = 0.6 0.071

a= 1, b=0.7, c = 0.3 0.073

a=O.25, b=0.35, c=O.15 0.076

HIGH TEMPERATURE Vol. 38

No. 3

2000

NUMERICAL INVESTIGATION OF SUPERSONIC FLOW


c~ cO

451

1.0 0.9~ 0.8 0.7 0.6 0.5 0.4 0.3 0.2

0.21 0.18 0.15 0.12i:..__ ~ \ 1

009 .i \
I I t I t

0.L ~ ~ ~ ~ ,5 ~ 0' ~ 3 '~ : 0.1 1.26 2.51


Fig. 4.

"*,,, ~

3.77

5.02

6.28

0.2

0.4
Fig. 5.

0.6

0.8

random error in preassigning the radius vector of the body surface. We have further analyzed the effect of the Mach and Reynolds numbers and of the parameters co and T w on the shape of shock wave departure and on the distribution of the longitudinal and circumferential pressure gradients under conditions of flow past a body of fixed shape for the following range of variation of the parameters: 102 < Re
<

106, 0.5 < CO< 1.0,

= 1.2, and T* = 0.1) and to determine the shape of shock wave departure and the longitudinal and circumferential pressure gradients. In the second stage, for any other set of parameters Re, M., co, and Tw, a solution to steady-state equations of viscous shock layer is found from the range of (9) using the marching (on the coordinate ~l) numerical method on the assumption that the shape of the shock wave and the longitudinal and circumferential pressure gradients are determined in the first stage of solving the problem. Figure 5 illustrates a comparison for relative heat flux between the solutions of the problem in complete formulation (curves) and the results of calculations performed using the suggested approximate approach (points). Here, the following parameters of the problem were varied as compared with the basic variant: (1) with Re = 102, (2) with M~ = 6, (3) with 7 = 1.4, (4) with Tw = 0.25, (5) with Re = 105 (curves 2 - 5 indicate the values of C~q multiplied by five). As demonstrated by comparison with the results of calculations performed for exact formulation of the problem, the suggested approach is characterized by a good accuracy and may be used to advantage for practical applications. In addition, the suggested approach is economical, because, in the case of parametric calculations of a flow past a body of preassigned geometry for all values of Re, M~, y, and T w from the range of (9), it makes sufficient a single application of the time relaxation method, which, from the computational standpoint, requires considerably more computer

0.05 < T w < 0.4.

(9)

It has been demonstrated that, for all treated shapes of the body, the variation of the above-identified parameters within the limits of (9) causes little variation of these quantities. On the other hand, it has been demonstrated in [21] that, if the shape of the shock wave and the distribution of the longitudinal and circumferential pressure gradients are assumed to be known functions, the marching methods of calculations may be used to advantage in solving steady-state equations of viscous shock layer. All this made it possible to suggest the following (consisting of two stages) approximate approach to the calculation of flow past a body of preassigned shape for a wide range of variation of the Re and M , numbers and of the parameters y and Tw. In the first stage, the time relaxation method is used to find a numerical solution of the input unsteady-state equations of viscous shock layer for a fixed set of parameters Re*, M*, T*, and T* (in our case, these are Re* = 103, M* = 10, HIGH TEMPERATURE Vol. 38 No. 3 2000

452

BORODIN, PEIGIN 7. Borodin, A.I. and Peigin, S.V., Izv. Akad. Nauk SSSR Mekh. Zhidk. Gaza, 1989, no. 2, p. 150. 8. Borodin, A.I. and Peigin, S.V., Teplofiz. Vys. Temp., 1993, vol. 31, no. 6, p. 925 (High Temp. (Engl. transl.), vol. 31, no. 6, p. 853). 9. Gershbein, E.A. and Shcherbak, V.G., Izv. Akad. Nauk SSSR Mekh. Zhidk. Gaza, 1987, no. 4, p. 134. 10. Peigin, S.V. and Tirskii, G.A., Itogi Nauki Tekh. VINITI Ser. Mekh. Zhidk. Gaza, 1988, vol. 22, p. 62. 11. Davis, R.T., AAIA J., 1970, vol. 8, no. 5, p. 843. 12. Tirskii, G.A., Izv. Vyssh. Uchebn. Zaved. Fiz., 1993, vol. 36, no. 4, p. 15. 13. Petukhov, I.V., Numerical Analysis of Two-Dimensional Flows in Boundary Layer, in Chislennye metody resheniya differentsial'nykh i integral'nykh uravnenii i kvadraturnye formuly (Numerical Methods of Solving Differential and Integral Equations and Quadrature Formulas), Moscow: AN SSSR (USSR Acad. Sci.), 1964, p. 305. 14. Abramov, A.A. and Andreev, V.B., Zh. Vychisl. Mat. Mat. Fiz., 1963, vol. 3, no. 2, p. 377. 15. Borodin, A.I., Ivanov, V.A., and Peigin, S.V., Zh. Vychisl. Mat. Mat. Fiz, 1996, vol. 36, no. 8, p. 158. 16. Grebennikov, A.I., Metod splainov dlya resheniya nekorrektnykh zadach teorii priblizhenii (The Method of Splines for Solving Ill-Defined Problems of Approximation Theory), Moscow: Mosk. Gos. Univ. (Moscow State Univ.), 1983. 17. Lyubimov, A.N. and Rusanov, V.V., Techenie gaza okolo tupykh tel (Gas Flow past Blunt Bodies), Moscow: Nauka, 1970, vol. 2. 18. Anderson, J., Tannehill, J., and Pletcher, R., Computational Hydrodynamics and Heat Transfer, New York: Hemisphere, 1984. Translated under the title Vychislitel'naya gidrodinamika i teploobmen, Moscow: Mir, 1990, vol. 2. 19. Kovenya, V.M. and Yanenko, N.N., Metod rasshchepleniya v zadachakh gazovoi dinamiki (The Splitting Method in Problems in Gas Dynamics), Novosibirsk: Nauka, 1981. 20. Gershbein, E.A., An Asymptotic Study into the Problem of Hypersonic Three-Dimensional Flow of Viscous Gas past Blunt Bodies with a Permeable Surface, in Giperzvukovye prostranstvennye techeniya pri nalichii fiziko-khimicheskikh prevrashchenii (Hypersonic ThreeDimensional Flows in the Presence of Physicochemical Transformations), Moscow: Mosk. Gos. Univ. (Moscow State Univ.), 1981, p. 29. 21. Peigin, S.V., Parabolic Viscous Shock Layer Theory for 3D Hypersonic Gas Flow: Shock Wave I, Proc. 19th Int. Symp. on Shock Waves, Marseille, 1993, Berlin: Springer, 1995, vol. 1, p. 139.

resources for its realization than the marching numerical algorithms. CONCLUSION The time relaxation method is used to perform a numerical investigation of a supersonic flow of viscous homogeneous gas at an angle of attack and slip angle past triaxial ellipsoids of different shapes. An efficient smoothing procedure is suggested that enables one to derive, with a good accuracy, a solution to the problem for a case of practical importance in which the shape of the body surface is preassigned in a tabular form. An analysis is performed of the effect of the determining parameters of the problem on the shape of shock wave and on the distribution of the coefficients of friction and heat transfer. An approximate approach is suggested to the calculation of flow past bodies of preassigned shape in a wide range of variation of the Re and M** numbers and of the parameters T and Tw. The accuracy and range of validity are estimated for a number of approximate approaches to the solution of the problem. ACKNOWLEDGMENTS This study received support from the Russian Foundation for Basic Research (project no. 98-01-00298). REFERENCES 1. Brykina, I.G., Gershbein, E.A., and Peigin, S.V., Izv. Akad. Nauk SSSR Mekh. Zhidk. Gaza, 1982, no. 3, p. 49. 2. Andreev, G.N., Tr. Inst. Mekh. Mosk. Gos. Univ., 1970, no. 5, p. 50. 3. Aleksin, V.A. and Shevelev, Yu.D., lzv. Akad. Nauk SSSR Mekh. Zhidk. Gaza, 1983, no. 2, p. 39. 4. Borodin, A.I. and Peigin, S.V., Teplofiz. Vys. Temp., 1987, vol. 25, no. 3, p. 509. 5. Gershbein, E.A., Shchelin, V.S., and Yunitskii, S.A., Numerical and Approximate Analytical Solutions of Hypersonic Three-Dimensional Viscous Shock Layer with Moderately Small Reynolds Numbers, in Giperzvukovye prostranstvennye techeniya pri nalichii fiziko-khimicheskikh prevrashchenii (Hypersonic ThreeDimensional Flows in the Presence of Physicochemical Transformations), Moscow: Mosk. Gos. Univ. (Moscow State Univ.), 1981, p. 72. 6. Gershbein, E.A., Krupa, V.G., and Shchelin, V.S., Prikl. Mat. Mekh., 1986, vol. 50, no. 1, p. 102.

HIGH TEMPERATURE Vol.38

No. 3

2000

Potrebbero piacerti anche