Professional Documents
Culture Documents
Butler Lighthill 1997 The Numerical Solution of Hyperbolic Systems of Partial Differential Equations in Three
Butler Lighthill 1997 The Numerical Solution of Hyperbolic Systems of Partial Differential Equations in Three
B y D. S. B utler
.E
D
R
A,Fort Halstead
(Communicated by M . J . —
,lihtgA
I Received 22 October 1959)
em ploying difference relatio n s along four b ic h aracteristics a n d one tim e-like o rd in a ry curve
thro u g h each p o in t. F ro m these difference relatio n s th e d eriv ativ es o f th e d ep en d e n t
variables a t th e unknow n p o in t are elim in ated . T he solution a t a n y p o in t can th e n be com
p u ted , w ith an error p ro p o rtio n al to th e step size cubed, w ith o u t referring to conditions
outside its dom ain of dependence.
The applicatio n o f th e m e th o d to th e system s of e q u a tio n s governing u n ste a d y p lan e
m otion an d ste a d y supersonic flow o f a n inviscid, n o n -co n d u ctin g fluid is discussed in d etail.
As an exam ple of th e use o f th e m eth o d , th e flow over a p a rtic u la r d e lta-sh a p ed b o d y h as been
com puted.
1. I n t r o d u c t i o n
There are several techniques available for th e num erical in te g ratio n o f hyperbolic
system s of p a rtia l differential equations in th re e ind ep en d en t variables. M any of
these techniques involve th e direct su b stitu tio n o f finite difference approxim ations
for th e derivatives in th e equations. The algebraic equations o b tain ed in th is way
can th en be solved in a step-by-step fashion. M ethods of th is ty p e h ave th e advantage
th a t th ey are com paratively sim ple a n d well suited for use w ith a u to m a tic digital
com puters, b u t u n fo rtu n a te ly it is difficult to a d a p t th em to problem s in which
boundary conditions are given on a fixed curved surface.
I t is well know n th a t there is a com plex of b icharacteristic curves associated w ith
each solution of a three-variable system of hyperbolic equations a n d th a t these curves
generate a fam ily of characteristic surfaces. F o r each o f these curves, th e system of
equations gives a relation betw een th e derivatives of th e dep en d en t variables along
the curve and th e derivatives in an y o th er direction lying in th e characteristic
surface containing it. These relations betw een derivatives in only tw o directions are
analogous to th e characteristic conditions, involving derivatives in only one direc
tion, obtained from hyperbolic system s in tw o independent variables. The num erical
technique developed in th is p ap er m akes use o f th is special p ro p e rty o f bicharac
teristic curves.
Several o th er m ethods using th is p ro p erty h ave been suggested for th e com pres
sible flow equations. Coburn & D olph (1947 ) suggest a m ethod, fu rth e r developed by
H o lt (1956), w hich involves th e use of difference relations set up along tw o o f the
bicharacteristic curves thro u g h each p o in t and one ord in ary curve. The ordinary
curve chosen is th e intersection of th e characteristic surfaces th ro u g h th e tw o bi-
cliaracteristics, an d therefore lies outside th e envelope of characteristic surfaces
through th e point. This m eans th a t in order to o b tain th e solution a t a n y p oint, it is
necessary to refer to conditions outside th e tru e dom ain o f dependence o f th a t point.
[ 232 ]
Hyperbolic systems of partial differential equations 233
ls a resu lt it is n o t clear how b o u n d a ry conditions a t shock w aves a n d sim ilar
ioundaries should be tre a te d .
Sim ilar m ethods have been described b y S auer (1950) a n d B ru h n & H aack (1958).
fruhn & H aack specify th e bich aracteristic curves w hich should be used b y relating
heir directions to th e direction o f th e p article velocity. T he bicharacteristic relations
an th e n be expressed in a n intrinsic form a n d it is possible to deal w ith th e boun d ary
onditions on a curved wall.
A different m eth o d is suggested b y T hornhill (1948). T his m ethod uses difference
elations applied along th re e b ich aracteristics or over th re e characteristic surfaces
hrough each p o in t. T he solution can th e n be c o n stru cted a t a m esh of p o ints which
re th e intersections o f th re e fam ilies o f c h aracteristic surfaces.
Downloaded from https://royalsocietypublishing.org/ on 18 March 2024
15 Vol. 255. A.
234 D. S. Butler
a space-like surface, bounded by its intersections w ith one or m ore non-space-like
surfaces on which additional boundary conditions are given. This situ a tio n is very
common in compressible flow problem s, since shock waves, co n tact discontinuities
and rigid boundaries are all non-space-like surfaces. I n these problem s th e general
m ethod described above cannot be used near th e non-space-like boundaries, since
com plete initial conditions are n o t available on some o f th e b icharacteristic curves.
Special m ethods can be devised for com puting th e solution a t p o ints on these
surfaces using th e boundary conditions in conjunction w ith difference relations
along bicharacteristic and o th er line elem ents. I n these special m ethods th e essential
feature of th e general m ethod is retained , b y again using a surfeit of difference
relations from which derivatives in directions outside th e dom ain of dependence can
be elim inated.
Downloaded from https://royalsocietypublishing.org/ on 18 March 2024
2. T h e g e n e r a l p r o p e r t ie s o f h y p e r b o l ic s y s t e m s in t h r e e v a r ia b l e s
A general system of w-quasi-linear differential equations in n -dependent variables
up and th ree independent variables xi is considered. These are w ritte n in th e form
a^ uvR xi) = (2*1)
where api a n d are know n functions o f a n d x {. I n e q u atio n (2-1) a n d th
m ainder of th is section th e dum m y suffix sum m ation convention is used, w ith
Greek suffixes /i, v, etc., running over th e first n integers a n d italic suffixes i, j, etc.,
taking th e values 1, 2 or 3.
A transform ation to independent variables * is m ade. E q u a tio n (2-1) th en
becomes %n(dujdxi) = bp ( 2-2)
7 MT a cx 3
det,v > § } (2-5)
k dxk\ li bxi dxj ’
where A t lji-
Hyperbolic systems of partial differential equations 235
The surfaces x 3(x1} x 2, x 3)= co n stan t, th ro u g h a n y p o in t w hich satisfy
0^ 04
0, ( 2 -6 )
lj dxidXj
ouch a q u a d ra tic cone a t th is p o int. T he problem is classified as hyperbolic, elliptic
>r parabolic a t th is p o int, if th is cone is real, im aginary or degenerate, respectively.
!t is assum ed th a t th e problem u n d er discussion is hyperbolic a t all points
onsidered.
A surface w hich satisfies condition (2-6) a t all its po ints is said to be characteristic
md th e envelope of all such surfaces th ro u g h a p a rtic u la r p o in t is referred to as th e
3haracteristic conoid o f th e p o in t. T he curves of co n ta c t o f th e characteristic
Downloaded from https://royalsocietypublishing.org/ on 18 March 2024
where 8 = 0 v'cos
| 0 vi—sin 0 ji{ ,
uftcc) = uv a t xi = ai
uv(f) =u v a t xi =xi( - </>)>et
(3-12)
+ i K * ,‘6 - ^ f e U +^ -
where At = 1 {Alv( a ) + A lv(})ha n d B* = + £ # )} .
E quation (3-7) is w eighted w ith th e fac to r h/t a n d in teg rated . F rom this integral,
equation (3-12) m u ltiplied b y tt is su b tra c te d . T his gives
■/*2i h A vd(J)
u„(a)
.Jo
—nA't, ju M np £ - nA U v(h)
f*2 n __
+ hB d<fi —nhB*+ 0( hz ). (3T3)
Jo
The th ree conditions (3-9), (3-10) an d (3-13) provide th re e equations for uv{oc), which
are independent if n ^ 3. I f n 3>, it is assum ed th a t a fu rth e r (n —3
be obtained from th e characteristic conditions along th e curve d xt — li d r. N one of
these conditions depend on th e d erivatives a t th e unknow n poin t xi = oq, so th ey
provide a basis for an ite ra tio n w hich determ ines u f a ) correct to
In practice th e integrals w ith respect to (j) are replaced b y sum s over four discrete
values of <j>. T he values ^ = 0, i\,r n a n d §
•°f uv, y^duJdXj) a n d v fd u fd xfj are th e n required a t th e intersections o f these four
bicharacteristics a n d th e curve dx{ = Ai d r w ith th e surface f{ x x, x 2, x 3) = 0. The
values of uv an d th eir derivatives in directions lying in th e surface are d eterm ined a t
these points b y in terpolating betw een a set of know n p o ints on th e surface. Since th e
surface is non-characteristic th e equations (2T) can th en be used to determ ine
ftiidujdxj) an d v ^ d u jd x^ a t these points. A lternatively, if it is sufficient to obtain
a solution correct to 0(h), th e term s in equations (3-9), (3-10) a n d (3*13), containing
hiidujdxi) and vi(duvldxi)can be ignored since th ey are 0 (h2).
238 D. S. Butler
4. A p p l ic a t io n o f t h e t e c h n iq u e t o t h e in t e g r a t io n o f t h e e q u a t io n s
OF PLANE UNSTEADY FLOW
The equations representing conservation o f mass, m om entum a n d entropy in
unsteady plane flow of an inviscid, non-conducting fluid are
dp dp dp 'du dv
0, (4-1)
£ + Ud i +Vd y +P dx dy_
'du du du dp
+ 0, (4-2)
ai + u t e + v ty.
'dv dv dv
Downloaded from https://royalsocietypublishing.org/ on 18 March 2024
+§£
>rs = W
o5 (4'3)
di + Wd i + VdyJ oy
dS Sa dS
0, (4'4)
~di + Ud~x+ V dy
where t is tim e, x, y are rectan g u lar C artesian co-ordinates, u a n d v are th e com
ponents of fluid velocity in th e x a n d y directions, respectively, p is density, p is
pressure and S is specific entropy.
E q u atio n (4-4) m erely sta te s th a t d $ = 0, along th e stream lines on which
dxfdt =u and dyfdt = v, an d m ay be used directly to determ ine 8 a t a n unk
point by tracing th e stream line th ro u g h th is p o in t back to th e sta rtin g surface. The
sound speed, a, in th e fluid is defined by
(dp/dp)s .
Along a stream line, d $ 0, therefore,
dp a 2 dp,
dp dp dp dp dp
(4-5;
dt JrU dx~'rV dt + USx + Vdy_
throughout th e fluid. The equation (4-5) is used to elim inate th e derivatives of in
equation (4’1). H ence
dp dp dp 'du
= 0. (4-6;
-dt+ U &c + Vd y +pa dx
The three equations (4-2), (4*3) an d (4-6) involve derivatives of only th ree variables
p, u and v. T hey m ay therefore be regarded as a system of equations for these thret
variables only.
The characteristic surface elem ents dx's(t,x, y
+ 2mv^ ^ + 2 v ^ ^ + 2 u ^ ^ } = 0 ,
ox dy cy dt dt d x )
and therefore touch a degenerate cubic surface which consists of th e streamlim
through th e point and th e quadric cone
( dx - u d£)2 + (dy - v£)2
d = a 2dt2.
(4*7
Hyperbolic systems of partial differential equations 239
F or sim plicity it is assum ed th a t a t some stage th e solution is know n in a region of
the plane t = t0-ahn d is req u ired a t th e p o in t (t0, x Q, yf). A system of co-ordinates
(r, 4>) is set up on th e ch aracteristic conoid v e rte x x 0, y 0), b y tak in g
t = T + t0- h , (4-8)
and in teg ratin g dx = (u+ acos
dy =
where initially, x = x 0, y = y 0, t = t0 a n d = S ubsequently 6 is defined b
condition (3-6), w hich reduces to
dx I dy
ta n 6 (4-11)
~ w df
Downloaded from https://royalsocietypublishing.org/ on 18 March 2024
and ijr
( 7 - 1 ) o'
The v ariab le ^ is a fun ctio n o f specific e n tro p y only.
I n term s o f th e s e ‘variables (4-12) becom es
fr dcr + cos 0 d u + sin ddv = —as dr. (4-13)
E q u ation (4*13) is applied in difference form along th e four bicharacteristics
<f>= 0, §77, 77 a n d §77, a n d conditions a t th e p o in t w here = 0 m eets th e plane
t = t0—h are d enoted b y th e suffix 1, those w here = §77 m eets th e plane b y th e
suffix 2, etc. Conditions a t (t0, x 0, y 0) are denoted b y cr0, u 0, v0, etc., if th e y are
correct to h(0)2a n d b y <f0, u 0, vQ, etc., if th e y need only be correct to 0(h). U sing this
notation,
pM o ~ o"i) + Ui K - uf) + Vx(vQ- vx) - ^ b a o ^ j ^ - ^ h a ^ + Oih3), (4-14)
Pi(&
o~<r*) + U*(uo- + J4 K “ Vi) - \ h a Qi ^ - \ h a ^ + °(h% (4-17)
where )»
^ = §(cos (j>i+ cos df),
and Vi = §(sin <j>i+ sin Of) for 1 ,2 ,3 ,4 .
240 D. S. Butler
Along th e stream line through (t0, x 0,y0)
da; =
d y = vdt,
and from equation (4*6),
du dv
dp = —pa2 dt,
dx dy
,i (du
or fd o - = - o ^ + ^ J d ( . (4-18)
^1 + ^2 + P3+
+ U2 + U3 + U^= 0 (h
Vi + V2+ V3+ V4 = 0(h),
Px-P
Ux—
Fi ~
P2 - P x = 0 (h),
U2- U i = 0 (h),
and V2—B — 2 +
Hyperbolic systems of partial differential equations 241
Therefore equations (4-20), (4-21) and (4*22) are readily solved approxim ately to
= - UR
1 - Rs) + 3),
8v 0 = —^(R 2 —Rf) +
vithout increasing th e order o f error.
Before eq uations (4-23), (4-24) a n d (4-25) can be used to determ ine cr0, u 0 a n d v0,
f and x i m u st be determ ined.
Qi s given to sufficient accuracy b y
Downloaded from https://royalsocietypublishing.org/ on 18 March 2024
0 i= <fi+hDi + (4-26)
vhere
5. A p p l ic a t io n o f t h e t e c h n iq u e t o t h e in t e g r a t io n o f t h e e q u a t io n s
OF STEADY SUPERSONIC FLOW IN THREE DIMENSIONS
Using dum m y suffix n o tatio n , w ith all suffixes tak in g th e values 1, 2 or 3 , th e
equations governing th e stead y m otion o f an inviscid, non-conducting fluid are, for
continuity, 0 s
“< 4 +^ ; = ° - <M)
242 D. S. Butler
for conservation of m om entum ,
(5*2)
0,
dx{ p
or U{ (dll I c‘Xj) = 0, (5-4)
dp
where q2 = ui ui and
JiS const. P
E qu atio n (5-4) gives d H — 0, along stream lines. Since th e
conserved thro u g h a stead y shock w ave, H will be c o n sta n t th ro u g h o u t m any
steady flows. H an d S can be com puted a t a n y unknow n p o in t b y tracin g back the
stream line thro u g h th a t point.
Defining th e sound speed a as in §4, eq u atio n (5*1) m ay be rew ritte n as
dp
ui + pa2 ^ = 0. (5-5)
dXi dxi
The four equations (5-5) a n d (5-2) are considered as a system of equations for
p, ux, u 2 an d u3. The b icharacteristic directions dxi of th is system satisfy
[(q2 —a2) —u{Uj] dxi d =0, (5-6)
and th e system is hyperbolic if q2 > a2. The bicharacteristic directions th ro u g h any
point can be expressed p aram etrically in th e form ,
dXj = (Uj + cocj cos 6 + c/^ sin 6 ) dr, (5-7)
where oq, and qjurae an o rthonorm al set of vectors, c is defined b y
direction norm al to th e shock an d 8/8/3 along th e shock. E q u a tio n s (4-6), (4-2) and
(4- 3) are applied im m ediately behind th e shock fro n t. I n term s of th e above notation,
th ey give 8p 8u . 8u 8v 8v
cos o) — z- s in w -^ + sm w -r + cos (o ^
8t doc 8/3 doc 8/3
. 8u
(6-3)
^ - « > t o - cosw t o + s m w p ’
8v ,_ x 8v . 8v 8p
p M =piQ- <I)s * - smidr x - COSMw
These equations express th e derivatives of th e variables p, u a n d v along th e shock
rays in term s of th eir derivatives in planes t = co n stan t, while th e bicharacteristic
relations (4T2) express derivatives along b icharacteristic rays in term s o f th e spatial
derivatives of u and v only. The spatial derivatives of m u st therefore be elim inated
from equations (6-3). 8p/8/3 is th e d erivative o f p along th e shock a n d since eq u a
tions (6T) hold everyw here on th e shock, th e y can be used to express 8p/8/3 in term s
of 8u/8/3 a n d 8v/8/3. The expression obtain ed is
8p ^
. (c o S W _8u + 8 .m 8v)
w __j> (6 -4 )
¥
where __^PsQ
(1 + 1 / i f 2)-
8p / 8cc is elim inated betw een th e th ree relations (6*3) to give tw o relations:
^ + p ( Q - q ) G o a o ^ + p ( Q - q ) s in w ^
8u . 8u 8v itd i 8u .
—pa* [coso)
t e - sin" p + s m " s +cosw ^ ) +p(° - 9) icos" t o + s m w teJ-
(6-5)
, 8uv
and - p s m w 1- + p c o s w 1-
of of
8p
p W -^ -s in ^ +c o s^ } -
¥
P(Q ~ q)j - sin oj^ + cos (o|^ J - A j c o s w ^ + s8mu •w ^ j . n\
(6-6)
Hyperbolic systems of partial differential equations 245
These relations (6*5) a n d (6-6) are sim ilar in form to th e b icharacteristic conditions.
A. fu rth e r relation is required along th e shock rays. T his comes from th e geom etrical
condition,
8 (0 8u .
" * - b [, (6-7)
~8t ¥
where th e shock relations (6-1) h ave been used to express in term s of S u /S /i and
and
B iT
w
fj=(1 + w -
Using th e first tw o shock relations (6-1) an d th e v ariable <r, defined in §4, we can
rewrite equations (6-5), (6-6) a n d (6-7) in th e form
Downloaded from https://royalsocietypublishing.org/ on 18 March 2024
^■dcr + e o sy d g
= -a(l
j - c o s 2ycos2w ) |^ - c o s 2ysinwcosGj + (1 - - cos2y sin2w) ™J df,
( 6- 8 )
where c o sy = (Q —q)ja,
Idu
qdoj = _ ( Q - q ) 2 sin (ocos o( ^ —(cos2 (o —sin2
2(1 + 1/ I f 2) \ f y + dx,
'dv
—2 sin o) cos w— ! + (6*9)
dx
and
(y+i) Idu 0v\
dw 2 sin o( os
c W r - - (cos2 co —sin2 co)
4(1 + 1/ M 2) \dy + dx)
The rig h t-h a n d sides o f equations (6-9) a n d (6*10) contain term s in th e v o rticity
{dv/dx— du/dy). The rig h t-h a n d side of th e corresponding form of th e bicharacteristic
relation (4-13) does n o t co ntain such a term . This term is elim inated betw een (6*9)
and (6-10) to give
_ ( y + l ) ( 3 + .M2) (Q (i ) ?
Idu 0v\
a co sy 2 sin (o cos l ^ + (cos2
o —sin2(o) + 2 sin a) cos (o dt.
\dy + dx) dy.
( 6 - 11)
‘ray ’ of propagation
of th e shock
surface t = t0—h
F igure 1. This shows th e segm ent of th e shock ra y and th e bicharacteristic segm ents used to
com pute th e solution a t a point on th e shock, (a) is a plane t = co n stan t on which the
solution is known, and (6) is th e life surface of th e shock.
w o) = 6 ,
M
,oQ
(f o) = 6
and to find Q0 and oj0 as a root of these equations. I f th e shock is w eaker still, so thal
(1 —1 1311)— 0 (h3) th e m ethod will no longer be applicable.
Hyperbolic systems of partial differential equations 247
7. C o n ta c t s u r f a c e s in p l a n e u n s t e a d y fl o w
contact surface
igtjre 2. This shows th e stream line a n d bicharacteristic segm ents used to com pute the
solution a t a p oint Oo
n a co n tact discontinuity, (a) is a plane t = constant on whic
solution is know n, (6) is th e life surface of th e discontinuity, 01, 02 and 03 are bicharac
teristic segm ents on one side of th e discontinuity an d 04, 05 and 06 are bicharacteristic
segments on th e other, 07 an d 08 are segm ents of th e stream lines through 0 , one on
either side of th e discontinuity.
Difference relations are used along three of the bicharacteristic elements through
l0, x0,y0) on each side of the contact surface. On the prim ed side the bicharacteristics
'»= co0, ojq+ \ tt and oj0—\ tt are used, and on the unprim ed side (f + co0 + \ tt
nd (o0 + \ i tare used.
In addition equation (4-18) is applied along two streamlines through (t0, x 0,y 0),
ne on either side of the surface.
Weighted sums of these eight difference relations are taken, represented sym
bolically by {a>0+ § 77} - {coQ+ r},
{(*(>+f r y - { v o - W * (7.2)
^{a>0 + f rr}+ %{coQ—^7r} -f {a>0 + 7r) —{4* 18},
+ f r Y + i{wo —fr } ' + fro}' ~ 18}', _
-hese relations do not contain ( dujdx )0, (dujdy + dv/dx)0 or (dv/dy)0.
248 D. S. Butler
The four difference relations represented by (7-2) an d th e tw o bou n d ary condi
tions (7-1) are used to determ ine p, u, v, p \ u' a n d b y an ite ratio n sim ilar to th at
described in § 4. D uring each cycle one of th e variables y 0) m u st be ad ju sted so
th a t th e new p oint lies on th e contact surface a n d q a n d are found by using the
conservation of entropy along stream lines on eith er side of th e surface.
Difference relations are used along th ree b icharacteristic elem ents th ro u g h (t0, x 0, y0),
corresponding to <p= o»0 -I- ,noj0 + \ n a n d oj0 + f7r. E q
difference form along th e stream line th ro u g h ( , x 0, y 0).
These difference relations are com bined as follows,
H + ifl-J -H + f tt},1
The m ethods described in th e preceding sections h ave been used to com pute th
flow over a p a rticu la r body. The p lan form of th is body is an isosceles triangle. Fo
the first 20 % of its length th e body is a rig h t circular cone. T he rem ainder o f th e bod;
has elliptic sections w hich becom e m ore eccentric as th e sharp trailin g edge i
approached.
The semi m ajor a n d m inor axes of th e sections are given b y
z
(0 < z ^ L),
« - l )
(0 < 2 ^ 0*2
V M -f)
z T 0-2LU
and (0-2 L < z <
V W > - i)L l i-
W here L is th e length of th e body, M 0is th e free-stream M ach n u m b er of th e flow and
is distance m easured in th e free-stream direction.
I n order to sim plify th e problem it is assum ed th a t M 0 is large so th a t hyperson
slender body th eo ry can be em ployed.
I he equations governing th e flow are th en analogous to th e equations o f unstead
plane flow (see, for exam ple, V an D yke 1954), a n d th e body is analogous to a n ellipt
Hyperbolic systems of partial differential equations 249
>iston w hich d istu rb s th e fluid during th e tim e in te rv al 0 ^ ^ 1. The semi m ajor
md m inor axes o f th is p isto n are
a= t (0 ^ < 1),
b= t (0 ^ f < 0-2),
r (£ —0-214
b=t (0-2 < f < 1).
l 0-8* } .
Figure 3. This illustrates th e delta-shaped body chosen for th e num erical example. The
curves draw n on its surface are its intersections w ith equally spaced planes z = constant
and x = constant. T he body is draw n for M 0 = ^/5.
S tarting values were fed in on th e plane t = 0-2. These were obtained from th e
Taylor-M accofl solution for flow a b o u t a n unyaw ed cone (Maccoll & T aylor 1933).
The solution was th e n com puted a t tw e n ty -th re e oth er tim e statio n s using th e high
speed, auto m atic, digital com puter AMOS. D etails of th e problem s encountered in
programming AMOS for th is work, a n d th e solutions adopted, are to be published
separately.
Some o f th e results are p resen ted in figures 4, 5, 6 an d 7. These figures show isobars
in q uadrants o f th e planes t = 0*2, 0-4, 0*6 a n d 0*8, respectively. These curves also
represent isobars in th e corresponding span-w ise sections o f th e analogous steady
How problem .
The isobars show n in figure 4 are concentric circles, corresponding to th e sym
metric T aylor-M accoll solution. The pressure grad ien t is tow ards th e body surface
indicating th a t th e shock is being followed by a com pressive wave. Before th e n e x t
16 Vol. 255. A.
250 D. S. Butler
stage has been reached (figure 5) m ost o f th is com pression w ave has been replaced by
a rarefaction which has reversed th e direction of th e pressure g radient n ear th e minor
axis of th e section. This rarefaction arises as a result of th e reduction of th e ra te oi
thickening of th e body near th e m inor axis. The m ajor axis o f th e body continues to
increase uniform ly and only sm all reductions in pressure h ave occurred here, due to
waves which have spread round th e body.
shock front
Downloaded from https://royalsocietypublishing.org/ on 18 March 2024
piston or body
surface
B y th e n e x t stage (figure 6) th e pressure grad ien t has been reversed over m ost <
th e region betw een th e shock and th e body a n d th e pressure n ear th e m inor axis hi
been reduced considerably; it is in fac t lower th a n th e free-stream pressure. A
interesting feature is th a t th e pressure m inim um is no longer on th e m inor axis. Th
is presum ably due to an over-expansion effect. F u rth e r dow n th e body (figure '
th is feature has become m ore pronounced, a n d th e isobars have becom e buncht
near th e edges of th e body. This indicates large pressure g radients n ear th e tip s ar
fairly uniform pressure over th e rem ainder of th e body.
Hyperbolic systems of partial differential equations 251
F or th e p o rtio n o f th e flow com puted very little change in shock-w ave shape
ecurs.
A m odel o f th e bod y chosen for th is exam ple has been te ste d in a w ind tu n n el
inning a t a M ach n u m b er of 3-5. P ressure m easurem ents w ere ta k e n from several
ressure holes drilled along tw o p a rtic u la r span-w ise sections o f th e body. These
Downloaded from https://royalsocietypublishing.org/ on 18 March 2024
a (deg)
F ig u r e 8 F ig ure 9
R e fe r e n c e s
B ruhn, G. & H aack, W . 1958 Z. angew. M ath. Phys. 9,
Coburn, N. & Dolph, C. L. 1947 Proc. lsZ Sym p. A p p l. M ath. Am er. M ath. Soc. p. 51
V an Dyke, M. D. 1954 N .A .G .A . Tech. Note 3173.
H artree, D. It. 1953 United States Atom ic Energy Commission Report N o. 2713.
H olt, M. 1956 J . F luid Mech. 1, 409.
Maccoll, J . W. & Taylor, G. I. 1933 Proc. Roy. Soc. A, 139, 278.
Sauer, R. 1950 Z . angew. M ath. Mech. 30, 247.
Thornhill, C. K . 1948 A .R .C . Rep. Memor. 2615.