Oomp*ttera Math. Applic. Vol. 24, N o . 11, pp. 45-59, 1992 0097-4943/92 $5.00 + 0.

00
Pr i nt ed i n Gr eat Br i t ai n. Al l ri ght s r eser ved Co p y r i g h t ~ 1992 Per gamon Press Lt d
RE AL GAS F L OWS I N A DUCT
P. GLAISTER
D e p a r t m e n t o f M a t h c ~ n a t i c a , U n i v e r s i t y o f Reacling~
P. O. Box 220, Whi t ekni ght s, Readi nz, RG6 2AX, U. K.
(Receieed Noeember 1991)
Ab s t r a c t - - A numer i cal scheme is present ed for t he sol ut i on of t he Eul er eqt mt i ons of ccanpressible
flow of a r eal gM i n a single spat i al coordi nat e. Thi s i ncl udes flow i n a duct of vari abl e croes-section
as well as flow wi t h cyl i ndri cal or spheri cal symmet ry, and can pr ove useful when t est i ng codes for
t he t wo-di mensi onal equat i ons governlnff compressi bl e flow of a real gas. The scheme is appl i ed wi t h
success t o a pr obl em i nvol vi ng t he i nt er act i on of convergi ng and di vergi ng cyl i ndri cal shocks for f our
equat i ons of st at e and t o a pr obl em i nvol vi ng t he refl ect i on of a conver~i' n~ shock.
1. INTRODUCTION
In 1988, Glaister [1] proposed an approximate Pdemann solver for the one-dimensional Euler
equations for compressible flow of a real gas, i.e., with slab symmetry. The resulting scheme
was applied with success t o a t est problem involving t he reflection of a shock, and a significant
improvement in t he efficiency of this scheme has recently been made [2]. For applications where
t he flow takes place in a duct of variable cross-section, it is again desirable t o have such a Pdemann
solver for such flows. In particular, it could be used t o compare with results obt ai ned from a
two-dimensional code for compressible flows when presented with a cylindrical or spherically
symmet ri c problem. In this paper we propose such a numerical scheme, and t est t he scheme on
a probl em involving t he interaction of converging and diverging cylindrical shocks.
In Section 2 we review t he differential equations governing the flow under consideration and in
Section 3 we discuss how these equations can be solved using flux difference splitting. In Section 4
we give the details of the resulting numerical scheme, while in Section 5 we look at the properties
of t he scheme. Finally, in Section 6, we describe two t est problems, and in Section 7 we present
some results for the problems of Section 6 using the scheme of Section 4.
2. EQUATI ONS OF F LOW
The ' one-dimensional' equations governing compressible flow of a real gas can be wri t t en as
1
p , + ~ ( r ) ( S ( r ) p u ) , = O , ( 2 . 1 a )
( p. ) , + = - p., ( 2 . 1 b )
1
e, + -f(~CSCr)uCe + p ) ) , = O , ( 2 . 1 c )
where the equations of st at e relating pressure, p, t o the density, p, and the specific internal energy,
i, is given by
p = p ( p , i ) , (2.1d)
and t he t ot al energy, e, is given by
1 2
e = pi J r ~pu , ( 2 . 1 e )
Types et by .A.A~-TFX
45
4 6 P . GLAI S TER
where u is t he velocity. Equations (2.1a)-(2.1c) represent a syst em of hyperbolic equations for
(p, p u , e ) - ( p, p u , e) ( r , t ) , where p u i s t he moment um, at a general position r along t he duct,
(of cross-sectional area S(r)), and at time t. The particular example S( r ) - r N , represent8 slab
symmet ry, ( N = 0), cylindrical symmet ry, ( N = 1), or spherical symmet ry, ( N -- 2). (N.B. The
coordinate r is given by r = z, y / ~ + y2 and ~ z 2 + y2 + z 2, in t he case of central symmet r y
with N = 0, 1 and 2, respectively, where z, y, z represent Cartesian coordinates.) An ideal gas is
represented by equation (2.1d) with p = (7 - l)pi, where 7 is t he ratio of specific heat capacities
of t he fluid.
In t he next section we find approximate solutions of equations (2.1a)-(2.1e); but first of all we
rewrite t hem in st andard ' conservation form' as
where
_Wt + ( F ( _ W) ) , = g( _W) ,
_ w= ,~ , _ F (W_ )- p+pu 2 ,
L*
t oget her with (2.1d), (2.1e), and
- - ~ p u
I r
g ( _ w )
S ~ r U( e
+,)
is a source t er m arising from t he geometry.
( 2. 2a)
(2.2b)
( 2 . 2 c )
3. FLUX DI F F ERENCE SPLI TTI NG
In t he case of slab symmet ry, for which S = 1, equations (2.2a)-(2.2c) reduce t o t he one-
dimensional Euler equations for a real gas in a single Cartesian coordinate, which can be solved
by flux difference splitting using t he approximate linearised Reimann solver developed by Glais-
ter [1]. We shall use t he similarity of equations (2.2a)-(2.2c) to t he Cartesian case t o develop a
corresponding met hod for duct flows keeping as far as possible t he valuable properties previously
found.
We consider a fixed grid in space and time with grid sizes At , At , respectively, and label t he
points so t hat rj = r j _l + At , t , = t n- z + At and ~ denotes t he approximation t o _W(r/,tn).
Consider also t he interval [rj -z, r j ] and denote by W/;, _WR t he approximations t o W at r j - z, r j ,
respectively. We now rewrite equations (2.23) as
OF
w_, + ~- _ (w_ ) w_ , = g_ (w_ ) ( 3 . 1 )
and solve approximately t he associated Reimann problem
w_, + 2(w_L , w_.)w_. = _ g ( _ w ) , ( 3 . 2 )
with dat a _WL, _WR either side of t he point rj _ , linearising by taking A(_Wz;,_WR) to be s constant
matrix. We shall use t he approximate form of equation (3.2),
- - - ~ Wj - 1
W~ +z W~ + .4(_WL, _WR) - ~ = ~ ( ~ " ) , (3.3)
A t - 7 x ~ -
whe r e A(_WL,_WR) is t he mat ri x (3.4), _~ is an approximation to _g and P may be L or R. The
mat ri x A(_WL,_WR) is an approximation to t he Jacobian mat ri x A - ~r(IZ) and is constructed
so t hat _FR- _FL -- A(_WL,_WR)(~/R- W) for any finite change of st at e and is given by
[ o 1 0 1
Real gas flows
T h e aver ages o f t h e fl ow var i abl es a r e gi ven by
v ~ ; + v ~ 7 '
wh e r e H = ~ is t h e t o t a l e nt ha l py, t o g e t h e r wi t h
P
A = u , i , H ,
47
( 3. 5a ) - ( 3. 5c )
( 3 . 6 )
(3. 7)
T h e a ve r a ge s o u n d s pe e d a is gi ven by
PPi (3. 8)
a 2 = / 3 ~ + ~ ,
wh e r e t h e aver ages / 3p, / 3i o f t h e der i vat i ves o f t h e e q u a t i o n o f s t a t e ( Pp, Pi ) are gi ven by
Pi- = Pi ( ~, 3) J i f --z--p _< 10 - m a n d --=--t - < 1 0 - m' ( 3. 9a)
I ~ v l 6p }
/3, = po(/~, 7) + [ A p [ + [ A i [ Ap o t h e r w i s e , (3.9b)
[ A i l 6p
/3~ = P ' ( P ' I ) + IAvl + J A i l A i
6 v = ~ p - v p ( ~ , , ) a p - v , ( ~ , ~ ) A i ,
w i t h
a n d wh e r e m is ma c h i n e d e p e n d e n t ( see [1,3]).
T h e ei genval ues o f A a r e
~1,2 "-- fi -I- a ,
wi t h c o r r e s p o n d i n g e i ge nve c t or s
(3. 9c)
As ---- fi, ( 3 . 1 0 )
~1,2 = _ ~ a , ~ = "
L ~ ~a ~- ~
(as i n t he st andar d Cart esi an case when S = 1).
(3. 11)
CN4J~ 24t11..0
Us i ng t h e above p r o p e r t i e s o f A we c a n wr i t e e q u a t i o n (3. 3) i n s uc h a wa y t h a t t h e l e f t - h a n d
s i de ha s i t s n a t u r a l c ons e r va t i on f or m, i. e. ,
W~p "!'1 - W / ~ + F j - F j _ 1 __~ ~ ( ~ v n ) , ( 3. 12)
A t A r -
wh e r e 0(_W") is a s ui t a bl e a p p r o x i ma t i o n t o t h e t e r m _~(W) on t h e r i ght h a n d s i de o f equa-
t i on ( 2. 2a) . We t h u s o b t a i n
__W~p + 1 - ~ = A t _ 0 ( W " ) - ~ r ( _ ~ A t _ _ _ ~ - 1 ) . ( 3 . 1 3 )
Bef or e we de s c r i be t h e me c h a n i s m us e d t o u p d a t e I~/" t o W~j +1 we l ook a t t h e a p p r o x i ma t i o n
0(LL "n) u s e d f or _g(W).
Gi ve n t h e f o r m f or _g(W) in (2. 2c), a n d n o t i n g t h a t e + p = p H , one s ui t a bl e a p p r o x i ma t i o n is
-0(L~Vn) = - P ' S - ~ r ' (3. 14)
us i ng t h e a ve r a ge s i n ( 3. 5a) , ( 3. 5c) a n d (3. 6), whe r e ~ r = r j - r j - i , A S = S ( r j ) - S ( r j - 1 ) , a n d
is a s ui t a bl e a ve r a ge f or S ( r ) over t h e cell [ r j - x , r j ] , e. g. , ,~ = ~ / S ( r j - 1 ) S ( r j ) . Th e r e ar e, o f
cour s e, o t h e r choi ces t h a t c oul d be ma d e .
4 8 P . GLAi STER
4. UP WI NDI NG
In t he homogeneous (g - 0 3 case the procedure now is t o proj ect A_F = _Fj - _Fj-1 onto the
eigenvectors of 2i. Each projection represents the contribution of one wave syst em t o A F. Here
we follow a procedure given in [3], and find t he projections bot h of A_F = _Fj - _~-1 and also
~(_W~). For linear conservation laws this procedure leads t o a correctly ' upwinded' t reat ment of
source terms. We t hen updat e W~ t o W~ +1. Suppose
3
A W = Wj - _Wj-1 = E &' ~ ' (4.1)
i-----1
so t hat
3
i = l
Since .4 has eigenvalues Xi with corresponding eigenvectors ~ , and
( 4 . 2 )
3
.~(_W.,~) = 1 E A ~ , ( 4 . 3 )
- Ar
i----1
for suitable/3i, we may write equations (3.13) as
3
At E ~i ~i ~, (4.4) _ ~ p + l = w~ -
i = 1
where
5 , = a , + _-3~ ( 4 . 5 )
Ai
and P may be L or R. To updat e W" t o V " + i , we use first order upwind differencing, i.e., for
each cell [ r j - i , rj], we
At - W~
add - ~ r Ai S i ~ t o _ j , when Ai > 0
o r
add - ~r Ai 7i ~ t o Vv'f_l, when Ai < 0.
This gives exact shock recognition for the Kiemann problem given by equations (3.2) within
t he resolution of t he grid, provided t hat we use the previously defined local averages given by
equations (3.5)-(3.9).
If we follow the algebra through, we obt ai n
1 Ap
a2
( 4 . 6 a ) - ( 4 . 6 c )
and
1 _ _ A S /~3 = 0 . ( 4 . 7 a ) - ( 4 . 7 c )
~ . ~ = ~ p ~ ~ '
(Not e t hat the quant i t y ~ is independent of time and therefore has t o be calculated only once.
If the area changes are due t o central symmetry, so t hat S( r ) = be ~, ~ = 1,2 for cylindrical or
spherical flow, then
r ~- - r j - t i f O" "-- 1,
AS (rj~j_~)~l= '
~ - - ~ r ~_r 2
S : i-~ if ~ = 2.
r j r j - x '
Pe a / s a s flows 49
Since these expressions are singular when a grid point is placed at t he origin, we use grids t hat
st raddl e t he origin.)
The expressions in equations (4.6) and (4.7) have been wri t t en in t erms of primitive variables
for simplicity. The expressions for &,, &=, &3 follow those given in [1] and signify wavestrengths
which would be generat ed by t he dat a 1 ~ - 1 , _ ~ ] in a parallel duct. The expressions for/ ~, ,
f12, fls represent modifications to those wavestrengths due t o area changes and only occur in t he
acoustic waves.
To solve equations (2.2a)-(2.2c) using t he finite difference approximation given by equa-
tions (4.4) and (4.5), we use t he met hod of upwind differencing on t he three waves with wavespeeds
A:, A2, A 3 and wavestrengths %, %, % (which will differ from t he usual wavestrengths in slab
symmet r y due to t he variation of S( r ) ) . This gives t he first order approximation mentioned ear-
lier. As in [1], we can modify this scheme to be entropy satisfying and define transfers t hat can
be limited using flux limiters to obtain an oscillation free scheme. Moreover, t he use of particular
flux limiters [4] will sharpen up certain features t hat would be smeared by using t he first order
met hod alone.
5. PROPERTI ES OF THE SCHEME
In this section we discuss briefly some properties of t he scheme proposed here for solving
equations (2.2a)-(2.2c) in relation t o t he features t hat we expect to occur in this t ype of problem.
( i ) If we consider t he special case of constant dat a p 7 = pn , i 7 - i n (so p7 - pn) and
n = u n = 0 ( n o f l o w ) , for all j at time t = n At , equations (4.4)-(4.5) reduce to uj
- w ? + - - o, (5.1)
At
yielding t he correct physical solution, namely u n+k = O, pn+k = pn, (pn+k = pn).
(ii) The scheme ' recognizes' st eady states. If t he dat a satisfies the compact residual equation
- - - t j _ 1 = A r # ( W ' ) j _ I / 2 , f o r .~]l j , ( 5 . 2 )
t hen ~ = ~2 = ~s = 0 for all pairs of cells and no updating takes place. Equation (5.2)
can be regarded as a two-point discretisation, second-order accurate at r j - t / 2 of t he or-
di nary differential equations governing steady, compressible flow. This propert y appears
especially valuable when shock reflection should leave behind a uniform unchanging flow.
(iii) The scheme also ' recognizes' shock-waves. At a shock A_F = SAW for some scalar shock
speed S and by equations (4.1) and (4.2), S is an eigenvalue of A. The projection of A W
onto t he local eigenvectors of A will be solely onto t he eigenvector which corresponds to S.
In this special case, the solution of t he linearised Riemann problem given by equations (3.3)
is exact.
The reason t hat t he formulation (3.3) recognizes shocks is t hat the right-hand side t er m g( W)
does not contribute t o t he shock wave, essentially because it does not contain any derivatives in
W, and therefore does not j ump when integrated with respect to r. To demonst rat e this point,
suppose t hat t he variable q j umps from qL to qR at r = ro > 0, t hen
lira / dr = l i m qL _~ S( r ) dr + qR S( r ) dr =0 ,
, - O J , o _ , S(r) , - . o , , , o
since S ' ( r ) / S ( r ) is continuous. Thus,
fro+~
lim / _g(_W) dr = O, (5.3)
(by choosing q = pu, q = pu 2 and q = u(e + p), in turn), i.e., t he shock speed is given by [ ~ ,
with t he right-hand side making no contribution.
50 P. QLAISTER
Moreover, in t er ms of t he t hree scal ar probl ems obt ai ned by dia4Fmalising t he s ys t em given by
equat i ons (2. 2a)-(2. 2c) wi t h .4 = ~ _ a const ant mat r i x, we have
OVi
~i OVi h~(W), i = 1, 2, 3, (5.4)
a t + 0 , =
where
Y -- X-I~w~ b = X - I g , (5.5)
and X is t he modal mat r i x consi st i ng of t he eigenvectors of A wi t h eigenvalues hi. Sol ut i ons t o
equat i ons (5.4) and (5.5) can be represent ed in t er ms of a ' shock sol ut i on, ' t hi s bei ng a sol ut i on of
t he homogeneous equat i on, i.e., equat i ons (5.4) wi t h _h(W) - O, and a ' source sol ut i on, ' bei ng t he
par t i cul ar sol ut i on of t he i nhomogeneous equat i ons (5.4) and (5.5). In effect t he scheme we have
devel oped solves equat i ons (5. 4)-(5. 5) approxi mat el y, and t hus t he i mpor t ant ' shock sol ut i on' is
model l ed as a consequence of t he const ruct i on of A.
6. T E S T P ROBL E MS
In t hi s section, we look at two t est probl ems used t o val i dat e t he previ ousl y descri bed al gor i t hm
of Section 4 for solving equat i ons (2. 2a)-(2. 2c).
The first t est probl em is concerned wi t h a converging cyl i ndri cal shock and we consider a region
0 < r < 200 for t he cyl i ndri cal l y symmet r i c case given by equat i ons (2. 2a)-(2. 2c) wi t h S ( r ) - r .
Initially, a cyl i ndri cal di aphr agm of radi us r : 100 separat es two uni f or m regions of gas at rest.
The i ni t i al condi t i ons are p -- 4, p -- 4 in t he out er region, and p -- 1, p - 1 in t he i nner region.
When t he di aphr agm is removed at t : 0, a converging shock wave followed by a converging
cont act di scont i nui t y moves t owards t he axis, r "- 0, and a diverging rarefact i on wave moves
out wards. The shock accelerates as i t approaches t he axis of s ymmet r y and is reflected f r om t he
axis, i nt er act s wi t h t he cont act di scont i nui t y (still converging), which results in a t r ans mi t t ed
shock, a converging cont act di scont i nui t y and a weak converging reflected shock.
The second t est probl em is concerned wi t h a converging and a diverging cyl i ndri cal shock in
t he same region. Initially, two cyl i ndri cal di aphr agms of radi i r -- 50 and r - 150 separ at e t hree
uni f or m regions of gas at rest . The i ni t i al condi t i ons are p - 4, p : 4 in t he inner and out er
regions, and p -- 1, p - 1 in t he mi ddl e region. When t he di aphr agms are removed at t -- 0, a
converging shock wave, followed by a converging cont act di scont i nui t y moves t owards t he axis,
r - 0, and a diverging shock wave, followed by a diverging cont act di scont i nui t y moves away
f r om t he axis. The shocks subsequent l y i nt eract , resul t i ng in a diverging shock wave weakeni ng
in st r engt h, t oget her wi t h a converging shock wave increasing in st r engt h. Each of t hese shocks
t hen i nt er act wi t h t he correspondi ng cont act di scont i nui t y, as in Pr obl em 1, and t hi s results, for
each i nt eract i on, in a t r ans mi t t ed shock, a weak reflected shock and a cont act di scont i nui t y.
Each pr obl em is t est ed for four equat i ons of st at e:
(a) ideal gas
p = ( 3 ` - 1 ) p i ,
where we t ake t he r at i o of specific heat capacities 3` t o be 1.4 correspondi ng t o a di at omi c
gas;
(b) stiffened gas (somet i mes used for a liquid)
p = B ( ~ - - 1 ) + (3` - l ) , i ,
where we t ake B = 1, Po = 1, 3' = 1.4;
(c) covolume (used in connect i on wi t h combust i on envi ronment s)
( 3 ` - 1 ) p i
P - - 1 - p b '
where we t ake b -- 0.2, 3` = 1.4; and
(d) t wo mol ecul ar vi brat i ng gas
( 3 ` - 1 ) p p = ( 3 ` - 1 ) p i ,
p "t" e x p ( p p / p ) - 1
where we t ake 7 = 1.4 and ~s -- 0.5902 which corresponds mol ecul ar oxygen at 2000K.
Real gas flows 51
P
3.99
2 . ~
m
-1.33,
L ~ T J .
I I I I I
4 0 8 0 1 2 0 I J O I r
I I I I I
~ 120 1 5 9 I
- 1 . 3 2
- 2. ~
U
o , w
0 . ( ~ ,
0 . $ 9 ,
. ! I I eJk I
J
-0.ll8 l
-0o~
-0.99 ,
At t i me t = 50.0
U
o . ~ 2
o . a 8
0.14
r
-o,14
-0.a8
-0.42.
/ -
I
i n ! ~ ltlO r"
P
S . 2 5 , ~,
-I ,?5
4, $9
At t i me t ---- 80.0
I I I I I
80 1 2 0 1 6 0 I r
U
0. I 6
0. ~
0 . 1 2
- 0 . 1 2
- 0. ~t
- 0. N
/ :
120 1 4 0 ~ r
/
At t i me t = 110.0
Fi gur e 1. Sol ut i on of Pr obl em 1 wi t h equat i on (a).
52
P
: t ~ o
- t . mt .
- . 2 . ( 5 6 ,
,___/
I I I I
4 0 8 0 I I O 140
P . GLAISTER
U
0.01
0. St
0. 27.
I
3OO r"
- 0, 2?.
-0.01
. I I I / ;
4 a O0 1 8 / I
/
I
3OO r
P
3 . 9 9
2 , 6 6 j .
-I , 3 3
~ t ~ ,
P
4.1L~.
I
40
- I . ~
At t i me t = 40. 0
I I I I I
W 120 I N I P
U
0 . ~
0 . 3 4
0 . 1 2
I I I I
80 I m 140 300 r "
- 0, 12
- 0 . ; ~
- 0. 36
At t i me t = 7010
U
0 . 3
0. 2
o. !
..0.1
- 0. 2
-0. 5
/
/ .
/
!
I I
m 13o
\
I ~ ... r
,a
!
i
J
I I I
1 8 I J O m r
I
At t i me t = 1 0 0 . 0
F i g u r e 2. So l u t i o n of P r o b l e m 1 wi t h e q u a t i o n ( b) .
I : )
lLd4.
2 , ~81..
I . I I
- 1 . 2 2
I I I I I
4O 8O 11o ItO
Re a l g a s f l ows
U
0~$.
0 . ~
r
-0, ~o.
- 0 . 6 6
-0.66
5 3
P
II.TII
2 I I ~ .
1 . 2 6
" 3 , 7 1 1 ,
41~.05
2 . 7 0
| ' I I I ,
- I o 1 5
-4.11l
I I ' 1 : :
I 0 I i O llll 2 o o
. _ _ . . , - - - , "
I I I I I
4 O 8 O l ~ O 1 4 0 ~ m O
At t i me t ~ 50. 0
U
0.S
0.1
r
- 0 . !
-0. 2
i - 0. 3.
At t i me = 80. 0
U
0 . 2 1
0,14,
O.OT
r
-0.07
- 0 . 1 4
-.0,11
" ; I "
. I I f i ; ~ "
r
At t i me t - - 100. 0
F i g u r e 3. So l u t i o n of P r o b l e m 1 wi t h e q u a t i o n ( c) .
54 P. GLAISTER
P
3.W
2,~16
-lA B .
-,%66
. - - - . . ~ . . ~
I I ! I I
w W I m 1 4 0 1 0 0
P
3 , 1 4 .
I . ~ 7
- I . ~ 7 ,
4 . 1 4 .
-4.71
I I I I I
4 0 8 0 I B 1 4 0 ~ r "
U
0 . 1 M ,
O. t
. I t , ' , I i J I
, . ~ m MO r"
i i '
At t i me ~ = 50.0
U
0. ~,
0.110,
0 . 1 5 .
-0.IS
- 0 . W
t ?
/ .
/
. I
4 O
! ! t r - I
10 Im y l i l 0 I~
J
P
' - i
1 . 8 4 ,
- I . 1 ~ ,
4.40.
I I I I
4 0 8 0 1110 1 4 0
At t i me t = 80.0
U
0.119
0.~t
0 , 1 3
I
P
4.13
- 0 , W
_f
~ - |
4O gO
I I I
I M 140 ~
/
t
t
r
At t i me t = I I 0 . 0
Fi gur e 4. Sol ut i on of Pr obl ~n 1 wi t h equat i on (d).
P
3. 99,
1. n
.-2~6
- 3 . 9 9
P
3. 99.
2 t ~ '
: I I I
40 IO 120 I m
Real gas flows
U
o . P J .
0. !0.
0 . 2 5 ,
!
2OO r
. . o . 2 5
. o . $ D
.o.75 ~
: ; -
4o go
.* I :
- I . 1
. ~, 99 ,
"At t i me
/
i I I t I
40 8O 120 140 20O r
--- 30.0
U
0. 61
--o54
i
t
4O W
, ,
l aO 14o r
i
#
5 5
P
2,aQ
1.21
- I . 2 1
At t i me t = 40.0
,
L
. _f
I ! ! !
t r
U
0 . 6 4 .
0 . ~ 6
o.m
. . om
- O * S ~ ,
- 0 , 8 6
\
I I I I :
o ~ I m t r
i
At t i me t --- 60.0
Fi gur e 5. Sol ut i on of Pr obl em 2 wi t h equat i on (a).
56
P
2;44
I . SS
I
40
- 1 . 8 .
/
i , w
I I I
I 120 140
- 0 4 4
P . GLAISTER
U
0 . 6 ,
0 , 4 ,
0 , 2
B
~ 0 r '
-0. 2
-0. 4
-0. 6
At t i me t = 20. 0
P
3 . 0 6 .
| , I ,
- I . I
4. 06
P
3 . 4 ,
1 = 2
- I . 2
-2.4
- . %6
|
/
I I I I !
40 I 120 I I 200
q i
/
I I I I I
40 I I I l t O 200
U
o.Ts
r '
-o.15
-O.W,
- 0 . ~ ,
At t i me t = 40. 0
U
0 . 5 1
0 . 0 6 .
0.17.
r"
- 0 . 1 7 ,
=0.06.
4. 51
I
J
-" : : -
4o m 120 l JO ; ~0
r"
&
o . . . / \
0. 25 . ; , :
I I I I !
I 120 IdO ~ r
/
/
: o
. . ' ~.
ol
i "
I I I I I
40 I I I lliO ~ i "
At t i me t --- 60. 0
F i g u r e 6. S o l u t i o n of P r o b l e m 2 wi t h e q u a t i o n ( b) .
Re a ST
- p
S . 4 2 .
1 . 1 4
- I , 1 4
. ~ . 2 1 i ,
..3.42
I I I I I
t o 80 I W 140 aO0 r
g M f l o w 8
U
O. W,
o . 1 6
o . 1 3
- 0. 13,
- 0. ; ~
- 0 . m
I I t
t o go
I I I
I W 140 " ago r
:
i j
1 . 1 5
At t i me t = 30. 0
U
_ ! !
4o g o
- I . I S
. 2 . 1 0
4 . 4 S
I I
laO 14o
0 1 1 6
0. a4.
0 . 1 2 ,
I
II00 r
- 0. 12
-0.;16.
- 0. 16.
I
4O
I I I I
P
3.43, rt
3.42, / "
1.21 - . , : " - -
-I . at
-it=
-.3.43,
At t i me t = 40. 0
I I I I I
40 gO 120 160 300 f "
U
0. 5
0 . 2 .
0 . 1
= 0 . I
- 0. 2
- 0. 3
.'t
! I I I I
r"
At time t --- 60. 0
F i g u r e 7. S o l u t i o n of P r o b l e m 2 wi t h e q u a t i o n ( c) .
58
P
S
I I I I I
W laO 160 I
P. C..-.LA ISTER
U
0, ?8
0 , 9 2
0, ~
r '
. . 0, w,
. . 0. 52
..0.N$
At t i me t = 30.0
P
3 . ~ 9 ,
\
- I . l m
/
I I I I I
U
0 , I t .
0 . ~
O . m
.-0,~
- 0. ~
-0,114
/ \
8
/
I I I ! I 0 r
y
P
3 . 6
2 . 4
- I . 2
--2,4
4, 6
At t i me t --- 40.0
0
p *
I
/ -
I I I I I
r
U
0 . 9 .
0 . 6 ,
0* 3 ,
- O i l ,
- 0. 6
-0.9
/ -
. ~ \
!
t
I
I
4O
, , o
80 I m I ~ r
S
At t i me t -- 60.0
Fi gur e 8. Sol ut i on of Pr obl em 2 wi t h equat i on (d).
Real gas flows 59
7. NUME RI CAL RE S UL T S
In t hi s sect i on we exhi bi t t he numer i cal resul t s obt ai ned for t he t est pr obl ems descr i bed in
Sect i on 6 usi ng t he scheme descr i bed in Sect i on 4.
Fi gures 1- 4 refer t o t he fi rst pr obl em of t he pr evi ous sect i on wi t h equat i ons ( a) - ( d) , respec-
tively, usi ng a second or der scheme wi t h t he ' super bee' l i mi t er [4] and 200 mesh poi nt s. Thr ee
out put t i mes have been f eat ur ed. For each equat i on of st at e, we observe t he devel opment of t he
shock and cont act di scont i nui t y, t he refl ect i on of t he shock, and t he i nt er act i on of t he shock wi t h
t he cont act di scont i nui t y. The resul t s for equat i on (a) compar e f avor abl y wi t h resul t s pr oduced
usi ng a mor e compl ex al gor i t hm [5].
The c. p. u, t i me r equi r ed t o comput e t he resul t s on an Amdahl V/ 7 was f ound t o be as follows:
wi t h 200 mesh poi nt s it t ook 0.04 c. p. u, seconds t o comput e one t i me st ep, and a t ot al of 8.64
c. p. u, seconds t o reach a real t i me of 54.0 using 216 t i me st eps.
Fi gur es 5- 8 refer t o t he second pr obl em of t he previ ous sect i on wi t h equat i ons (a)--(d), respec-
tively, agai n usi ng 200 mesh poi nt s and a second or der scheme wi t h ' s uper bee' hmi t er . Thr ee
out put t i mes have agai n been f eat ur ed. For each equat i on of st at e, we observe t he devel opment
of t he shocks and cont act di scont i nui t i es, t he i nt er act i on of t he shocks, and t he subsequent i nt er-
act i on of each shock wi t h a cont act di scont i nui t y, resul t i ng in t r ans mi t t ed and refl ect ed shocks
and cont act di scont i nui t i es.
A mi nor modi f i cat i on t o our al gor i t hm allows for a vari abl e ( adapt i ve) t i me st ep and gives t he
abi l i t y t o decr ease t he t ot al amount of comput i ng t i me used.
8. CONCL US I ONS
We have ext ended t he one-di mensi onal versi on of Gl ai st er ' s scheme [2] t o obt ai n a new scheme
t ha t i ncl udes cyl i ndri cal l y and spheri cal l y s ymmet r i c probl ems. We have shown how t hese pr ob-
l ems br eak away f r om s t a nda r d ' conser vat i on' f or m and t hus give rise t o source t er ms, and t hat
wi t h t he appr oach out l i ned in Sect i on 4 we can achieve sat i sf act or y resul t s on t wo s t andar d t est
pr obl ems. The mai n f eat ur es of t hi s scheme are
(i) i t can be used for cyl i ndri cal l y or spheri cal l y s ymmet r i c pr obl ems wi t h confi dence, and
(ii) i t can be used as a compar i son wi t h resul t s f r om t wo-di mensi onal schemes by choosi ng a
large number of mesh poi nt s for accur acy and not be expensi ve on comput i ng.
Th e scheme can also be used for general duct flows.
REFERENCES
1. P. Glaister, An approximate linearised Riemann solver for real gases, J. Comp. Phys. 74,382-408 (1988).
2. P. Glaister, An efficient Ftiemann solver for unsteady flows with non-ideal gases, Comp~tera M~th. Applic.
24 (3), 77-93 (1992).
3. P. Glaister, Second order difference schemes for hyperbolic conservation laws with source terms, Numerical
Analysis Report 6-87, University of Reading, (1987).
4. P.K. Sweby, High resolution schemes using flux limiters for hyperbolic conservation laws, SI AM J. Numer.
A~a/. 21,995-1011 (1984).
5. M. Ben-Artzi and J. Falcovitz, A second order Godunov-type scheme for compressible fluid dynamics,
J. Cornp. Phys. 55, 1-32 (1984).