Download as doc, pdf, or txt
Download as doc, pdf, or txt
You are on page 1of 11

GENERAL EQUATION SET SOLVER FOR COMPRESSIBLE AND

INCOMPRESSIBLE TURBOMACHINERY FLOWS


Douglas L. Sondak
*
Boston University
Boston, MA
Daniel J. Dorney

NASA Marshall Space light !enter


Marshall Space light !enter, AL
ABSTRACT
"ur#o$achines %or propulsion applications operate &ith
$any di%%erent &orking %luids and %lo& conditions. "he
%lo& $ay #e inco$pressi#le, such as in the li'uid
hydrogen pu$p in a rocket engine, or supersonic, such as
in the tur#ine that $ay drive the hydrogen pu$p.
Separate codes have traditionally #een used to solve
inco$pressi#le and co$pressi#le %lo&s. "he (eneral
)'uation Set *()S+ $ethod can #e used to solve #oth
inco$pressi#le and co$pressi#le %lo&s, and it is not
restricted to per%ect gases, as are $any co$pressi#le,%lo&
tur#o$achinery solvers. -t also contains preconditioning
%or acceleration o% steady,state si$ulations or rapid
eli$ination o% initial transients in unsteady si$ulations.
An unsteady ()S tur#o$achinery %lo& solver has #een
developed and applied to %lo&s through tur#o$achinery
co$ponents operating in #oth li'uids and gases. -t has
#een sho&n to #e a versatile techni'ue %or solving a &ide
range o% tur#o$achinery %lo& pro#le$s.
NOMENCLATURE
(reek Sy$#ols.

trans%or$ed coordinate directions


trans%or$ed ti$e
/
trans%or$ed pseudo,ti$e
0o$an Sy$#ols.
), , ( %lu1 vectors
h enthalpy per unit $ass
J Jaco#ian o% coordinate trans%or$ation
p pressure
2 conserved varia#les
2p pri$itive varia#les
0e 0eynolds nu$#er
" te$perature
u, v, & physical velocity co$ponents
U, 3, 4 contravariant velocity co$ponents
Superscripts.
k pseudo,ti$e step inde1
n physical ti$e step inde1
v viscous
Su#scripts.
t tur#ulent
1, y, 5
partial di%%erentiation, physical
coordinates
6ther sy$#ols.
*
7
+ trans%or$ed coordinates
di%%erence
/
Senior Scienti%ic 8rogra$$er, 6%%ice o% -n%or$ation
"echnology, Senior Me$#er A-AA.

Aerospace )ngineer, Space "ransportation Directorate,


Associate ello& A-AA.
!opyright 9:;;< #y the A$erican -nstitute o% Aeronautics and
Astronautics, -nc. No copyright is asserted in the United States under
"itle =>, U.S. !ode. "he U.S. (overn$ent has a royalty,%ree license
to e1ercise all rights under the copyright clai$ed herein %or
(overn$ental 8urposes. All other rights are reserved %or the
copyright o&ner.
INTRODUCTION
"ur#o$achines %or propulsion applications operate on
$any di%%erent %luids and under a &ide range o% %lo&
conditions. "he &orking %luid $ay #e air, li'uid or
gaseous hydrogen, li'uid or gaseous o1ygen, kerosene,
etc. "he %lo& $ay #e inco$pressi#le, such as in the
A$erican -nstitute o% Aeronautics and Astronautics
=
%uel pu$p in a li'uid,%uel rocket engine, or supersonic,
such as in the tur#ine that $ay drive the %uel pu$p.
Both co$pressi#le and inco$pressi#le %lo&s are
governed #y the Navier,Stokes e'uations. ?o&ever, in
a nearly,inco$pressi#le %lo& there is a great disparity
in &ave speeds, since the speed o% sound approaches
in%inity %or a truly inco$pressi#le %luid. A co$pressi#le
%lo& solver &ill encounter nu$erical sti%%ness i% applied
to a nearly,inco$pressi#le %lo&, and the algorith$ &ill
%ail. Because o% this, it is co$$on practice to use one
algorith$ %or inco$pressi#le %lo&s and a di%%erent
algorith$ %or co$pressi#le %lo&s.
"he (eneral )'uation Set *()S+ $ethod
=
has #een
developed to handle #oth co$pressi#le and
inco$pressi#le %lo&s. -t can #e used to solve the %ull,
unsteady, <,D Navier,Stokes e'uations, and &ith the
introduction o% a single input %lag can reduce to the
pseudo,co$pressi#ility $ethod co$$only used to
solve inco$pressi#le %lo&s. 4hen used in con@unction
&ith a dual ti$e step, it can #e used %or ti$e accurate
si$ulations.
4hen the Navier,Stokes e'uations are cast in their
general %or$, &ithout applying the per%ect gas relation,
they can #e solved %or any &orking %luid %or &hich
properties are availa#le. "he ()S techni'ue in
con@unction &ith the e'uations cast in this %or$ results
in a general %lo& solver applica#le to $ost conditions
encountered in tur#o$achinery applications.
A code has #een &ritten to solve <,D, unsteady
tur#o$achinery %lo&s using the ()S $ethod. "est
cases have #een per%or$ed %or tur#o$achines operating
in lo&,speed air *inlet Mach nu$#er o% ;.;>+, transonic
gaseous o1ygen, li'uid o1ygen, and kerosene *08,=+.
0esults have #een co$pared &ith test data &here
availa#le, and &ith the results o% a &ell,developed
ideal,gas, co$pressi#le tur#o$achinery %lo& solver
:
.
EQUATIONS
"he <,D, unsteady, Navier,Stokes e'uations $ay #e
e1pressed in generali5ed, curvilinear coordinates in
dual,ti$e,step %or$ as

( )
/
=
A A A A A
A A A
0e ;
v v v
Q Q E F G
E F G

+ + + +
+ +

&here
[ ]
=
A
T
Q J u v w E


is the standard set o% conserved varia#les. "he %lu1es
also have their standard de%initions. "he varia#le
represents the trans%or$ed physical ti$e step, and
/

represents the trans%or$ed pseudo,ti$e step %or


su#iterations.
De%ining the vector o% pri$itive varia#les as
[ ]
=
A
T
p
Q J p u v w T


the Navier,Stokes e'uations can #e re&ritten as
( ) ( )
( )
/
=
A A A A A
A A A
0e ;
e p p p
v v v
Q Q E F G
E F G


+ + + +
+ +

&here
; ; ;
; ;
; ;
; ;
=
A
A
p T
p T
p T
p T
o o
p p T T
e
p
u u
v v
w w
h h u v w h h
Q
Q




+ +

1
1
1
1
1
1
1
]
-n order to pre,condition the e'uations,
p

and
T
in
)'n. are re,de%ined as the pseudo,properties
p

and
T

. "he values o% the pseudo,properties can #e


chosen to opti$i5e convergence %or di%%erent types o%
%lo&s, such as inco$pressi#le, su#sonic co$pressi#le,
supersonic, etc. "he preconditioned $atri1 is given #y
A$erican -nstitute o% Aeronautics and Astronautics
:
; ; ;
; ;
; ;
; ;
=
A
A
p T
p T
p T
p T
o o
p p T T
p
p
u u
v v
w w
h h u v w h h
Q
Q










+ +

1
1
1
1
1
1
1
]

NUMERICAL METHOD
)'uation &ill #e solved using appro1i$ate
%actori5ation. -n the dual,ti$e,step $ethod, the
solution is iterated to convergence in pseudo,ti$e at
each physical ti$e step. "he physical ti$e di%%erencing
&ill have 6*:+ accuracy, and the pseudo,ti$e
di%%erencing &ill #e 6*=+. Second,order accuracy is not
re'uired %or pseudo,ti$e since it is driven to
convergence each step. Since t&o di%%erent ti$e
coordinates are #eing used, physical and pseudo, the
%ollo&ing notation &ill #e used to di%%erentiate #et&een
nu$erical ti$e di%%erences.
=, = n k n
n

+ +

=, k n k n
n

+

and
=, = =, n k n k
k

+ + +

&ith n indicating the physical ti$e and k indicating the
pseudo,ti$e. Note that values at ti$e level n do not
have a k inde1 since the inner iterations have #een
converged at ti$e level n.
"he 6*=+ pseudo,ti$e di%%erence ter$ in )'n. is
appro1i$ated #y
( ) ( )
/
=, = =,
/
=
A A A
n k n k
p p p
Q Q Q


+ + +


and the 6*:+ physical ti$e di%%erence is appro1i$ated
#y
( )
=, = =
< =
A A A A
:
: :
=
n k n n
p p p p
Q Q Q Q

+ +
+
_

,

"he ,direction inviscid %lu1 derivative is
appro1i$ated #y
=
A A A A
n n n
p n p
E E A Q

+
+
and other inviscid and viscous %lu1 di%%erences are
appro1i$ated in a si$ilar %ashion.
Applying )'ns. , to )'n. yields
( )
( ) ( ) ( )
( ) ( ) ( )
=, = =,
/
=, = =
=
=
A A
= < =
A A A
:
: :
A A A
A A A A A A
A A A
0e
A A A A A A
;
B
C
n k n k
p p p
n k n n
e p p p
n n n
n n n
p n p p n p p n p
n n n
v v v
n n n
v v v
p n p p n p p n p
Q Q
Q Q Q
E F G
A Q B Q C Q
E F G
A Q B Q C Q



+ + +
+ +

+ +

+ + +
+ + +
+ +
+ + +

_

,
_ _

, ,
4e &ant to &rite this e'uation in ter$s o%
A
k p
Q ,
since that is the varia#le %or &hich &e &ill #e solving
on the le%t hand side. Adding and su#tracting
=,
<
A
:
n k
p
Q
+
&ithin the parenthesis in )'n. and applying )'n.
yields
( )
=, =
= <
A A
:
= < =
A A A
:
: :
p k p
n k n n
p p p
Q Q
Q Q Q

+ +

_

,
_ _

, ,

Applying )'n. to )'n. yields
A$erican -nstitute o% Aeronautics and Astronautics
<
( )
( )
( )
/
=
=, =
=
=
= < =
:
A A A
A A A A
0e
= < =
A A A
:
: :
A A A
A A A A
0e
A A A
A
0e
B
C
C
B
p e
p p p
v v v
p p p k p
n k n n
e p p p
p p p
v v v k
p p p n p
v
p
A B C
A B C Q
Q Q Q
A B C
A B C Q
E F G
E






+

+ + +
+ +
+

+ +
+ +
+ +
+ +
_ _

, ,
_ _

, ,
( )
A A
v v
p p
F G

+

All %lu1 Jaco#ians are at ti$e level n. "he last t&o
ter$s on the right hand side can #e co$#ined i% desired.
( )
( )
/
=
=, =
= =,
= < =
:
A A A
A A A A
0e
= < =
A A A
:
: :
A A A
A A A A
0e
B
C
B
C
p e
p p p
v v v
p p p k p
n k n n
e p p p
p p p
v v v n k
p p p p
A B C
A B C Q
Q Q Q
A B C
A B C Q




+
+
+

+ + +
+ +
+

+ +
+ +
_ _

, ,
_ _

, ,

Since &e are solving the overall e'uation %or values at
pseudo,ti$e level k+1,
=,
A
n k
p
Q
+
re$ains on the right
hand side.
Be%ore per%or$ing appro1i$ate %actori5ation, it is
convenient to de%ine
/
<
:
p e
S

"he %actori5ed e'uation is given #y


( )
( )
( )
( )
/ =
= / =
= / =
/
=, =
/
= =,
A A
0e
A A
0e
A A A
0e
< =
A A A
:
: :
A A A
A A A A
0e
B
C
v
p p
v
p p
v
p p k p
n k n n
e p p p
p p p
v v v n k
p p p p
S A A
S S B B
S S C C Q
Q Q Q
A B C
A B C Q



+ +
+
+
+
+

+ +
+ +
1
]
1
]
1
]
_
_

,
,

0ight ?and Side
"he ti$e ter$ on the right hand side is straight%or&ard
to i$ple$ent, since it contains no di%%erences.
Solutions are re'uired at t&o previous ti$e steps in
order to get second order te$poral accuracy. or the
viscous ter$s, standard second,order central di%%erences
are used, as is co$$on practice. -nviscid ter$s are
co$puted using the appro1i$ate 0ie$ann solver o%
0oe
<
. "he eigenvalues o% the syste$ are re'uired %or
0oeDs sche$e. ?ere, the relevant eigenvalues are those
o% the $atri1
=
p p
A

, &hich are given #y


=,:,<
U

and
( )
E,F
:
: : : :
=
=
: G
= E
G G
H
I
T
x y z
U
h
U

t + + +

_

,
_

,
&here
T p T
h +

G G G
T p T
h +

=
p
h

Si$ilar e1pressions can #e &ritten %or the other t&o
coordinate directions #y su#stituting V or W %or U, and

or %or in )'ns. and .


A$erican -nstitute o% Aeronautics and Astronautics
E
Le%t ?and Side
-deally, %astest convergence is o#tained &hen the sa$e
$ethod o% di%%erencing is used on #oth the le%t and right
hand sides o% )'n.. "his is possi#le %or lo&,order
sche$es since the #lock tridiagonal structure o% the
e'uations can #e $aintained. ?igher order sche$es
re'uire larger di%%erence stencils, and using the$ on the
le%t hand side precludes the use o% #lock tridiagonal
solvers. ?ere, Steger,4ar$ing
E
%lu1,vector splitting is
used on the le%t hand side. "his does not a%%ect the
accuracy o% the solution, since the le%t hand side is
driven to 5ero during the pseudo,ti$e iterations, and it
has #een sho&n to #e a ro#ust techni'ue.
PRECONDITIONING PARAMETERS
)'uation contains t&o yet,to,#e,de%ined para$eters,
G
p
and
G
T
. "he para$eter
G
T
has li$ited e%%ect on
convergence rate %or $ost %luids, and is si$ply set
e'ual to
T
here.
"he s'uare o% the speed o% sound can #e e1pressed as
: T
h
a


&here is de%ined in )'n. . -% in )'n. is replaced
#y
G

as de%ined in )'n. , an arti%icial sound speed can


#e de%ined,
:
G
T
p
h
a


Su#stituting )'n. into )'n. , noting that
G
T
has #een
set e'ual to
T
, and solving %or
G
p
yields
( )
G
:
=
=
T p
p
p T
h
a h



6nce the arti%icial sound speed is de%ined, the
preconditioning para$eter is kno&n. 3enkates&aran
and Merkle)rror. 0e%erence source not %ound have
developed a $ethod %or esti$ating
p
a
#ased on the
calculation o% inviscid, pressure,gradient, and unsteady
velocity scales.
"he inviscid velocity scale is si$ply the physical
velocity,
: : :
inv
V u v w + +
"he pressure,gradient velocity scale is given #y
pg
p
V


"he unsteady velocity scale is given #y
$a1 , ,
y
x z
un!
"
" "
V
# # #
_



,
&here
x
" ,
y
"
, and
z
" are characteristic unsteady length
scales in the x$ y$ and z directions.
"he %inal velocity scale is then given #y
( )
$in $a1 , , ,
p inv pg un!
V V V V a 1
]

&here a is the speed o% sound.
CODE STRUCTURE
"he general code structure is #ased on a &ell,
esta#lished co$pressi#le, <,D, unsteady
tur#o$achinery %lo& solver
:
. -t e$ploys a syste$ o%
overset 6,grids and ?,grids, &ith the values on the 6,?
#oundaries #eing updated each ti$e step #y trilinear
interpolation %ro$ the ad@acent grid. "he Bald&in,
Lo$a1 tur#ulence $odel
F
is used %or tur#ulence
closure.
RESULTS
:,D =,=J: Stage "ur#ine
"he %irst test case is the :,D si$ulation o% the %lo&
#ased on e1peri$ents on a =,=J: stage Lo& Speed
0esearch 0ig *LS00+ tur#ine at United "echnologies
0esearch !enter
6
. "he inlet Mach nu$#er is ;.;>, and
the ratio o% e1it static to inlet total pressure is ;.KF;F.
"he actual tur#ine contained ::, :L, and :L #lades in
#lade ro&s one, t&o, and three respectively. -n the
A$erican -nstitute o% Aeronautics and Astronautics
F
si$ulation, the %irst,ro& geo$etry &as scaled #y ::J:L
to allo& the si$ulation to #e per%or$ed &ith one #lade
per ro& &hile $aintaining the correct #lockages.
"he grid di$ensions in each o% the three #lade ro&s are
=;K1EF in the ?,grid and :F=1E= in the 6,grid. "he
grid is sho&n in ig. =.
Figure 1 C!"#u$%$i!&%' gri() 1*1+, -$%ge LSRR
$ur.i&e
Figure , Pre--ure /!&$!ur-) 1*1+, -$%ge LSRR
$ur.i&e
8ressure contours are sho&n in ig. : to de$onstrate
'ualitatively the e1pected pressure %ield. "he do$ain
has #een repeated in the circu$%erential direction to
clari%y the contours. "he characteristic accelerations
and decelerations are seen through the passages.
"i$e,averaged sur%ace pressures are sho&n in igs. <,
F, co$paring the co$puted values &ith e1peri$ental
data %or each #lade ro&.
Figure 0 Pre--ure #r!1i'e) -$%$!r 1) 1*1+, -$%ge
LSRR $ur.i&e
Figure 2 Pre--ure #r!1i'e) r!$!r) 1*1+, -$%ge LSRR
$ur.i&e
A$erican -nstitute o% Aeronautics and Astronautics
M
Figure 3 Pre--ure #r!1i'e) -$%$!r ,) 1*1+, -$%ge
LSRR $ur.i&e
"he $atch is good %or the stators, &ith so$e
discrepancy on the suction sur%ace o% the rotor. "he
suction,sur%ace $is$atch has #een o#served in a
nu$#er o% previous si$ulations o% this con%iguration #y
other researchers.
igure M sho&s entropy contours. "he #oundary layer
gro&th and &ake shedding can #e seen in each #lade
ro&. "he &ake %ro$ the %irst stator convects into the
rotor passage and takes on the characteristic hairpin
shape as it #eco$es entrained in the rotor #lade sur%ace
#oundary layers. Both the stator,= &ake and the rotor
&ake then interact &ith the do&nstrea$ stator, creating
a co$ple1 pattern o% high and lo& entropy %luid
Figure 6 E&$r!#4 /!&$!ur-) 1*1+, -$%ge LSRR
$ur.i&e
:,D Supersonic,-n%lo& "ur#ine
"he ne1t test case is a si$ulation o% the %lo& through a
gaseous o1ygen *(6N+ tur#ine &ith supersonic in%lo&
to the rotor, typical o% those #eing proposed %or reusa#le
launch vehicles. "his case &as chosen to test the shock
capturing a#ility o% the sche$e in order to validate the
coding o% the %lu1,di%%erence splitting.
"he tur#ine has a design inlet Mach nu$#er o% MO;.:F,
an inlet static pressure o% :::F psia, and an inlet static
te$perature o% appro1i$ately "O:::F 0. "he rotor
rotates at <:,;;; 08M, the 0eynolds nu$#er #ased on
the inlet conditions and the rotor a1ial chord is
appro1i$ately =.:1=;
M
and the ratio o% the rotor e1it
static pressure to vane inlet total pressure is ;.=:F. "he
tur#ine has := vanes and F: rotors. -n the current
con%iguration a :;,vaneJF;,rotor,#lade appro1i$ation
has #een $ade, resulting in a si$ulation #lade count o%
: vanes and F rotor #lades. "o keep the pitch,to,chord
ratio *#lockage+ constant, the %irst,stage vanes &ere
scaled #y a %actor o% :=J:; and the rotor #lades &ere
scaled #y a %actor o% F:JF;. -n the vanes, each ?,grid
contained MF1F= points, and each 6,grid contained
=>=1<= points. "he rotor #lade ?,grids each contained
>F1F= points, and the 6,grids each contained =>=1:=
points. "he grid is sho&n in ig. >.
Figure 5 Gri() GO6 $ur.i&e 7i$8 -u#er-!&i/ r!$!r
i&1'!7
8ressure contours are sho&n in ig. L. "he pressure
decreases rapidly through the covered portion o% the
vane passage. "he potential e%%ect o% the rotor on the
pressure %ield o% the vane can #e clearly seen #y the
di%%erences #et&een the uncovered portions o% the
suction sur%aces o% the t&o vanes. "he pressure
increases across the shocks at the rotor leading edges,
and then the %lo& re,accelerates.
-n order to get a #etter vie& o% the shock &aves, &akes
and e1pansions, Mach nu$#er contours are sho&n in
ig. K. "he %lo& enters the vane passage at a lo&
su#sonic Mach nu$#er. "he %lo& accelerates through
the vane passage, reaching sonic *choked+ conditions at
A$erican -nstitute o% Aeronautics and Astronautics
>
the throat. "he %lo& #eco$es supersonic in the
diverging portion o% the vane passage, reaching a peak
Figure 9 Pre--ure /!&$!ur-) GO6 $ur.i&e 7i$8
-u#er-!&i/ r!$!r i&1'!7
Mach nu$#er o% nearly :.;. A strong e1pansion &ave
syste$ e$anates %ro$ the pressure side o% the trailing
edge. "he e1pansion &aves $ove across the vane
passage and interact &ith the suction sur%ace o% the
ad@acent vane. "he vane &akes convect into the rotor
passage &here they interact &ith the rotor #o& shock.
"his interaction creates a su#sonic %lo& region near the
leading edge that &eakens the #oundary layer and
causes rotor suction sur%ace #oundary layer separation.
"his condition e1ists until the rotor passes through the
vane &ake. "he %lo& di%%uses as it e1its the rotor
passage, decelerating #ack to su#sonic %lo& conditions.
:,D and <,D L6N "ur#ine
"he ne1t test case is the si$ulation o% %lo& through a
lo&,pressure o1idi5er tur#ine &ith L6N as the &orking
%luid. "he actual tur#ine &as e1hi#iting cracks, and it
&as thought that the cracks $ight have #een caused or
e1acer#ated #y trailing edge vorte1 shedding. "he
si$ulations &ere per%or$ed to study the unsteady %lo&
through the tur#ine. Both :,D and <,D si$ulations &ere
per%or$ed in order to deduce &hether <,D e%%ects
&ould alter the $idspan shedding %re'uency.
"he actual tur#ine had <K vanes and M= rotor #lades.
"he co$putation &as per%or$ed on t&o vanes and three
rotor #lades to si$ulate a E;,vane, M;,rotor,#lade
Figure : M%/8 &u".er /!&$!ur-) GO6 $ur.i&e
7i$8 -u#er-!&i/ r!$!r i&1'!7
tur#ine, &ith appropriate scaling to keep #lockages the
sa$e as in the actual tur#ine. "he inlet Mach nu$#er
&as ;.;E, and the 0eynolds nu$#er #ased on rotor a1ial
chord &as
M
;.F =;
. "he ratio o% e1it static pressure to
inlet total pressure &as ;.>LL. "he :,D vane grid &as
=>=1F= %or the 6,grid and MF1E= %or the ?,grid. "he
rotor 6,grid &as =>=1F=, and the ?,grid &as L;1<=.
"he grid, &hich has a total o% FM,<>F points, is sho&n in
ig. =;. -n the <,D si$ulation, the vane 6,grids &ere
=F=1E=1:=, and the vane ?,grids &ere ==M1F=1:=.
"he rotor 6,grids &ere =:=1E=1:= and the rotor ?,
grids &ere =:<1<=1:=. "he average value o% yP &as
appro1i$ately =.F %or #oth ro&s. -n addition, the
sur%ace #oundary layers &ere discreti5ed &ith =;,=F
grid points. "he co$putational grid %or the three,
di$ensional si$ulation contains =.=E $illion grid
points.
-nstantaneous te$perature contours are sho&n in ig.
==. "e$perature contours &ere chosen #ecause they
clearly sho& the vorte1 shedding %ro$ the vanes and
the interaction o% the vortices &ith the rotor #lades.
So$e o% the vortices i$pinge directly on the rotor #lade
leading edges, and are split across the suction and
pressure sur%aces. 6thers have their shape distorted due
to the potential interaction #et&een the #lade ro&s.
8ressure contours %or the <,D o1idi5er tur#ine case are
sho&n in ig. =:. "he potential interaction is sho&n #y
the di%%erences #et&een the pressure contours on the
A$erican -nstitute o% Aeronautics and Astronautics
L
Figure 1; Gri() ,*D !<i(i=er $ur.i&e
Figure 11 I&-$%&$%&e!u- &!&*(i"e&-i!&%'
$e"#er%$ure /!&$!ur-) ,*D !<i(i=er $ur.i&e
rotor #lades, particularly over the %irst hal%,chord.
Upstrea$,propagating pressure &aves can #e seen
s$oothly e1iting the inlet #oundary.
-nitial :,D si$ulations &ere per%or$ed &ith an ideal,
gas code, yielding a shedding %re'uency o% E<.L k?5.
"his di%%ered su#stantially %ro$ the e1peri$ental value
o% <M k?5. "he present si$ulations resulted in
shedding %re'uencies o% <M.K k?5 in :,D and <M.F k?5
in <,D. "his sho&s that itDs i$portant to use the correct
%luid properties to accurately capture the shedding
Figure 1, Q Mi(-#%& i&-$%&$%&e!u- &!&*(i"e&-i!&%'
#re--ure /!&$!ur-) 0*D !<i(i=er $ur.i&e
%re'uency. -t also sho&s that, %or this con%iguration, <,
D e%%ects on the shedding %re'uency &ere $ini$al.
<,D 08,= -nducer
"he %inal validation case involves the %lo& o% kerosene
*08,=+ through a lo&,pressure %uel pu$p *L88+
inducer. "he inducer has an inlet pressure o% :: psi, an
inlet te$perature o% <:; 0 and the $ass %lo& is =L;
l#$Jsec. "he inducer contains three #lades, each o%
&hich has appro1i$ately <;; degrees o% &rap. "he
co$putational grid, &hich $odels one o% the inducer
#lades, consists o% an inlet duct %ollo&ed #y the inducer
#lade. "he duct grid *?,grid only+ has di$ensions o%
KM1E=1:=. "he inducer 6,grid has di$ensions o%
:F=1=K1:=, and the inducer ?,grid has di$ensions o%
==>1E=1:=. "he sur%ace grid is sho&n in ig. =< *&here
all three #lades are sho&n %or clarity+. igure =E sho&s
sur%ace pressure contours through the inducer. Although
the solution is not co$pletely periodic, the proper
trends are displayed and the kerosene properties #ehave
as e1pected.
CONCLUSIONS AND RECOMMENDATIONS
"he (eneral )'uation Set $ethod has #een
i$ple$ented, along &ith general %luid properties, into a
code %or tur#o$achinery applications. A series o% test
cases have #een run de$onstrating that the code
per%or$s &ell on the %lo& o% lo&,speed air, transonic
gaseous o1ygen, li'uid o1ygen, and kerosene *08,=+. -t
has #een sho&n that the correct $odeling o% %luid
A$erican -nstitute o% Aeronautics and Astronautics
K
properties is i$portant %or the accurate prediction o%
vorte1 shedding %re'uencies.
Figure 10 Q C!"#u$%$i!&%' gri( 1!r $8e LPFP
i&(u/er>
Figure 12 Q I&-$%&$%&e!u- &!&*(i"e&-i!&%' #re--ure
/!&$!ur- i& $8e LPFP i&(u/er
"he dual ti$e step algorith$ re'uires speci%ication o% a
para$eter *the ratio o% pseudo,ti$e,step to physical
ti$e step+, and the choice &ill a%%ect the convergence
rate. !urrently, this ratio can #e set glo#ally or a local
pseudo,ti$e step can #e used. -t is reco$$ended that a
series o% si$ulations #e per%or$ed on typical
con%igurations using a range o% values %or this
para$eter. "his &ill help to deduce a range o% the
para$eterDs value to opti$i5e convergence rates. "his
&ill #e particularly %or co$puting inco$pressi#le and
nearly,inco$pressi#le %lo&s.
6ne issue that has #een encountered in running certain
cases has #een that the %luid can encounter non,physical
phase changes during startup transients. "his reduces
the sta#ility o% the code, since a very %ine grid &ould #e
re'uired in the vicinity o% the phase di%%erence in order
to resolve it. "his pro#le$ has #een $itigated #y
li$iting the pressure and te$perature changes that can
occur during initial transients. "hese li$its are re$oved
once the initial transients have settled out. uture &ork
&ill involve the develop$ent o% a techni'ue to handle
the phase changes in a $ore elegant %ashion.
REFERENCES
A$erican -nstitute o% Aeronautics and Astronautics
=;
=
3enkates&aran, S., and Merkle, !. L., RAnalysis o% 8reconditioning Methods %or the )uler and Navier,Stokes )'uations,S
3on Tar$an -nstitute Lecture Series, March L,=:, =KKK.
:
Dorney, D.J., Davis, 0.L., )d&ards, D.)., and Madavan, N.T., UUUnsteady Analysis o% ?ot Streak Migration in a "ur#ine
Stage,GG A-AA Journal o% 8ropulsion and 8o&er, 3ol. L, No. :, March,April, =KK:, pp. F:;,,F:K.
<
<
0oe, 8. L.,SAppro1i$ate 0ie$ann Solvers, 8ara$eter 3ectors, and Di%%erence Sche$es,GG Journal o% !o$putational
8hysics, 3ol E<, No. :, =KL=, pp. <F>V<>:.
E
E
Steger, J. L., and 4ar$ing, 0. ., UUlu1 3ector Splitting o% the -nviscid (asdyna$ic )'uations &ith Application to inite,
Di%%erence Methods,GG Journal o% !o$putational 8hysics, 3ol. E;, =KL=, pp. :M<,:K<.
F
Bald&in, B. S., and Lo$a1, ?., UU"hin Layer Appro1i$ation and Alge#raic Model %or Separated "ur#ulent lo&s,GG A-AA
paper >L,:F>, =Mth Aerospace Sciences Meeting, Jan. =M,=L, =K>L, ?untsville, Al.
M
Dring, 0. 8., Blair, M. ., Joslyn, ?. D., 8o&er, (. D., and 3erdon, J. M., UU"he )%%ects o% -nlet "ur#ulence and
0otorJStator -nteractions on the Aerother$odyna$ics and ?eat "rans%er o% a Large,Scale 0otating "ur#ine Model,GG NASA
!ontractor 0eport E;>K, May, =KLM.

You might also like