https://ntrs.nasa.gov/search.jsp?

R=19900004433 2019-01-11T01:17:39+00:00Z

NASA Technical Memorandum 102469

AIAA-90-0643

Analysis of Rotary Engine Combustion

Processes BaSed On Unsteady?
Three-Dimensional Computations

M.S. Raju
Sverdrup Technology, Inc.
NASA Lewis Research Center Group
Cleveland, Ohio

and

E.A. Willis
National Aeronautics and Space Administration
Lewis Research Center
Cleveland, Ohio

Prepared for the

28th Aerospace Sciences Meeting
......... spons_ore d_by the America n Institute of Aer_onauti_c_sand A_stronautics ....................
RenD, Nevada, January 8-11, 1990

_(NASA-TM-IO2469) ANALYSIS OF ROTARY ENGINE N90-13749 _

COMBUSTION PROCESSES BASED ON UNSTEADY,
THREE-DIMENSIONAL COMPUTATIONS (_IASA) 37 p
CSCL 200 Unclas
G3/36 0252608
 k_

Analysis of Rotary Engine Combustion Processes

Based on Unsteady, Three-Dimensional Computations

M. S. Rajut
Sverdrup Tech., Inc.,NASA Lewis Research Center,Cleveland, Ohio

E.A. Willis:_
NASA Lewis Research Center, Cleveland, Ohio

Abstract

A new computer code has been developed for predictingthe turbulent,and chemicallyreact-
ing flows with sprays occurringinsideof a stratified-charge
rotary engine (SCRE). The solution
procedure is based on an Eulerian-Lagrangianapproach where the unsteady, three-dimensional
Navier-Stokesequationsfora perfectgas-mixturewith variablepropertiesare solvedin generalized,
Euleriancoordinateson a moving gridby making use of an implicitfinite-volume,
Steger-Warming
fluxvectorsplittingscheme, and the liquid-phaseequations are solvedin Lagrangian coordinates.
Both the detailsof the numerical algorithm and the finite-difference
predictionsof the combus-
tor flowfield
during the opening of exhaust and/or intake,andalso during fuelvaporizationand
combustion, are presented.

I. Introduction

The rotary combustion engine (RCE) would be desirableas a powerplant for lightaircraft,
drones (includinghigh-altitudeapplication),auxiliaryand ground power units,and alsofor ma-
rine and industrialapplication,if only itsefficiency
could be improved closerto that of diesel
engines.Ithas inherentadvantages over reciprocatingengines in terms of higher airflowcapacity,
higher power-to-weightratio,and lessvibration,among others. An initialattempt to introduce
a gasoline-fueled
rotaryengine intothe generalaviationmarket was unsuccessfulbecause of poor
fueleconomy, uncertain availability
of avgas, and marginal weight advantage over contemporary
reciprocatingengines.Subsequent researchsponsored by industry,NASA, and the Navy has ledto
the development of the stratified-charge
rotaryengine (SCRE) concept,which has demonstrated
multifuelcapability.Current R&D sponsored by NASA isaimed at reducing the cruisebrake spe-
cificfuelconsumption (BSFC, Ib/bhp-hr) from a currentvalue of about 0.42 to 0.35 or lessby the
end of 1992. The expected improvement willbe enabled by furtherCFD - driven fuelinjection,

tSenior Research Engineer

_Chief Engineer, Propulsion Systems Division,Member AIAA

spray,and nozzleoptimizations,rotor pocket and nozzlerelocations,
and relatedmodifications.In
thispaper, we address the CFD aspectof the technologydevelopment program forpredictingthe
complex flow patternsoccurringinsideof a Wankel engine.
Early modelling efforts
on the Wankel enginewere based on thermodynamic models 1'2and also
on one-dimensional modelling of premixed-charge combustion,s Multi-dimensionalmodels of the
Wankel engine are very recentin origin;Grasso et al.4have presentedthe first
three-dimensional
computations of a SCRE during the earlystagesof flame propagation. Subsequent computations
performed by Abraham and Bracco5,ehave led to some important design changes in the rotary
engine development at John Deere & Co., especiallyin the fuel injectorconfiguration.Their
code, REC-3D-FSC-86, is a modifed versionof the KIVA code developed at Los Alamos National
Laboratory7 forthe modelling of reciprocatingengines.KIVA makes use of a conditionallystable
algorithm,and thestability
ofthe KIVA scheme isimproved by making use ofan acousticsubcycling
stepin order to alleviate
thestiffness
problems arisingfrom compressibility
effects.
There appears to
be considerableroom forimprovement in the code, sinceitneglectsthe spatialgradientswhenever
the gridspacingbecomes smallerthan some predefinedvalue and alsorequiresexcessiveCPU time
when the enginespeed becomes small.Shih et al.s presentedthe first
two-dimensionalcomputations
of a motored Wankel engine in the absence of combustion. Their code, LEWIS-2D, is based on
the Beam-Warming type of ADI method. Their computations have subsequentlybeen extended to
three dimensions in Steinthorssonet al.
9 Linear stability
analysishas shown that the ADI method
is unconditionallystablein two dimensions but is unconditionallyunstable in three dimensions.
Although artificial dissipation has some stabilizing effect, an excessive amount can impair stability
and reduce accuracy and convergence. Recently, Li et al.lo have modified their LEWIS-3D code
based on upwind schemes together with the incorporation of a k - _ turbulence model.
The present solution procedure differs from both REC-3D-FSC-86 and LEWIS-3D in many ways
in terms of the numerics, and also the submodels used for turbulence, combustion, and sprays.

II. Physical Description

A schematic of the Wankel engine is shown in Fig. 1. The Wankel engine is composed of a
peripheral housing with provisions for the intake and exhaust ports, fuel injector and spark igniter,
a three-flank rotor, and a crank shaft. The contour of the inner surface of the outer casing of the
Wankel engine is composed of a two-lobe peritrochoid. _I The contour of a rotor revolving along
an outer housing is represented by a peritrochoid inner envelope. The geometric analysis of the
rotor and housing surfaces c_ be found in Yamamoto. 11 The rotor surface is further modified by
the formation of a rotor pocket ` The presence of a rotor pocket not only alters the expansion
and/or compression ratio of the engine but also plays an important role in modifying the fl0wfield,
and mixing and combustion characteristics of the combustor. The present rotor configuration is
adopted from Steinthorsson et al.0

2
 The rotor turns eccentricallyat one third of the crank shaft speed. The three combustion
chambers of the Wankel engine are the three regions enclosed between the three rotor faces and
the peritrochoid housing, two side housings, two side seals, and lead and trail apex seals. In
the present calculations, only one of the three combustion chambers is considered, since leakage
through the seals is assumed to be negligible. As the rotor revolves around the crank shaft, each of
the combustion chambers is continually deformed. This produces the necessary compression and
expansion of the fluid for the required engine performance during each one of the cyclic operations.
The intake and exhaust ports, spark igniter, and the fuel injector are located along the peritrochoid
housing as shown in Fig. 1.
In the present study, the computations are initiated before the opening of the exhaust port
for the combustor formed with the second rotor flank as shown in Fig. 1. The initial conditions
correspond to the conditions of quiescent air at pressure, P_, = 1 atm, and temperature, T_,
= 300K. As the rotor moves in the clockwise direction the exhaust port opens and the residual
gas moves out of the combustion chamber, since the normally imposed pressure in the exhaust
remains lower than the interior engine pressure during most of the compression cycle. The exhaust
conditions are given by

Op OF_ 8e
On On On
P = Pe=h, u = w = O, (1)

v = --Cdc [2 (P- -pP.=h)] J°'s

where p, e, and It are the fluid density, internal energy, and mass fraction, respectively; u, v, and w
are the velocity components in Cartesian coordinates; C& (= 0.9) is the discharge coefficient; and
subscript i and n represent species and the normal component of the boundaries, respectively.
As the rotor moves further in the clockwise direction, the intake port opens and fresh air moves
into the combustion chamber. There is an overlapping region during which both the exhaust and
intake ports are simultaneously open before the exhaust port closes completely. During this process
not only the residual gas but also some of the fresh intake might escape through the exhaust port.
Most of the intake occurs during the expansion stroke of the engine. The inport conditions are
given by

P = P_.t,T = T_.,,,
yi = It_,o_,
P = Pi._,u=w=0, (2)

v =-C,_ [2 (P_nt- P)] °'5

P

3
 As the rotor turns further, the intake port closes and the liquid fuel is injected into the chamber
during the compression stroke before the top-dea_l center (TDC) is reached. Spark injection pro-
vides the initial energy needed for the early droplet evaporation and also helps to promote ignition
of the vaporized fuel and air mixture. Most of the combustion is completed as the rotor moves past
the TDC, and most of the residual combustion products are eventuMly driven out of the combustion
chamber through the exhaust port. The whole combustion performance is determined by a very
complex interaction of various engine parameters including the location of the exhaust _nd intake
ports, shape of the rotor pocket, injector and spark timings, fuel properties, and many others.

III. Gas-Phase Equations in Generalized Coordinates

The governing unsteady equationsbased on the conservationof mass, momentum, energy,and

speciesforturbulent,reacting,and compressibleflowsarepresentedinstrongconservationlaw form.
The exchanges of mass, momentum, and energy through liquid-phaseinteractionare consideredby
the inclusionof appropriatesource terms. The Reynolds-averaged equations are formulated in
generalizedcoordinatesto accommodate the time-variation
of the complex combustor geometry.
-=4 "-4 -=t "==t

aq +aF aG aH • aF,
0-7 _T+-_-_+ o---{
o_

OG,
+--_+-_-
OH,d_-"
s,+so (3)

where

p D (U + _,)

( ouipD po. (_+ _,)+PD_,

/po_/
q= /PDw/ F=

IPD_I p D • (U + _t) +P D U
! po_II
\ pD_o I p Dy! (U + _,)

p Duo (U + _,)
 . ,,.(,.+_,)+pD,.
,.,o (v+_,/

H= ,,_.(w+,,/+p.,.

.0,_(w+,,/
0
f Ll _,

1:) _= r, v + D _.v rvv .I- D _, rv,

D _, I"-=+ D _v r,, + D _, r.
_Io,+,lo,+wI,,+_(c,_+c,_+c,_,)
/,6 J
_.LTJ
, Do,,
(_.,_+o,,_+o,_)
{ gvl "
gv2

.-(, gv3
Gv-- gv4 =
gv5
gv6
 0
f hvl "_
D _= rz= + Dfv rzv + D _. rzs
Ilv2
D fz rz_ + Dfv rvv + D _. r v.
hva
D _, Tzz + Dfv rus + D _, rzs
Hv = hv4
u h,,,+ v h,,3+wh.+k( C ,.{_T_ +Cs_a-_ +C_)
hv5
[_vg
\/iv7 J
, p.-(o,_+o._ +o._)
0
0
0
0
So=

-_,.,.(_)%(_)__.(-_,
(n_z_ o._s (m_l _.5e(-¢l,)
-D Wo uo A kwl/ _,.'o/

)-_k nk mk

._,.,(h,.-t,,,,,)
>-:_
_k nk mk
0

c, = D (_.,. + _.,. + _,,,)

c, = o (_.,..+_.,. + _.,.)
c, = D(,:+,,+
C. = D (t/.g. + _/t,_', + q,_")

c, = D(¢,.+¢,_,
+¢,.)
 -
_z 2 (2o,, a,, a,,,') /

rzW

rzz

rvv
= 2
_J'k(=a,,
a_ a,,
a= aw)
rvz = _ (o;.+

2 f,,Sw au
_z
= _' V'_ a=
= Pe_ + I_t
k = k_ + pCp_
Prt
_m
Sc_p

and x, y, and z are the Cartesian coordinates in the physical space; _, r/, and _ are the coordinates
in the computational space; D is the determinant of the matrix, J in Eq. (6), and is also a
measure of the volume of a computational cell; t/_ is the mass fraction of the I'th species; g, is
a vector representing the source terms arising from the finite-rate chemical reactions; Wi is the
molecular weight of the species; u_ is the etoi_iometric ratio of the ith species participating in
a given reaction step; A and Ea are the pre-exp_nential coefficient and activation energy of a
given Arrhenius reaction-rate term; _ is a vector representing the source terms arising from the
liquid-phase interaction; nk is the number of droplets in a kth characteristic representing a group
of droplets; mk is the vaporization rate of a droplet belonging to the kth characteristic; rk is the
droplet radius; hi, and lk,ef! are the enthalpy of the fuel vapour at the droplet surface, and the
effective latent heat of vaporization; /_t is the turbo.lent viscosity; kl,. and /_lm are the thermal
conductivity and laminar viscosity of the gas mixture and are determined using Wilke's mixing
rule with fourth-order polynomial fits based upon temperature dependence12; Cp,r, is the specific
heat of the gas mixture at constant pressure and is also determined from fourth-order polynomial
fits involving temperature dependence; Prt (= 0.90) is the turbulent Prandtl number; Sot (= 0.90)
is the turbulent Schmidt number; the subscripts f, o, l, £, c, rn, and k represent fuel, oxidizer,
liquid-phase, laminar, chemical reaction, gaseous mixture, and characteristic, respectively.

The pressure and temperature are calculated iteratively from the following procedure:
 _=;
N.
e=_yih_---+
P 2
u2
+
v:
+ wz
) (4)

where

hi = h°y, + T Cp_y, dT
JT,.!

P'_A
cp_= _-_( ;_+ A:_T + As_T= + A4_Ts + As_T4)
N.
P = pP,_T _ y_ (5)
,-1

where h_ is the heat of formation of ith spec.;es, and P_ is the universal gas constant. Equation
(5) is the equation of state for a gas mixture of ?7, species. The Jscobians of the coordinate
transformation are given by

j= y_ y_ y_ (6)
z_ z, z_

and

J-;= ,t, ,Ty ,1, (7)

_ffi fv fz

The metric coefficients resulting from the coordinate transformation are evaluated from the
following identities:

1
e.,= _ ((_): - (y_z),)
1
_v = _ ((",_')_ -
1

1
 1

r/t = - (rl=:rt + rlvyt + rl,,zt)

_,= - (_x, + _y, + _._,) (8)

It is noteworthy that the following equatio_.s represent the metric invariant terms arising from the
coordinate transformation:

Dt + (D_,)_ + (D_t)._ + (Dft): = 0 (9)

(D_,)_ + (Dw=), + (Dg=): = 0 (10)

(D_)_ + (D_,). + (D_)_ = 0 (11)

(D_,)_ + (Dr/,),-t- (D_',)_ = 0 (12)

When the governing equations are formulated in strong conservation form, it is essential that
the left-hand side of Eqs. (9) to (12) vanish identically when the derivatives are approximated by
finite-differences; otherwise spurious source terms may result from geometrically induced errors, is
Equations (10) to (12) are satisfied identically when central differences are used to evaluate the
spatial derivatives. This is true since the metric identities in Eq. (8) are written in conservative
form. However, the determinant of the coordinate transformation is computed numerically from
the solution of Eq. (9) in order to avoid grid-motion induced errors, is

IV. Details of the Spray, Combustion, and Turbulence Models

Here we provide a brief description of the spray model as it is adopted from Raju and
Sirignano. 14,15 The solution of the liquid-phase equations is extended further in the present study
from the two-dimensional to three-dimensional computations. Also, several modifications are incor-
porated into the interpolation procedure between the Eulerian and Lagrangian coordinate systems
as the gas-phase computations are performed in the generalized coordinates as opposed to the
Cartesian coordinates used in Raju et al. 14,16 The interaction between the two phases is taken
into account by the following procedure: (1) In order to obtain the solution of the liquid-phase

9
equations,it isfirstnecessaryto know the gas-phase propertiesat the particlelocations.In the
present computations, the gas-phasepropertiesare evaluatedby using a second-orderaccuratein-
terpolationmethod involvingvolume-weighted averaging;(2) The ordinary differential equations
describingparticlesize,position,and velocityare solvedby the second-orderaccurateRunge-Kutta
method. The partialdifferentialequationdescribingthe transienttemperature variationwithin the
droplet interioris based on a simplifiedvortex model and issolved by an implicitmethod. The
formulation for the droplet vaporizationrate is based eitheron a simplifiedgas-phase boundary
layer analysisor on a simplifiedcorrelation,
Is depending upon the droplet Reynolds number; (3)
Finally,afterthe liquid-phaseequations are solved,the source terms evaluated at the particlelo-
cationare redistributedamongst the eightcomputational nodes surrounding the particleby using
volume-weighted averaging.
The success of the spray model also depends a great deal on the correctspecification
of the
injectorexitconditions.The injectorisan eight-holesconfigurationand islocatedalong the middle
of the peritrochoidhousing as shown in Fig. 1. The timing of the injectoropening and closingis
determined by the given engine operatingconditions.The fuelemerges in a fan shape consistingof
eight streams as shown in Figs. 5 and 6. Both the initial
dropletvelocities
and temperatures are
assumed to be known, and the dropletsizesare determined by the Rosen-Rarnlerdistribution. 16
The droplet injectiontiming is determined by the resolutionof the computational cellsused in
the gas-phase computations.14 The present model does not take into account detailsof the liquid
filamentbreakup and itssubsequent effecton the conditionsat the injectorexit.
The solutionprocedure could perhaps be improved with the considerationof dropletdispersion
due to turbulence. However, the effectsof turbulent dispersionin the modelling of combusting
sprays were found to be small in a previous study, in comparison to the uncertaintiesin the
specification
of the initial
conditionsat the injectorexit.
Iz The present model isbased on a dilute
spray approximation where the spray characteristics
are based on an isolateddropletbehaviour.
O'Rourke and Bracco,Is Greenberg and Tambour, 19and Asheim et al._0have modelled liquidsprays
includingdropletcollisions. The importance of dropletcollisionand breakup in the overallspray
behaviour is not well established,
especially
in the regionsof the spray where the dropletloading
is low. In the present computations, the effectof variablepropertiesin the liquid-phaseis not
considered,though thisfactorbecomes very important when the dropletsvaporizenear the critical
conditions.21,
22
The combustion model isbased on an analogoustreatment of laminar diffusion
flameswith the
assumption that no envelope flame exists.The reactionrate isdetermined based on a single-global
kineticmechanism of Westbrook and Dryer.2s For n-decane,the kineticmechanism isgiven by

C10irlr22
-Jr-
15.5(O2+ 3.76N2) --*10CO2 + 111120 + 58.28N2 (13)

By assuming equal binary diffusivities

for allthe speciesin the mixture, the concentrationsof
N2, C02, and 1120 can be determined from simple algebraicrelationships
based on the atomic

10
balance of the constituent species, once the mass fractions of fuel and oxidizer are known from the
solution of the two gas-phase equations based on the conservation of fuel and oxidizer.

YH20 -- K2 - I_IK2yo2 - K2YC,oH2_

Yco2 = K2K3 - K, K2K3yo2 - K2Ksyc,oH22 (14)

I/N2 = 1 - K2 - K2K, - yo2(1 - K,K_ - KxK2K,) - Ye,oH,2(1--K2 - K2Ks)

where K, = 4.29, K2 = 0.08723, and Ks = 2.222. Note that the effect of turbulence on the
reaction rate can be very important, but is not considered in the present solution procedure since
a realistic model is not currently available. In the near future, we are planning to implement the
turbulence-reaction model used in Raju and Sirignano 14'1s which is based on the eddy break-up
model of Spalding 24 requiring the solution of an additional equation involving the square of fuel
concentration fluctuations. The eddy break-up model provided some useful results in the modelling
of premixed flames, however, its appl;cabiiity in a spray environment is uncertain.
The turbulence model used is a constant eddy viscosity model of Steinthorson et al. 9 where the
turbulent ditfusivity is given by

Pt = aTf_P (15)

and a T is a function of the crank angle O, and fl is the crank speed. One obvious discrepancy of
this model is its failure to satisfy the condition of/Jr - 0 at the walls. ° The k - e turbulence model
of Launder and Spalding 25 will soon be incorporated into our solution procedure.

V. Details of Flux Vector Splitting

The present finite-difference formulation is based on an upwind scheme because of its superior
numerical stability, and efficiency properties compared to those of a centered difference scheme. 26
The most widely used flux vector splitting methods are those of Steger and Warming, z° van Leer, 2_
and Roe. 2s Recently, there is a considerable interest in extending these methods for the modelling of
reactive flows to solve problenm emerging from the design of the National Aerospace Plane (NASP),
and Air-assisted Orbital Transfer vehicles (AOTV). _9's° Details of the generalization of the Steger
and Warming flux vector splitting for a perfect gas mixture with variable properties are presented
in this section. Since the flux vector F(Q) of Eq. (1) retains its homogenous property for the
equation of state considered, the flux vector can be split into two parts,

F = F÷ + F- (16)
where F + is the subvector associated with the non-negative eigenvalues of A, F- is the subvector
associated with the non-positive eigenvalues of A, and A is the Jacobian matrix, aL_.The eigenvalues

11
of the matrix, A are given by

A4 -- U + _, + aA_ (17)

A5 = U + _= - aA_

21 1(o_
where_ = (_ + _; + _.),, = - V_,o..j _, i, the sp_d ofsound,_,, = _g'l y,C_, _._ =
Cp,n - R, and /} = Ru _i=l
N, _s;. It can also be shown that

F + = MA+M-1Q,, F- = MA-M-IQ (18)

where the diagonal elements of the matix, A±, are given by

_ = _ = _ = _ = A_= 1(u+ _,+ lu + _,1)

_= 1 (U + _, +aA_ + [U + _, + aA_[) (19)

_=
and

1 0 0 0 0
,, 6,,p o _(,, + 6==) _(,,- 6==) o o
,, -6=: _,p _(,, + 6,=) _(,,-6,=) o o
,,, o -6: _(,,, + 6:) _(w- 6:) o o
- _o_T p(6:, - 6:) p(6,,,- 6,,,,,)_(h + 0=) :;_2=(h-0=) ,_,1 _,_,2
Yl 0 0 p o
Y2 0 0 2__
.,/3= o p
M -1 =

12
 _-_+&o_ _ _h -/d_ o o o

_ h - _-_--_ o o o

---x-,[oo , {&a _ {6,_ _[_,,,_

"
+C'r- 1)ho] -(,-i- 1)u] -(_- 1)v] -(-_- 1)w]

! " 1
_[__ __1 _ _ -_ -_hl -_ - __:x-
+C-y- 1)_o] -('r- 1),,1 -(-1- 1),,] -('r- 1)_1
-_ 0 0 0 0 1, 0
p P
-_ 0 0 0 0 0 1
p #

and

_o = _o+ f _oaT- _a:o¢_.T-1(_' + _' + _2),

Z Tr,! l_u
N

and also "t = Cpm/Cv,n, it = a-_' _= a_' _r= a-_¢' and O = _zu+ _'_v+ _w.

The resulting components of the split fluxes F + are given by

13
 (2o)

In this section, we have presented the derivation of the the split fluxes associated with the flux
vector F(Q). The corresponding split fluxes associated with the vectors G(Q) and H(Q) can be
derived in a similar way.

VI. Details of the Numerical Method

Solution for the gas-phase equations is obtained by making use of a finite-volume, Lower-
Upper (LU) decomposition scheme. The governing equations are linearized in a delta form where
the nonlinear terms associated with finite-rate chemistry and convection are treated implicitly,
while the diffusion terms and the source terms arising through liquid-phase interaction are treated
explicitly. The time-linearized governing equations in delta form are

= -Q"AD + Aff (21)

where A + = _' B+ = _aQ ' C+ = _aQ ' L = _aQ, At is the time step size, 6 + and 6- are forward
and backward differences, respectively, and

(22)

14
 Upon factoringEq. (20)we obtain the followingsequence:

(23)

(24)

It is noteworthy that to be consistentwith the objectiveof deriving a finite-volumecode,

the split-flux
differences
in Eq. (21) are implemented accordingto Monotone Upstream-Centered
Schemes for Conservation Laws (MUSCL)-type differencing, st's2The fluxesat the cellfaces are
first
obtained by a fullyupwind first-order
accurateinterpolation,
and then centereddifferencesare
used for both the forward and backward spatialoperatorsevaluatedat the cellcenters.Centered
differences
are alsoused forevaluatingthe spatialoperatorsassociatedwith the viscousterms.
For the dynamic grid calculations,
the metric quantitiesare evaluated at time leveln+l, and
D n+1 is evaluated from the solutionof Eq. (9)by using an explicitmethod. The numerical grid
is generated by an algebraictechniques with the help of the grid-generation
code taken from the
LEWIS-3D code.9
By adopting an algorithm taken from the RPLUS-3D code,12 the present code is vectorized
ratherefficientlyby operatingon allpointsin a diagonalplane of the computational space,simul-
taneously.The diagonalplane isone on which i+j+k = constant.The integrationproceedsduring
the backward and forward substitutionsteps from one corner node of the computational grid and
ends at a corner node which isfarthestfrom the initial
cornernode.
The boundary conditionsare implemented explicitly
by defininga layerof phantom cellsoutside
the boundaries of the computational domain. Itisalsonoteworthy that the left-handsideoperators
of Eqs. (23) to (24) requireblock diagonalinversions.

VII. Results and Discussion

Here we present the resultsof our preliminarycomputations for a singlecase corresponding

to the operating conditionslistedin Table 1, where subscriptso, c, r, and h representopening,
closing,rotor,and housing, respectively,
V_,iis the initial
droplet velocity,
and Tg,iis the initial
droplettemperature. Isothermalwall conditionsare implemented in the case considered.

15
 Table 1. Operating Conditions
Engine Parameters Generating Ra_iius(R)= 0.1064 m
(see Fig. 1) Eccentricity(E)= 0.01542 m
Clearance(C) = 0.004 m
Chamber Width(W)= 0.07 m
Port Width(Wp)= 0.05 m
Engine Speed 4000 rpm
Intake Port 00 = -1.26 rad, 0v = 5.96 rad,/_ = 1.25 atm,
Tint = 300 K, YI,_,, = 0
Exhaust Port 0o = -5.96 rad, 0c = 1.07 rad, Pezh = 0.85 atm
Fuel Injector 00 = 8.3rad, 0v = 8.75 tad,Vdj = I00 m/s,
Td,_= 300 K
Spark ignition Timings 0o = 8.35 rad, 0c = 8.475 rad
Temperature of Rotor
and Housing Surfaces Th = T, = 400 K

The computations are performed with a variable time-step corresponding to a maximum CFL
number of 20, and on a grid with a mesh size of i=31, j=16, and k=20, where i, j, and k represent
the coordinate surfaces in the direction extending from the trailing-edge surface to the leading-edge
surface of the combustor, from the rotor to housing surface, and from the side wall to the symmetry
plane of the domain between the end-to-end side walls, respectively.
As described in Section II, the computations are initiated before the opening of the exhaust port,
and are terminated after the completion of combustion process. Figure 2 shows the variation of
the engine volume versus crank angle. The indicated engine volume is obtained from the numerical
integration of the individual computational-cell volumes. The figure demonstrates the ability of
the numerical method to accurately reproduce the combustor volume changes corresponding to
a maximum and minimum displacement volume of 750 c.c. and 115 c.c, respectively, yielding a
compression ratio of about 6.5.
Figure 3 shows flow patterns (particle traces) during the middle of the exhaust (Fig. 3a), of
the intake (Fig. 3b), and of the simultaneous opening of both exhaust and intake ports (Fig. 3c)
together with the schematic of a Wankel engine (Fig. 3d). The particle traces are coloured based
upon the local value of the internal energy e. Fig. 3a shows clearly that the bulk fluid motion of the
residual gas is essentially directed out of the engine chamber through the exhaust port. The fluid
motion is clearly seen to be influenced by the combined effect of the positive pressure difference
that exists between the chamber and back pressure, and of the rotor motion on the fluid near the
rotor surface. Figure 3c shows that while part of the residual gas is escaping through the exhaust
port, fresh air is entering the chamber through the intake port. Because of the difference that exists
between the intake and exhaust pressure (1.25 and 0.85 atm), part of the fresh air emerging from
the intake is drawn towards the exhaust before it recirculates in the rotor pocket.

16
 Figure 3b shows a complex flow patterncreated by a strongjetof freshairemerging from the
intakein a crossflow. The cross flow is createdby the clockwisemovement of the rotor surface.
The flow pattern revealsthe existenceof two clearlydefinedrecirculation
(low-pressure)
regionson
both sidesof the intake along the ith direction.Upon the impingement of the intakejet on the
rotor surface,the jet,near the leadingedge, ispartlydrawn intothe low-pressureregion,and part
of the fluidiscarriedover backwards through the openings between the outer edge of the jet and
the side-walls.
Figure 4, which issimilarto Fig. 3b, shows an experimentalflow-visualization
resultobtained
by Hamady et al._ using a transparent-sidedrotary engine motoring testrig. _Microballoon_
seeding materialwas illuminatedby a pulsed laserlightand recordedby high-speed photography.
At the leading edge of the jet,a recirculatoryflow pattern quite similarto that shown in Fig.
3b is clearlyseen. Although quantitativecomparison has not yet been attempted, the degree of
qualitativeagreement noted here isquite encouraging.The trailing
vortex,which isclearlyshown
in Fig. 3b, isbeyond the fieldof view of Fig. 4.
The droplet trajectories
at 0 = 8.5 and 8.75 tad are shown in Figs. 5 and 6. The polydisperse
characterof the spray is representedby different sized circleswhich are indicativeof the sizeof
the initial
droplets.The initial
dropletsizesrange between 10 #m __ rkj __ 30 pro, and the initial
dropletReynolds numbers vary between 75 __ Rekj _ 600. The wide disparityin Rekj isa resultof
the steep risein the chamber pressuredue to combustion from 6 to 35 atm during the time of fuel
injection.Because of the largeinitial
momentum associatedwith these particles,they retaintheir
initial
path as describedby theirinitialconditions.It takes about 1.5ms for the largestparticles
to vaporize.The deflection
of the particles
in the directionof the gaseous flowisevident from Fig.
6 as the dropletsbecome smallerdue to evaporation.
The temperature distributionwithin the combustion chamber during the earlystagesof flame
propagation at 0 = 8.5 rad, and aftercombustion at 0 -- 10 rad is shown in Figs. 7 and 8,
respectively.In Fig. 7, the highest temperature region (2900 K) is confinedto the region near
the rotorpocket, where the liquidfuelisinjected.The temperatures are lower near the wallsand
in the clearanceregionsnear the leading and trailingapex seals,where the heat transferto the
walls isgreatestbecause of the high surface-to-volumeratio.In Fig. 8, the highest temperature
regionextends allthe way from the regionnear the rotor pocket to the region near the leading
apex seal,while the temperatures are lower in the regionnear the trailingapex seal.During the
expansion stroke,the region near the leadingapex sealbecomes wider, causing a decreasein the
heat transferrate to the walls in that region,while heat transferto that regionwithin the chamber
interiorincreasesdue to convectionof the fluidas influencedby the rotor rotation.An opposite
trend is observed in the regionnear the trailingseal.
The gaseous-fuelmass fractioncontours at 0 = 8.5 and 8.75 rad are presented in Figs. 9
and 10. It isnoteworthy that in presentingsome of the resultsinvolvingthe iso-contourlinesof
fuelconcentrationin Figs. 9 and 10, and alsoof pressurein Fig. 13, some of the contour lines
representingthe near-maximum valuesare shown in dotted lines.Thus, Figs. 9-10 show that the

17
regionnear the fuelinjectorlocationisfuelrich.Diffusionof the fuelconcentrationwith time and
the influenceof convection on the distribution
of the fuelconcentrationmore towards the leading
regionisalsoevidentfrom the comparison of thesefigures.While the stratified charge givesriseto
a diffusionflame, a carefulexamination alsorevealsthat part of the evaporated-fueland oxidizer
mixure burns likea premixed flame. This is demonstrated by an absence of fuel concentration
in the region near the rotor pocket surfacewhere the fuelconcentrationresultsotherwise from
the presence of liquidfuelin that region,as shown in Figs. 6 and 7. It is more likelythat the
combustion characteristicsin that regionmight be influencedby an isolated-combustingdroplet
behaviour.
Fig. 11 shows the angular variationof the amount of theevaporated-fueland alsothe amount of
reacted-fuel.The resultsareobtained by integratingthe sourceterm contributionsof the gas-phase
equations arisingfrom the production of fueldue to evaporationand also from the consumption
of reactantsdue to combustion. It isnoteworthy that the totalamount of liquidfuelinjectedis
determined based on an equivalenceratioof 0.7.The resultsshow that the totaltime forcomplete
vaporizationand alsocombustion islessthan 2 ms. The slopeof thesecurvesindicatesthat most of
the fuel,afteritevaporates,reactsquicklywith oxidizerto form products.This in turn impliesthat
most of the fuelburns in a premixed-flame environment. Since the vaporizationrate and ignition-
delay characteristics
of thismodel were not known in advance,however, the fuelinjection
and spark
timings were arbitrarilychosen to be 66° and 63 °, respectively,
before TDC. These conditions,
which in retrospectare clearlynon-optimal,correspond to very advanced fuelinjectionand spark
timings. Under more optimal engine operatingconditions,the fuelinjectionevidentlysho_.:_d not
begin before30° to 45 ° from the TDC. Both the vaporizationand combustion characteristics
might
be quite different
ifthese timings are chosen according to optimized operatingconditions.Future
work willaddress the optimizationof these timings in terms of the overallcombustion behaviour.
Note alsothat because of theseadvanced timings,additionalenergy has to be suppliedforthe work
to be performed during the remainder of the compression processfrom the time aftercombustion
to the time beforethe TDC isreached.
Figure 12 shows the velocityvectorplotsat fourdifferent
crank angles:Fig. 12a at the beginning
of fuelinjection,
Figs. 12b and 12c during combustion, and Fig. 12d aftercombustion. In these
velocityvectorplots,only threedifferent
sizesof arrow symbols areused to distinguishthe variation
between the maximum and minimum valuesin magnitude. These plotsindicatethat the direction
of fluidmotion near the symmetry plane is mainly determined by the rotor motion; however, for
a briefperiod during the early stagesof flame propagation as shown in Fig. 12b, the expanding
gases do create a motion in a directionoppositeto the main bulk flow. The non-uniformity in
pressurebetween the leading and trailingregionsas shown by the pressurecontours in Fig. 13
betterexplainsthe reason forthe strongbulk fluidmotion created by the rotor movement. Recall
that,as in Figs. 9 and 10,some of the near-maximum contoursare shown in dotted lines.These
figuresdo show a more or lessgradual decreaseof pressurefrom the trailingto the leading apex
seals,and the pressuredecreasesin the directionof the fluidmotion. The higher pressuredrop, as

18
expected from the substantialfrictionlossesin the regionscloserto the apex seals,is alsoevident
from these figures.

VIII. Concluding Remarks

We have presented a descriptionof a new computer code developed for the modelling of
stratified-charge
rotaryengineperformance based on thesolutionof the unsteady,three-dimensional
Navier-Stokesequations,with the use of convenient submodels for turbulence,combustion, and
sprays. The detailsof the rotary engine flowfieldduring exhaust and/or intake processesand
compression stroke,and also the detailsof the mixing, vaporizationprocessesduring and after
combustion have been presented fora singlecase with advanced fuelinjectionand spark ignition
timings.The salientfeaturesof thiswork are summarized below:

1. The code takesapproximately 3 CPU-hours, when the calculations

are performed on a gridwith
a mesh sizeof 31x16x20 on a CRAY Y-MP, for a non-reactingcase,and it takes about 7.5 to 10
CPU-hours fora reactivecase with sprays.For the non-reactivecase,the solutioncan be marched
in time non-iteratively,
but, fora reactivecase,the solutionisobtained by an iterative
procedure.

2. One apparent findingof our study isthat vaporizationappears to be more rate-controllingthan

mixing during the combustion process,at leastin the case that we have studiedwith advanced fuel
injectionand spark ignition.

3. There is a good degree of qualitativeagreement between the predictionand an experimental

flow-visualization
pattern of Hamady et al.
s3 obtained during the intakeprocess.

4. The presentsolutionprocedure makes use of an extremely simplifiedconstant diffusivity

turbu-
lence model. The k - _ turbulencemodel of Launder and Spalding_5 willsoon be incorporatedinto
our solutionprocedure.

5. The present combustion model is based on laminar kinetics.While recognizingthe fact that
no realistic
model would be availablein the foreseeablefuturefor a proper treatment of the ef-
fectof turbulence on combustion reactionrates,the eddy break-up model of Spalding_4 willbe
incorporatedinto our solutionprocedure in order to account forsome effectof turbulence on the
combustion processes.

6. After implementing the above-mentioned modificationswe willconduct a parametric study in

order to optimizethe locationand alsotiming of the fuelinjectorand spark igniter.

19
 Acknowledgements

The work of the first author (MSR) was supported under contract NAS3-26266 from the NASA
Lewis Research Center. The authors would like to thank Dr. J.-S. Shuen and Mr. J.M. Barton for
some helpful discussions held during the course of this work.

References

1. Danieli, G.A., Keck, J.C., and Heywood, J.B., "Experimental and Theoretical Analysis
of Wankel Engine Performance," SAE Paper 780416, 1978.
2. Roberts, J.A., Norman, T.J., Ekchian, J.A., and Heywood, J.B., "Computer Models for
Evaluating Premixed and DISC Wankel Engine Performance," SAE Paper 860613, 1986.
3. Bracco, F.V. and Sirignano, W.A., gTheoretical Analysis of Wankel Engine Combus-
tion," Combustion Science and Technology, Vol. 7, No. 3, 1973, pp. 109-123.
4. Grasso, F., Wey, M.-J., Bracco, F.V., and Abraham, J., "Three-Dimensional Computa-
tions of Flows in a Stratified-Charge Rotary Engine," SAE Paper 870409, 1987.
5. Abraham, J., Wey, M.-J., and Bracco, F.V., "Pressure Non-Uniformity and Mixing
Characteristics in a Stratified Charge Rotary Engine," SAE Paper 880604, 1988.
6. Abraham, J. and Bracco, F.V., "Comparisons of Computed and Measured Pressure in a
Premixed-Charge Natural-Gas-Fueled Rotary Engine," Rotary Engine _ Analysis
and Developments, SAE SP-768, SAE, Warrendale, PA, 1989, pp. 117-131.

7. Amsden, A.A., Ramshaw, J.D., O'Rourke, P.J., and Dukowicz, J.K., "KIVA: A Com-
puter Program for Two- and Three- Dimensional Fluid Flowsw_th Chemical Reactions
and Fuel Sprays," LA-10245-MS, Los Alamos National Laboratory, Feb. 1985.
8. Shih, T. I.-P., Yang, S.-L., and Schock, H.J., "A Two-Dimensional Numerical Study of
the Flow Inside the Combustion Chamber of a Motored Rotary Engine," SAE Paper
860615, 1986.
9. Steinthorsson, E., Shih, T.I-P, Schock, H.J., and Stegman, J., _Calculations of the Un-
steady, Three-Dimensional Flow Field Inside a Motored Wankel Engine," SAE Paper
880625, 1988.
•10. Li, T., Steinthorsson, E., Shih, T. I.-P., and Nguyen, H.L., "Modeling & Simulation of
Wankel Engine Flow Fields," SAE Paper 900029, 1990.
11. Yamamoto, K.,Rotary Engine, Tokyo Kogyo Co., Ltd., Tokyo, Japan, 1969.

2O
12. Yu,S.-T.,Tsal,Y.-L. P.,and Shuen, J.-S., _Three-Dimensional Calculation of Supersonic
Reacting Flows Using an LU Scheme," AIAA Paper 89-0391, 1989.
13. Thomas, P.D. and Lombard, C.K., aGeometric Conservation Law and its Application
to Flow Computations on Moving Grids," AIAA Journal, Vol. 17, No. 10, Oct. 1978,
pp. 1030-1037.
14. Raju, M.S. and Sirignano, W.A., _Spray Computations in a Centerbody Combustor,"
The 1987 ASMFrJSME Thermal Engineering Joint Conference, Vol. 1, P.J. Marto and
I. Tanasawa, Eds., ASME, New York, 1987, pp. 61-72.

15. Raju, M.S. and Sirignano, W.A., "Multicomponent Spray Computations in a Modified
Centerbody Combustor," AIAA Paper 88-0638, 1988.

16. Kuo, K.K., Priniciples of Combustion, John Wiley & Sons, Inc., New York, 1986.
17. Faeth,G.M., _Mixing, Transport and Combustion in Sprays," Progress in Energy and
Combustion Science, Vol. 13, No. 4, 1987, pp. 293-345.

18. O'Rourke, P.J. and Bracco, F.V., _Modeling of Drop Interactions in Thick Sprays and a
Comparison with Experiments," Stratified Charge Automotive Engineering Conference,
Institute of Mechanical Engineering, London, England, 1980, Paper C404.

19. Greenberg, J.B. and Tambour, Y., aFar-Field Coalescence Effects in Polydisperse Spray
Jet Diffusion Flames," Twenty-First Symposium (International) on Combustion, The
Combustion Institute, Pittsburgh, 1986, pp. 655-663.

20. Asheim, J.P., Kirwan, J.E., and Peters, J.E., _Modeling of a Hollow-Cone Liquid Spray
Including Droplet Collisions," _ of Propulsion and Power, Vol.4, No.5, Sept.-Oct.
1988, pp. 391-398.
21. Chiang, C.H., Raju, M.S., and Sirignano, W.A., UNumerical Analysis of Convecting,
Vaporizing Fuel Droplet with Variable Properties," AIAA Paper 89-0834, 1989.

22. Haywood, R. and Renksizbulut, M., "On Variable Property, Blowing and Transient
Effects in Convective Droplet Evaporation with Internal Circulation," Heat Transfer
1986, Vol. 4, C.L. Tien, V.P. Carey, and J.K. Ferell, Eds., Hemisphere Publishing
Corp., New York, 1986, pp.1861-1866.
23. Westbrook, C.K. and Dryer, F.L., _Chemical Kinetic Modelling of Hydrocarbon Com-
bustion," _ in Energy an__dd
Combustion Science, Vol. 10, No. 1, 1984, pp. 1-57.
24. Spalding, D.B., "Mathematical Models of Turbulent Flames: A Review," Combustion
Science and Technology, Vol. 13, Nos. 1-6, 1976, pp. 3-25.
25. Launder, B.E. and Spalding, D.B._Le, II_,i.c_I Models of Turbulence, Academic press,
London, 1972.

21
26. Steger, J.L. and Warming, R.F., aFlux Vector Splitting of the Inviscid Gasdynamic
Equations with Application to Finite-Difference Methods," Journal of Computational
Physics, Vol. 40, No. 2, Apr. 1981, pp. 263-293.
27. van Leer, B., UFlux-Vector Splitting for the Euler Equations, m 8th International
Conference on Numerical Methods in Fluid Mechanics, (Lecture Notes in Physics, Vol.
170), Springer-Verlag, 1982, pp. 507-512.
28. Roe, P.L., _Approximate Riemann Solvers, Parameter Vectors, and Difference Schemes,"
Journal of Computational Physics, Vol. 43, No. 2, Oct. 1981, pp. 357-372.
29. Liu, Y. and Vinokur, M., _Nonequilibrium Flow Computations. I. An Analysis of Nu-
merical Formulations of Conservation Laws," Journal of Computational Physics, Vol.
83, No. 2, Aug. 1989, pp. 373-397.
30. Shuen, J.-S., Liou, M.-S., and van Leer, B., aInviscid Flux-Splitting Algorithms for Real
Gases with Non-Equilibrium Chemistry," Journal of Computational Physics, in Press,
1989.

31. Belk, D.M., UUnsteady Three-Dimensional Euler Equations Solutions on Dynamic

Blocked Grids," Ph.D. Thesis, Mississippi State University, 1986.
32. Anderson, W.K., "Implicit Multigrid Algorithms for the Three-Dimensional Flux Split
Euler Equations," Ph.D. Thesis, Mississippi State University, 1986.

33. Hamady, F., Schock, H., and Stueckon, T., _Airflow Visualization and LDV Measure-
ments in a Motored Rotary Engine Assembly-Part 1: Flow Visualization," SAE Paper
900030, 1989.

22
 Side Housing

/
X

Rotor Housing

i/
,,,.;///
=/ ,/ ,/ /:/ /X/W//////
I Z

Apex SeaJ
lY

Exhaust
\
Port Port
_-_ Intake Rotor
Housing

Rotor
Edge t
3

Apex
SeaJ Apex
Sead

t_- ""l _ Fuel Injector

L ( Also Spark Ignitor)

Fig. 1 Schematic of the Wankel engine that was studied.

23
 750 750

700 - 700

650 650

600 600

550 550

500 500

U
U

U
450 450 u

E 400 400
D
D

350
350 o>

300 300

250 250

200 200

150 150

100 I I i I • I J I I I _ I , I_ i I i 100
-6 -4 -2 0 2 4 6 8 10 2

crank angle (radlans)

Fig. 2 Variation of the engine volume versus crank angle.

24
 OR;G"'"" m,,,"_r
COLOR PHOTOGRAPH

25
 ORIGINAL PAGE
BLACK AND WHITE PHOTOGRAPH

Fig. 4 Experimentally obtained flow pattern during intake

at 1170 rpm in a supercharged rotary engine.

26
 DROPLET TRAJECTORIES'
CRANK ANGLE = 8.5 RAO, K=19

DROPLET TRAJECTORIES,
CRANK ANGLE= 8.5 RAD, 1:15
O J

DROPLET TRAJECTORIES,
CRANK ANGLE= B.5 RAD. J= I

Fig. 5 Droplet trajectories at 8 -- 8.5 rad.

27
 CRANK ANGLE = 8.7 RAD. K:19
DROPLET TRAJECTORIES,

© ©

© ©

°_i_° DROPLET
CRANK ANGLE: TRAJECTORIES,
8.7 RAD. 1=15

DROPLET TRAJECTORIES,
CRANK ANGLE= 8.7 RAD. J= I

Fig. 6 Droplet trajectoriesat 8 -- 8.75 rad.

28
 ORIGINAL PAGE
COLOR PHOTOGRAPH

29
 ORIGINAL PAGE
COLOR PHOTOGRAPH

II

3O
 FUELCONCENTRATION
CCNTOURS, //
CRANK ANGLE= 8.5 RAg. K=19 / /

CMAX:_

FUEL CONCENTRATION CONTOURS,

CRANK ANGLE= 8.5 RAD. I=16
CMAX= 0.0781,CMIN= 0.0000,0C= .00A

mllII1

FUEL CONCENTRATION CONTOURS,

CRANK ANGLE= 8.5 RAD. J= 7
CMAX= Q.0290,CMIN= 0.0000.DC= .002

Fig. 9 Fuel mass fraction contours at 8 = 8.5 rad.

31
 FUEL CONCENTRAT[ON CONTOURS,
CRANK ANGLE= 8.7 RAO. K=19 ////
CMAX= . , = . . = ,

FUEL CONCENTRATION CONTOURS,

:RANK ANGLE= 8.7 RAD. I=15
e.1017.CMIN= 0.0000.1]C= .005

FUEL CONCENTRATION CONTOURS,

CRANK ANGLE= B.7 RAD. J= 7
CMAX= 0.0355.CMIN= 0.0000.DC= .002

Fig. I0 Fuel mass fraction contours at 6 = 8.75 rad.

32
 .040 I .040
I I I t I i 1 I I r I ' i ' I _ T -T- I '

.035 .035

.030 •030
E
E

.025 • 025 -_
N
E
(_
0
O. C
11 I/ 0
,'I}
> u
.020 .020
i I
3
:4-
#
o
O
.015 .015
--vaporlza[Ion

O O
E E
q_ ..... con, bu_tlon
.010 .010

.005 .005

0 I
| I Jl I I I I I I I I I ] I I ] 0
8.2 8.3 8.4 8.5 8.6 8.7 8.8 8.9 0.0 g.1

crank angle (radians)

Fig. 11 Variation of amounts of fuel evaporated and

reacted versus crank angle.

33
 V_'LOC_TY
vECTOR
P_.OT, (a) .h

._.__ M e _ _ __-
__.....v

VELOCITY VECTOR PLOT, ['1-.,1 ._

CRANK ANGLE= 8.5 RAD. K=19 k'-"/ _.7'
'-" •

VELOCITY VECTOR PLOT, / \ _I_

CRANK ANGLE: B..7 RAD. K:lg I_C)

VELOCITY VECTOR PLOT,

(d_ CRANK ANGLE: 10.e RAO, K:Ig

Fig. 12 Velocity vector plots.

34
 m_mum---_
(a] PRESSURE CONTOURS, _ /L_
CRANK ANGLE = 8,3 RAD, K= 19 / /
PMAX= .57E+g6,PH[N=

(hi PRESSURE CONTOURS,

CRANK ANGLE= 8.7 RAD, K= 19 //')

(c) PRESSURE CONTOURS,

CRANK ANGLE= 10._ RAD, K= 19

Fig. 13 Pressure contours.

35
 i

National Aeronautics amd Report Documentation Page

Space Administration
2. Government Accession No, 3. Recipient's Catalog No.
1. Report No.
NASA TM- 102469
AIAA-90-0643
4, Title and Subtitle 5. Report Date

Analysis of Rotary Engine Combustion Processes Based on

Unsteady, Three-Dimensional Computations 6. Performing Organization Code

7. Author(s) 8. Performing Organization Report No.

M.S. Raju and E.A. Willis E-5249

10. Work Unit No.

505-62-51
9, Performing Organization Name and Address
11. Contract or Grant No.

National Aeronautics and Space Administration

Lewis Research Center
Cleveland, Ohio 44135-3191 13. Type of Report and Period Covered

Technical Memorandum
12. Sponsoring Agency Name and Address

National Aeronautics and Space Administration 14. Sponsoring Agency Code

Washington, D.C. 20546-0001

15. Supplementary Notes

Prepared for the 28th Aerospace Sciences Meeting sponsored by the American Institute of Aeronautics and
Astronautics, Reno, Nevada, January 8-11, 1990. M.S. Raju, Sverdrup Technology, Inc., NASA Lewis Research
Center Group, Cleveland, Ohio 44135; E.A. Willis, NASA Lewis Research Center.

16. Abstract

A new computer code has been developed for predicting the turbulent, and chemically reacting flows with sprays
occurring inside of a stratified-charge rotary engine (SCRE). The solution procedure is based on an Eulerian-
Lagrangian approach where the unsteady, three-dimensional Navier-Stokes equations for a perfect gas-mixture
with variable properties are solved in generalized, Eulerian coordinates on a moving grid by making use of an
implicit finite-volume, Steger-Warming flux vector splitting scheme, and the liquid-phase equations are solved in
Lagrangian coordinates. Both the details of the numerical algorithm and the finite-difference predictions of the
combustor flowfield during the opening of exhaust and/or intake, and also during fuel vaporization and combus-
tion, are presented.

18. Distribution Statement

17. Key Words (Suggested by Author(s))

Rotary engine Unclassified- Unlimited

Combustion Subject Category 34
Fuel injection
Computational fluid dynamics

20. Security Classif. (of this page) 21. No. of pages 22 Price*
19. Security Classif. (of this report)

Unclassified Unclassified 36 A03

i

NASA FORM 1626 OCT 86

*For sale by the National Technical Information Service, Springfield, Virginia 22161