.f_ / i ._f

NASA Contractor Report 195478AIAA-95-2968

lfj i I -_.J

A Transient Model of the RL 10A-3-3A

Rocket Engine

Michael E Binder

NYMA, Inc.

Engineering Services DivisionBrook Park, Ohio

June 1995

Prepared forLewis Research Center

Under Contract NAS 3-27186

National Aeronautics and

Space Administration

,#

o_

!

o_Z

...I

Qo,_uJ

z

ZC._uJZJ"_W

_LU

0

,,m_ i,f,1

_0

P'# p"# i_

I ..i e_i

l_i

UC

00_

o0

o

A Transient Model of the RL10A-3-3A Rocket Engine

_tchael Bind_NYMA, Inc.

2001 Aerospace ParkwayBrook Park, Ohio 44129

Abstract

RL10A-3-3A rocket engines have served as the main tm3Pulsion system for Centaur upper stagevehicles since the early 1980's. This hydrogen/oxygen expander cycle engine continues to play amajor role in the American launch industry. The Space Propulsion Technology Division at theNASA Lewis Research Center has created a computer model of the RL10 engine, based on detailed

component analyses and available test data. This RL10 engine model can imxlict the perfcmnanoeof the engine over a wide range of operating conditions. The model may also be used to predictthe effects of any proposed design changes and anticipated failure scenarios. In this paper, theresults of the component analyses are discussed. Simulation results from the new system modelare compared with engine test and flight data, including the start and shut-down transientclmtaoeristics.

1.0

The RL10A rocket engine is an important componentof the United States space infraslntctme. Two RL10engines form the main propulsion system for theCentaur upper stage vehicle, which boosts commercial,scientific, and military payloads from a high altitudeinto Earth orbit and beyond (planetary missions). TheCentaur upper stage is used on both Atlas and Titanlaunch vehicles. The initial RL10A-1 was developed inthe 1960's by Pratt & Whitney (P&W), under contractto NASA. The RL10A-3-3A, RL10A-4, and RL10A-4-1 engines used today incorporate componentimprovements but have the same basic configuration asthat of the original RL10A-I engine. RL10's havebeen highly reliable servants of America's spaceprogram for over 30 years. The RLIOA-3-3A engine isrepresented schematically in Figure 1.

The Space Propulsion Technology Division (SPTD) atthe NASA Lewis Research Center began developing acomputer model of the RL10 in 1991. This model wasintended for government use in engine system research,_-analysis and flight failure investigations. Thefirst version of the model was created using dataprovided by Pratt & Whitney, and the ROCket EngineTransient Simulator (ROCETS) I system analysis

program. This model could accurately p_llct the steady-

state performance of the RL10A-3-3A, but the predictedlime required for the engine to reach a specified thrustduring engine start (time-to-accelerate) showedsignificant differences with test data 2. It is believedthat these discrepancies were due to errors inexuapolating the available component perfmmance datato cover engine-start conditions, as well as errors in the

physical models used for heat umsfer and two-phaseflow. Analysis of each RL10 engine component wasundertaken in order to verify the origin of the dam

lXovided by P&W, and to improve the accuracy of themodels at far off-design conditions. These analyseswere also used to benchmark our ability to acowalely

model new rocket engine designs forwhich test data arenot yet available; the RLIO engine system provided testdata to validate the available component and system

modeling tools.

In this paper, the RLIOA-3-3A rocket engine and itsvarious components me described briefly. The analysisn:mlts for each component are then discussed, includingcomparisons with existing component test data. Thenew engine system model, which includes the results ofselected component analyses, is described and

pn_dictionsof me model_ mmparedto gn_-test andflight data. For a more detailed discussion of themodeling work summariz_ here, the reader is referredm

theRL10A-3-3ARocketEngineModelingProjectF'mMReport3.

As the simulation results will show, the new RL10

model ccuectly predicts variation in engine transientbehavior due to inlet conditions, initial thermal

conditioning, and ignition delay.

2.0 RL10A-3-3A Engine Description

The RL10 engine design (all models) is based oll a fullexpander cycle, as shown in Figure I. Hydrogm fuel isused to cool the thrust chamber and nozzle, and the

thermal eaezgy Wansfetmi to the coolant is used to drivethe aubopumps. Warm hydrogen gas is injected withcryogenic liquid oxygen into the comlmslkm chamberand burned to provide thrust. During enghg sum, fueltank iaessme and the initial ambient heat in the coolingjacket metal are used to start rotation on the mrbiue.After ignition, the heat of combustion is used toaccelerate the tmbopumps to full power. Because theCentaur upper stage vehicle uses two RL10 engines, itis important that the engines start simultaneously (tominimize thrust imbalances). For the purpose of

providing a quantitative measure of the engine starttimes, we shall refer to the time between the startsignal and the chamber pressure reaching 200 psla asthe t/me_/erate.

During engine shutdown, the fuel inlet, fuel shut-off,and oxidizer inlet valves are clmed. The oumbustion

process stops mul the fuel and oxidizer drain flora theengine system; LOX drains out through the thrustchamber, and the fuel drains out through the pumpcool-down valves.

3.0 Turbomachinerv Analysis

3.1 Turbonumu Background Information

The RL10A-3-3A Uubopump includes a two-stage fuelturbine which drives a two-stage fuel pump ou a

shaft, rex! a single-stage LOX pump tla-ough agear box. At the engine's normal operating point, afuel flow of 6 lb/sec is pumped to a wessure of 1100psla, and 30 Ib/sec of LOX is ptmaped to 600 psia. Thenormal operating speed of the fuel pump is 32000 rpm,and the LOX pump speed is 12800 rpm.

Pratt & Whimey provided the NASA SPTD with testdata maps of head coefficient and efficiency for each

pump stage as functions of flow coefficient, andincluded a speed correction factor f_" efficiency. Thesemaps do not cover the entire range of operating

conditions exlgdenced by the pom_ during engine tortand shutdown. P&W had also wovided the SPTD withtest data maps of turbine efficiency and flow zesistanceas fanctlom of overall pressure ratio sad velocity ratio(u/Co). These maps do cover a range of conditions

suitable for engine start and shutdown sinudatious.

3.2 Detailed Pumn AnalysesTwo different analysis codes, PUMPA 4 and LSISO 5,were used to model the RL10A-3-3A fuel and LOX

pumps. Tan pamp head coefficients _ by eachcode agree with test data to within five percent (5%)over the engine's normal steady-state operating range.The PUIvIPA and LSISO efficiency predictions,however, differed from test data by as much as fifteenpercent (15%), and could therefme not be used in theengine system model. PUMPA was also used topredict the performance of the RL10A-3-3A pumps atthe engine start conditions. The results of theseanalyses were used qualitatively to help extrapolate thehead maps beyond the available test data provided, asdiscussed later in this section. It sJz3uldbe noted that a

subsequent version of the PUMPA code was recentlydeveloped which better wedicts the RL10A-3-3A pumpdesign point efficiency, without affecting the headpredictions. The new version of PUMPA wascompleted too late to allow a comprehensive a_tlysls ofstart coalitions to be perforated again for this project.

In addition to the PUMPA and LSISO analysesabove, a third analysis was peffcmned which

was spec_xcally designed to estimate the lOW speedpump head (as experienced during start). This methodwas suggested by Rostafmski 6 and requires that thedesign point perfommuce of the pump be known. This

when cambiued with a separate model of thepump exit diffuser, appears to match well with thelimited test data available at engine start conditions.Although Im3mislng, this modeling technique provedimpractical for transient system simulation (slowexecution, numerical instabilities, etc.) and was

therefcce not included in the new RLI0 system model

Using available engine test data and informatim gainedfrom the analyses discussed above, the pumpperformance maps provided by Pratt & Whitney wereexlrapolaled to include _nditions at engine start andshutdown (zero speed, zeverse flow, ca_ etc.). Inorder to represeut such a wide range of operatingconditions, a map format suggested by Chaudl_y 7 isused. The new map format defines normalized head

parameter(h) and torque parameter (13)as functions of athird parameter, 6 (them) as described below. The new

pumpperfmmancemapsfortheengine system modelme shown in Figures 2 and 3.

h= I_=N Q + N

whe_ ffihead (in feet)

N = shaftspeed (inrpm)Q = volmnetricflow(ingpm)andthesubscriptd denotesthedesigncondition.

The results of the pump analyses descn'bed aboveindicate that it shouki be possible to predia the general

performance characteristics of new pump designs.Results from this type of analysis are valuable forconceptual engine design and simulation activities.Such component Ixedictions may not be sufficientlyaccurate for use in engine start-transient simulations,especially if no test data is available with which toanchor the new pump models.

3.3 Detailed Analysis of Fuel Turbin_

The RLIOA-3-3A turbinewas alsoanalyzedruingthe

TURBA code s, which is cmrently being developed atthe NASA Lewis Research Center. TURBA is a one-dimensional mean-line code which combines basic

physics (vdocity trianglesand isentropicrelaX'ms)with empirical cot_lations derived from existinguubine designs. The turbine perfmamnee predictiouscould notbe directly compared with the maps providedby P&W. Instead, both sets ofmaps were used asinputs to a simple turbine simulation, and the resultingoverall efficiencies and flow rates were compared.Although the overall performance trends i_edicted by1XJRBA are similar to those indicated by the P&Wdata, a more quantitative comparison shows thatsignificant differences exist. The predicted overallturbine effgietgy, for ex_m_ diffe¢_bymore than5%ffmn the P&W data, especially at low speeds. It hasbeen fmlh_ noted that relatively small variations in theturbineperfotmaige at low speeds can profoundly affectthe RL10 engine time-to-accelerate. Possible

explanations for the poor match between TURBA

output and test data have not been explored; theTURBA code is still considered to be in the

development phase. The performance maps providedby P&W have flmrefore been retained in the new systemmodel.

The turbine analysis performed in this study indicatesthat it is possible to estimate the design pointperformance of a new turbine to within a few percenLIt is also possible to predict the overall trends inperfmmance at off-design conditions. As with thepump analyses discussed above, however, the accmacyof the turbine lm_iictious may not be sufficient for usein transient or deep-throuling simulations of a newengine. When component test data is available for anew turbine design, it might be possible to adequatelyadjust the model based on only a few test data points.

4.0 Thrust Chamber and Cooline Jacket

4.1 Thrust Chamber Backtwound Informationwalls of the RL10A-3-3A thrust chamber are

consmgted of stainless steel tubing. Hydrogen fuel ispumped throegh these robes in order to cool the walls

of the thrust dmmber and provide thermal en_gy to theturbine. The robes are brazed together and reinforcedwith bands on the outside, as well as a metal girdlearound the throat sectim. A silver throat insert is cast

in place to increase the nozzle area ratio and specificimpulse. The thrust chamber normally operates at apressure of 475 psi& a mixture ratio (O/F) around 5.0,a thrust of 16500 lbf, and a specific impulse of 445seconds.

The analysis of the RLIOA-3-3A thrust chamb_ wasdivided into three basic areas: 1) cooling jacket heattransfer, 2) combustion chamber performance, and 3)nozzle performance. Each analysis is described below.

4.2l_etailfd Analysis of Coolint Jacket HeatT_mreroriginal model of the RL10 cooling jacket had only

five heat transfer cakulatm nodes distributed along thecooling circuit This model was considered to be toocoarse and amore detailed model was c_mted for this

project.

CET93, a _ equih'laium program 9, wasused to refine the table of hot-gas properties. TheRocket Thmnal Evaluator (RTE) code lo was used topredict the flow resistance of the cooling jacket and theeffects of tube curvatt_e ou heat transfer rate. Heat

transferbetween the combustion gas and chamber walls

waspnxk_ ustagan euthalpy-driven Banz correlatimI t. The euthalpy gradient was used instead of thetemperature gradient because this more accuratelypredictsvariationinheatIransferatdifferentmixtureratios. A Colbum correlatioa 12was used to detetlBi_the heat transfer from the chamber wall to the coolant

flow. It was clisc_vered that cembining twenty hot-gasand metal property nodes with five (rather than 20)coolant nodes couM significantly increase thecomputational efficiency of the transient system model

without loss of accuracy overall. This theconfigurmkm was used in the new RL10A-3-3A systemmodel.

Figure 4 shows the predicted heat flux, walltemlmauue, coolant _ and wesm_ along theaxial length of the thrust chamber cooling jacket. Testdata show_ ux_ _ ta tempenm_ and _are not available for comparismx. The accma_ of thenew heat Wausfer model can only be judged by theoverall _ rise and wesuue drop across thecooling jacket. Based on these parameters, anempirical _ of 1.08 was added for the hot-gasheat transfer coefficient and a fact_ of 0.94 applied tothe predicted jacket flow resistanoe. These empiricalcorrection factors represent average values, since theactual heat _ metlicient _ to vary somewhatfiem one RLI0 engine to another. These variationsmay be due to small mmmfactming diHe_aces; they arenot c_nsidered critical as loag as the engine hassufficient starting pow_.

A simple cme-dimeasiem_ film boiling model was alsoadded to the oooling jacket heat trm_fer model. Fdm-boiling effects have been suggested as the cause of thefour to eight Hertz pressure oscillations oftenex_ dsmg the RLI0 engine startseqne_e. Teenew model still does not show these pressureoscUlations; they may be due to two-dimensionaleffects not modeled here or to local choking within thetwo-phase fluid.

The analyses wes_ted hem demmsuale the capab_tyof one-dimensional models to pmdi_ the effects ofvarious oondifims on heat transfer. Depending on theaccuracy required for system simulations, some

adjustment to the heat transfer _ts using testdata my be required. Test results ate also useful indefining the variability in heat trm_e_ characteristicsdue to manufacturing tolerances and other factors.

4.3Detailed Analysis of CombustionChamber Performance

In addition to revising the ¢ombmtion gas propertytables f(g the new model, several other imwovementswere made. In the original RL10 engine model, thethrust chamber was treated as a _ingle point, withoutconsidering axial variation. In reality, there mem,wnmmwn losses and eJ_ager in static pressure alongthe hot-gas flow path which will affect performance.These effects were relatively simple to wedict and wereadded to the model. Au analysis of the RL10A-3-3Athrust chamb_ _bly was also perfonned usingthe

ROCCID code 13, which provides a c_pability ofmodeling the propellant injectors, atomization andcombustion processes. The objectives of this analysiswere to validate onr _ty to wedict ¢*_usingRL10 data from P&W, and to extend the range ofmixture ratios represented in that data set. The RL10injextef proved difficult to model using ROCCID;several aspects of its design are not found in the morecontemporary designs which ROCCID was intended tomodel. As a result, the results of the ROCCID

analyses did not show a good match with the P&Wdata. 2_e R_ model also _ numea'iealconvergence woblems at low pressures (below 160psla), where the c*-eff'tciency changes significantly.Tbe data maps wovided by P&W have beea retained inthe new system model

During the engine start sequmce, heat transfeg in theinjector can play a discernal_ role in the system'sdynamic behavior, prima_y by clumgingthe densityofthe injected LOX. Simple models of heat ttm_er intheinjectorekmems endinter-_t bulkheadwereadded to the new engine model. Although the_ isinsufficienttestdatato validate tbe modeis, the results

al_earmasmable.The additionofthesemodeledeffects

delaysthetime-to-accelerateby apwoximatelyI00

milliseconds,Considered over all engine start

transients simulated, this delay results in a moreaccurate predktion of time-to-accelerate. Figure 5shows the _ heat tnmsfef rate in the injoc/_ as afunction of time during a typical engine start.

4.4Detailed Analvsta of No_,.zle PerformanceThe RLIOA-3-3A nozzle perf_ affects themmbustion chamber Wesmre and flowrate, as well asthe specific impulse and _-ust of the engine. Prau&Whitney had originally ixovided nozzle perfm_ancedata in the form of specific impulse (Isp) tables withadditional corre_ons for various kinetic losses.

Analyses were performed at Lewis using a TwoD:on_l$ion_l _es (TDK) pl'ogram 14in O_ to

better understand the P&W data. Figure 6 shows theoutput of the TDK analysis compared with the P&Wdata. The results match well at the engine's normaloperating point of 475 psia and O/F = 5.0. Thewedicted and P&W values differ mote significantly atlow pressures and mixture-ratios, however. Thepredicted Isp maps have been included in the newRL10A-3-3A system model.

Several different _ were takm to detmaine the

nozzie discharge o3efficknt (CA). P&W had specified aCA of approximately 0.98. A Navief-Stokes analysis 15

was performed at Lewis which wedicted the dischargecoefficient to be 0.979, a remarkable agreement.Trimming the CA value used in the engine simulationto match lxedicted chamber laesmre with test data gavea value of 0.975. The TDK analysis descn3_ abovehad further indicated that the CA may ch_mge somewhatwith chamber pressure and mixture ratio. Afterconsidering these various results, a constant CA of0.975 was selected fog use in the new system model.

$.0 Miscellaneous Comnonents

In general, the ducts, valves, and manifolds in the RL10engine were not analyzed in detail Many of thesecomponents have complex geometries that wouldrequire f'mite-element methods to model im31gdy. Inthe case of the fuel pump cool-down valves, oxidizercontrol valve, madLOX injectar elements, however, themodels for two-phase flow contained in the originalmodel required improvement. We also attempted toverify the resistance of a single duct as specified byPratt & Whitney using generic one-dimensionalmetho&.

During the engine start, several components experiencetwo-phase critical and mrJmked flow conditions. Thefnel-pump cool-down valves, which vent liquidhydrogen overbom_ ale always clinked and thehydrogenflashes to vapor as it is vented. The oxidizer controlvalve and LOX injector elemmts experience two-phaseflow for only a small period of time during start,transitionin8 at some point between choked andunchoked conditions. The challenge was to devisemodels which allow a relatively continuous transitionbetween the various flow conditions during start.

A number of different _ were cousidered 26 J718. Ultimately, a model was derived which treats theflow as incompressible, but limits the assumeddowusueam wessure to either saturation og isenlropiccritical pw._sme, depending on the value of the pressure

upstream of the orifice or valve. This modelingapproach was used fog the LOX injector elements andfuel cool-down valves. Two-phase flow in the oxidizercontrol valve is modeled as incemweuibie, limited bythe saturation pressme of the fluid until the flow

becomes entirely gaseous, after which it is Ireated asisentropie flow of an ideal gas. These models agreewell with avaflabie test data.

The fluid resistances of ducts and tubes are typicallydetermined by flow testing those components. For

new rocket engine systems, empirical data ofthis kind may not be available during the analysisphase. A simple ol_ff-dil3_e_onal analysis19 of flow inan RLI0 duct (from the turbine discharge to the mainfuel shut-off valve) was peffmmed and the results wea'e

compared with the resistmwe _ by P&W. Tneinflate roughness on the interior of the duct was notknown, so we considered a range of options fromsmooth commercial steel pipe to drawn tubing. Theanalyses indicated a range of possible K valnes19 from0.928 to 0.487; the value of g given by P&W was0.648. Our estimates the_fore define a range of

possible values which bracket the suggested value withm each" of 25 to 43 %. The _ce provided byP&W has been retained for the new RL10 enginemodel, but this analysis suggests that we c_mwobablyestimate the resistance for a new (untested) duct towithin 4./- 30%. Better estimates might be possible ifthe surface roughness of the intended duct is welldefined.

It is evident from the discussion above that accurateone-dimmsioml models of ducts and valves in a new

engine design will require at least some flow testing.Befme such data is available, it would be prudent to

_tsider the effects of uncertainty in engine systemsimS. In the case of valves and ducts where two-

phase flow might exist, it is advisable to test thecomponents over d_Jr entire operating range, sincetwo-phase effects caa often lead to unexpec_ behavi_.Flow models which include two and three dimensional

effects may also woduce more accurate resistance

predictions.

6.0 _lew RL10A-3-3A En2ine System Model

Tne new RLIOA-3-3A engine system model includesthe results of several of the detailed componentanalyses, as described above. In addition to thesecomixagnt model changes, several improvements waemade in the structnre of the system model itself.

Tracking of the total-to-static conversions for pressure

and mthalpy was iml_ved in the new system model,

for example.

It became necessary to create two difterent models of

the RL10 engine: one for ¢imzdatlng Start transient

behavior and steady-state peff_ and the other for

simulating shut-down transient behavior. During shut-down, the ducts and manifolds in tbe engine me emptied

into space, and dynamk: volumes had to be added to themodel to allow the simulation of these effects.

Including these dynamic volumes in the start transient

and steady-state model changed the predicted start

transient behavior significantly, in disagreement withtest dam. These differences could not be resolved, and

so two separ_ modeLs w_e developed.

6.1Effects of Modelin_ Uncertainty

Before discussing the output of the new system model,

it is important to note several unresolved sources of

uncertainty in the model which will affect our ability toaccurately simulate a given RLI0 engine firing. Tlw.se

uncertainties can be divided into four categories: 1)

enceminty in hardware _ 2) _ties

in valve dynamic behavior, 3) _ties in engine

initial conditions, and 4) uncertainty in the main

chamber ignition delay. _ is also a great deal ofnon-linear interaction between RL10 engine

components 2o. Charactaizing the _,nction betweenthe various operating lXtmneters with uncertainty was

beyond the u:ope of this study.

6.1.1 Uncertainti_es in Hardware

eh_wtt,ri_tiot

There LSsome mceminty in the actml value of the

discharge coefficient for the fuel-immp cool-down

valves. In the RL10 model, the discharge

coefficimt is set at 0.6 for grom_test and 0.8 forflight. These values were chosen based on

discussims with enginec_ at Pratt & Whitney butno real caWntion data is available to verify thesevalues, l_e res/stmce of the cool-down valves is

.n important factor in the engine time-go-accelwate.

'1"ne drag tmque (due to bearings, seats, gears, etc.)

of the RLIO Imbopump (fuel and LOX combined)

is a known f_wce of mgine-to-engine variation.

Tbe valne LSnot genially mea._d for ea,'h engine

but past studies have shown that the torques v&y

from 8 to 36 Ibf-in (with resla_::t to tbe fnel pumpshaft)2o.A constantnominal value of 20 lbf-in

has been used for all simulaficms nm for this study.

It is uncertain what the actual values of running

torque were for the lest and flights considered but it

is unl/kely that the values were all Wecisely 20 Ibf-in.

6.1.2 Uncertainties in Valve movement

"Fne transient behavior of the engine in both start

and shutdown is largely determined by the opening

and closing of valves. Variations in valve and

actuator behavior actually do occm for a variety of

reasons. In some cases, the opening and closingtimes of valves can be infen'ed f:n:an test data. In

most cases,however, this is not posm'ble because

of the limited number of engine sensors and their

dynamic response rates. Valve data provided byP&W has been cmnbined with information infm'ed

from avaUable test data to define 'typical" valve

movement schedules for the new system model.

This single set of typical scigdules was used for allsimulations performed in this study.

6.1.3 Uncertain Initial conditions

The temlga'atme of the combustion chamber,nozzle and cooling jacket at the beginning of the

engine start sequence is an importmt factor in theengine timc-to-acccler_. Unlike the engine inlet

and tempetatm_, there is no foible

meamtemmt of initial jacket _ for any

given test or flight. Temperatures that are

measured on the engine genea_ily show false

readings before start due to ambient conditiot_,metal _ with other c_a_ and the

lack of pmpeUant flow at that time. Tha initialtemperature of the coolingjacket, ducts, manifokis,

and ether components must be estimated, often

based on limited information from past testing.

In the RLI0 model diacussed here, tbe_

of the cooling jacket is asmanecl to be a uniform540 R for furst butas and 350 R for second bums.

Tha _oling jacket ialet manifoM is asmmod to be

at 200 R became the inlet manifokl is extmm_ to

some of the fnel pump toM-down flow before start.

All other _ in the systan m'e assumed to

be inthermalequflflx-imnwith thewopellantflows

at start. Because these assumptions are somewhat

arbitrary, fl_ey me h'kely to be in avor to some

degree for my given firing.

6.1.4 Uncertainty in hmition delay

For the simulations considered here, the ignition

times were set mammlly to agxee with the meamnxl

data. In order to simulate m engine start for which

data is not yet available, a model of the ignition

Igocess would be required. 131is model could be

based on theorelkal analysis, or might be derivedfrom test dau_ NASA does not currently have anignition model for the RL10.

6.2RL10 Steady-state Engine Performance

Ten test cases are considered for the steady-_aateperformance predictions. Five tests are based ondiffe_ntquiescentoperatingpointsf_ a single ground-testrunofa singleengine(EaglneP2087,Run 2.01,

Ocloi_ 4,1991). The other five tests are based on thefinal states of five sturt-Umsient data sets (five different

ground-testruns) of a singleengine(P2093). Flight

data has not been included in this comparison becauseinsufficient data exists to determine the mixture-ratio

and trim position of the oxidizer control valve (OCV)

for those firings. For the ground-test runs considered,the OCV position has been trimmed in the simulationto achieve the steady-stale mixture ratio indicaledby thetest data. Since the OCV position is not a measuredparameter, the simulated trim position could not beverified directly with test data. A comprehensiveperformanee wediction for a typical case b shown inTable 1. In general, only a few parameters are actuallymeasmed on engine firings (14 parameters on grmmd-tests, 8 in flight). Of the fotmeen pmmneters measuredin ground-tests, five are used as inputs to the model(inlet conditions and chamber lZessure), and so onlynine _ _ are cmapm_ with test data foreach case.

Figure 7 shows the distribution of error between the

measured and predicted parameter values in the tenground-test cases. The model lX'edictions match themeastm_ values to within 10% fcf all pmmnetefs on alltests (a total of 90 values). Most wedictions are within3% of the test resuRs. The most significant errms arein the turbine inlet temperature and the pump discharge_esmres. The difference between the wedicted andmeasuredturbineinlettemperaturevariesfremengine

toengine,asdiscussedinsection4.2of thispaper.

The en_ in the pump discharge ptesmres appems to beassociated with mfl)cpump speeds that are consistentlylower than measm_. This _ in speed is mnstlikely due to small errors in the turbine maps andcooling jacket model; these errors cannot be easily

corwxted for without adversely affecting the wedictedstm behavior. The turbine performance maps wovidedby P&W for transient simulation are not the same as

those originally provided for use in the steady-statemodel. The original maps do not work well insimulating the start transient but the new maps do notmatch as well at the engine's design operating

conditions. The new system model's steady-stateare therefore slightly less acowale than those

of the original system models. It was decided, however,that the turbine maps suggested for start transientmodeling would be used throughout, and the associatedsteady-state mor accepted.

6.3RLI0 Start Transient SimulationsThe results of RLI0 start transient simulations were

compared with both ground-test and flight data. Figure8 (a - e) shows the predicted and measured startWamients of a single ground-test first-burn. Figure 9shows chamber pressure and pump speed data for anAtlas/Centaur flight (AC-72), while Figure 10 showssimilar data for the second burn (restart) of a differentflight (AC-74). In each of these runs, the ignition timehas been setin the model based on examination of the

testorflightdam. 1_e differencebetweengronnd-testand flight engine simulations is the value chosen forthe fuel cool-down valves discharge coefficient (whichreflects differences between the vehicle and test-standductwork). The difference between first and second bcm

simulations b the a_umcd initial temperatm_ of thecombustion chamber metal. These assunwM variationswornalso discnssed in section 6.1 above.

The start model generally matches the measured tim¢-to-accelmme of the engine to within approximately 230milliseconds, using only estimates for initial_ bearing fxicti_ valve u:heduies and otherfactors which may vary from nm to run and fromengine to engine. Table 2 gives the predicted vs.measured finw-to-a_x_ate for six ground-test and threeflight-engine firings. One of the flight simulations isoff by 280 msec (rather than 230 reset), but thisappears to be an aberration relative to other flight-engine starts. Comparing the results of this startUan._ent with those from other frights, it appears likelythat the conditions f_ this flight were differeat in waysoth_ than their inlet conditkms alone. The model

correctly predicts start variations due to different engineinlet conditions, initial thermal conditions, anddiffeamces betwem gnmnd and flight hardw_.

The reader may note from Figures 8-10 that there aresome transient differences between the predicted andmeasured chamber pressures which occur after theengine bootstraps but before it reaches the quiescentstate. The small oscillations evident in the test data medue to oscillations of the Thrust Control Valve CI'CV)se_vo-mechanisnL The simulation does not include a

model of the actuator dynamics, but the TCV is

assumed to open as a simple linear function of

combustionchamberpressure. The simulationthereforeoverMmotsthedesiredchamberpressureanddoesnotoscillate,htseveralcases,tbesimuiatimdoesshowsceneunusualuausiemsbeforereacbtagsteady-state;theseapsgartobedueto volume dynamics tn theLOX pe_ iuiet ducc As the OCV suddmly opens andthe LOX system _ the simulatim pmlktso,cmaems musodby nutd inerea, andphase changm. Thesemmsients,which Ke not evidentin the test data, may occtw in the simulations becanseOCT serve dynamics m'e not included in the model.Thesetramientdifferencesbetweenpredictkmand test

ate not considered significant; they would be minimizedif models af tbe TCV and OCV actuatms are deveicj_in the futm_.

To demmsuate one potmtial applkation of the systemstart model, Figure 11 shows the predicted metal

of the combustion chamber just upstreamof the thront (its hottest point). This _ is notmeasm-ed, even in ground tests. The temperatme in thiscase peaks at amend 1875 R, which is a few hundw_degrees below the melting point of the silver throatinsert. Infoanation of this kind can be used to helpdetermine conqmtzmt wear and to _ the impaa of_ or hmdwa_ chaqes to tbe engine.

6.4RLIO Shutdown Transient Simulations

Two firings have been used for comparison betweenmodel predictions and measured data. RLI0 eagineshutdowns do not appear to have any distinct featureanalogous to the time-to-accelerate for start transients.Although there are subtle variations in the rate ofdeceleration, the nature of these differences is not as

well understood as in the case of engine start.

Figure 12 (a-d) shows the wedicted vs. measuredshutdown for a gronnd-test engine. The RLIOshutdown model has capturedmany interesting effectsthat occur during shutdown. In Fignre 12c, for

example, the simulated and measmed venturi pressmesboth show a characteristic dip, rise and then falioff inthe fuel venturi upstream pressure. This feature iscaused by the dynamic interaction of the fuel pumpcool-down valve opening gad main fuel shutoff valveclosing. Anotber _g charactedstic of Ibe RLI0shutdown transient (as shown in Figure 12c) is thejump in fuel pump inlet wesstwe due to reverse flowthrough the fuel pump.

7.0 Concludina Remarks

The major goals set for this pmjeot were to create atramient model of the RLIOA-3-3A rcdwA eagine for

government use, to betu_r understand the engine and itsmmlmnems, m_dto beaclanmk tha available cmnponemmalysis tools using an existing mgine design. ThesegoershavebernacmmpUstet

Tha new RL10 start trmsieat model accmm_y Igedicts

tbeenginet e-t , me whm to ground-test and flight data. The model can granulate enginestart transients over a wide range of inlet conditions,initial themml conditious, and ignition delays. Thismodel also paedicts steady-state ped'ommance valueswhich are within 10% of the meamred values in all

cases, and within 3% for most _. The newRLI0 shutdown model successfully reproduces the

eDgi_'s transieat behavior after main eogine cut-off.These new system models muld be used in the future topredict the effects of cJumges in the mgine design, andto simulate off-nmninal opmuing conditious.

i,!performingthedmanedcomponentanalysesdescribedin this patsy, a great deal has been leamod about theRLI0 e_gine. This activity has also provided amoplmaunity to compare the output from availablecomponent modeling tools with test data from anexisting engine design. Comparison of themmlysisresalts with data provided by Pratt k Whitneyindicates that at least some empirical correction must bemade to the results of the component models. Suchcomponent models are nonetheless valuable inpredicting the off-design _ of the engine

components, especially once _ cozre_ons havebeen included. Detailed three-dimensional

computational-fluid-dynamic models may also beconsidered in the future to improve the _tncy ofmmponentperformancept fictious,thoughevenughadvancedtechniqueswillinvolvesome uncertainty,

especiallyfornew cx_aponentde&igus,The capability

may notyetexisttoweci_y predictthebehaviorof

new componmts of engines for which no test data isavailable. In inch cases, the best that can be expec_dis to define a range of performance and transientbehavior based on malysis. This type of infomationcan be extremely valuable in the design anddevelopment of new ccmpommts or systems, especiallyin combination with probablistic and uncertaintyanalysis techniques.

Acknowledgements

'Ibiswork was Ix_fonned undexconlractNAS3-27186.

The anth0f would like to acknowledge the participation

of the following individuals, who performed the

component analyses which were incorporated into the

new RL10A-3-3A system model.

Tom Tomsik (NASA LeRC) - Heat Transf_

Joseph Veres (NASA LeRC) - Turbine and Pumps

Ken Kacynski (NASA LeRC) - Combu.qion and Nozzle

Doug Rapp (Sverdrup / NYMA) - Hot gas properties

Dean Scheer (Sveldrup / NYMA) - Pumps

Albert Pavli (Sveadmp / NYMA) - Inject_William Tabata (NASA LeRC) - General RL10

information and expert_.

Pratt & Whitney (West Palm Beach, Florida) also

participated in this work, supplying us with designinfoflnation and test data forvalidation.

References

1. Pratt & Whitney Government Engines,

System Design Specification for the

ROCETS System - Final Report, NASA

CR-184099, July 1990.

2. Binder, M. - Sverdrup Technology Inc.,

An RLIOA.3.$A Rocket Engine Model

Using the ROCETS Software, AIAA

Paper 93-2357, June 1993.

3. Binder, M., Tomslk,T, Veres,J,

RLIOA-3-3A Modeling Project- Final

Report, NASA TIM (number pending),1995.

4. Veres, J, A Pump Meanline AnalysisCode (PUMPA), NASA TM.106745,

October 1994.

$. Gulbrandsen, N. Centrifugal Pump Loss

lsolatiou Program (LSISO), COSMIC

Program # MFS-13029, April 1967.

*(note that the version of LSISO used in

this study is not publicly available, this

reference gives a general description of

the program in its original, publiclydisseminated form).

6. Rostafinsld, W, An Aulytical Method

for Pr#dictiag the Performcmce of

Centrifugal Pumps During Pressurized

Startup, NASA TN D-4967, January1969.

7. Chaudhry,H, Applied HydraulicTransients - 2ml Edition, Van Nostrand

Reinhold, New York 1987.

8. Veres, J., A Turbine Meanliae Analysis

Code (TURBA), NASA TM (number

pending), 1995.

9. McBride,B. and Gordon,S. CETg$ andCETPC: An Interim Updated Version of

the NASA Lewis Computer Program for

Calculating Complex Chemical

Equilibria witk Applications , NASATM-4557, March 1994.

10.Nnragbi, M.H.N, RTE - A ComputerCode for Tkree-Dimensiolud Rocket

Thermal Evaluation, (NASA Grant NAG

3.892), July 1991.

11.Bnrt_ D.R, Smrvey of Relationships

between Tkeory aml Experiment for

Convection Heat Transfer in RocketCombustion Gases, Advances in Rocket

Propulsion, AGARD, Technivision

Services, Manchester, England, 1968.

12. Holman,J.P, Heat Transfer - FourthEditiom McGraw Hill, 1976

13.Muss, J. and Nguyen,T_ User's Mauual

for Rocket Combustor Interactive Design

(ROCCID) arid Aualysis Computer

Program, NASA CR-1087109, May1991.

14. Nickerson,G, Two-Dimeasioual

Kinetics (TDg) Nozzle Performauce

Computer Program. Users Manual,

(NASA Contract NAS8-39048), March1993.

15. Kacynski, K., Calculut/ou of Propulsive

Nozzle Flowfieid in Multidiffusing,Ckemicallyreacting Environments,

NASA TM 106532, 1994.

16. Henry, R. and Fauske, H, Tkc Two-

phase Critical Flow of Oue-Compouent

Mixtures in NooJes, Orifices, amd ShortTubes, ASME Journal of Heat Transfer,

May 1971.

17. Applied Physics Laboratory, JANNAF

Rocket Engine Performamce Test Data

Acquisitiou aml laterpretattou Manual,

CPIA Publication 245, April 1975.

18.Shnonean, R. and Hendricks, R.,

Generalized Clwzrts for the Computation

of Two-phase Choked Flow of Simple

Cryogen Fluids, Cryogenics, February1977.

19. Crane Co. and ABZ Inc., Crane

Compauiou to Flow of Fluids Through

Valves, Fittings, and Pipe, (software),1992.

20.RLI0 Start Capability Working Group

(government/industry cooperative), AC-71 Failure lnvestigatiou - Final Report,

(no contract or report number given),December 1993.

10

OxkberCoatm/Vdve(OO0

MBind_ / M.Milis

Figure 1RLIOA-3-3A Engine System Schematic

1!

1.50

1.00

0.50

-'I

I.

0.000

EmL,

'P -0.50

iio

:]::

-1.00

-1.50

/"'m

m l, _ L

FP Ist Stage

FP 2rid Stage

........ Ox P.mp

-2.00

0.00 0.50 1.00 1.50 2.00 2.50

Speed/FlowParameter (the,-)

3.00

Figure 2

Extrapolated Pump Head Maps for RL10 Model

3.50

0.70

0.600.50 /

! °..°,,,. .---I °''° II/ --........I 0.20

o.lo

o.oo l:P 2rid Stake

.o.lo ........ Ox P.mp _

-0.20

0.00 0.50 1.00 1.50 2.00 2.50 3.00

Speed/Flow Parameter (theta)

3.50

Figure 3

Extrapolated Pump Torque Maps for RLI0 Model

12

2O

18

14

lO

!'|

4

2

4

-24

]h,_dl,elod lleel Plu deq Jmelml

J

ri,J il

•-J IXI\i %I

-14 0 14 24

m

60 4O SO

2000

_844

.1440

1444

f200

1000

404

_ 404

2OO

4

-20

]P_41mmlIiml Taqmnm_ mlmqau_m, Wdl

e,I

,_ !It._

V-\

!%+.I -

-14 4 14 24 64 44

a.Vd I_dJm - Hed. mldpdmm (im+b_ _ 11n_

SO

S44

4S0

4OO

644

| _54J,,

200

1S0

140

SO

4

-24

FT_lhmd CLdml T_NSMIT _lll J_

l 1464

K. 0 I08O

_ 464

"% tl

i-TJ °-1O 4 14 20 50 44 SO

J_d r'd'_" • Ned. ladpdmb (kud._ m_ _)

764

*s+O

/

//

-14 4 '14 14 54 44

&dd e'ud_. Ne.dt.]r,Jdtm_ C.udn. -'n 'n,n._

Figure 4Predicted Axial Variation in Heat Transfer with New Model

60

]3

8O

7O

.0 //I /

a_ 4o

JlO

oro.oo 0.SO 1.oo I .SO 2.o0 2.So s.oo s.S0 4.00

Time from MES (mc)

Figure $Predicted Heat Transfer in Injector Piena during Engine Start

45O

4OO

350

LJ

300

25O

2OO

x/l�.:�

I!

n

---- - NtWIqlmWdi_-lS

---4--- PaW mSllmml _

II_ib °

%

"1%

m

0 2 4 8 8 10

]m_ st_, (o_

Figure 6

TDK/ODE Predictions of Isp vs. P&W Suggested Values

]4

0

0

Is

100%

90%0I-,

80%

"_ 70%

© 60%=1

50%1:

•- 40%¢=

30%a-- 20%:l

E= 10%

O%

Percent Error relative to Measured Values

Figure 7Steady-state Predictions vs. Measurements from

Ground-Test Engine Runs (all runs, all parameters)

]6

8OO

SO0

400

SO0

J lO0

r_

lOO

o_o4 o.so

J?t

I=. ....

I I I

__)

(a)

14400

14000

20OO

tO000

10@0

i 8000

4000

24100

J//

J i I_ mOmxmd 11ncD_

i

0oO0 0.64 1.00 1 .SO 2.00 2.S0 $.00 S.S0 4.00

Time 8rim _ (me)

_)

1200

I ooo _r_.

f_ soo

Jj I00 ,,

_ '°° //I/f

0.00 OoSO 1.00 I .SO 2.00 2.S0 3.00 S.50 4.@0

I I I

_ Gmad Tin4 Doll

| I !

(c)

44;0

700

400

500

400$00

i200

1oo

0.1_ O.H

/Y

I I I

1 .(HI l.Se _.llO 2.M :I.O0 3+S0 4.00

(d)

800

f

,oo f,oo y_

i Jsoo I..

i 400 "'_

!.o//1_ 200

too

\

............. Ground Tul OmM

I I I

1.04 ! .80 2.00 2.50 3.00 S.SO 4.00

11me Imm _s_ _)

(c)Figure 8

RLIOA-3.3A Start Transient Simulation Output for Ground Test Conditions

17

SO0

$00

400

j:lO0

J200

_100

i

/I l J

0 O.S 1 1.5 2 2.S

_k,i, _d_ (re, I)

IN00

14000

II,000

i iI0ql+ 0

4000

I I

--,v:-_ IgslsDm (/(,.il_?mspbmnl_q)

I I I /jr

//

0.5 I

__.Jl_

1.5 2 2.5 3

'lnil m,,ml MBS fJ,O

Figure 9RL10A-3-3A Start Transient Simulation Output

for Atlas/Centaur Flight AC-72 (first burn)

4100

SO0 -- " -

4oo _"_'+"+I I

J 300

200

I_ 100

il-l I

0.04} 0.SO 1.10 $ .SO 2.04 2.S0 3,.00 _1.$4}

arm ]dgl.S (lie)

,-i...._

J

i

4.N

II0_

140_

Figure 10RLI0A-3-3A Start Transient Slmulatio_ Output

for Arias/Centaur Flight AC-74 (2rid burn)

18

Table 2Comparisonof Measuredand Predicted

RL10 Engine Time-to-Accelerate

Type of Run Simulation Time

(sec from MES)

Measured Time

(sec from MES)

diFFerence(msec)

Cm3m_ Test 2.09 2.26 170 (early)

Grm_ Test 1.80 1.90 100 (early)

C._md Test 1.51 1.43 80 (late)

Cam_ Test 1.72 1.70 20 (late)

Grmmd Test (Relight) 1.91 1.84 70 (late)

Caotmd Test (Religh0 2.00 2.08 80 (early)

Flight 1.98 1.90 80 (late)

Flight 1.95 1.67 280 (late) *

_t O_ugho 2.562.33 230(early)

* Note : Although this run had inlet conditions similar to other flights, these engines started about 300msec earlier relative to MF.S. This may indicate a difference in the engine other than inlet conditions (seesection ofthisreporton uncertainty).

ZI

J 1500

1SO@

JJ| 11oo

O00

!i 700

X

I000

1700 _ _

500

r

O.O00.SO 1.001JO2.002.SOS.005.f_4.00

Tim, ZrmmDam _m,)

Figure 11Predicted Maximum Cooling Jacket Metal Temperature

during Engine Start

19

S_ 104)00

"_1 I I450 11HlO0

I_00

j,. ; 1..

SO S@OO

0.2 0.4 0.1 O.I 1

TS/e hm Mdn helm C_eJ..dr lisnd OiIK_O) Ime)

(a)

I I

............ i_.md TIll OlI

I

0.2 0.4 O.I 0.8

tSmtmmJkdnhsJ_Cea-ddmd(l_CO)(m)

(b)

I00 7@0

I -.---._,-- (s.tn.)Lolml m

J--_ --_-_-.- .

1.°° j_ 300

L.\\ i,oo20O

,. _'_ ,% _,.0 0

0 I).2 0.4 0.1 O.I I

"nine tn,m Mde _ C:w-e8 dpd (JdU:O) (me)

k

\

((:)Figure 12

RLIOA-3-3A Shutdown Simulation for Ground Test Conditions

" Jlkm

_Qmmml TIIII Bill

_,,_.----_ _.qqW_ _

0.2 0.4 O.I O.l

_m _ u-L. icqh, (:_ dpd IMICO) ira)

(d)

2O

Form Approved

REPORT DOCUMENTATION PAGE OMBNo.0704-0188

Public reporling burden for this collection of information is estimaled to average 1 hour per response, including the time for reviewing instructions, searching existing data sources.gathering and maintaining the data noecled, and cornple!ing a.nd reviewing the co.lleclion of inJormation:. Send _ts r,egar.ding th_bu_enestimateorany ot_her.,1Sa_jeff_h_scofleclion o| informalion, including suggestions for reducing this burden, to Wash=ngton Headquarters :_ervees, Uweczorale lot mTormal_on ul_eratlunr* cmu net-_,_:,. --:_ ,Jw _,-,Davis Highway. Suite 1204, Arfington. VA 22202.4302. and to the Office of Management and Budget. Paperwork Reduction Prolect (0704-0188). Washinglon. DC 20503.

1. AGENCY USE ONLY (Leave blank) 2. REPORT DATE 3. REPORT TYPE AND DATES COVEREDJune 1995 Final Contractor Report

5. FUNDING NUMBERS4. TITLE AND SUBTITLE

A Transient Model of the RL10A-3-3A Rocket Engine

6. AUTHOR(S)

Michael P. Binder

WU-242-70--04C-NAS3-27186

7. PERFORMING ORGANIZATION NAME(S) AND ADDRESS(ES)

NYMA, Inc.

Engineering Services Division2001 Aerospace ParkwayBrook Park, Ohio 44142

9. SPONSORING/MONITORING AGENCY NAME(S) AND ADDRESS(ES)

National Aeronautics and Space Adminisu'ation

Lewis Research Center

Cleveland, Ohio 44135-3191

8. PERFORMING ORGANIZATIONREPORT NUMBER

E-9702

10. SPONSORING/MONITORING

AGENCY REPORT NUMBER

NASA CR-195478

AIAA-95-2968

11. SUPPLEMENTARYNOTESPrepared for the 31st Joint Propulsion Conference and Exhibit, cosponsored by AIAA, ASME, SAE, ASEE, San Diego,California, July 10-12, 1995. Project manager, Joseph A. Hemminger, Space Propulsion Technology Division, NASA

Lewis Research Center, organization code 5320, (216) 433-7563.

12a. DISTRIBUTION/AVAILABILITY STATEMENT

Unclassified - Unlimited

Subject Category 20

This publication is available from the NASA Center for Aerospace Information, (301) 621---0390.

12b. DISTRIBUTION CODE

13. ABSTRACT (Maximum200 words)

RL10A-3-JA rocket engines have served as the main propulsion system for Centaur upper stage vehicles since the early1980's. This hydrogen/oxygen expander cycle engine continues to play a major role in the American launch industry.The Space Propulsion Technology Division at the NASA Lewis Research Center has created a computer model of theRL10 engine, based on detailed component analyses and available test data. This R10 engine model can predict theperformance of the engine over a wide range of operating conditions. The model may also be used to predict the effectsof any proposed design changes and anticipated failure scenarios. In this paper, the results of the component analyses arediscussed. Simulation results from the new system model are compared with engine test and flight data, including the

start and shut-down transient characteristics.

14. SUBJECTTERMS

RL10; RL10A3-3A; Centaur; Transient system model; ROCETS Program

17. SECURITY CLASSIFICATIONOF REPORT

Unclassified

NSN 7540-01-280-5500

18. SECURITY CLASSIFICATION

OF THIS PAGE

Unclassified

19. SECURITY CLASSIFICATION

OF ABSTRACT

Unclassified

15. NUMBER OF PAGES

2216. PRICE CODE

A0320. LIMITATION OF ABSTRACT

Standard Form 298 (Rev. 2-89)Presc_'iloecl by ANSI Std. Z39-18298-102

Cl_ •I_ -L ill Q.O CD

O

1,10

C_

3