nationaladvisorycommiti’ee “foraeronautics/67531/metadc56680/m...nationaladvisorycommiti’ee...
TRANSCRIPT
-
.- —--
NATIONALADVISORYCOMMITI’EE“FORAERONAUTICS
TECHNICAL NOTE 3926
EXPERIMENTAL COMPARISON OF SPEED - FUEL -FLOW
SPEED -AREA CONTROLS ON A TURBOJET
FOR SMALL STEP DISTURBANCES
By L. M. Wenzel, C. E. Hart, and R. T.
Lewis Flight Propulsion Laboratory‘Cleveland, Ohio
Washington
~c)~ REFEWNCE March 1957
ENGINE
Craig
LIBRjuIYCOPYMAR 81957
LANGLEYAEkONAIjTICALLABORATORYUBRAfW,HACALAFJG~EY r!=. u. VIIWINIA
ltxcToEE-Hm-
-
— ..
NATIONALADVISORYCOMMITTEEFORAERONAUTICS
TECHNICALNOTE3926
EXPERIMENTALCOMPARISONOF SPEED- FUEL-FLOWAND SPEED-AREA
CONTROLSON A TURBOJETENGINEFORSMALLSTEPDISTURBANCES
By L. M. Wenzel,C. E. Hart,andR. T. Craig
SUMMARY
Optimumproportional-plus-integralcontrolsettingsforcontrolofenginerotationalspeedwithfuelflowweredeterminedby minimizationofintegralcriteria.Theoptimumsettingsthusdeterminedcorrelatedwellwithanalyticallypredictedoptimumsettings. —...—
At theoptimumcontrolsettings,thespeedovershootratiowas 1.28;simulatedturbinebladetemperaturedidnotovershoot.Theoptimum-con-
. trolsettingswerenotappreciablyaltere:by thelimitationof therate.-..[ of changeof fuelflow.
,- The frequencyresponseof enginespeedto exhaust-nozzleareacouldbe representedby a linearlag,specifiedby therotortimecon6ta&,plus‘“a deadtime. The linearityof thesystem,however,is Umited to theex-tentof thephysicallyrealizableareachange,andthusholdsonlyfor —
smalldisturbances.
Overa limitedrange,speedwas satisfactorilycontrolledby exhaust-nozzlearea. Withinthisrange,thissysta was foundto possesscertainadvantagesoverthespeed- fuel-flowcontrol.
INTRODUCTION
Controlof therotationalspeedof a turbojetengine@ a primarymeansof regulatingtkust andmaintainingsafeoperatinglimits.Opti-mizationof speedcontrolis thereforeof considerablepracticalimportanceandhasbeentheobjectof muchendeavor. .—
Severalcriteriaforuse in optimizingclosed-loopcontrolsystemsaresuggestedin references1 and 2. E general,thesecriterias ecifyoptimumcontrolsettingswhenan errorfunction,suchas
$3(error)dt,
where t is time,isminimized.h reference3, a met”=”‘isdeveloped. foranalyticallydeterminingtheoptimumcontrolfora linearclosed-loop
.
-
2 NACATN 3926 .
processbased.onthiscriterion.Thismethodis thenappliedto thespe- ““ .cificcaseof a speed- fi-el-flowcontrolon a tur%ojetingine.Foracontrolledenginewithno constraintson speedor jmnperature,theanalyt-icallydeterminedoptimumcontrolwasa proportio~-plus-integralcontrol -(neglectinga small.derivativeterm). The loopgainwas equalto theen-gine---timeconstantdividedby theenginedead‘cbn&andthe “control . —
integraltimeconstantwas equaLto theenginetinieconstant.———
Sincetheproportional-plus-integralcontrol3.sin wideuse,..the.ln-._ .._+dicatedrelationsfortheoptimumcontrolsetting6_wouldhavegreatutility “-Pprovidedtherelativelysimplecriteriausedforo@imizationresultin Fsatisfactoryresponseof enginespeedandtemperature.In thisreport,resultsof an investigationof enginespeedcontro~arepresentedwithemphasison therelationsbetweentheresponsesof enginevariablesandthevalueaof twodifferentoptimizingcriteria.A.comparisonof theex-perimentallyobtainedvaluesforoptimumcontrolsettingswiththosede-terminedinreference3 is alsomade. Thiscomparisongivesan indicationof thevalidityof theanalyticalresultsforuseas a fileof thumbfordeterminationof theoptimumcontrolsettingsfora proporttonal-plus-integralcontrol. —T- .— ,-‘+.-
In additionto-theconventionalmethodof con%rollfigspeedby fuel —
flow,speedmay alsobe controlledby exhaust=nozzlearea. It iS, of .
course,recognizedthatby Itselfa speed-areacon7&olis notfeasible”.However,considerableattentionhasbeendirectedtowardthissystemwiththeideaof usingit as oneloopof a two-loopconirol.
In orderto buildup a backgroundforspeed-tiescontrol,theresponse- “--of enginespeedto exhaust-nozzleareawaaobtained.The investigationofthespeed-arealoopwasmadeon a single-loopbasig. Twgnonlinearcontrol_.._schemesarepresentedand discussed,andtheirperformancedataaregiven.
.APPARATUS
Engine
An axial-flowturbojetenginewithannularprevaporizingfuelinjec-turswas runin a sea-levelstaticteststand. Theenginewas equippedwitha variable-areaexhaustnozzle,whichis shownschematicallyin fig-ure1. The electricinputsignalis convertedto a hydraulicsignalintheelectrohydraulicservomotor.Thissignalis airplifiedandconvert%dto rotarymotionby thehydraulicpoweramplifier.-.Therotarymotionthusimpartedto theuppervaneis transmittedto thelowervaneby the .interconnectingmechanicallinkage. *
The frequencyresponseof thissystemwas dependenton theampLitudeof thecalledforchange.Theresponseof theareavalveto a airiusoidal -inputfora peak-to-peakamplitudeof 15-percent~Stedareaispresentedin figure2. — -———-—
-
NACATN 3926 3.
Thisvariable-areasystemwasdesignedto havea fastresponsewith. littleconcernaboutitseffecton thrust.As a result,itsresponseis
muchfasterthanthatof a practicalsystem,anditscoefficientofdischargeis probablymuchlower.
FuelSystem
A research-facilityfuelvalvewasusedb regulatefuelto theen-ginemanifold.Thevalvehadan aplituderesponseflattobeyond10“-cyclesper secondanda deadtimeof 0.005second.Thedescriptionofthisvalveis givenin reference4.
Fromtheenginemanifoldthefuelwas distributedto flowdividersmountedcircumferentiallyaroundtheenginefuelinjectors. *
* Computer;.
+ An electronicanalogcomputerwasused& operationsnecessaryto makeup thedesired. andto obtaintheintegralcriteria.
INSTRUMENTATION
Speed
andfinallyto thevaporizing
to performthemathematicalcontrol-loopconfigurations
.-
A magneticpickupwasmountedh thecompressorcaseoppositea rowof steelcompressorrotorblades. Thepassingof a compressorbladedis-turbeda magneticfieldproducedby thepickupandgeneratedar”electricpulseat theoutputterminals.
In orderto obtaina transient-speedsignal,thesepulseswerefedto an electroniccircuitwhichproduceda voltageproportionalto thentier of pulsesperunittime. No dynamicscouldbe measuredW thissysta withtithefrequa~ rangeof ccmcern.Thepulseswerealsofedto a digitalpulsecounterto measuresteady-statespeed.
.—-—_.-
TailpipeGasTemperature
Transienttailpipegastemperaturewasobtainedby measuringtheout-putof 16 thermocoupleswhichweredistributedaroundtheperipheryof the. tailpipe.Pairsof thethermocoupleswereconnectedin series,aridthe‘“”—outputsof thepairswereparalleled.The thermocoupleswereconstructedof 22-gagechromel-alumelwireandhada timeconstant”o”fapproximately
...--——.
-
4 — NACATN 3926.
0.30secondat theoperatingconditions.Theoutputof foursimilarpairs-- ‘-of thermocoupleswas connectedto a self-baIancingpotentiometerto read - —-steady-statetailpipegastemperatures. —
IWelFlow.-
A signalindicativeof fuelflowduringa ti%nsient-wasobtained””bymeasuringthepositionof thethrottlevalvein thefuelControlby meansof a lineardifferentialtransfo?.mer.It is“high~yprobablethat”higher- - ‘-l.orderdynamicswereaddedto thesystemby thef&l manifoldandflowdi- 7
viders.However,becauseof theextremedifficultyinmeasuringactual ufuelflowintotheengine,thepositionof thetbottlevalvewas ch&en _ _as thesignahtobe.used.Tf theadditiond-higher-orderdynamicsb-ythedistributionsystemof theengineis neglected,thenfuelflowis pro- “~portionalto throttle-valveposition,because,theresearch-facilityfuelvalvewas designedtomaintaina constantpressur~edropacrossthethrottlevalve. Steady-stxrtefuelflowwasmeasuredon ayotemeter.
.Exhaust-NozzleArea -
.— --A
A measureof exhaust-nozzleareawas‘obtainddby connectinga posi- - ~tionsensorto oneof therotatingshafts,as indicatedin figure1.Althoughtherotationof thisshaftis not exactl..jproportionalto exhaust- “nozzlearea,onlysmallerrorsresultif linearityisassumed.
Recorder --.-
A six-channeldirect-writingrecorderwasusedto recordtransi&ntcontroldata. Therecorderhada“frequency.respo~seeszentkllyflat—io ‘- “--100cyclesper second.A galvancnnetricoscilIogr&phrecorderwas used” to “-–:recordsinusoidalfrequency-responsedata. The&der rtigeof ch&irtY_speedsavailableon thisrecordermadeibmore afilicableforthesepur-poses. Thefrequencyresponseof thisrecorderwi% essentiallyflat‘ibeyond100cyclesper second.
— ..-—
DYNAMICS,CONTROLS,ANDOPTIMIZINGCRITERIA
EhgineDynamics -.
Shownin figure3 is thefrequencyresponseuf enginerotationalspeedto exhaust-nozzlearea. Thesedatawereobtained-%ysiimsoidallyva~ing “- j-theexhaustnozzleandmeasuringtheresultantspeedfluctuations.Thedataweretakenover.thesamespeed-rangeas was ~bsequentlyusedin the :controlsstudy. The datamaybe representedby a“””first-orderlag:with”a 1-timeconstantofiO.83secondplusa dead”timeof CL021second,as shownby
—.
thecurvesin figure3.—
-
NACATN 3926 5.
Theresponseof engfierotationalspeedto a sinusoidalinputto. thefuelvalveis showninfigure4. Intuitively,thisresponseshould
havethessmebasictimeconstantas the speed-arearesponse.However,[email protected] tofigure4 showsthatadditionaldynamicsarepresentin thissyst&.””-–Thesedyasmicsareattributedto thefueldistributionandcmbustionprocesses.(Thephase-shiftcurvesdrawnin fig.4 willbe discussedin
~ the section“Selectionof optimumcontrolsettings.”)Thedataof fig-* ure4 weretakenoverthe speedrangeusedinthe controlsstudy.
-— —
A transient-datatraceshowtigtheresponse.of enginevariablesto aburstin fuelflowis presentedin figure5. Thedeadthe betweentheinputandthebeginningof theresponseof speed,is 0.079second. (Deadtimeismeasuredto thebreakin accelerationbecauseit ismoreevid-&t
—.—
thanthespeedbreak.) .-.
Speed- Fuel-FlowControl,
. A blockdiagramof thespeed- fuel-flowcobtrolloopispresentedin-— figure6(a). Thecontrolis of proportional-pl~-integralconfiguration
and itsactionmaybe representedby thetransferfunction. ..(.++).
(Symbolsare definedin appendixA.) The loopgain KL is definedas theproductof thefrequencyinvariantgains of theloopcomponents.Sincethegainsof allthecomponentsexceptthetherangeof testconditions,KL willcontrolsetting.
Theoperationof thecontrolloop
control&e cori6ideredconstantoverhereinafl%rbe referredto as a ‘“———
is thus: Measuredspeed Nm iscomparedwithsetspeed Nfi,and thedifference,:or speederror Ne, iSappliedto theinputof thecontrol.The error;signalismodifiedaccord-ingto thecontroltransferfunctionandfedtothe fuelv~lve~”’The fuel. _valvethenmanipulatesfuelflowuntil Ne becc@eszero. —_....—
Speed-AreaControl ,
In thisinvestigation,theoperatingpointof theexhaust-nozzleareawas closeto itswide-openposition.Thus,a limitedamountof additionalareateal,area
,
tem,.
was availableforacceleration.Whenoperatedas a closed-loopsys-thespeed-arealoopbecamenonlinearwhen$hecontrolcalled“formoreopeningthanwasphysicallyrealizable. ~
If a linearproportional-plus-integralcontrolwereusedin thissys-adverseeffectswouldoccurwhenthesysttibecsmenonlinear.At
-
6 NACATN 3926.
theseconditions,an excessivechargewouldbuildup on theintegralpor-tionof thecontrol,holdingtheareawideopenaftertheerrorhadbeenreduced.Thisactionwouldcauselargespeedover6100ts.In orderto ob-taingoodclosed-loop”performance,somerneah6wasr%qufredto preventthisexcessivechargefroq,buildingup whiletheareawaLwideopen. Twonon-linearcontrolschemeswereinvestigatedandaredescribedsubsequently.
A blockdiagramof speed-areacontroltypeI is presentedin figure6(b). Theoperationof thiscontrolduringan acce~erationis as follows:Inputsa, b, andc occurstiultaneously.Inputa tiitiatesa stepchange,m8 ; inputb openstheswitch,breakingthe cc)ntro~loopj and inputcsuppliesa signalta theareavalve,causingit to opento itsmaximumposition.
The engineaccelerateswhentheexhaust-nozzleareaiswideopen.Duringtheacceleration,theswitchingcircuitcomparesthedecreasingNe to a predeterminedvalue,setto about5 percen%of theinitialerr-or.When Ne reachesthislevel,theswitchingcircuitcausestheswitchtoclose,returningtheengineto normalclosed-loopcontro~
A blockdiagramof speed-areacontroltypeIIIs presentedin figure6(c). Theoperationof thissystemduringan accelerationis thus: Astepchange ANs is initiated.The switchingcircuitcomparesthecon-troloutput,or thesignalto theareavalve,to a j?reset..levelcorre-...spendingto thellmit.oftheareamovement.Whenthecontroloutputex-ceedsthislimit,theswitchingcircuitopenstheswitch,preventinganyfurtherchargefrombuildingup on theintegrator.Thecontrolthenhasproportionalactiononly. When Ne decreasesto a valuesuchthatthecontroloutpulmnolongerexceedsthislimit;”-thestitchingcircuitclotie-stheswitch,restoringpro~rtional-plus-integralcontrol.
, .-—
.
.-
—
..—
?
.
..——
Ibis emphasizedthatthesecontrolsin set–speed.Forusewithotherinputs,changesin setspeed,itmaybe necessary
weredesignedforstepincreasessuchas decelerationsor rampto modifythecontrols.
OptimizingCriteria — — .—
Desirablepropertiesof a controlledsystemare speedand stabilityof response,characterizedby overshoot,solutiontree,timeto firstzero,and so forth. Unfortunately,speedand stabilityof responseareoftenindirectconflictwitheachother. .— .—.
In a linearclosed-loopsystem,as loopgainis increasedto improvethespeedof response,overshootincreasesandthesystembecomesmoreOS- .cillatory.The designeris calleduponto makesomecompromisebetweenthesecharacteristics.Thechoiceof an optimumcontrolwouldbe greatly
.
-
NACATN 3926 7
simplifiedif thedesignerhadavailablea singlefigureof merit-whichtookall or mostof theseconflicti~characteristicsint”o”-account..—_ Somefunctionof theerror,whichwouldincreaseas theerrorsbecamelargerandof longerduration,wouldbe applicable.Thisfunctionwouldbemintiizedforan optimumcontrol.
‘~’j~~=./,eldt ‘becauseSeveral criteriah#e beens gestedf -J2),smongwhichare o e2 dt, ‘h~~2~0~~~fE”~e~;.Chosenforthisinvestigationwere o 0of theirsimplicityandtheeasewithwhichtheycsnbe mechanizedonan analogcomputer.The controlledsystemforwhichthesecriteriaattainedminimumvaluesis definedheretias optimum.A blockdiagrmrepresentativeof the systemusedto obtainthevaluesof the criteriais shownin figure7.
PROCEDURE
A steady-statemap of enginespeedagainsttailpipegastemperatureis presentedin figure8. Accelerationstithspeed- fuel--flowand speed-areacontrolsare indicated;theinitialpointforall8c6elerationsis at,94 percentratedspeed,andthefinalpointis at 97 percentratedspeed.
The testingprocedurewas the ssmeforallthreecontrolloops. Theloops(fig.6) weresetup, smd stepincreasesin setspeedwereinitiated.Carewastakento maintainboththesizeandthe initialpointof allaccelerationsuniform.
Controlactionwasvariedby changingtheloopgainandthecontrolintegralttieconstant.Therangeof loopgainsinvestigatedwas fromavalueapproachingthepointof instabilityto approximately10 percentofthat vaiue. The controlintegraltimeconstantapproximatelyone-halfto fourtimestheengine
RESULTSAND DISCUSSION
wasvariedfrom —t5meconstant.
Speed - Fuel-FlowControl
Transientdata.- Transientdatatracesarepresentedin figure9 forvaryingvaluesof loopgainandcontrolintegraltimeconstant.There-sponsefora low gaincontrol(KL= 1.5)and Tc = 1.5secondsis shownin figure9(a). The initialincreasein fuelflowis onlyslightlygreaterthanthefinalvalue,caustigtheerrorto dtiinishslowlytowardzero-.Thisbehavioris similarto an uncontrolledsystem.
Increasingtheloopgainto 7.4,as shownin figure9(b),improvesthespeedof performancegreatly.The speederrorrisestowarditsfinal
.—
-
8 NACATN 3926 .
valuequickly,reachingit forthefirsttime0.33secondafterthe_ini- .—
tiationof thestep. It thenovershoots31 percent.Tailpipegastemper-aturehasan overshoofl.ratioof 3.1. (Performsmceparametersof overshoot,overshootratio,and timeto firstzeroaredefinedon thisfigure.)
A furtherincreasein theloopgatito 10.3givessomewhatfasterperformance,as shownin figure9(c). Reductionof thethneto firstzeroto 0.29secondfrom0.33second(fig.9(b))is doneat theexpenseof alargespeedovershootratio,whichis now1.50. Tailpipe-gas-temperatureovershootratiohasalsoincreasedandis now4.0. +P
EChanging~c to 0.5second,witha loopgainOf 7.4,makesthesys-
temmoreoscillatory.In figure9(d),theovershootis 62 percent,twicethatof figure9(b). Thethe to firstzerois 0.31second,andthetemperatureovershoot-~atiois 4.1.
Increasing%C to 4.0secondsat thisloopgainhasa stabilizing.effecton thesystem,as shownin figure9(e). The speedovershootis re-ducedto 27 percentmd thetimeto firstzerois sllghtlyincreasedto”0.34second;thetemperatureovershootratiois reducedto 3.0.
Selectionof optimumcontrolsettings.- A plotof timeto firstzeroagainstovershootratioobtainedfromtheseandotherdatais preserited-infigure10. Fromthiscurve,whichtakesintoaccoun%theportiono~thetransientonlyup to thefirstovershoot,it appears.that.itwouldbedesirableto operatewithveryhigh KL and Zc.—
—
●
✎
—
If,however,figure9(e)is examinedcloselyaftertheinitialoscil-lationshavesubsided,it is seenthat-asmallerror”persistsforarelativelylongtime. Thistypeof erroris verydifficulttQ evaluate.
Theusefulnessof theintegralcriteriais apptient.Thechoiceof-themostfavorablesystemwouldnotbe an easyone,“ifit“werebasedonthesesmalldifferencesinperformance.Throughtheuseof theintegralcriteria,theproblemis reducedto simplypickinga-minimumvalue.
Thevaluesof theintegralcriteriaare.plottedagainstspeedover-shootratioin figure11. Thecriteriavaluesarelargeforslowsystaishavingsmallovershoots,reacha minimum,andthenficreaseagainas theYsystembecomesmoreoscillatory.Boththecriteriaattaintheirminimumswitha controlintegraltimeconstantof 1.5secondsandam overshootratioof 1.28. In figure10,thispointon thecurvefor %C = 1.5secfidsOccursat a loopgainof a~roximately7. ‘“ “~ ‘
Zf.thesecontrolsettings,chosenby theintegralcriteriaas optimum,areto be comparedwiththeoptimumsettingsindicatedin reference3, tl-etwosystemsmustbe similar.
—
.
.
-
NACATN 3926.
The dynamicsof thetheoreticalsystemconsistedofa deadtime. Xn theresponseof speedto exhaust-nozzle
9
a linearlagandarea(fig.3),
theexperimentaldatafitresponsecurvesof a linearlaganda deadtimeverywell.
In theresponseof speedto fuelflow(fig.4),thisis not so._~e-causeof theadditionaldynamicspresentin thefueldistributionand com-bustionprocesses,thedatado not fitthefirst-orderresponsec-s.Theamplituderatiois attenuatedmorerapidlythanthelinearlagtrans- ““ferredfromfigure3, andmorephaseshift(curveA) is exhibitedthan
.—
wouldbe expectedfromthelinearlagplusthedeadtimemeasuredinfigure5.
In orderto interprettheexperimentalresults,thisadditionalat-tenuationandphaseshiftshouldbe combinedwiththatdueto thelinear _lagand deadtimeto forman effectivelagandan effectivedeadtime.In evaluatingtheeffectivelag,a curveis fairedthroughthedataof theupperportionof figure4 in theregionaboutan amplituderatioof 0.7.,The fairedcurveintersectsthelineof 0.707amplituderatioat a frequen-cy of 0.15cycleper second.Thisgives 1‘e,eff = 1.06seconds,‘-
“i as comparedwith Ze = Q.83secondmeasuredfromtheresponseof speedto3 area.
The effectivedeadtimeis determinedby plottingthephaseshiftthatwouldresultfroma linearlagwitha timeconstantof 1.06seconds(curveB, fig.4). The differencebetweenthiscurveandthedatais due “-”-to theeffectivedeadtime. The effectivedeadtimeis evaluatedby add-ingto curveB thephaseshiftdueto variousamountsof deadtime. Thebestpossiblematchis shownas curveC, whichis obtainedby”the””-additionof a deadtimeof 0.15second.Thisvalueof td,eff= 0.15secondiSconsiderablylargerthanthevaluefordeadtimemeasuredfromfigure5, ___whichis 0.079second.
—.
Accordingto reference3, theoptimumcontrolsettingsare~c,Opt /= Ze,and KL = Ze td. Applyingtheserelationsto thissystem
gives = 1.06seconds,and ‘e eff—
‘L,opt= td}eff= 7.1. These=C,opt= ‘e,eff .valuescomparefavorablywiththoseobtainedexperimenl%lly,whichwerezc,opt= 1.5 secondsand KL,opt= 7“
Temperatureovershoot.- The-ov=shootratioof tailpipegastempera-ture T is plottedagainstspeedovershootratioin figure12. Fortheopthnuncontrolsettings,the T overshootratiois about3.0. This.temperaturewas sensedby a thermocouplehavinga timeconstantof 0.3second.
..
-
10
Becauseof thelarge T overshootsexcursionsof turbinebladetemperaturelate Tk, the T signalwas compensated
NACATN 3926.
obtainedlan investigationof theTb wasmade.”In orderto simu- ‘“-to elimi~tethelagof thesens- —
ingthei%ocoupleandwas thenpassedthrougha lagof 1.5seconds,which _is conservativelysmallforsimulationof turbine%ladedynamics(ref.5).A datatraceshowingthetransientbehaviord.this variable,f~ra repeatrunof thetransientshownin figure9(b),ispresentedin figure13....
The turbine-blade-temperatureovershootratios,obtainedwithva51ZtousKL’S at ‘GC= 1.5seconds,areplottedinfigure12. It .isseenthatat :theOptti~ controlsettings(KL= 7.4,~c = 1.5)no Tb OVerShOOtOCCUrS. :
Thisis importantin thata compromisein optimumspeedcontrol,forthepurposeof limitingtemperatureexcursions,is apparentlynotiequired.
Limitationof fuel-valveresponse.- In orderto obtaindatamorerepresentativeof a practicalcontrolsystem,a circuitwasaddedbetweenthecontrolandthefuelvalvewhichlimitedtherateof changeof thesignalto thefuelvalve. Thus,therateof changeofifuelflowwaslimitedb approximately150percentof ratedflnw-per10thsecond.
—.
Transient-datatracesforthissystemarepresentedin figure14 for .-
KL’s and TctS comparablewiththosein figure9: Of particularinterest “i-n thesedatais thefuel-valvepositiontrace,whichr~seson a rampinit-ially. (Thisis notclearin fig.14(a)becau~ethechengein fuelflowis small.)Comparisonof thetracesof fuel-talvepositionandcon-troloutput(recordedbeforetherate-of=changeltiitingcircuit)showsthatthefuelvalvecannotfollowtheinitialrisebut followstheslowerreturnexactly.Thisactionis typicalof a motor-drivenfuelvalve.
Theprimaryeffectthattheadditionof thiscircuithadon theper-fmmanceof thesystemwasthe increaseof timeto firstzero(fig.15).Thisincreasewas,on theaverage,about15 percentfora givensetofcontrolconstants.
The integralcriteria,plottedin figure16,minimizeat highervaluesthanwereobtainedwithoutthelimitationon thefuel-valveresponse.Theminimumvaluesoccurat Tc = 2.0secondsand.an overshoot-ratioof about1.28. On figure15,thispointcorrespondsh-a loop-gab-of--about6.5.R is importantto notethatthecontrolsettingsnecessaryto obtaintheminimumintegralvaluesremainessentiallyunchangedaltloughtheres&nse —
of’thefuelvalvehasbeenlimited.
Thetailpipe-gas-temperatweovershootiatiofor thissyste?hisplot-tedin figure17. For theoptimumcontrolsettings,theovershootratiois about3.2. Althoughsimulatedturbinebladetemperatureswerenot in-
.
vestigatedforthissystem,it-maybe assumedthatno overshootoccurr-ed,since the tailpipe-gas-temperatureovershootratio is approximatelythe .same. .-—
-
NACATN 3926 U
Speed-
The increasinguse of
Exhaust-Nozzle-AreaControl
afterburnerson modernturbojetengineshasemphasizedtheneedfortwo-loopcontrolsin whichboth-speed-andtailpipetemperaturearecontrolled.Oneof thepossibletwo-loopconfigurationsis thatinwhichspeedis controlledby exhaust-nozzleareaandtetiperature,by fuelflow. ,
It is realizedthatthespeed-areacontrolhasa limitedutilityasa single-loopsystem.However,an investigationwas carriedouton thisbasisto avoidthecomplicationsthatwouldbe causedby a simultaneousmanipulationof fuelflow.
Transientdata,typeI control.- Transient-datatracesforspeed-areacontroltypeI arepresentedin figure18 forvaryingvaluesof KL “fid‘c“ Eugineresponseforcontrolsettingsof Zc = 0.1 secondand_..KL = 7.5is shownin figure18(a). Speedovershoots24 percent,and itstimetofirstzerois about0.75second.Sincetheareais openedto acceleratetheengine,tailpipegastemperaturedecreases;andthepossibilityofenginedfiagedueto temperatureovershootsis eliminated.
ticreasingthegainto 17.5,as shownin figure18(b),reducesthespeedovershootto 10 percent.The tme to firstzerois unchanged(0.75see). IncreasingTc to 0.5second(fig.18(c))increasesthespeedovershootto 22 percent.The timeto firstzerois again0.75second.
Thereasonfortheseminutedifferencesin timeto firstzeroisclear. As seenin figure18,almostalltheaccelerationis accomplishedwiththeareawideopen. Sincecontrolactiondoesnotbeginuntilthe--errorapproachesZt?roj the initialportionof thespeedresponseis,forallpracticalpurposes,independentof controlsettings.
Selectionof controlsettings,typeI control.- In figure19 axeplottedvaluesof speedovershootratioagainstloopgain. Overshootratiois seen to decreaseas KL becomeslargerand %C becomessmaller.Pasta loopgainof abut 25,howev~, onlysmallimprovanentsarerealizedforfurtherincreasesof loopgain. Also,amplificationof noiseincreases,andtheareatendsto followtheseunwantedsignals.In viewof thesefacts,it is deemedundesirableto operatewitha loopgaingreaterthanabout18.
In thissystem,theprimaryfunctionof thecontrolis to supplyasignalto closetheareafromwide-openpositionto itsnewoperatingpoint. Thisshouldbe doneas quicklyas possiblein orderto minimize
. overshoots.As anticipated,thetestdatashowthata low Zc anda highKL wouldbestaccomplishthis.
-
12 .— NACATN 3926 .
The integralcriteria,plottedin figure20~followthesametrends .-as doesspeedovershootratio. Thesetrendsindicatethatthebestoperationis obtainedby usinga controlhavinga“iargeKL anda smtiil ““ “–‘c“
Transientdata,typeII control.- Transient-datatracesforspeed-areacontroltypeII arepresentedin figure21 forvaryingvalueso?? ‘CC.Engineresponse’forcontrolsettingsof ‘cc= 0.1secondand KL = 7.5 isshownin figure21(a). The speedovershootratiois1.20,andthetimeto
.+
firstzerois about0.78second.---5w
IncreasingTc to 0.5second(fig.21(b))reducesthespeedover-shootto a verysmallvalue,about3 percent-.Thetimeto firstzero@as _increasedto 0.92second.Forthissystem,thear=ais notheldopenaslongas it was in figure21(a).
A furtherincreasein %C to 1.0second(fig.21(c))resultsin anoverdampedresponsehavingno overshoot.Theareahereremainsopenfor .a stillshortertime. It is seenin thesedatathattheperformanceofthesystemismoredependenton Zc thanthesyst~ withtypeI control.
.—
As ~c is decreased,thecontributionof theintegraltermbecomesof greatersignificance.If ‘cc is toosmall,as-~nfigure21(a),alargechargeisbuiltup on theintegralbefore-thE.errorreacheszero.Sincethischargeholdstheareaopenaftertheerrurhasreachedzero;theenginecontinuesto accelerate,causinga speedovershoot. —
If,on theotherhand,‘cc ismadeblarg~; theresponsesuchasshownin figure21(c)results.Therateat whichtheintegralchargesisreduced.Hence,unless KL is increased,a sustat>ederror:snecessary
—
to buildup the integralchargerequiredforoperationat thenewconditions.
Selectionof controlsettings,typeII control.- SpeedovershootratioisplottedagainstKL in figure-22. It is.seenthatovershootmaybe reducedby increasingKL or %C. In figure2~jvaluesof th-integral _=criterionareconstantbeyonda KL of 7.5andPracticaiiyindependentofTc. If-similarperformancecanbe obtainedby operatingat-ahighor lowvalueof KL, it ismoredesirableto operateat thelowervaluebecauseof thenoiseamplificationpreviouslymentioned.
In viewof thesefacts,therecommendedcontrolsettingsforthis ,controlare ‘CC= 0.5 and KL = 7.5. However,sincetheperformanceof-thesystemis dependenton theresponseof theareaservomotorandthephysicallimitof theareaopening,thisrecommendationcannotbe .generalized.
-
NACATN 3926 --13.
Comparisonof Speed- Fuel-Flowand,Speed-AreaControls
rlwl-l*
Overtemperature.- As pointedoutpreviously,thetrendtowardtur-binebladedamagedueto highenginetemperaturesduringan accelerationis notpresentin thespeed-areacontrol.
Speedovershoot.- Speedovershootfortheoptimumspeed- fuel-flowcontrolis 28 percent.Thebestoperationof thespeed-areacontroliswithlittleor no overshoot.
Accelerationcapabilities.- Theadditionalexhaust-nozzleareaavail-ableforaccelerationabovetheoperatingpointwas 13 percentratedarea;theadditionalfuelflowavailableforaccelerationatmvetheoperatingpointwasapproximately100percentratedfuelflow. Therefore,theac-celerationcapabilitiesof thespeed- fuel-flowcontrolaregreaterthanthoseof thespeed-areacontrol.
Timeto ftistzero.- Becauseof thelargeraccelerationcapabilitiesof thespeed- fuel-flowcontrol,thetimeto firstzerofor thissystemis smaller.
Solutiontime.- Becauseof the speedovershootobtainedwithspeed- fuel-flowcontrol,thetimerequiredfor Ne to reachsmdwithin10 percentof itsinitislvalueis aboutthesameforbothcontrols. —
thestay
Integralcriteria.- Thevaluesof theintegralcriteriaat theirminimumss.reaboutthesameforbothcontrols.
Dependenceon controlsettings.- The controlintegraltimeconstantandtheloopgainmustbe heldto closetolerancesforoptimumoperationwiththespeed- fuel-flowcontrolbut considerablevariationis-allowableforthespeed-areasystems.
Simplicityof control.- Thepresenceof thelimitin theareaneces-sitatedtheuseof a nonltiearspeed-areac~tro~~*ich ~volvedaddi- .tionalcircuitry.Sincethespeed- fuel-flowloopwas ltiear,thecontrolwas shple in comparison.
The comparisonsbetweenthetwo systemsare summarizedin thefol-lowingtable. ~ thistable,+ denotesmorefavorableaction;-, lessfavorableaction;ando, similaractionby bothsystems.
.
.
-
14 -, – NACATN
Speed- f’uel-Speed-areaflowcontrcll.control
Overtemperatme +
Speedovershoot. +
Accelerationcapabilities +
Timeto firstzero +
Solutiontime o 0
Integralcriteria o 0
Dependenceon controlsettings -. t
Simplicityof control +
CONCLUSIONS
3926
Theoptimumspeed- fuel-flowcontrolobtainedin thisinvestigation,basedon minimizationof errorintegralcriteria,canbe satisfactorilypredictedby applicationof lineartheory.Agreementwasgood,although.higher-orderdynamicswerepresentin theenginet~t-werehot considtiedin thelinearanalysis.Thetheoreticaloptimumcontrolis OLproportional-plus-integralconfiguration,anditsa~tion.rn.aybe represented
( h).by thetransferfunctionKL 1 + The loopgain KL is eqml to the
ratioof theenginetimeconstantto-theenginedeadtimelandthecontrolintegraltimeconstant‘ccis equalto theenginetimeconst~tj s istheLaplaceoperator.
At theoptimumcontrolsettings,thespeedover.shoot.ratiowas 1.28.Simulatedturbinebladetemperaturedidnotoversho6t,indicatinga corn- —
—.
promisein optimumspeedcontrol,forthepurposeof limitingtemperatureexcursions,is apparentlynotrequired.Limitationof therateof changeof fuelflowdidnotappreciablyalterthevaluesof theoptimumcontrolsettings.
The frequencyresponseof enginespeedto exhaust-nozzleareacould
———
—.-
be representedby a linearlag,specifiedby therotiora deadtime. The linearityof thesystem,however.,istentof thephysicallyrealizableareachangeandthusdisturbances.
timeconstant,pluslimitedto theex-holdsonlyforsmall , :
-
NACATN 3926 15.
Whenusedas a controllingvariable,theareais frequentlyopenedto itsmaximumposition.If a proportional-plus-integralcontrolis tobe usedin thespeed-arealoop,someprovisionmustbe madeto preventsaturationof the integralcomponentof thecontrolwhileIiTieareais onitslimit.
By usinga.nonlinearcontrolwiththischaracteristic,speedwas sat-isfactorilycontrol-ledby exhaust-nozzleareaovera limitedrange. Withinthisrange,thespeed-arealoopwas foundto possessadvantagesoverthespeed- fuel-flowloopwithrespectto speedandtemperatureovershoots.
LewisFlightPropulsionLaboratory -.NationalAdvisoryCommitteeforAeronautics
Cleveland,Ohio,Decmber 13,1956
.
.
.
-
16-..-. -. —.. —---NAGATN 3YZ6
a,b
e
KL
N
s
T
t
t~
Wf
APPENDIXA
SYMBOLS :=
constantsassociatedwithintegralcriteria._
error
loopgain
enginerotational
Laplaceoyerator
speed,percentrated
tailpipegastemperature,percent-rated
time,sec
deadtime,sec
fuelflow,percentrated
Subscripts:
b turbineblade
c control
e error
eff effective
m measured .:,
opt optimum .-.-
8 set
REFERENCES
—
—.
—.-r
-.
1. Hall,AlbertC!.:TheAnalysisandSynthesisof-LinearServomechanisms.TheTechnologyPress,M.I.T.,Cambridge(Mass.),1943.
2. Graham,FrankD.,andLathrop,RichardC.: Synthesisof “Optimum”TransientResponse- CriteriaandStandardFo~s. WADCTech.Rep.No.53-66,WrightAir Dev.Center,AirRes.andDev.Command,Wriglrt-PattersonAirForceBase,Aug.1953. (RDONO.206-11.) ~
-
NACATN 3926 17
3. Boksenbom,AaronS.,Novik,~vid, andHeppler,Herbert:Opttium. ControllersforLinearClosed-LoopSystems.NACATN 2939,1953.
4. Otto,EdwardW., Gold,Harold,and Hiller,KirbyW.: ~Si~ andPerformanceof Throttle-TypeFuelControlsforEngineDynamicStudies.NACATN 3445>1955.
5. Hood,Richard,andPhillips,WillismE.,Jr.: DynamicResponseofTurbine-BladeTemperatureto Exhaust-GasTemperatureforGas-TurbineEngines.NACARM E52~4, 1952.
-
Hydraulic Electro-
T *E
1’)1
PoBition
Inputsignal
Figure1. - Schemtic diagramof variable-areaexhaust-nozzleeyd.em.
.1, Ii:
-
NACATN 3926 19 —.
dU3l-l-Y
M4!
Frequency, CPU
Figure2. - FrequencyreaponaeOr exhaust-nozzleposltlOntoZimzvldz+l-Put@ ti-pe~=ntratedarea. .-.
.
-
20 NACATN 3926
1 .
.8
.6
.4
.2
.1.08
.06
.04
.02
.01
3
6
9
12 W.-.6 \
\ ~Timecon”stant,0.63qeo;deadtime,0.021sec / ‘Q
1s0 I I I 1 I III I I I [ I I Iltt180-. -- ---- ---.U1
\
.v+ .U6.Uu.1 .2 .4 .6 .8 1 2 4 6 a 10Frequency,opa
.
.
.Figure3. - Frequenoyreeponeeorengine rotational speed to elnusoldelvariationorexheust--nozzlepoeition.Operatingrange, 95.5+3percentratedspeed.
-
NACATN
.
3926
.1.08
.06
.04
w >[ I
I 11111 n111111 1 1’N
.02
.01
0
so
m
90
120
H
lea
210
240
270
Frequenoy,CPS
Fi*ve4. - Frequencyrespame of engfrierotationalspeedto mhumidal input ta fueloneratlmranxe.95.5+3Dercentratadspeed.
21
--
-
------ . .
.-_–_.T -..N
—-” “-” “-””j N%P,+. —/
i .’”‘dk%
+-0.01 aec
M“”’,e\-”
P-f’
Tallplpepressure
~“
~’-.; “’..”\
Caupre8.9m-di8_ pressure
,, I ,-_J.--.-J:h’-’
---- 1
,.i”i,.
k
i“!.,“1Ii
I :,,,, , I
Figure 5. - Responseofuncontrolledeng- toburstinfuelflovillwtrat~deadtimepresen’t. UI :m
i ‘1’!
-
, 4161 , ,
(a) Speed - fuel-flow control.
Figure 6. - Block diagram of control loops.
t%
-
.
Speedsensor
t=
Switchingcfrcuit
%iff
Ha swi.t#ch Control~. Area Enginevalve
AN*
t )
----- ——. - --—- ____ _
)a b c
(b)Speed-areacontrolt~ I.
Speed asensor
t
— --. .
‘*
* Area * Enginevalve
.
(c)Speed-areacontroltypeII.
Figure6. - Concluded.Blockdiagramof controlloops.
.
-
“i!J-4, ,
Ne Squaringb Gain
Inte-- device grater
4161
J’a N: ilt
Inte-gl’’ator ‘Lrb lNeldt
Figure 7. - Block diagramof systemused for obtainingintegralcriteria.
N(n
I I i.i
-
100
90
80
7C
60
,
txhaust-nozzleerea,percent rated 50
90-
30 40100. \
113\ \
) 60 70 8
Fuel flow,—Wf.?
percent rated
90
— ‘80i-Accelerationwith speed -fuel-flowcontrol
— 70.-Rase point for allaccelerations
60
speed-area contr 1
\
/ \
90 100 rEngine speed,N, percentrated
Figure 8. - Steady-statemap of engine
,
Vsxiablefi.
T9-C*
-
I ,:3-4 ;ack 4161
I
1 (a) Cmtrollntegmlt imomntmt, 1.60a0cmb; 10cgIgrIin,1.6.
BiSUIW9. - Trm-mientdatarorO* -fuel-flowocatrcd for smp inomaw in engine mat weed.
I
-
.-
N03
s2!$
(b) CmtrOl htwml tho oomtant, 1,5 seomds; low @n, 7.4. w
Fl&lree. - Calthmd. Trmnhnt datafm n~ad - fuel-flowcontrolrti B* Inarmae h englnanet apred. 1%m
-
1
,1 , , i 4161 , t
(e)ControlIntagraltirceaonstent,1.5seoonde;loopgain,103.
Figure9.- Continued.Transientdataforspeed.fiel-flowoontrolforstepInoreaseInengineeetspeed.
-
1
-
I 4161
(e)Cmtml intigral tlw cumtmt, 4.0 seO@B; loop* 7.4.
n&?JI’a9. - cmoluded. hwdont data for npee-a- ruel.rlm oontml rOi-w,ep inomam in engine set speed.
maN9)
-
32
.7Controlintegraltimeconstant,-
‘cC,sec1
0.50
.6
1.0_l.5\4.0
.5\
.4 \
\
.3 .— ——
.21.0 1.2 1.4 1.6 1.8 2.0
Speedovershootratio
Figure10.- Variationof timeto firstze?mwithspeedover-shootratioforspeed- fuel-flow control.
.
-
NACATN 3926 33
C!ont;ol in+egraltimeconstant,
14– TCJsec
.
0.75
12 ,.5
\
.5
,1.0
10
4.0
81.5
61.0 1.2 1.4 1.6 1.8 2.0
Speed
(a)
FigureU.. - Variationofshootratiofor speed-
overshootratio
integral criteria with speedover-fuel-flow control.
-
34 NACATN 3926
1.0 1.2 1:4 1.6 1;8 2.0
Controlintegraltimeconstant,
~c)sec
4.0
.5
h
1.5.75
i1.0
\
\
-1 KL.”
1.0
—
.
.
Speedovershootratio
fll(b)b = Ne dt.oFigure11.- Concluded.Variationof integralcriteriawithspeedovershootratioforspeed- fuel:f’lowcontrol.
.
-
. 35
1+co.-l-+
6. , i , ,Tailpipegastemperature
5.
Controlintegraltimeconstant,
4.– ~c)sec
4.0- .1.511.0Y
3 .75- ..5 >
2 . /Simulatedbladetemperature;
/ controlintegraltimeconstant,1.5sec
1.1.0 1.2 1.4 1.6 1.8 2.0
Figure12. -ratiowithcontrol.
Speedovershootratio
Variationof tailpipe-gas-temperatureovershootspeedovershootratiofor speed- fuel-flow
-
(Aal
>,,. ,, I “Em”: .’ ‘
-
4161
(8)Ommal impal tb Ocmtant, 2.0 Immldsl locm Sdn, 1.s.
F&m 14. - Tz=uimt data for sund - fud-flw amkml with lW.tad,mb of N of M ml. for sw timun in ui@n -S ~.(A-1
-
I
1
, ! -c9-rP ,, ,I
-
I , #4161 1
g
2W
1%m
CNco
-
-.
l!-0
1
!,
19-E*,,
!3$JI
mlm
-
C{-6 ,I
4161 4
,: ,,
-
.7
.6
.5
.4
3..
1 IControlintegraltimeconstant,
TC9cec
;.5
1.0
2.04.0
.
\
%\ .
\
F- -——
1.0 1.2 1.4 1.6 1.8 2.0
.
,-
Speedovershootratio —
Figure X5. - Variation of speedovershootratio with time tofirst zero for speed- fuel-flow control with limited rateof changeof fuel valve. r-, .,-. --
-
NACATN 3926 43
dcol-l+
8
\o
d
Controlintegraltimeconstant,
Tcjsec
14 0.5
12.1.0
10–4.0
2.0\
81.0 1.2 1.4 1.6 1.8 2.0
Speedovershootratio
(a)aJ
m ~2 dt
0=”
Figure16.- Variationof integralcriteriawithspeedover-shootratiofor speed- fuel-flowcontrolwithlimitedrateof changeof fuelvalve.
-
44
&
i IControlintegral I
Tc>aec
I
timeconstant,
4.0~2.0
\ !
I
%
2.0
,.0
I I
1.0 1. 2 1.4 1.6 1.8 2Speedovershootratio
/1 Im
(b) b Ne dt.o —
.0—
Figure16.- Concluded.Variationof integralcriteriawithspeedovershootratioforspeed- fuel-flowcontrolwithlimitedrateof changeof fuelvalve.,
.
.
-
NACATN 3926 45
0%-l-PaM
IControlintegral
6 timeconstant,Tcjsec
4.0
5. /
4
3
2 / {
71.0 1.2 1.4 1.6 1.8 2.0
Speedovershootratio
Figure17. - Variationof tailpipe-gas-temperatureovershootratiowithspeedovershootratiofor speed- fuel-flowcon-trolwithlimitedrateof changeof fuelvalve.
-
.-. .
(8) ccnwol inte~l * wawut, 0,1 nemndi k-w pin, 7.5.
~~ 1S. - ~nient d4* fm emad-ama oontml type I for ntm inomam h .ngine set opead.
T9iw , ,1
-
, , $ 4161 aI
1
(b)c~ml M.awal tti .msksnt,0.1aoooml;lmp gal.,1?.s.
Fi.sorO18.- Continued.Franalentdatafarapaad+reaoontroltypaI faruterilnoreaseIn.znglnnmetapaed 11-4
-
.
(0) control lntegra time Oommt, 0.5 moond; loop gsln, 17.5.
- le. - Cmoluded. !I’ramlantchtY fm weed+- Cmh’d tyrm I for at+ horease h ●I@M set npeml
T9H’
-
*
1.6
1.4
1.2
l,C
0.5
.25 \
o
Figuretrol
4
,CJ-7 ,
—
— _
iz 16hop @iII, KL
—
—
47161
—
2
—
.
19. - Variationof speed overshootratio with loop gain for speed-axeacon-type I.
I
-
ao
Controlintegraltime constant,
see
112
10\ 1
.10
8-
6.0 4 8 12 16 20 24 28
IOOp gain, KL
(a) aJ
N: dt.0
Figure 20. - Variationof integral criteriawith loop gain for speed-areacontroltype I.
t
CN
1%0)
I1 191P
, #
-
, ,M-7 :Rck 4161 , ,
I Controlintegraltime constant, I I I I I IIsec
12 0.10 .25 .50
10
\ \\
\
8
6
!4CA
%O-I
—
—
o 4 8 12 16 20 24 28I.oopgain, KL
J’11.
(b) b Ne dt.
Figure 20. - Concluded. Variationof integralcriteriawith loop gain for speed-areacontroltype I.
I
-
(a) Cantrol integral time conatar$, 0.1 flee.nd;Imp @n, 7.5.
Fi~e 21.- Trarmientdetafor speed-areaoontroltypeII for nte~inareasein em aet ped.
.-C9T7 .
-
. .“1
4161 * e
u1%m
(b)controllntagral”tlmo.onstant,0.5 e.ea.nd; loop gain, 7.5.
Figure 21. - Cmtmmd. !Mneiont data fm swad-=ea control type II fm ste9 ~AaSa ~ en~e set sWd.
-
. .
(a)Controlintegralthe oormtmt, 1.0 sec!ond;loop gain, 7.5,
F@ure 21. - Concluded.Trensientdatafor speed-areaocmtroltne II for stepIncreasein engineset speed.
. -Em‘1
-
I, > . 4161 1 .
1.4- ,Control integraltime constant,
~c>sec
o.i1.2. \
.25
.ko
1.0.0 4 8 12 16
Mop gain, KL
Figure 22. - VBJ?ia_biOUof speed overshoot ratio withtrol type II.
20 24 28 ~
loop gain for speed-areacon-
1m(n
I
I
,, ,, I 1 -.
-
.
olm
12,Controlintegraltime constant,
sec
i.o10.
I.5
.i
8.\
60 4 8
Figure 23. - Vaxiationofspeed-areacontroltype
12mop gain,
IntegralcriterionII.
KL20 24
. .I
1‘“2a Ne dt with bop @n for28