NASA Technical Memorandum 4705
July 1995
Flight Assessment of the Onboard Propulsion System Model for the Performance Seeking Control Algorithm on an F-15 Aircraft
John S. Orme and Gerard S. Schkolnik
Technical Memorandum 4705 July 1995
Flight Assessment of the Onboard Propulsion System Model for the Performance Seeking Control Algorithm on an F-15 Aircraft
John S. OrmeGerard S. SchkolnikNASA Dryden Flight Research CenterEdwards, California
FLIGHT ASSESSMENT OF THE ONBOARD PROPULSION SYSTEM MODEL FOR THE PERFORMANCE SEEKING
CONTROL ALGORITHM ON AN F-15 AIRCRAFT
John S. Orme* and Gerard S. Schkolnik†
NASA Dryden Flight Research CenterEdwards, California
Abstract
Performance Seeking Control (PSC), an onboard,adaptive, real-time optimization algorithm, reliesupon an onboard propulsion system model. Flightresults illustrated propulsion system performanceimprovements as calculated by the model. Theseimprovements were subject to uncertainty arising frommodeling error. Thus to quantify uncertainty in the PSCperformance improvements, modeling accuracy must beassessed. A flight test approach to verify PSC-predictedincreases in thrust (FNP) and absolute levels of fan stallmargin is developed and applied to flight test data.Application of the excess thrust technique shows thatincreases of FNP agree to within 3 percent of full-scalemeasurements for most conditions. Accuracy to theselevels is significant because uncertainty bands may nowbe applied to the performance improvements providedby PSC. Assessment of PSC fan stall margin modelingaccuracy was completed with analysis of in-flight stalltests. Results indicate that the model overestimates thestall margin by between 5 to 10 percent. Because PSCachieves performance gains by using available stallmargin, this overestimation may represent performanceimprovements to be recovered with increased modelingaccuracy. Assessment of thrust and stall marginmodeling accuracy provides a critical piece for acomprehensive understanding of PSC’s capabilities andlimitations.
Acronyms
ADECS Advanced Engine Control System
*Aerospace Engineer. Member AIAA.†Aerospace Engineer. Member AIAA.Copyright 1995 by the American Institute of Aeronautics and
Astronautics, Inc. No copyright is asserted in the United States underTitle 17, U.S. Code. The U.S. Government has a royalty-free license toexercise all rights under the copyright claimed herein for Governmentalpurposes. All other rights are reserved by the copyright owner.
1American Institute of Aero
AEDC Arnold Engineering Development Center, Tullahoma, Tennessee
CDP Component Deviation Parameters
CPSM Compact Propulsion System Model
DEEC Digital Electronic Engine Control
DEFCS Digital Electronic Flight Control System
DFRC Dryden Flight Research Center, Edwards, California
HIDEC Highly Integrated Digital Electronic Control
Kf Kalman filter
MDA McDonnell Douglas Aerospace, St. Louis, Missouri
OFP Operational Flight Program
PSC Performance Seeking Control
PW Pratt & Whitney, West Palm Beach, Florida
S/MTD STOL Maneuver Technology Demostrator
TTW total temperature and weight flow method
VMSC Vehicle Management System Computer
Nomenclature
AJ nozzle throat area, in2
CDecs environmental control system bleed air drag
CDpb plenum bleed drag
nautics and Astronautics
CDtrim combined inlet spillage and trim drag
CIVV compressor inlet variable guide vane angle, deg
D total aircraft drag, lb
DINL incremental inlet spillage drag, lb
DNOZ nozzle drag, lb
DSTAB incremental stabilator and inlet cowl trim drag, lb
EPR engine pressure ratio, PT6 /PT2
Fex excess thrust, lb
FG gross thrust, lb
FN net thrust, lb
FNP net propulsive force, lb
FPR fan pressure ratio, PT2.5 /PT2
FPRCIVV off schedule CIVV factor
FPRdis inlet distortion factor
FPRi installed stall line FPR
FPRo operating FPR
FPRRe Reynolds factor
FR ram drag, lb
H pressure altitude, ft
M Mach
N1 fan rotor speed, rpm
N1C2 fan rotor speed, corrected to station 2, rpm
N2 compressor rotor speed, rpm
Nx longitudinal acceleration, g
PD percent difference, percent
PLA power lever angle, deg
PS2 static pressure at engine face, psi
PT total pressure, psi
RCVV rear compressor variable vanes, deg
SMF fan stall margin
SMHC high-pressure compressor stall margin
SSVM steady-state variable model
SVM state variable model
TSFC thrust specific fuel consumption, sec–1
TT total temperature, °R
VT true airspeed, kn
WACC DEEC-calculated corrected fan airflow, lb/sec
WCfan corrected fan airflow, lb/sec
WChpc corrected high pressure compressor airflow, lb/sec
WF gas generator fuel flow, lb/hr
WFA/B afterburner fuel flow, lb/hr
Wfan fan airflow, lb/sec
Wt aircraft weight, lb
Wt0 zero fuel aircraft weight, lb
Wtf total indicated fuel weight, lb
α angle of attack, deg
β angle of sideslip, deg
ρ inlet cowl angle, deg
∆3 inlet third ramp angle, deg
η random effects
Subscripts
meas measured value
model modeled value
Superscript
' postflight model-predicted value
Prefix
∆ perturbation or difference
2American Institute of Aeronautics and Astronautics
EC90 312-3
Fig. 1. The F-15 highly integrated digital electroniccontrol aircraft.
Suffix, PW1128 Engine Station Numbers (fig. 2)
2 fan inlet or engine face
2.5 compressor inlet
3 compressor discharge
4 combustor exit
4.5 low-pressure turbine inlet
6 augmentor inlet
7 exhaust nozzle throat
Introduction
Optimizing the integrated control variables is animportant way to improve the performance of anairplane. For example, the Advanced Engine ControlSystem (ADECS), flight tested by NASA in the early1980’s, was the first to use digitally communicatedinformation between the flight control and enginecontrol computers to increase engine performance.1 Ashortcoming of the ADECS design was that it was basedon predetermined control schedules. This shortcomingmade the system unable to distinguish and optimize forunique operating characteristics between differentengines of the same class (for example, two F100engines with different levels of degradation).
Recently, an advanced optimization routine, calledPerformance Seeking Control (PSC), was flownonboard a NASA F-15 aircraft.2 The PSC is amodel-based, real-time adaptive, onboard propulsionsystem optimization algorithm with in-flight thrustcalculation capability. The PSC method of in-flightthrust determination closely resembles the totaltemperature and weight flow (TTW) method firstdeveloped during the XB-70 flight test program.3
Limited assessment of PSC modeling accuracy has beenreported for calculated internal engine temperatures,pressures, and rotor speeds as well as for static thruststand measurements.4–6 Accuracy of the PSC in-flightFNP calculation needs to be assessed to evaluate thefuture value of the PSC concept.
The PSC performance improvements6–10 derivedfrom reducing stability margins are based on modelcalculations that may be in error by an unknownamount. During flight test of the PSC algorithm, engineperformance was improved by reducing the calculatedfan stall margin (SMF), thus operating closer to thepredicted surge line. Accuracy of the PSC SMF modelneeds to be assessed.
3American Institute of Aer
Flight test techniques and analysis were designed byNASA Dryden Flight Research Center (DFRC) to assessthe PSC modeling accuracy of net propulsive force(FNP) and SMF. A methodology similar to the“accelerometer method” derived by Beeler, et al.,11 andapplied by Ray for evaluating thrust calculation methodsduring throttle transients12 was adopted for analysis ofquasi-steady state FNP estimations during levelaccelerations. Accuracy of modeled increases in FNPwas determined by comparisons with measuredincreases of excess thrust. The PSC stall marginmodeling was assessed by intentionally stalling the fanin-flight and analyzing the SMF estimate. At the pointthe fan stalls, the estimate should be zero.
This paper describes the results of an independentassessment of the PSC thrust calculation and SMFmodel accuracy. Flight tests were flown from Mach 0.5to Mach 2.0 at intermediate and maximum afterburningpower settings.
Aircraft and Engine Description
The PSC algorithm has been tested on thetwin-engine, high-performance McDonnell DouglasAerospace (St. Louis, Missouri) (MDA) F-15 airplane(fig. 1). The DFRC modified the F-15 airplane to theHighly Integrated Digital Electronic Control (HIDEC)for the purpose of integrated control research. UniqueHIDEC features include a digital electronic flightcontrol system (DEFCS), two Pratt & Whitney (PW)(West Palm Beach, Florida) F100-PW-1128 engineswith digital controls, and the digital data buses.
onautics and Astronautics
Additional information on the F-15 airplane can befound in ref. 1.
Figure 2 shows the F100-PW-1128 engine. This low-bypass ratio, twin-spool, afterburning turbofan engine isa derivative of the F100-PW-100 engine. The engine iscontrolled by a full-authority digital electronic enginecontrol (DEEC) that is similar to the current productionF100-PW-220 engine controller. The DEEC providesopen-loop scheduling and closed-loop feedback controlof corrected fan speed (N1C2) and engine pressure ratio(EPR) by way of the fuel flow (WF) and the nozzle area(AJ). The compressor inlet variable guide vane (CIVV)and rear compressor variable vane (RCVV) positions arescheduled with rotor speeds using open-loop control.The DEEC software has been modified to accommodatePSC trim commands; however, the normal DEECcontrol loops, such as N1C2 and EPR, have notbeen modified. Reference 13 gives a more detaileddescription of the PW1128 engine.
Performance Seeking Control Algorithm
The PSC as developed by MDA and PW and installedon the NASA F-15 HIDEC aircraft is a model-based, adaptive algorithm which performs real-timeoptimization of the propulsion system during quasi-steady-state operation. Information available from theonboard airdata, flight control, and engine controldigital computers is shared with the PSC software.
Essentially, the algorithm consists of an estimationroutine to update propulsion models and an optimizationroutine to optimize the controlled variables.
The PSC algorithm and software was hosted on aVehicle Management System Computer (VMSC).Figure 3 summarizes implementation of the PSCalgorithm into the HIDEC aircraft. The three majoralgorithm elements as they reside in the VMSC consistof the identification, modeling, and optimizationcomponents. Four optimization modes were used duringthe flight testing. They are the (1) maximum thrust modefor enhanced aircraft acceleration, (2) minimum fuelmode for improved fuel efficiency during cruise, (3)minimum turbine temperature mode for extendingengine life during cruise, and (4) rapid decelerationmode for reductions in time to decelerate fromsupersonic to subsonic conditions. Each of these modesproduced beneficial performance improvements,6–10 butall the accrued benefits equally rely upon on accuratein-flight thrust calculation. Modes 1 and 4 seek tomaximize or minimize FNP, or thrust, respectively.Meanwhile, modes 2 and 3 are required to maintain aconstant referenced level of thrust as an optimizationconstraint. Thus, errors in the FNP calculation willaffect the results for all the optimization modes.
The entire PSC algorithm is duplicated for left andright propulsion systems, and no cross communicationexists between the models, identification, or
4American Institute of Aeronautics and Astronautics
Fig. 2. The F100-PW-1128 engine with digital electronic engine control.
Nozzle7
PSC controlled variables
AfterburnerLow-pressure turbine
64.54
High-pressure turbine
Combuster
Compressor
FanStation
numbers
32.52
DEEC sensorsPS2 TT2 N1 N2 PT4 TT4.5 PT6
CIVV RCVV WF WFAB AJ
950200
Fig. 3. The PSC implementation and process flow diagram.
1 2 3 4 5 6 7
Background processingForegroundprocessing
Kalman filter
Identification Modeling OptimizationLooping
Optimal propulsion trimsfrom the VMSC
Vehicle Management System Computer
Flight measurements
PSC control effectors:1. Inlet cowl2. Third ramp3. Variable vanes4. Variable vanes
5. Core fuel flow6. Afterburner fuel flow7. Nozzle throat area
950201
Compact PropulsionSystem Model
Linear programming
optimization. The PSC trims are applied to thepropulsion system approximately five times a secondsubsonically and two times a second supersonically. Thereduced trim rate application during supersonicoperation is the result of including the large inlet model.Detailed descriptions of the PSC algorithm have beenreported,7, 8, 14 so only selected aspects related to modelvalidation are described in the following subsections.
Identification
The adaptive feature of the PSC algorithm isprimarily provided by a Kalman filter (Kf), whichattempts to match an onboard compact propulsionsystem model (CPSM) to the characteristics of theactual engine. The filter estimates five componentdeviation parameters (CDP) that represent deviationsfrom nominal engine operation. These parametersconsist of low spool efficiency adder, high spoolefficiency adder, fan airflow adder, compressor airflowadder, and high turbine area adder.15 Input to the Kf wasconstrained to the suite of sensors for the standardDEEC measurements (fig. 2).
Kalman filter estimates of the CDP comprise a set ofintermediate variables passed to the modeling portion ofthe PSC algorithm. The model is incrementally adjustedwith the CDP to more nearly accurately matchmeasured engine operation. This adaptive feature allows
5American Institute of Aer
for the PSC optimization to be applied to any F100series engine independent of state of degradation.
The CDP are defined to be zero for a nominal engine(that is, an engine exactly like the model). The CDP aresensitive to any measured difference from the nominalengine including engine-to-engine variations, enginedeterioration, measurement bias, Reynolds effects, andmodeling errors.5 The influence of any single effect onthe CDP cannot be identified because of the limitednumber of measurements available from the standardDEEC.16
Modeling
The second major component of the PSC algorithm isthe CPSM. The CPSM integrates separate steady-statemodels of the engine and inlet. The engine modelconsists of the piece-wise linear Steady-State VariableModel (SSVM) and nonlinear engine model. Thecurrent linear point model is scheduled with flightmeasurements. The CDP are included as input to theSSVM to improve modeling accuracy.
The SSVM is the only part of the modeling to beadjusted with input from the Kf. Engine parametersensitivity to changes in control effectors and CDP isstored for each point model. Because the PSC controllerusually trims the engine off its nominal operating line,
onautics and Astronautics
the SSVM also accounts for off-nominal engineoperation.*
Following completion of the linear SSVMcalculation, additional nonlinear calculations are made.Nonlinear calculations use a combination of analyticalequations and empirically derived data tables. Input tothe nonlinear routines includes measurements andSSVM output. If a variable is both measured andestimated, the flight measurement is used in thenonlinear calculations. Among the nonlinear output areFNP, its force components, and SMF. Net propulsiveforce and SMF are important parameters for the PSCoptimization.
A compact inlet model provides calculated inletperformance for the variable three-ramp F-15 inlet.During subsonic operation, inlet performance ischaracterized by inlet pressure recovery and inletspillage drag. Because the inlet ramps are consideredoptimized for subsonic operation, the inlet ramps are notcommanded by PSC subsonically. At supersonicconditions, however, PSC commands the inlet ramps toachieve the desired performance goal. For supersonicoperation, inlet performance includes incrementalstabilator trim drag and the inlet stability margins ofshock displacement ratio and percent critical mass flow.Because of the additional performance considerations atsupersonic conditions, the CPSM includes a morecomplicated and detailed version of the F-15 inletmodel during supersonic operation. Integration of theengine and inlet is accomplished by passing the SSVMoutput of fan airflow to the inlet model as an input, andinlet model-estimated pressure recovery is sent to theSSVM and nonlinear engine model.
Optimization
Information of the CPSM-modeled plant is passed tothe optimization logic for optimal trim determination.Certain CPSM output variables are treated asconstraints. Others are treated as an objective or costfunction. The FNP is used as an equality constraint forthe minimum fuel and minimum turbine temperaturemodes. Maximum FNP and minimum FNP are theobjectives of the maximum thrust and rapid decelerationmodes, respectively. Certain important constraints, suchas SMF, must not be allowed to become negative. Afeasible solution (that is, one that does not violate any of
6American Institute of Aer
*Yonke, William A. and Nobbs, Steven G., Performance SeekingControl (PSC) Final Report, MDC 94B0003, McDonnell DouglasAerospace, St. Louis, Missouri, Jan. 1994. This report is not availableto the public. Contact the authors with queries regarding this report.
the constraints) is sought by the optimization logic withsuccessive calls to the CPSM between trim application.
Optimal trims are applied when the objective functionimproves and after a predetermined number of calls tothe CPSM, referred to as looping. Optimal trim ratedepends upon flight condition and the number of loops.At subsonic conditions, there are six loops to eachoptimization cycle. At supersonic conditions, there arethree loops per cycle. It takes longer to complete threeloops supersonically than it does to complete six loopssubsonically. This difference is primarily caused byinclusion of the larger compact inlet model and theadditional controls for inlet ramps and afterburner fuelflow. Timing of the six-loop subsonic operation wasbetween 0.2 to 0.3 sec. For the supersonic three-loopprocess, timing was between 0.5 and 0.7 sec.
Net Propulsive Force Assessment Technique
Because the FNP calculation is given special attentionin this report, a closer look at the calculation iswarranted. Referring to figures 4(a)–4(e), calculationsfor gross thrust (FG), ram drag (FR), nozzle drag(DNOZ), inlet drag (DINL), and incremental stabilatordrag (DSTAB) are combined in equation 1 to define FNP.
FNP = FG – FR – DNOZ – DINL – DSTAB (1)
The different propulsion-related forces in equation 1are assumed to act along the flightpath. This assumptionis reasonable for low angles of attack.
The PSC method of in-flight thrust calculation, likethe TTW, is derived from an energy balance through theengine and uses measured parameters as well as knownengine characteristics. As adopted for application to theF100-PW-1128 engines, inputs to the TTW calculationwere limited to available production instrumentation.Thus, a number of inputs to the TTW calculation aretaken directly from the SSVM outputs. Theseparameters are the model predicted fan airflow (WCfan),augmentor inlet total temperature (TT6), combustor exittotal pressure (PT4), and engine face total pressure(PT2). Table lookups, analytical gas dynamic equations,and empirically derived equations are used in thecomputation of FG.
The FR is defined and calculated as the product of trueairspeed, VT, and fan airflow, Wfan. The DNOZ is modeledas a table look-up of wind-tunnel jet effects data andscheduled with flight-measured input. The DINL iscomposed of three drag terms: CDtrim, CDpb, and CDecs.The inlet spillage drag combined with trim drag ismodeled and referred to as CDtrim. The plenum bleed
onautics and Astronautics
(a) Gross thrust.
CVNOZNozzleanalysis
WFA/B
WFcore
Total enginemass flow x Vexit
Ideal FG
++
Afterburner ducttotal pressure loss,gamma calculation,
and afterburner fuel-to-air ratio
VexitPT6PT4
P0
W fan
WCfan
PT2
TT2
+ Calculate FG FG
CalculatedMeasured
TT6
Corrections tofan airflow
950202
(b) Ram drag.
Wfan
WCfan
TT2
Correctionsto
fan airflow
CalculateVT
CalculateFR
M
950203
FR
PT2
VT
(c) Nozzle drag.
M
Aj
P0
Calculate q
CalculateDNOZ
q
950204
PT6
Tablelook-ups
ofwind-tunneljet effects
data
q( )DNOZ
DNOZ
Fig. 4. The PSC onboard in-flight thrust model.
drag (CDpb) calculation involves complicated geometricrelations between the variable inlet ramps and flightconditions. Calculation of the final inlet drag terminvolves table lookups to account for environmentalcontrol system drag (CDecs), drag which is associatedwith the bleeding of air from the inlet. The incrementalstabilator drag component of FNP accounts for thecoupling of the propulsion system with the airframe. Ifthe inlet is trimmed to an off-scheduled position, anincremental lift is produced, thus imparting a pitching
moment on the airframe. To offset the pitching momentand maintain level flight, the stabilator is used.Stabilator is repositioned to produce a counteractingpitching moment; however, stabilator drag is alsoaffected. The DSTAB is defined as the incrementalstabilator trim drag plus the incremental inlet drag. Bothof these drags are associated with moving the cowloff-schedule. Note that the stabilator is not directlycontrolled by PSC but rather depends on the autopilot orpilot to trim out the pitching moments. Wind-tunnel
7American Institute of Aeronautics and Astronautics
(d) Inlet drag.
(e) Incremental stabilator drag.
Fig. 4. Concluded.
DINL
M
P0
WCfan
α
ρ∆3
Table lookupsfor ECS drag
Wind-tunneldata trim
drag tablelookup andspill drag
Calculate qq
950205
Computed plenumbleed drag
Calculatetotal
inlet drag
CDecs
CDpb
CDtrim
Hp
CalculatedMeasured
M
P0
WCfan
ρ
950206
Calculate qq
DSTABDSTABqWind-tunnel
data tablelookup
Calculate incrementalstabilator trim drag
caused by off-schedule inlet
( )
CalculatedMeasured
data of DSTAB are computed with table lookups andscheduled with measurements and SSVM variables.
Excess Thrust Method
The accuracy for the onboard PSC FNP estimate wasassessed with an analysis of aircraft excess thrust, Fex.The analysis technique consists of a comparisonbetween modeled and measured increases inlongitudinal force from maximum thrust modeacceleration tests. For 1-g flight, FNP relates tolongitudinal acceleration, Nx, through the followingexpression:
Fex = FNP – D = Nx × Wt (2)
where D represents total aircraft drag excluding thepropulsive drag terms.
The method chosen to evaluate modeled FNPaccuracy compares back-to-back, 1-g accelerations withand without PSC engaged in the maximum thrust mode.A significant increase in acceleration was experienced.The first acceleration was completed as a baseline, andthe second was completed with the PSC maximumthrust mode engaged. With PSC engaged, predictedincreases in FNP from the baseline were also reflectedas measured increases in Nx. From equation (2) duringsuch PSC operation, only the aircraft drag term, D, isunknown. However, because none of the PSC trimsdirectly influence airframe drag, changes in drag
8American Institute of Aeronautics and Astronautics
between the two accelerations are assumed to be zero;that is, ∆D = 0. Then, the following expression relatesan increase in FNP with an associated increase in Nx:
∆Fex = ∆FNP = ∆(Nx × Wt) (3)
During in-flight tests, a conventional linearaccelerometer measures Nx, but there is no instrumentavailable to measure FNP. For this reason, twocalculations of ∆Fex will be made, one based onmeasured Nx, ∆Fexmeas, and one from modeled FNP,∆Fexmodel according to equations 4 and 5.
∆Fexmodel = ∆FNP = FNPon – FNPoff (4)
and
∆Fexmeas = ∆(Nx × Wt) = (Nx × Wt)on – (Nx × Wt)off (5)
where on and off refer to acceleration data with andwithout the PSC maximum thrust mode engaged.Aircraft gross weight, Wt, is calculated as the sum ofzero fuel weight, Wt0 and total indicated fuel, Wtf.
Differences between the measured and the modeled∆Fex represent modeling error. Ideally, modeling errormay be presented as a percentage of measured values;however, any percentage error calculation containing adenominator that approaches zero may be misleading.As described in the Results and Discussion section, the∆Fexmeas levels approach zero at some flight conditions.Thus, calculations with ∆Fexmeas in the denominator willbe inflated. To alleviate the potential for misinterpretingthe results, modeling accuracy will be presented interms of percentage of full-scale measured Fex. TheFexmeas value is taken from the test completed withoutthe PSC optimization selected, or (Nx × Wt)off. Keep inmind the percentage difference (PD) value does notrepresent conventional modeling error. Note also thatPD values are less than percentage error values. The PDwas calculated according to equation 6.
PD = [(∆Fexmodel – ∆Fexmeas)/(Nx × Wt)off] × 100 (6)
A secondary benefit of reporting modeling accuracyin terms of full-scale Fex is that PD analysis results canbe directly applied to the prior PSC findings. Forexample, prior findings identified a 10 percent ∆Fexincrease for maximum thrust mode operation at Mach0.7 and an altitude of 30,000 ft. The PD results indicatea 2 percent level of ∆Fex modeling accuracy at thiscondition. Thus, bounds on the optimization results maynow be given. The maximum thrust mode achievesbetween an 8- to 12-percent increase in Fex.
Test Conditions
Accelerations were conducted for the PSC maximumthrust mode at two power settings and three altitudes.Flight test data are acquired from onboardinstrumentation of PSC model outputs and sensorreadings. Table 1 summarizes the conditions for thetests.
Table 1. Test conditions.
Data Standardization
Data standardization and corrections to referenceconditions are addressed before performance data are tobe analyzed. The FNP validation test points were flownin sets of two through the same air mass at an equivalentaltitude. This procedure was used to reduce theinfluences of changing atmospheric conditions onpropulsion system operation and aircraft drag; however,aircraft total drag differed noticeably between the twoaccelerations because of differences in aircraft weight.These drag differences require standardization ofmeasured test day excess thrust to a standard weightcondition. A simplified performance model of the F-15aircraft was used to accommodate changes in aircraftdrag resulting from differences in aircraft mass.17
Error bands associated with this method previouslydeveloped for quantifying acceleration performanceimprovements10 should be less than 1.4 percent at analtitude of 45,000 ft and 3.4 percent at an altitude of30,000 ft. The PSCoff weight was selected as thestandard weight to which the PSCon test data werecorrected. The following correction was applied to testday (Nx × Wt)TEST with PSCon to obtain (Nx × Wt)on:
(Nx × Wt)on = (Nx × Wt)TEST + ∆DCORR (7)
where ∆DCORR = D'off – D'on. The D' is the postflightmodel predicted drag. Errors in the drag correction areexpected to be small. Only the difference in modeleddrag is being applied as a correction.
Testpoint
Powersetting
Altitude,ft
Mach numberrange
1 Mil 30,000 0.50 to 0.95
2 Max 30,000 0.60 to 1.60
3 Max 45,000 0.80 to 2.00
9American Institute of Aeronautics and Astronautics
Uncertainty Analysis
Measurement uncertainty of ∆Fexmeas must be knownif Fexmeas is to be considered the true value ofcomparison for ∆Fexmodel. To that end, an uncertaintyanalysis was performed to gauge the significance ofmeasurement errors on the ∆Fexmeas calculation. Table 2lists the measured inputs to the ∆Fexmeas calculation.Measurement of Nx has a manufacturer’s quotedprecision of ±0.50 percent full-scale. The precision ofthe total indicated fuel weight, Wtf was ±3.50 percentfull-scale. Aircraft empty weight, obtained in specialweight and balance tests, has a ±0.07 percent full-scaleprecision.
Measurement uncertainty of ∆Fexmeas was less than1.15 percent at test condition 1. This test representstypical data to be analyzed for excess thrust modelingaccuracy. Levels of uncertainty are expected to besimilarly small for the other test conditions.
Table 2. Measurement ranges and uncertainties of∆Fex input parameters.
Fan Stall Margin Assessment Technique
Engine stability margin is expressed in terms of stallmargin remaining. Stall margin is defined as thedifference between the stall and operating pressureratio, at constant airflow, divided by the reference(operating) pressure ratio. Stall margin may be definedfor the fan and for the compressor, but only the SMF isdescribed in this report. Fan stall margin is one of thecritical parameters output from the CPSM. This marginis used as a constraint during the optimization process.Calculated stall margin must never be allowed tobecome negative while optimizing the selectedperformance index. A negative calculated stall marginmay cause the engine to stall. The intentional fan stalltest was conducted to quantify the overall accuracy ofthe PSC SMF model.
Much of the performance gain achieved by PSC atsubsonic speeds is accomplished by uptrimming theengine pressure ratio (EPR), effectively increasing the
Parameter RangeUncertainty, percent
of full scale
Nx –1 to 1 g ±0.50
Wt0 0 to 280,000 lb ±0.07
Wtf 0 to 10,800 lb ±3.50
1American Institute of Aer
fan pressure ratio (FPR). If FPR is increased at aconstant airflow, then stall margin is reduced (fig. 5(a)).Reduced stall margin was observed and reported forthree of the PSC optimization modes,7 especially atsubsonic flight conditions. Individual components of thestall margin calculation can not be assessed because ofthe insufficient information available during flight test.Even so, reviewing the methodology of the PSC SMFmodel to understand the possible sources of error isworthwhile.
The PSC incorporates the PW stability auditmethodology that assesses the effects shown infigure 5(a) to predict SMF. Base stall line data obtainedfrom engine testing at Arnold Engineering DevelopmentCenter (AEDC) (Tullahoma, Tennessee) of a one-sixthscale F-15 model in 1970–1971 were tabulated andstored onboard the VMSC.1 A base stall line definition isdetermined in real-time with SSVM-corrected fanairflow as input to a table lookup. The base stall line isadjusted by a Reynolds factor (∆FPRRe), CIVV factor(FPRCIVV), random effects, η, and distortion effects(∆FPRdis) to calculate the installed stall line. The∆FPRRe accounts for off-design airflow temperature anddensity. A decrease in Reynolds number lowers the fanstall line by decreasing its fan pressure ratio, FPR. TheFPRCIVV describes the effects of off-schedule guidevanes. Opening the variable guide vanes of the fanreduces the FPR of the stall line. The random effectsinclude a 1-percent margin for engine-to-enginevariations and a 1-percent margin for control tolerances.
Because there are no known means of determiningrandom effects in-flight, η is fixed at a constant 2percent. Distortion effects are made to adjust the stallline for distorted airflow entering the engine face. ThePSC uses an inlet distortion model which takes intoaccount angle of attack, α , angle of sideslip, β, Machnumber, and compressor inlet variable vane angle.Tabulated results from the AEDC tests wereincorporated as part of the PSC inlet distortion model.An increase in ∆FPRdis also causes the stall line FPR tobe reduced. All decrements to the base stall line aredetermined in real-time by the PSC model.
Besides adjustments to the base stall line, a fixedamount of stall margin buffer is added to the nominaloperating line. Pratt & Whitney suggested this additionas a safety precaution to prevent stalling because ofuncertainties in the audit methodology and to increasethe buffer against the statistical nature of stalls.Implementation of the minimum allowable stall marginremaining has been as a fixed value of 4 percent at alldry power settings and 6 percent at maximumafterburner power. At partial afterburner power settings,
0onautics and Astronautics
the limit is increased to 10 percent because of stabilityconcerns about augmentor sequencing. For example atmaximum afterburner power, at an airspeed of Mach0.9, and an altitude of 30,000 ft, an approximately18-percent stall margin is available for uptrimming(fig. 5(b)).
The final installed stall line is defined as
FPRi = FPRCIVV × [1 + ∆FPRRe] × [η] – ∆FPRdis (8)
Figure 6 shows the inputs and calculation process forFPRi. The operating fan pressure ratio is defined as
11American Institute of Aeronautics and Astronautics
(a) Typical operating conditions.
(b) Stall margin available for maximum afterburner power at Mach 0.9 and an altitude of 30,000 ft during cruise.
Fig. 5. Fan stall margin audit for the F100-PW-1128 engines during the PSC program.
Minimum allowable remaining
Reynolds effects
Engine-to-engine variation (random)
Control tolerances (ra
ndom)
Minimum case inlet distortion
Off-nominal vane effects
FPR
WCfan
Installed stall line
Nominal operating line
Base stall line
950207
Available for uptrim
Random
Worst case inlet distortion
Augmentor sequencing
Remaining 18 %
3 %
2 % Random
Inlet distortion
Remaining
With PSCWithout PSC
950208
Margin availablefor uptrim
5 %
6 %
2 %
2 %
12American Institute of Aeronautics and Astronautics
Fig. 6. The PSC onboard in-flight installed stall line calculation.
α
β
Distortion effects, ∆FPRdisdis
Random effects,η
950209
WCfan
PT2
TT2
M
N1C2
Base stall line
Reynolds factor, ∆FPRReR e
CIVV
CIVV factor, FPRCIVVIVV
(Assumed constant)
Engine-to enginevariations
Control tolerances
Installed stall line, FPRi
CalculatedMeasured
FPRo = PT2.5/PT2 (9)
where PT2.5 is determined from the SSVM, and PT2 isoutput from the compact inlet model. Fan stall margin isdefined as
SMF = [FPRi – FPRo] /FPRo (10)
Intentional Fan Stall Method
Procedurally, fan stalls must be intentionally inducedbecause the PW F100 class of engines are proven to berobust with respect to encountering stall and are veryunlikely to stall without some assistance. Errors inestimated SMF will be determined by comparing SMFvalues at the time of a stall event. To reduce thelikelihood of a single bad data point, the test wasrepeated so that a total of four stalls were evaluated.
Test Conditions
A steady-state cruise flight condition of Mach 0.9 andaltitude of 30,000 ft was chosen as the test condition tominimize the effect of outside factors and allow for avery controlled test. For the test engine to be stalled,power was set at maximum afterburner, while the other,nontest engine was modulated to maintain constantMach number. A preprogrammed set of progressivelyincreasing EPR trim values was selected until a fan stallwas detected. Mach 0.9 and an altitude of 30,000 ft isthe engine design point. At this condition, no ∆FPRReadjustment is made. Inlet distortion is expected to verysmall at this condition. The variable vanes are onschedule and not trimmed so that no FPRCIVVadjustments will be made. With these conditions, theSMF model provides its most accurate estimates.
Results and Discussion
Test results of the excess thrust test and theintentional fan stall test are discussed in the followingsubsections. The tests were designed to assess the FNPand SMF modeling accuracy of the PSC algorithm.
Excess Thrust Test
Figure 7 shows a comparison of two back-to-backaccelerations for test condition 1 with and without thePSC maximum thrust mode selected. The curve labeledPSCon is with maximum thrust mode selected. Thecurve labeled PSCoff is without maximum thrust modeselected. Time to accelerate from Mach 0.5 to Mach0.95 is reduced by 12 sec or about 15 percent with PSC,as illustrated by the PSCon curve of figure 7(a). Theroughly 0.03-g increase of Nx (fig. 7(b)) results fromsignificant increases in FNP (fig. 8(a)). The FNP gains
1American Institute of Aer
are achieved by uptrimming the engine and operatingwith reduced SMF (fig. 7(c)).
(a) Measured Mach number.
(b) Measured longitudinal acceleration.
(c) Modeled fan stall margin.
Fig. 7. Maximum thrust mode optimization results attest condition 1.
The model and the measured values of Fex showsustained increases from PSC maximum thrust modeapplication for most of the run. Measured increase in Fex
12 sec
Machnumber
.9
1.0
.8
.7
.6
.5
0 20 40Time, sec
950210
60 80
PSC offPSC on
Nx, g
.4
.3
.2
.1.5 .6 .7
Mach number950211
.8 .9 1.0
PSC offPSC on
Left enginestall margin,
percent
Right enginestall margin,
percent
30
20
10
0
30
20
10
0.5 .6 .7
Mach number950212
.8 .9 1.0
PSC offPSC on
3onautics and Astronautics
is nearly 1000 lb at Mach 0.7 (fig. 8(b)). At the sameMach number, the model predicts 850 lb of increase inFNP (fig. 8(a)). Throughout the acceleration, ∆Fexmodelis predicted less than ∆Fexmeas. A maximum temperaturelimit causes the Fex increases to taper off somewhat nearthe end of the acceleration.
Figure 8(c) illustrates the PD between measured andmodeled ∆Fex. Model accuracy is at 2 percent or less ofmeasured baseline excess thrust for most of the Machnumber range. As Mach 0.95 is approached, PDincreases to nearly 4 percent. Modeling accuracy isexpected to degrade above Mach 0.95 where the valueand uncertainty in propulsion system drag increase.
(a) Modeled increase in FNP.
(b) Measured increase in Fex.
(c) Percent difference of modeled to measured increasein excess thrust.
Fig. 8. Excess thrust results at test condition 1.
.5 .6 .7 .8 .9 1.0
∆FNP, lbor
∆Fexmodel
FNP,lb
Mach number
15,000
10,000
5,000
1000
500
0
950213
PSC offPSC onDifference between PSC off and PSC on
.5 .6 .7 .8 .9 1.0
∆(Nx x Wt), lbor
∆Fexmeas
Nx x Wt, lbor
Fexmeas
Mach number
15,000
10,000
5,000
1000
500
0
950214
PSC offPSC onDifference between PSC off and PSC on
.5 .6 .7 .8 .9 1.0
PD,percent
Mach number
10
5
0
950215
1American Institute of Aer
Figures 9(a)–9(c) present comparison plots of themeasured to modeled ∆Fex and PD for test condition 2.The ∆Fex levels lie between zero near Mach 1.0 and2000 lb at higher Mach numbers. The level of accuracyis less than 2 percent from low subsonic until aroundMach 1.3. At which point, modeling accuracy worsensto reach about 10 percent above Mach 1.45.
(a) Modeled increase in FNP.
(b) Measured increase in Fex.
(c) Percent difference of modeled to measured increasein excess thrust.
Fig. 9. Excess thrust results at test condition 2.
Results for test point 3 show Fex increases of about1000 lb for most of the acceleration (figs. 10(a)–10(c)).Modeling accuracy is generally within 3 percent belowMach 1.7. From Mach 1.7 to Mach 2.0 the level of
.6 .8 1.0 1.2 1.4 1.6
∆FNP, lbor
∆Fexmodel
FNP,lb
Mach number
50,000
25,000
0
2000
1000
0
950216
PSC offPSC onDifference between PSC off and PSC on
.6 .8 1.0 1.2 1.4 1.6Mach number
50,000
25,000
0
2000
1000
0
950217
∆(Nx x Wt), lbor
∆Fexmeas
Nx x Wt, lbor
Fexmeas
PSC offPSC onDifference between PSC off and PSC on
.6 .8 1.0 1.2 1.4 1.6
PD,percent
Mach number
15
5
0
950218
10
4onautics and Astronautics
accuracy steadily deteriorates with a maximumdifference of over 10 percent at Mach 2.0.
(a) Modeled increase in FNP.
(b) Measured increase in Fex.
(c) Percent difference of modeled to measured increasein excess thrust.
Fig. 10. Excess thrust results at test condition 3.
Numerous possible sources of error in the ∆FNPcalculation exist. The FNP calculation represents themost complicated parameter output from the PSC modeland relies upon more measurements and model inputsthan any other PSC calculated parameter. Contributionsto absolute FNP error caused by specific measurementinaccuracies and unmodeled physical phenomena areextremely difficult to quantify, especially from datagathered in-flight. This analysis makes no attempt to
∆FNP, lbor
∆Fexmodel
FNP,lb
.8 1.0 1.2 1.4Mach number
1.6 2.01.8
50,000
25,000
0
2000
1000
0
950219
PSC offPSC onDifference between PSC off and PSC on
PSC offPSC onDifference between PSC off and PSC on
∆(Nx x Wt), lbor
∆Fexmeas
Nx x Wt, lbor
Fexmeas
.8 1.0 1.2 1.4Mach number
1.6 2.01.8
50,000
25,000
0
2000
1000
0
950220
PD,percent
.8 1.0 1.2 1.4Mach number
1.6 2.01.8
10
15
5
0
950221
1American Institute of Aer
identify and specify distinct sources of error. Such ananalysis is beyond the scope of current activity;however, absolute FNP error will be partially offsetwhen taking differences between absolute levels of FNPand making comparisons of ∆FNP or ∆Fex.
All three of the tests analyzed demonstrate modelingaccuracy to be within 3 percent or less for the majorityof each acceleration. Each test case displayed similarmodeling error behavior. Toward the end of eachacceleration, modeling accuracy deteriorated beyond the3-percent levels to as much as 10 percent. The errorsdiscovered with the excess thrust technique representcombined model errors of the numerical representationsof the unique left and right propulsion systems. As aresult, excess thrust model error is not be expected to beexactly the same for other propulsion systems with thesame class of engines. On the other hand, overall trendsare expected to be similar.
Intentional Fan Stall Test
Using the intentional fan stall methodology, fourengine stalls were induced, three on the left engine andone on the right. Figure 11 depicts time traces ofestimated stall margin from before the EPR uptrim isapplied until the moment of stall at which time modelexecution is halted. The model predicts a zero stallmargin when the fan operating point crosses theadjusted stall line. At the test conditions, adjustments tothe base stall line are made only for inlet distortion andrandom effects.
Left engine stalls occurred when EPR was uptrimmedby 26 and 30 percent. As shown in figures 11(a)–11(c),calculated SMF varied between –9.5 and –10.2 percentfor the left engine at the point stall occurred. Comparingthe stall events, differences of the EPR uptrim requiredto induce stall and of estimated stall margin at the pointof stall may be explained by the statistical nature of stallphenomena. Taken together, results from each stall eventshow the SMF model for the left engine contains errorsof approximately 10 percent at this test condition.
(a) Left engine with a 26 percent EPR uptrim.
Fig. 11. Fan stall margin test results.
0
– 10
– 20
20
10Fan stallmargin,percent
10Time, sec
20
Point of stall SMF = – 10.2%
950222
0
5onautics and Astronautics
(b) Left engine with a 26 percent EPR uptrim.
(c) Left engine with a 30 percent EPR uptrim.
(d) Right engine with a 30 percent EPR uptrim.
Fig. 11. Concluded.
Figure 11(d) shows the right engine stall event. Thestall occurred much sooner after EPR uptrim applicationthan was the case with the left engine. Detected SMFerror at the time of stall was –5.6 percent or aboutone-half of the error seen with the left engine model.Modeling differences were expected because of theoperating differences between these engines.
The possible sources of SMF model error includecalculated fan airflow, inlet and exit total pressures aswell as measurement errors. The assumption of constantrandom effects introduces some level of error, however,not at levels high enough to resolve the detected error.Modeled PT2.5 has been previously reported on to bewithin 2 percent of measured at this flight condition, soPT2.5 model error will contribute some of the modelerror in SMF. Fan airflow, WCfan, is very sensitive tounmodeled sensor bias and the Kf CDP estimates. Themodel estimate for fan airflow has been reported to beas high as 5 lb/sec higher than the DEEC-calculatedairflow. Pratt & Whitney believe the PSC modelprovides a more accurate value.4, 5 Errors in model
0
– 10
– 20
20
10Fan stallmargin,percent
30
Point of stall SMF = – 9.5%
950223
0 10 20Time, sec
0
– 10
– 20
20
10Fan stallmargin,percent
10Time, sec
20
Point of stall SMF = – 10%
950224
0
0
– 10
– 20
20
10Fan stallmargin,percent
10Time, sec
20
Point of stall SMF = – 5.6%
950225
0
1American Institute of Aer
airflow will appear as errors in stall margin through thebase stall line table lookups.
Differences between the left and right stall marginmodel errors may be explained between physicaldifferences between the engines. For example, the leftengine generally operated at higher turbine temperaturefor a given fan speed, thus indicating a moredeteriorated engine. Nonwhite noise and biascharacteristics differ between individual sensors locatedon the left and right engines and will contribute to theeffects of unmodeled sensor errors.
Results showing that SMF error is always negativeimply that at given level of calculated stall margin, thereis in reality an extra amount of stall margin availablebefore the fan will stall. This built-in conservatism mayhave prohibited PSC from achieving optimalperformance improvements. Three of the PSCoptimization modes drive the SMF to its constrainedminimum stall margin remaining value.7 Pratt &Whitney suggested that a 5-percent reduction in SMFdesign requirements could lead to thrust-to-weightincreases of 3 percent or translate to a 1.2-percent fuelburn reduction. Clearly, one way to achieve increasedperformance at levels exceeding the performanceimprovements demonstrated with PSC is by improvingmodeling accuracy.
Conclusions
The Performance Seeking Control (PSC) algorithm, amodel-based, adaptive control algorithm, was flighttested with built-in models of the engine and theinlet. Flight results show substantial performanceimprovements from the F-15 PSC algorithm. Becausethese benefits are determined from the PSC modeling,demonstrated results are only as accurate as the models.As a result, assessing and quantifying the modelingaccuracy of the PSC algorithm is critical.
A method of determining PSC model accuracy for theestimation of increases in net propulsive force, FNP,was developed and applied to flight test data. Thismethod, the excess thrust technique, has been applied toassess the FNP model. Results show that measuredincreases of FNP are generally accurate to within 3percent of full-scale excess thrust. Accuracy to theselevels is significant to the estimated performanceimprovements provided by PSC in all of its optimizationmodes. The FNP is used in each of the optimizationmodes. With an assessment of changes in FNP accuracy,uncertainty bands may now be applied to theoptimization results.
6onautics and Astronautics
The second test to evaluate PSC modeling accuracyinvolved intentionally stalling the fan. Results indicatethat the PSC model is excessively conservative, and themodeling accuracy is within the –5- to –10-percentrange for fan stall margin. With improved stall marginmodeling capability, PSC may have achieved largerperformance improvements than were demonstrated.
References
1Highly Integrated Digital Electronic ControlSymposium, NASA CP-3024, 1987.
2Goecke Powers, Sheryll, comp., An ElectronicWorkshop on the Performance Seeking Control andPropulsion Controlled Aircraft Results of the F-15Highly Integrated Digital Electronic Control ResearchProgram, Proceedings of the Electronic Workshop,NASA TM-104278, 1995.
3Beaulieu, Warren, Campbell, Ralph, and Burcham,William, “Measurement of XB-70 PropulsionPerformance Incorporating the Gas Generator Method,”J. Aircraft, vol. 6, no. 4, July–Aug. 1969, pp. 312–317.
4Maine, T., Gilyard, G., and Lambert, H., APreliminary Evaluation of an F100 Engine ParameterEstimation Process Using Flight Data, AIAA 90-1921,July 1990.
5Orme, J.S. and Gilyard, G.B., Subsonic Flight TestEvaluation of a Propulsion System ParameterEstimation Process for the F100 Engine, NASATM-4426, 1992.
6Conners, Timothy, R., Thrust Stand Evaluation ofEngine Performance Improvement Algorithms in anF-15 Airplane, NASA TM-104252, 1992.
7Gilyard, G.B. and Orme, J.S., Subsonic Flight TestEvaluation of a Performance Seeking Control Algorithmon an F-15 Airplane, NASA TM-4400, 1992.
8Lambert, H.H., Gilyard, G.B., Chisholm, J.D., andKerr, L.J., Preliminary Flight Evaluation of an Engine
1American Institute of Aer
Performance Optimization Algorithm, NASA TM-4328,1991.
9Orme, J.S. and Gilyard, G.B., PreliminarySupersonic Flight Test Evaluation of PerformanceSeeking Control, NASA TM-4494, 1993.
10Orme, John S. and Conners, Timothy, R.,Supersonic Flight Test Results of a Performance SeekingControl Algorithm on a NASA F-15 Aircraft, AIAA94-3210, June 1994.
11Beeler, De E., Bellman, Donald R., and Saltzman,Edwin J., Flight Techniques for Determining AirplaneDrag at High Mach Numbers, NACA TN-3821, 1956.
12Ray, Ronald J., Evaluating the Dynamic Responseof In-Flight Thrust Calculation Techniques DuringThrottle Transients, NASA TM-4591, 1994.
13Burcham, Frank W., Jr., Myers, Lawrence P., andWalsh, Kevin R., Flight Evaluation Results for a DigitalElectronic Engine Control in an F-15 Airplane, NASATM-84918, 1983.
14Smith, R.H., Chisholm, J.D., and Stewart, J.F.,“Optimizing Aircraft Performance with Adaptive,Integrated Flight/Propulsion Control,” J. Engineeringfor Gas Turbines and Power, vol. 113, Jan. 1991,pp. 87–94.
15Luppold, R.H., Roman, J.R., Gallops, G.W., andKerr, L.J., Estimating In-Flight Engine PerformanceVariations Using Kalman Filter Concepts,AIAA-89-2584, July 1989.
16España, Martîn D. and Gilyard, G.B., On theEstimation Algorithm Used in Adaptive PerformanceOptimization of Turbofan Engines, NASA TM-4551,1993.
17Orme, John S., Digital Performance SimulationModels of the F-15, F-16XL, F-18, F-104, TACT F-111,X-29, and Hypersonic Research Vehicle, NASATM-104244, 1992.
7onautics and Astronautics
REPORT DOCUMENTATION PAGE Form ApprovedOMB No. 0704-0188
Public reporting burden for this collection of information is estimated to average 1 hour per response, including the time for reviewing instructions, searching existing data sources, gathering and maintaining the data needed, and completing and reviewing the collection of information. Send comments regarding this burden estimate or any other aspect of this col-lection of information, including suggestions for reducing this burden, to Washington Headquarters Services, Directorate for Information Operations and Reports, 1215 Jefferson Davis Highway, Suite 1204, Arlington, VA 22202-4302, and to the Office of Management and Budget, Paperwork Reduction Project (0704-0188), Washington, DC 20503.
1. AGENCY USE ONLY (Leave blank) 2. REPORT DATE 3. REPORT TYPE AND DATES COVERED
4. TITLE AND SUBTITLE 5. FUNDING NUMBERS
6. AUTHOR(S)
8. PERFORMING ORGANIZATION REPORT NUMBER
7. PERFORMING ORGANIZATION NAME(S) AND ADDRESS(ES)
9. SPONSORING/MONOTORING AGENCY NAME(S) AND ADDRESS(ES) 10. SPONSORING/MONITORING AGENCY REPORT NUMBER
11. SUPPLEMENTARY NOTES
12a. DISTRIBUTION/AVAILABILITY STATEMENT 12b. DISTRIBUTION CODE
13. ABSTRACT (Maximum 200 words)
14. SUBJECT TERMS 15. NUMBER OF PAGES
16. PRICE CODE
17. SECURITY CLASSIFICATION OF REPORT
18. SECURITY CLASSIFICATION OF THIS PAGE
19. SECURITY CLASSIFICATION OF ABSTRACT
20. LIMITATION OF ABSTRACT
NSN 7540-01-280-5500 Standard Form 298 (Rev. 2-89)Prescribed by ANSI Std. Z39-18298-102
Flight Assessment of the Onboard Propulsion System Model for the Perfor-mance Seeking Control Algorithm on an F-15 Aircraft
WU 533-02-03
John S. Orme and Gerard S. Schkolnik
NASA Dryden Flight Research CenterP.O. Box 273Edwards, California 93523-0273
H-2060
National Aeronautics and Space AdministrationWashington, DC 20546-0001 NASA TM-4705
Performance Seeking Control (PSC), an onboard, adaptive, real-time optimization algorithm, relies upon anonboard propulsion system model. Flight results illustrated propulsion system performance improvements ascalculated by the model. These improvements were subject to uncertainty arising from modeling error. Thus toquantify uncertainty in the PSC performance improvements, modeling accuracy must be assessed. A flight testapproach to verify PSC-predicted increases in thrust (FNP) and absolute levels of fan stall margin is developedand applied to flight test data. Application of the excess thrust technique shows that increases of FNP agree towithin 3 percent of full-scale measurements for most conditions. Accuracy to these levels is significant becauseuncertainty bands may now be applied to the performance improvements provided by PSC. Assessment of PSCfan stall margin modeling accuracy was completed with analysis of in-flight stall tests. Results indicate that themodel overestimates the stall margin by between 5 to 10 percent. Because PSC achieves performance gains byusing available stall margin, this overestimation may represent performance improvements to be recovered withincreased modeling accuracy. Assessment of thrust and stall margin modeling accuracy provides a critical piecefor a comprehensive understanding of PSC’s capabilities and limitations.
Airbreathing engines (aircraft); Aircraft engine testing; Aircraft propulsion systems; Engine control systems; Flight tests; Turbofan engines AO3
21
Unclassified Unclassified Unclassified Unlimited
July 1995 Technical Memorandum
Available from the NASA Center for AeroSpace Information, 800 Elkridge Landing Road, Linthicum Heights, MD 21090; (301)621-0390
Presented as AIAA 95-2361 at the 31st AIAA/ASME/SAE/ASEE Joint Propulsion Conference, San Diego, California, July 10–12, 1995.
Unclassified—UnlimitedSubject Category 07
Top Related