Table Of ContentNASA Technical Memorandum 105254
AIAA-91-2792
Application of an Integrated Flight/
Propulsion Control Design
Methodology to a
STOVL Aircraft
Sanjay Garg
National Aeronautics and Space Administration
Lewis Research Center
Cleveland, Ohio
and
Duane L. Mattem
Sverdrup Technology, Inc.
Lewis Research Center Group
Brook Park, Ohio
L
:! i " ' _ i * t !',/I "_i 'f .... ] ¢ ," _
Prepared for the
Guidance, Navigation and Control Conference
sponsored by the American Institute of Aeronautics and Astronautics
New Orleans, Louisiana, August 12-14, 1991
APPLICATION OF AN INTEGRATED FLIGHT/PROPULSION CONTROL
DESIGN METHODOLOGY TO A STOVL AIRCRAFT
DuancL.Mawm
_y G_g
NationaAle_nau_cs m_dSpaceAdminLutm6on and Svev_mp T,m'mology, Inc,
Lewis ResearchCenter Lewis Resemch Ccme_Group
Clevelm_, Ohio 44135 BrookPark,Ohio 44142
AbstrLct
Results are presented from the application of an indicating the major steps in the methodology. The
emerging Integrated FLight/Propulsion Control (IFPC l IMPAC design approach evolved from an evaluation of the
design methodology to a Short Take-Off and Vertical two IFPC design methodologies that were developed under
Landing (STOVL) aircraft in transition _light. The steps in an Air Force sponsored program called Design Methods for
the methodology consist of designing a centralized IntegratedControl Systems (DMICS) ([2,3]). The IMPAC
controller, partitioning the centralized controller into methodology strivestocombine thebeataspectsofthetwo
separate subsystem controllers and designing command DMICS methodologies.
shaping prefilters to provide the overall desired response to The overallstructureforthe linearcontroldesign
pilot command inputs. A previously designed centralized portionofIMPAC (correspondingtoBlocks2and 3inFig.
controller is first validated for the integrated 1)isshown inFig.2.The stepsinthelinearcontroldesign
are:(I) Design of a centralizedfeedback controllerto
airfraane/engine plant used in this study. This integrated
plant is derived from a different model of the engine providecommand trackingand stabilityand performance
subsystem than the one used for the centralized controller robustness considering the fully integrated
design. The centralized controller is then partitioned in a airframe/propulsion model as one high-order system; (2)
decentralized, hierarchical structure comprising of airframe Partition of the centralized controller into a decentralized,
ateral and longitudinal subcontroUers and an engine hierarchical form compatible with implementation
subcontroller. Command shaping prefilters from the pilot requirements; and (3) Design of command shaping
control effector inputs are then designed and time histories prefilters from pilot control effectors to commanded
of the closed-loop IFPC system response to simulated variables to provide the overall desired response to pilot
pilot commands are compared to desired responses based inputs. In this paper, results are presented from the
on handlin_ qualities requirements. Finally, the propulsion application of these steps of the methodology to IFPC
system sstety and nonlinear limit protection logic is design for a linear integrated airframe/propulsion model of
a STOVL aircraft in transition flight to demonstrate the
design steps. The centralized controller used in this study
for transients that encounter the propulsion surge margin was designed previously in Ref. [4] and the details of the
Limit. centralized control design step are discussed in that
reference. So only those features of the centralized control
Introduction design step which are relevant for a thorough
The trend in future military fighter/tactical aircraft understanding of the overall IFPC design and evaluation
design is towards aircraft with new/enhanced maneuver results are presented here. Also presented in this paper are
capabilities such as Short Take-Off and Vertical Landing some preliminary results from the evaluation of the linear
(STOVL) and high angle of attack performance. An control design in the presence of the propulsion system
integrated flight/propulsion control (IFPC) system is operating schedules and the propulsion system
nonlinearitiedsue tosafetyand Limitprotectionlogic.
required in order to obtain these enhanced capabilities
with reasonable pilot workload. An integrated approach to Inthe followingsections,a briefdescriptionofthe
control design is then necessary to achieve an effective vehiclemodels tobe used forcontroldesignand evaluation
IFPC system. Such a design approach is currently being isfirstpresented.The applicationofthethreestepsinthe
linearcontrol design processis then discussedfor the
developed at NASA Lewis Research Center as part of an
vehiclemodel under study and intermediatedesignresults
ongoing STOVL controls research program. This
arepresented.The complete pointcontroldesignwith the
methodology is referred to as IMPAC - Integrated
propulsion system operating schedule and limit protection
Methodology for Propulsion and Airframe Control [1]. The logic included is then evaluated for sample pilot control
significant features of the IMPAC methodology are the
inputs and the response is compared with that of an "ideal
design of a centralized controller considering the airframe
response model" which is derived from Level I handLing
and propulsionsystems as one integratedsystem and the
partitioningofthe centralizedcontrollerintodecentralized qualities requirements.
subsystem controllers for state-of-the-art IFPC
Vehicle Model
implementation. Here partitioning means representing the
The vehicle considered in this stud)" is
high--order centralized controller with two or more lower
representativeof the delta winged E---7D supersonic
order subcontrollers which approximate the input/output
STOVL airframe powered by a high bypass turbofan
behavior of the centralized controller. The centralized
engine [Sj..The aircraftis equipped with the following
control design accounts for all the subsystem interactions
controlettectorse:jectorstoprovidepropulsiveliftatlow
in the design stage and the partitioning results in easy to
speeds and hover; a 2D--CD vectoring aftnozzle with
implement subcontro21ers that allow for independent
afterburnerforsupersonicflight;a vectoringventralnozzle
subsystem validation. The entire IMPAC methodology is
forpitchcontroland liftaugmentation during transitionl
presented in detail in Ref. [1]. A flowchart of "the
and jetreactioncontrolsystems (RCS) forpitch,rolland
methodology, is shown in Fig. 1 with the numbered blocks
yaw controlduring transitionand hover. A schematic
diagram oftheaircraftwith relativelocationofthevarious
controleffectorsmentioned above isshown in Fig. 3.
zj_
I Nonlinear J
•(r.,__d,), _ v=t=,u, :D/CD
[dynamicmodelsI
I State
ICon,rmoolde.l.l--I Cenlr=i=Ied
controldesignI
! generation) _ (2)1 ' VectoraJ_e RallRCS
Pitch RCS VemtralNo_le Thrusters
Thruner
linearfeedback
II ¢Coerndlrfoalkizred Figure 3. Control l_"ton for E-?D Aircraft
En_ne compressor bleed flow is uK.d for the RCS thrusters
rI Pc.o_nitlriool.._]_ I pCarotitinon,irnogl,Jer the mixed engine flow is used u the primary ejector
[ structure_j I (3) flow. Detailed ducting d/sgrams of the mgine and
Plu1_tioned d/scussion of the ejector STOVL concept are svailsble in
Ref.
lubsyl_em
linear_nlmhrs [6JThel procedure for generating linear integrated
air&sine/engine models for control dedgn and evaluation
Subsystem control "design from the separate nonlinear airframe and propulsion
system simulations is discussedinRef. [5]. The integrated
linear design model used in this study is of the form
x=A_+B_ ; y=C_+D_ (I)
subsystem subsystem
where the state vector is
control design controldesign
J Fullenvelop(e4) r Fullenvelop(e4) = [u,v,w,p,q,r,¢,0,N2,N25,
Tmhpc,Tmpc,Tmhpt ,Tm]pt] T (2)
I Fsuubllsoynsvteelompceontrollers with
u = Axial Velocity, R/s
v Lateral Velocity, ft/s
nonlinear I I nonlinear w = Vertical Velocity, ft/s
d coSnutrboslydselesmign II II coSnutbroslydsetesmign p Roll Rate, rad/s
(S)l I <s)
q Pitch Rate, rsd/s
Nonlinear I r Yaw Rate, r&d/s
¢ = Roll Attitude,rad
I subsystemcontrollers J 0 ffiPitch Attitude, ra_]
t ! I )
Fullsystem N2 -- Engine Fan Speed, rpm
controllerassembly N25 ffi High Pressure Compressor Speed, rpm
Tmhpcffi High Press. Compressor Metal Temp., OR
i. Tmpc ffi Burner Metal Temp., OR
Tmhptffi High Pressure Turbine Metal Temp., OR
t evaluation J
Tmlpt= Low Pressure Turbine Metal Temp., OR.
Figure i. IMPAC Methodology Flowchaxt
IYa-"__ Airframe u=_
, l--"_contr°lle!r -J
Z, k.--
! _'-..-_I,'opu,su,o._n!l
--Co-;,_--,T-jr
Pilot I^ . J I partilioning
...... I _ommano [ Zc(+)_, e I Centraliz_ [ u Jlntegraled I z
'_ prefilter _._ _ controller _ model I
Figure 2. Block Diagram for Linear Control Design
PortionofIMPAC
The control inputs are consistent with the actual implementation. Details of
control blending for other control eft'tots based on
= [6e,6a,&,AQR,AYR,ARR,
open-loop control effectiveness studies and designers'
WF,Ag,ETA,A78,ANG79,ANGg] T (3) knowledlge of system dynamics are dii_-us_, in Ref. [4].
with Note that there is an absolute nonlinearity in the
& = Elevator Deflection, deg relationship from commanded RCS areas to compressor
_a = Aileron Deflection, deg bleed flow demand in that although the RCS area may be
& = Rudder Deflection, deg positive or negative depending on the de, red direction of
AQR - Pitch RCS Area, in2 RCS thrust, the compressor bleed flow demand (WB3) to
generate the thrust is always positive. For a linear model
AYR = Yaw RCS Area, in2 trimmed about sero RCS are_, this relationslfip is of the
ARR ffi Roll RCS Area, in2 form WBB/ffiKilAfl_ I where WB3i is the demanded bleed
WF = Fuel Flow Pate, Ibm/hr flow due to A_ command with i representing Q, Y or R
A8 ffi Aft Nozzle Throat Area, in2 for pitch, yaw or roll RCS, respectively, Ki is an
ETA = Ejector Butterfly Angle, deg appropriate constant and ]. I represents absolute value.
A?8 = Ventral Nozzle Area, in2 All the results presented in this paper include the ARCS to
WB3 nonlinearity in the closed--loop evaluation system.
ANG79= Ventral Nozzle Vectoring Angle, deg
The flight phase considered in this study is the
ANG8 = Aft Nozzle Vectoring Angle, deg.
decelerating transition during approach to hover landing.
During this Right phase, the control of the aircraft Is
The modal outputs are
trsmsitioning from aerodynamic control surfaces to
= [V,V,0,q,%¢,p,_,r, propulsion system generated forces and moments. For this
study, the linear design model was obtained for a
N2,_,FG9,FGE,FGV] T (4)
with steady--state level Right at a trim speed of Voffi 80 Knots
V = True Airspeed, ft/s and a trim Right path angle of "fo= --3 deg, with
V = Total Acceleration, ft/s 2 propulsive lift supporting approximately 60 % of the
= Longitudinal Flight Path Angle, deg aircraft weight and with adequate distribution between
B ffiSideslip Angle, deg ejector and ventral nozzle thrust to provide pitch trim.
= Rate of change of Sideslip Angle, deg/s Although the airspeed span for transition flight phase is
FG9 -- Aft Nozzle Gross Thrust, lbs from 120 Knots to 50 Knots, open-loop analyses indicated
FGE -- Total Ejector Gross Thrust, lbs that the 80 Knot model provides a *'good average'* of the
FGV = Ventral Nozzle Gross Thrust. dynamic behavior of the aircraft in transition flight. Thus
the 80 Knot integrated model is used as the nominal
The other outputs are as discussed under state description controldesign model. Eigenvalue analysisof the design
model indicatedan unstableshortperiodmode atA=1.3.
except that the angular positions and rates are in degrees.
The current study did not include actuator and sensor
dynamics. These dynamics and the estimators for the three Control Design
thrusts will be included in future nonlinear control Centr_ Control l)elip
evaluation work. Recent advances in He control theory [8] and
The propulsion system state variables described
computational algorithms to solve for He optimal control
above differ from those used in the integrated models of
the previous studies (Refs. [4,5]). The propulsion system laws [9] have made this theory a viable candidate to be
linear models used in the previous studies were derived applied to complex multivariable control design problems.
from a cetailed non-tea]time aero-thermo cycle-deck In general terms, this technique provides the designer the
simulation of the turbofan engine. The propulsion system means to synthesize a controller for *'best" guaranteed
models used in this study are derived from a simplified performance in the presence of "worst case" disturbance
real-time Component Level Model simulation of the (or command). Proper formulation of the control design
turbofan engine. During preliminary IFPC design studies problem using H® theory provides for building in stability
for the E---TO aircraft, some discrepancies were noted in robustness and obtaining an adequate trade-off between
the transient and steady--state gross thrust and fan speed performance and allowable control power in the resulting
response between the CLM and the detailed cycle---deck controller. The results of the preliminary application of H
simulations. Since the CLM simulation is to be used in the av
control design techniques to IFPC design for the E--7D
piloted evaluation of the control augmented aircraft
STOVL aircraft, reported in Refs. [4,5], have been very
because of its realtime execution capability, it was decided
to use the CLM based linear models for further control encouraging. So the It® control synthesis technique is being
designand evaluation used for the centralized control design portion of IMPAC.
As mentioned earlier, a centralized IFPC design
Several of the control inputs u, listed above in (3),
was presented in Ref. [4] for the E-7D transition phase
include blending of the actual aircraft control effectors. integrated model with the propulsion linear model derived
For instance, only 3 RCS areas, AQR, AYR and ARR, are
from the detailed cycle deck simulation. Detailed
used in the linear design model whereas the full nonlinear
robustness analyses [4] showed that this controller provides
model has 5 controlled RCS areas. The reasons for this are
closed---loop system robustness for large variations in the
that the nose pitch RCS and the two yaw RCS thrusters
engine rotor dynamics and rotor speeds (N2, N25) response
pro_-Jdethrustin only one directionas shown in Fig.3,
to fuel flow (WF) input. Since the main difference between
and the wing tip rollRCS thrustersare to be used
the integrated model being used in this study and that
differentially for roll control and collectively for pitch
used previously is that the CLM based engine model has
control. Since yaw RCS thrusters provide only forward
slower rotor dynamic response, it was decided to first
thrust, leh yaw RCS is used for right yaw and right yaw
investigate whether the previously designed controller will
RCS is used for left yaw in the nonlinear model. Using
provide adequate performance for the integrated model
both left and right yaw RCS areas in the design model can
being used in this study. Fig. 4 shows an example result of
resuh in a control design that uses the two areas
dosed-loop system evaluation using this 14th order
d_fferential]y to enhance yaw control which will be
centralized controller on the CLM based integrated design "m.the pres._.ce of the RCS bleed flow nontinesnty,
model. The open-loop fan speed response to a step fuel discusseaearlier. This type of clo6ed--Ioop system provides
flow input of WF=100 lbm/hr is compared in Fig. 4(a) for independent control of acceleration, pitch, /light path
the detailed cycle-.deck and CLM based linear propulsion angle, roll and sideslip from the various pilot control
models corresponding to the g0 Knot trim condition, *nd e_ectors such as stick, throttle and rudder pedals etc., thus
the closed--loop fan speed response and fuel flow reducing pilot wor]doad, and also control of the propulsion
requirements for a step fan speed command of 200 rpm, system operating point (N2) independent of the aircraft
using the centralized controller, are compared in Figs. 4(b) motion. Independent control of roll (Pv) and sideslip angle
and (c), respectively. From Fig. 4(a) we note that the (_) will result in a control system that provides automatic
CLM based model response is slower, with a 10 % increase turn coord/nat/on thus further reducing pilot workload.
in tr9 0 (90 % rise time) ms compared to the cycle-deck
Coetn/ler Psrtit/oning
based model, and the perturbation steady-state fan speed • In ms overall aircraft design, traditionally the
response is lower by 20 %. Fig. 4(b) shows that inspite of engine manufacturer needs a separate engine controJler to
the differences in the open-loop model response, there are be able to independently perform extensive testing to
no noticeable differences in the clc4ed-loop fan speed assure an adequate design and engine integrity. Also the
response. As is to be expected, Fig. 4(c) shows that the centralized IFPC controller might contain many feedback
fuel flow requirement will be higher for the CLM based paths which are not desirable _om practical
model in order to track the same fan speed command as implementation considerations. To address these
with the cycle-deck based model. The comparison between difficulties, the idea of partitioning the centralized
the CLM based integrated mode] and the cycle-deck controller (Block 3 in Fig. 1) into separate airframe and
model for closed-loop response to step commands in other propulsion system subcontrollers was introduced in Ref.
controlled variables using this centralized controller was [2]. The desired structure of controller partitioning will
equally as good as the fan--_peed response comparison. depend on the coupling between the various subsystems
Therefore it was decided to use this controller as the
and on practical considerations related to integration of
centralized controller for the current IFPC design. the independently controlled subsystems. A decentralized,
Although the centralized controller is discussed in detail in hierarchical control structure as shown in Fig. 5 was
Ref. [4], some information on the control design philosophy cho6en for controller partitioning in IMPAC. In Fig. 5, the
and the controller structure is presented in the following to subscript "a°' refers to airframe quantities, "e" refers to
assist in understanding the controller partitioning and propulsion system quantities, and "c" refers to commands.
prefilter design steps.
The centralised controller structure is consistent The intermediate variables, _ea' represent propulsion
with that shown in Fig, 2 with the controlled variables system quantities that affect the airframe, for example
selected to be propulsive forces and moments. Fig. 5 is simplified in that
only the feedback paths £rom the controlled output errors
= [Vv,Qv,%Pv,fl,N2]T (5) are shown. The controller partitioning problem can be
stated as follows :
with Vv=V+0.1V, Qv=q+0.30, Pv=p+0.1¢ and the
others as discussed under plant outputs. The controller "Given: K(s)s.t. =z(s).[
i t_'(s)J'
inputs are the tracking errors e = zc- z, the plant outputs
Y asin (4)but without the gross thrust(FG9, FGE and
FGV) measurements, and with the RCS bleed flow where _ = Ue
demand WB3. The blendingof controlledvariablesasin
(6) was chosen to provide the response types that are
desirableforgood handling qualities[10,11]in transition
andy= Ye'
flight.The choice of Vv corresponds to designing an
accelerationcommand system with velocityhold,and the
choice of Qv and Pv correspond to designing a rate and a particular set of _ea'
command-attitude holdsystem.
The centralized controllerprovided decoupled
command trackingofthe controlledvariables_ up tothe
desiredbandwidth foreach individualcontrolledvariable
Z7 810 ose
Z4
ISO moo
ZtO! //,/f IS¢ -- CU,I sse
:/ _C_ "E IlO ...... ¢¥cJe SO@
".. _ ¢11cle
11 ii ..... cvc|t | so _ 4R,@
9
io
_e
1 3@
o .... z.... ,.... i.... | ....
i • 3 4 S 1 & 3 4 $ ! s .t 4
Y (s) I (s) I Is]
(a) Open-Loop Fan Speed (N2) (b) Closed-Loop N2 Response to (c)Closed-Loop WF ReouiremenT
Response to Step Fuel Flow N2 Command for N2 Command Tracking
WF=100 lbm/hr for Cycle-Deck
and CLM Based Propulsion
Models
Figure 4. Example Response with Centralized Controller of Ref. [4].
Find: Ka(s), Ke(S) with
. C4niillltzlid oorill_lM, K(ii}
is U
.,.,,d: +_eea ue
Sothat: The closed-loop performance and
robustness with the subcontrollers
Figure 5. Simplified Block DiqFam for _tralised,
Ka(s) and Ke(S) match those with HierarchicalControllerPartitioning
I'"l
the centralized controller K(s) to a
desired accuracy. Furthermore the
engine subcontroller should provide YK-- Ylon and _= "_io/n
tracking of the interfacevariable (6)
Yeng U_mg J
commands, Zeac- to allow for with
independentsubsystem check-out. Ylat= [epv,e_,¢,P,_,r,TB]
A state--space parameter optimization based
procedure to solve the above controller partitioning Ylon--[evv'eQv'e_,'V'V'e'q]T
problem is currently being developed. Some preliminary
resultsusingthisapproach are availableinRe[.1221.The Yeng= [eN2,N2,WB3IT (7)
controllerpartitioningforthisstudy was performed using and
straightforward steps, starting from the centralized
controller,which exploitthe designers'knowledge oflhe U-lat= [6a,6r,AYR,ARR]T
coupling between the airframe and propulsion system
dynamics. The controllerpartitioningstructureused for Ulon= [6e,AQR,ANG79,ANG8] T
thisstudyisshown inFig.6.The partitioningwas done in
two major steps:firstthe centralizedcontrollerwas U'-eng=[WF,A8,ETA,AT8] T (8)
partitionedinto decoupled lateraland longitudinalplus
enginesubcontrollersand then thelongitudinalplusengine The variousquantitiesin (7) and (8) above have been
subcontroller was further partitioned into separate defined earlier under the vehicle model description. The
hierarchicallongitudinaland engine subcontrollers.The nozzle vectoring angles, ANG79 and ANGS, were included
det_ded partitioningsteps,discussedinthefollowing,are:
is pazt' of z_e longitu_tinaJ controls is these mainly a_fect
(I) Partitioningof centralizedcontrollerinputs and
the pitch dynamics of the aircraft and have very little
outputs intothreeinput/output subsetscorrespondingto effect on the propulsion system dynamics.
•the lateral,longitudinaland engine subcontrollers;(2)
The interface from the propulsion system model to
Partitioningof the centralizedcontrollerinto lateraland
the airframe model was defined by the gross thrust from
longitudinal-plus-engine subcontrollers; (3) Obtaining the the three engine nozzle systems - aft nozzle, ventral nozzle
_e command trackingportionof the enginesubcontroller
and the ejectors. Therefore, the intermediate variables _ea
from the longitudinal-plus---enginseubcontroller; (4)
for hierarchical partitioning from the airframe
Designingthe_ea command trackingportionofthe engine subcontroller to the engine subcontroller were chosen to be
subcontroller(;5)Obtaining theportionofthelongitudinal z"ea= [FG9,FGE,FGV] m (9)
subcontrollerthat approximates the desired_ea response
with the longitudinal-plus-enginesubcontroller;and (6) Ylai -- [ Lateral Controller Ui-t
Designing lead compensation for the longitudinal --I K (s)
lid
I
subcontrollerportionofpreviousstepto generatethe _ea
Longitudinal Controller: Kto.(s )
command taking into account the finite_ea tracking
bandwidth. The emphasis in the controllerpartitioning - ................ !
presented herein ]s on matching the closed-loop Y_on
performance with the centralized controller. The
robustnessofthe closed-loopsystem with the partitioned
controllerhasnot been evaluatedand willbe addressedin
lll_i _llli¢
thefuture.
+i
Zu_( Engine Controller: 1%.9(s )
The controller inputs _;K' consisting of the tracking
1.) , , ,i,,,*r_
errors e and parts of the integrated plant outputs y as
discussed earlier, and the controller outputs Y were first Yer_ '
partitioned into three subsets corresponding to the lateral
I
anti longitudinal aircraft dynamics and the propulsion I
system dynamics. This partitioning of controller inputs L ...... .-. ......
and outputs is as follows:
Figure 6. PartitionedControllerStructure
S_eD _: order approximgtion of the H controller. Note that this
Since the integrated models are derived about Q
steady-state level/light, there is very tittle coupling from
Kzea(S ) controller when combined with the Keng(S)
the lateral controls Ulat to longitudinal and propulsion
controller of Step 3 as shown in Fig. 6, results in an engine
system dynamic response. However, there is strong subcontroller which provides decoupled command tracking
of fan speed and the three grols thrusts.
coupling from the engine controls Ueng to the longitudinal
response. Therefore the centralized controller K(s) was Ste_s:
first appro_mated by decoupled lateral _nd For the longitudinal---plus--engine subsystem, the
longitudinal-plus--engine subcontrollers i.e.
Ylon" Ueng portion of the Kl+e(S ) subcontroller provides a
K(s)_ [K0lat(s)K0l+e(S)]j (101 desired response for the interface variables ¢-_ from the
longitudinal controller inputs Y-Ionto be able to track the
with the lateral subcontrolier Klat(S ) such that longitudinal controlled variable commands. The transfer
_lat(S)= K1at(S).Ylat(S) (11) function matrix from the longitudinal controller inputs
YlontO the longitudinal controller outputs Ulon and the
and the longitudinal-plus--engine subcontrolier Kl+e(S )
desired interface variable outputs - was obtained by
such that Zeades
considering the _eng-_ Ueag loop closed _ shown in Fig. 7.
lon(=S)l r lo.I(S)
Ueng(S)J Kl+e(S)" LYeng(s)j (12)
State-space representations of the partitioned
subcontrollersKlat(S) and Kl+e(S) were obtained by IJIOn
Eno,rm D
reduced order approximations of the corresponding
input/output portions of the 14th order centralized Comroller
controller. Application of the internally balanced
realizationbased controllerreduction approach [131 _ LoKr_I+lt_u(dst)rml
resultedin a 4th orderlateralsubcontrollerand a 10th
Yer_ I
order longitudinal-plus---enginesubcontroller. The
dosed-loop peffomance, includingthe decouplingofthe
lateraland longitudinalresponses,with this contrgller
Figure7. Block Diagram forDetermining Klon(S)Part of
partitioningapproximation compared very well witL the
theLongitudinalPartitionedController
performanceobtainedusingthecentrahzedcontroller.
Step3:
The Klon(S ) portion d the longitudinal subcontroller, with
The portion of the engine subcontrollertransfer
the structure shown in Fig. 6, was then obtained as a 9th
functionmatrixfrom Yeng to Ueng' Keng(S) in Fig. 6, was order approximation of this transfer function matrix.
obtained as a 4th order approximation of the
_z__:
corresponding portion of the longitudinal-plus-engine
Since the engine subcontroller provides limited
subcontrollerKl+e(S)usinginternallybalanced controller
bandwidth for Zea command tracking, some lead
reduction.This Keng,(s)portionoftheenginesubcontroller
approximates the ]'anspeed command tracking,gross
compensation is needed on the Zeades generated by the
thrust regulationand bleed flow disturbance rejection
propertiesofthecentralizedcontroller.
Klon(S) portion of the longitudinal controller in order to
__¢._:
The closed-loop gross thrust responses for the have _ea= ¢-eadse so as to be able to maintain the
longltudinal-plus---enginesystem, using the Kl+e(S) centralized control performance with the partitioned
subcontroller,were analyzed for longitudinalcontrolled controllers. The lead compensation, Klead(S )in Fig. 6, was
variable commands to determine the requirements on chosen to be of the form
thrustcommand (Yeac)tracking.This analysisindicated
Klead(S) ffidiag(Kl(S),Kl(s),Kl(S)) ,
thattrackingbandwidths of4.5rad/sforeach ofthethree
thrusts,FGg, FGE and FGV, would be adequate to with Kl(S) = s+4.5 12
maintain high levelsystem performance with partitioned 4.5 s+12
controllersU.sing amixed sensitivityH formulationwith
er resulting in an effective bandwidth of 12 rad/s in the
the engine subsystem as the design plant, a controller was
Zeades-,Zearesponses.
designed to provide decoupled command tracking of Yea
Extensive comparisons were made between the
with fan speed regulation up to the desired bandwidths_
closed-loop system responses with the centralized
The Yea command tracking portion of the engine controller and with the partitioned subcontrollers
describedabove.An example comparison isshown inFJg._.
subcontroller, K (s) in Fig. 6, was then obtained as a 3rd 8 and 9 fora transientpitchrateand steady---statpeitch
Zea
$ .I .S
.IS
@
_ -- Cen_._alized
m *.1 \ ...... P•_ll:loned
,"_ "** Comm&nd
$
.I
• I ,: •nl:r811zed
/
.OS /
1 -.3
y ..... p_u'tiTt on •d •-% ...°*" i -- p_5:Jt:ien• d
.............. _':'r.--='?".._........ '..........................
0 ,, ,, .... J.... I.... *.... *.... -.4
O 1 2 $ 4 5 I I $ 4 • • 1 I J 41 9.
I"(e) I" (•) • (*)
(a) (h) (c)
Figure 8.
Comparison of Closed-Loop Response to Pitch Attitude Command (0c) with
Centralized and Partitioned Controllers
Z40
attitude hold command. Fig. 8(a) shows the 0c and the 0 ,..,,............
Z]O
response with the centralized and partitioned controllers
and Figs 8(b) and (c) show the corresponding velocity (V) 180
and flight path angle (_,)response. There is a small
tSO
degradation in pitch tracking and a slight increase in
velocity and flight path response coupling in going from
centralized to partitioned controller. Similar small levels of tzo
performance degradation were observed for the other
m 90
longitudinal commands while there was no noticeable ...... Partit|oned
degradation for the lateral commands, so
Analyses of the gross thrust requirements to track
30
the pitch command indicated large transient and
steady-state FG9 and FGE requirements with little o
change in FGV. Fig. 9 compares the response of the FG9
• and FGE gross thrusts for the 0c command with -3o
0 1 z 3 4 5 6
centralized controller and with partitioned controller. The T (.)
FGE response with the partitioned controller is very close
to that with the centralized controller and there is
300
agreement between the FG9 responses also.
The parameter optimization approach of Ref. [12]
O
can be applied to the partitioned subcontrollers developed
m
herein to more closely match the performance with the
centralized controller. However, the levels of performance -300
degradation evident from the evaluation of the partitioned i_ .......-- PCeAntrratliiztedioned
subcontrollers are quite small. Therefore it was decided to £ -$00 \
proceed with further development and evaluation of the v
IFPC design with these partitioned subcontrollers. -900
f_
Prefilter Design -IZO0
As mentioned earlier,it is desirableto provide
_
decoupledcontroloftheaircraftmotion inthevariousaxes
-1500
from thepilotcontroleffectorstoreducepilotworkload in
demanding tasks such as the deceleratingapproach to
-|800 *..I .... I.... I.... I .... _....
hoverlandingtaskbeingconsideredhere.Typically,in an
o ] z 3 4 5 6
integratedcontrolmode forthe transitionphase,thepilot
• (,)
would have independent controlofpitchrate(q),rollrate
(p),sideslipangle(_), accelerationalongflightpath (V) Figure 9. Gross Thrust Response to Pitch Attitude
and flightpath angle (7)through the longitudinalstick. Command (0<:) with Centralized and Partitioned
lateralstick,rudder pedal,throttleand a thumb whe_._ Controllers
eitheron thestickor thethrottleassembly),respectively
14].The "idealresponsemodels" forresponseineach of
variable of interest.The above choice of the "ideal
thesevariables(zi)to pilotselectedcommand (Zisel)was
response model" resultsin a fast response with little
chosentobeoftheform
overshoot. The t{}0ivalues for the 5 pilotcontrolled
zi riw_(s+l/r i)
variableswere selectedbasedon LevelIhandlingqualities
]sel= s2+2(_:is+_ (13) requirements[10,11]and arelistedinTable I.
The partitioned controllerprovides reasonable
decouplingin the closed-loopresponseto commands, so
= 0.89,wi--1.78g/tg0iand ri= 0.625.t90i
only separate single-input single---outputprefiltersare
needed to compensate forany deficiencieisn thetracking
ofeach individualcommand. Analyses ofthe closed--loop
where tg0iisthedesired90 % risetime forthecontrolled
command tracking response with the partitioned controller
indicated that simple prefilters of the form
s+a i b i
Pi(s)= ai Z ........ Selected
i _ Desired
providing lead compensation in the frequency region of
desired bandwidth of control (_) for each pilot D
e
commanded variable (zi) will be sufficient to "dose]y" qm 0 :
v
match the ideal response models. The lead prefilter B'
parameters used for this study are listed in Table 1. No
lead compensation was needed for the roll rate because the °Z
r_l_nse obtained with the feedback controller itself was
-3 ': ..
sufficiently close to the desired response
An example comparison of the ideal response model
and the dosed-loop response achieved with the lead "40 ". = . =1I .... ZJ , , , i 3| .... 4I = t i , Sj .... 8I , . . l
precompensation of Table 1. for a pitch rate command is
X(=)
shown in Fig. 1O. Acceleration (V) and roll rate (p)
response comparisons also showed an excellent match
between the desired and achieved responses. There was Figure 10. Pitch Rate (q) Response to Pilot Selected
some discrepancy for the flight path angle (7) and sideslip Command (osel) - Ideal tad Achieved with Partitioned
angle (_ responses in that the achieved responses had an Controller &ridLead Preeompensation
initial response delay and an overshoot as compared to the
ideal response. However, both the initial response delay
and the overshoot appear to be small enough to not have elements of Z-ea, are usumed to be measured variables.
an)- significant deterioration in the piloted system
performance. The fan speed schedule generates a fan speed command as
a function of the total gross thrust command, which is
calculated based on the individual thrust commands (_ea).
IFPC EvLlu&tion C
For this application, total gross thrust is defined relative
In order to evaluate the closed loop system response
before going to a full nonlinear simulation, the to the aft nozzle (i.e., it will be the thrust obtained if the
configuration in Fig. 11 was established. This total engine mass flow were to exit from the aft nozzle).
configuration is comprised of five major sections: pilot The engine limit logic modifies the engine controller
command shaping, the partitioned, hierarchical controller, commands to the engine actuators based on whether a
engine fan speed schedule, engine limit logic, and model of limit condition exists. These modified actuator commands
the integrated plant which includes the ARCS to bleed
axe passed on as U-eng to the integrated plant model.
flow demand nonlinearity. The command shaping block
consists of the prefilters and blending to create Vv, Qv and The limit logic block protects the engine from
temperature extremes and from exceeding the fan surge
Pv commands from V, q and p commands. The
margin. The core engine temperatures (and other
partitioned,hierarchicalcontrollerconsistsofthe lateral, variables [15]) are limited by the acceleration/deceleration
longitudinal and engine subcontrollers with the (accel/dece]) schedule which limits the fuel flow as a
input/output structuresdiscussedinthe previoussection.
function of high pressure compressor exit static pressure
Fo: the current evaluation, the three gross thrusts. (PS3), the core rotor speed (N25) and the high pressure
Cano.._m._m.._d a$phin._..99 I [
Prefilters _ Blend,ng _m-._b.,"_. _ .J d L._i I ....
pilot -- i,i Ji i _ _", _ _I r_r • . I u Io,n ,,.-JInlegralea
, L _ I J c I Engine ---,4_- Y
/ (+) _ • [ Model ,
r 2 Qa
e - i Yeng _ I zeFiJla
Figure 11. Block Diagram for Overall IFPC System Evaluation
8
