Professional Documents
Culture Documents
Electro-Hydrosta,tic .$ctuat,or*
5 46
rigid. T h i s is B valid assumption) since the iiatiiral
freqnency of t,he hinge foi a well drsigiied system
is ~iiricligieater thaii its bairdwidtli. The a ~ ~ i i i i 1 1 ~ -
tiori of rigidity breaks down at Iiigh freqiiriicies.
but can safely hr ignored for tlie controller debigii
procedure since it is well above the bandwidth fre-
queiicy.
The piston and load dynamics are still es-
Figure 2: Sirriplified A c t u a t o r C o n t r o l Systein pressed as
- S o t D r a w n to Scale
547
Figure 3: Bare EHA Block Diagram
of air above arid below the flight control surface period approximation of the aircraft’s longitudinal
applies a torque on the hinge assembly, which i n channel is extracted and is written as,
turn adds back-pressure t o the piston. This hinge
moment is modeled as fol!ows:
of attack and surface deflection. For the frequency of interest, the short pe-
T h e aerodynamic load’s dynamic pressure p is a riod approximation closely resembles the full s t a t e
function of airspeed VC,and air density p , as shown model. As seen in Fig. 4, the phase and attenu-
in Eq. (18) [TI. The air density in a standard at- ation characteristics are closely matched at. high
mosphere drops expoiientially {sith iiicreasiiig al- frequency. Tlie solid lines represent the full st.ate
titude. Tlie tlyrramic pressure q increases as the model frequency response and ’+’
lines represent
51acli number gets liiglier and altitude gets lower. the short period model frequency response. T h e
approximat,ion, as seen in the figure. is n o t valid
for frequencies below 0.5 rad/sec d u e to t.lre effects
q = ;pr; (18)
from the slow phugoid mode.
T h e hinge moment for niodern fighters with st.a-
bilator, where Ch= ;= Clra, is modeled by
3 Compensator Design
In order to proceed with t h e QFT design, t h e
block model as shown in Fig. 3 most first be
2.5 Short Period Approximation
transformed t o a state space model Using block
If t h e aircraft’s forward speed is assumFd constant manipulation rules, Fig. 3 is rnanipuldted to form
(i.e., U E 0), the X force equation can be neglected the standard inner and outer QFT loop structures
since it does not significantly contribute to the shown i n Fig. 5 . The inner loop controls the angu-
short period oscillation [l]. Thereby, t h e short lar rate of the motor while the outer loop coritrols
548
;p...<{
: e '.,. ............................................
utcr L o o p
.........
,; .:qFT.Tn n c r L,.-
,'up .........................
-1:
..................................................
.................................................................
/i
,jets)
inner loop robustness effectively shrinks the re-
gion of uncertaiury !,i.e.. the template size) of the
oiiter loop. while also increasing the outer loop role. Two flight conditions ac41, (1.000 Et alti-
gain margin. Tile inner loop controls the angular tude and 4lach l.l), and adS. (:?O,OOO ft d t i t n d e
rate of the electric motor. T h e direction of the and Mach O.58), tleterniine the perimeter of t,he
motor rotation timeterrninesthe direction of How plant templates. Sirice the flight condir,ions ac18
o i the Iivdrariiic fiiiitl. nntl iiltimately t l i e piston anti ac41 result in r.he largest template iize. the
direction. T h e rise time is t,he dominant liesign EH,\ parameters are varied to enlnrse the tern-
constraint in the inner loop. plates tiescribed by these flight conditions. The
Figure 6 shows the flight conditions consitl- largest Lempiate is a t t,he freqnency of 5 r;itl,/sec.
ereti. Both siibsonic and supersonic points are with about 22 degrees in width im(i 6 t i B in height,.
tested to determine t h e maxirnnm template width. The inner loop compensator dionltl tie o f low
IVliilc t,iir dynamic p r e s s u r e parameter (7 piays the order. A iinity forward gnnii is siitficieut to iar,isfy
most significant role in enlarqing t h e template size, the inner loop optimal boiincls, bill it also recjiiires
the aircraft damping m t i nasirral frequencies for the outer loop gain to he Iiiqli. 'Tliroiiqii s r ~ v e r a l
given nititudr a n t i airspeeti also play a siqnificant desigri ireratioii. ii gain of 10 is irsed i n the inner
549
Figure 7 : Inner Loop C o m p e n s a t o r Design - Figure S: O u t e r Loop Compensator Design
Tlle loop transniissioii is steered t h r o u g h t lie
t r o u g h , as required.
It is reasonable to conclude that the actuator
is q u i t e stiff from the load end: hence, the ester-
loop ~vliiclireduces t h e gain required i n t h e outer rial disturbance can be ignored (i.e., d z ( t ) = 0 ) .
loop. .4pole at the origin is introduced t o create a T h e FHA should provide adequate internal dis-
type 1 system. t.0 eiisiire tracking of a step inpiit. turbance (i.e.. d l ( f ) = U-,(!)) reject.ion by atten-
This pole also allows the nominal loop to fall be- uating t h e disturbance input by -20 d B or more.
tween the crevices of the optimal bounds: as call Hence. t!ie disturbance bound specification is de-
be seen in Fig. 7. T h e iiomiiial plant transmission fined as
a t 1 rad/sec can be situated as low as 48 d B by us-
ing the phase and magnitude information. instead
of 74 d B by using t h e magnitude information only.
This equates t o a gain reduction of about 20. T h e fiiglit conclitions c~efille the perime,er
inner loop compensator transfer function is of the plant templates are ac28 and a c q l . a t the
opposite ends of t h e envelope. ,4 DC gain of 60 is
G m ( s )=
1000(s + 15) required t o bring t h e noininal open loop trairsrriis-
s(S + ison) (32)
sion above the respective optimal b o u ~ i d sshown
in Fig. 8. X pole a t 8.5 radjsec aiid a zero at
14.5 rad/sec are added to meet the optimal and
stability boiiiitls. A pair of cornplrs poles are
3.2 O u t e r Loop Control . .
a t i d ~ dat -200 i 3200 t o swing the traiisinission
iine across t.lie -180 degree phase line. Fignre 8
AA cOmlnOn lnodel of convelitiona, a i r c r a f t actil-
e r ~ tile tlorninarlt pole
a t o r s is of ~ o l i r t ~ i ~ o r d,vith s h o \ ~ that
s the nomirial loop trniisrriission satibfies
lorated at &bout -20 [6]. ~l~~~ all tile required bounds. T h e resulting coriipen-
of refer-
eilce is used to build tile upper and lower t,rackillg sator's transfer function is shown in Cq. (26).
~ O U R ~ SThus,
. tracking bounds that satisfy both
t h e time and frequency specifications a r e defined G ~ (= ~ )
2.813793 * 106(s 14.5) +
as (s +
4- 8.5)(sz 400s 80000) ( 2 6 ) +
270(s + 5O)(s + 400) A prefilter with t.liird order zeros over third
T,"(S)
=
( 8 + 4 5 ) [ s + 60)(s + 2000)
(23) order poles is required t o satisfy the bouitds. A
z e r o and t w o poles at high frequency are required
and t o satisfy the frequency specifications, magnitude
'Loo as well a phase. T h e resulting prefilter transfer
T,i(s)= (s + lo)(. -+ 30) ("1 function is slioivii i n Eq. (27).
550
F i g u r e 9: O u t e r Loop Prefilter Design F i g u r e 10: O u t e r Loop Time D o m a i n U n i t
S t e p Disturhaiice R e s p o n s e
=-
(s + 3 . 6 ) ( s + 8)(s + 4444.4) (27) .i--
1 '
FO(5)
(ti + 3 . 2 ) ( s + 200)(s + 200) ' '
551
4.1 Random Sensor Noise Error
T h e advantages of tlie Q F T design caii
seen i n Figs. 14 tlirorigii 16, wliicli sininlate the ef-
fects of L\'DT sensor noise (random error) i n t h e
Q F T structure as slioivii in Fig. 13. T h e masi-
m u m error tolerat,ion level of t.ypical aircraft qual-
ity actuator Rr\M LVDTs is about one percent [8].
This uncertain feedback quantity can then cause
chattering, or noise, on t h e output as the feed-
back control system tries to zero o u t the error.
Figure 13. Addition of LI'DT Sensor Soise 111
l'nit step responses with 1% sensor uncertainties
of both tlie high forward gain and QFT designs
QI'T Fect1l)acli S t r u c t u r e
are shown in Fig. 14. I n Figs. 14 and 15, t h e
figures (a) are t.he high forward gain design re-
sponses, while the figures (b) are the Q F T design
4 Sensor Noise Effects responses. Tlie zero-noise responses for the re-
In this design, two observable states, t h e piirnp spective designs are shown as dashed lines in all
iiiotor's angular rate and the piston positionj are the fignres. As can be seen from t h e figures, the
measured and iised for t h e synthesis of feedback QFT design substantially outperforms the high
controls. Tachometers are used to seiise t h e rota- gain design. T h e effect of noise is almost negli-
tional velocities: while Linear \;ariahle Differential gible in t h e QFT design, while noise effects are
Transformers (LL'DT); embedded inside t h e actu- quite visible jn tlie high gain design.
ator piston chamber, are used to sense t h e piston T h e differences are even more startling as t h e
RA11 movement [2]. X11 sensors introduce some signal to noise ratio is decreased from 100 t o 1.
degree of inaccuracy. There are two basic types of Even with t h e sensor noise level equal t o t.hat
~neasrireineiiterrors tliat are dealt. with in this pa- of the co~nniandedinpnt, t,he Q F T design nom-
per: randoin and bias. IVitli random errors, the iiially tracks the unit st,ep command. But with
signals can take on any values within a band of the high gain design, the ont.put is fairly chaotic,
finite size from tlie nominal \.dues. Sensor bias with t h e output amplified t o about six times t h e
errors entail offsets from tlie nominal value by a co m man ded sign al.
.\ noise error function shown of IAE type [3]
constant amonnt. 'The senbor errors can either be
introduced gradually o w r the lifetime of t h e sen- in Eq. (28) is used t o qiiant.it.at.ively compare t,he
sor (i.e., {war and tear) or introdnced abruptly controller performances.
tliie t o compoiieiit failures. Tlie degree of bia.s er-
rors 01 random noise determines the width of the
error band.
E = l1 nbs(ynomtnal(t) - ~notre(i))dt (28)
It is a well k n o w n fact t h a t a lrigli system gain T h e noise error function measures tlie deviation
tencis t,o amplify the effects of sensor noise; hence, of tlie non-zero noise response from the nominal
t.he most effective connterineasure t o sensor noise (i.e.; zero noise) res;)onse.
is to decrease t h e control system's loop gain as
much as possible, wliiie still meeting specifica-
tions, e.g., tracking and distnrhance rejection [TI.
4.2 Sensor Bias Error
QFT is an ideal design niethod I O econorriically The QFT deqign handles sensor bias errors better
satisfy this balance. T h e troughs in the conipos- than tlie high gain de5ign. If the feedback sig-
ite bounds on the Nichols chart can be exploited nal is offset from t h e actual output level due t o a
during loop shaping, and they allow for substan- sensor bias, control systems will chase after an off-
tial gain rediictions. set steady s t a t e value rather than t h e conimanded
552
ko:vy/
i OG . . . . .
i 04 . .. . . . . .. .
02
07 Id)
0 01 02 03 09 05 06 07 08 09 1
'0 01 02 03 04 05 O b 07 08 09 1
-I ' '1
$
1 0O 4
8
02
0
0 01 02 03 01 05 06 07 08 09 1
llmslae)
~ i ~ 1~1: , , ~ s ~ ~R , ~~ \\,it],
~ ~
LVDT ~Figure ,17: Unit. ~ with ~Sensor ~
~ S t e p ~Response
Sigiial t o floise R a t i o of 100. Bias Error of il%
015
v
lT
-
05
i(:
o 01 02 03 01 05 06 07 on 09 1
'0 01 02 03 0 4 05 06 07 08 09 1
2 0 8
06 $06
100 I O 4
02
02 I bl
0 0 01 02 03 04 05 06 07 06 09 1
0 01 02 03 04 05 06 07 08 09 1
llmO[Se2l
llnolra)
553
QFT design is better at handling sensor bias than
the high gain design.
i s t con^. 1
i
' t,
o 36 1
1 .Up
i 022
t,
(sec)
0.85
t,
(sec)
I 10 -
EH.4 0 36 1044 j 0 75 0 95
T h e first-order actuator model is frequently
~
554
nrathematical modrl of an ad\~ancetl EIIiZ is de-
rived from basic device principles. Tlte niztlie-
niatical inodel that cornpleteiy describes tlie E I I A
dynamics is required to design a control system.
T h e nse of QFT for the control of t.lre actLiator
establishes a rigorous design procediire antl facil-
itat,es other EHA control system dejigiis i n the
fu t,n re.
T h e QFT controller t1es;gn is compared against
the high gain design for sensor noise handling
characteristics. T h e effrcts of sriisor noise are
Figure 20: Time Domain S i m u l a t i o n s of t h e almost negligible in the QFT control system re-
P i t c h Control System at M a c h 0.7 and 30,000 sponse while the effects of sensor noise dominate
ft A l t i t u d e t!ie high gain control m response. Tlte QI'T
controller nominally t t h e unit step col~1rniIrlti
-r-- ' I even wit,li the signal to noise ratio of 1, while the
high g a i n design fails to track the command.
T h e E H A model used in a siniple flight control
system is compared against the same flight coil-
trol system with tlie first-order arid foiirth-order
conventional iiycir aulic actuator models. Some
lead is added to the E H A control m to
meet. the given fiiglit control syst,ern frequency
domain specifications; specifications which con-
s,.q,.-., /!._.Cl ventional fourth-order act.uator (model) can not
meet. thus^ the flight control system with the
Figure 21: Frequency Domain Siinulatio~isof EHA model is more stable \vit.li respect t o gain
t h e Pitch Co~itrolSystem a t AIsch 0.7 and variations than the flight control system wit11 a
30,000 ft Altitude foiirt 11-order c o n ~ e niorial
t actuator iiiodel. Fii r-
thermore, the amount o f lag at t,he upper band-
widtli frequency is reduced \vit!i tlie E H A niociel.
qiiencies t o within specifications. T h e main short- whicli minimizes the chances for actuator rate sat-
fall of using t h e iirst-order rnotlc! is that i t does not uration. Hence, the coritrolled EH.,\ hchaves like
accuratcdy describe tlie frequency domain behav- a first-order actuator model. antl ?et tl,e iicielity
ior of the actuator. 'The EHX model. composed of the actuator model has not sacrificed..
of the actuator and compensator components. be-
haves like t tie first-order actuator model, yet fully
describes its beliavior. References
Blakelock, John €1. Azitonicitic Control of.42,-
6 Conclusions crclft a n d illissiles, Kew York: John iViley k
Sons, Inc., 1.991.
T h e Electro-Hydrostatic Act.uator offers easier
d e Silva. Clarence W. Control Sensors und
maintainability and higlier degree of combat snr-
Actuntors, Englewood Cliffs: Prentice-Iiall,
vivahility over conventional act,uators.
Inc., 1989.
The sub-component dynamic models are used
t c construct the required EHA model. r e s u l h g in D'Azzo, Jolin J. arid Hoiipis, Constantine
a complex feedback structure shown in Fig. 3 . A 1%.L i n e a t Control S y s t e m Analysis a n d De-
555
s i g n - CO>li,c1ttlor2cll o n t i .I~ut/c7.ri.New York:
1I c C; r a w - II il I Ti1c . , 1'3 68.
[ I ] LIorse, ;\IIeii C.CIecII.olrytli.ciir/ic St-i~t~onitch-
a n ~ s r t j ,K e w j ~ o r k :~ l c C ~ ~ ~ ~ ~ Inc..
v - I I i196:s.
ll,
[51 I<an g , I\ i IT o . E /cc f TO - f I ~ / d ~t notic
s il c t U ci t or
COII t r.o/lfi. Dc.qigii Lis ir13 Q it (i n t i t at i e:e Fc ed-
bock I'hcory, AIS Degree Tlic.sis, School of
Eiigineeriiig, Air Force Iiist.itiite of Technol-
ogy (At:).\\'riglit-Pattcrson XFB OH, De-
ceinb er i 9 '31 h FIT/ G E/ ES C,/ 9 4D-16.
I
556