1997 pachterQFT

INTERNATIONAL JOURNAL OF ROBUST AND NONLINEAR CONTROL, VOL.
7, 591—608 (1997)
MODELLING AND CONTROL OF AN

ELECTRO-HYDROSTATIC ACTUATOR
M. PACHTER, C. H. HOUPIS * AND K. KANG

Department of Electrical and Computer Engineering, Air Force Institute of Technology, Bldg 640,
2950 P St, Wright-Patterson AFB, OH 45433-7765, U.S.A.
SUMMARY
The control of a novel aerospace actuator is considered, viz., an electro-hydrostatic actuator. It offers a high
degree of maintainability and combat survivability of the aircraft’s flight control system because all the
actuator’s elements are collocated. The quantitative feedback theory robust control method is used to design
a controller for this novel actuator. Parameter variation, sensor noise and flight condition variability are
explicitly considered in the design process. The resulting design is not only robust with respect to actuator
parameter variations and flight condition and insensitive to sensor noise, but, in addition, the controlled
actuator’s phase lag is significantly reduced, thus improving the performance of the overall flight control
system. ( 1997 by John Wiley & Sons, Ltd. This paper was prepared under the auspices of the U.S.
Government and it is therefore not subject to copyright in the U.S.
Int. J. Robust Nonlinear Control 7, 591—608 (1997)
No. of Figures: 21 No. of Tables: 1 No. of References: 8
Key words: actuator compensator design; quantitative feedback theory; robust control theory with
parametric uncertainty
1. INTRODUCTION
Actuators are a critical subsystem in any flight control system; however, robust automatic control
of actuators has not fully been explored in the past. This research focuses on the modelling and
control of a new electro-hydrostatic actuator (EHA). Electric motors when used as actuation
elements have a limited torque-to-mass ratio, due to the finite and limited magnetic flux density
that can be generated.1 High pressure hydraulic systems, with system pressure of 2 000 to
5 000 psi, can generate high forces resulting in higher torque-to-mass ratios than in electric
motors. In addition, high pressure hydraulic systems are stiffer against the load than electric
motors. Hence, the novel EHA actuator employs a hybrid approach, where a brushless DC motor
drives a hydraulic pump to circulate high pressure fluid into the piston chamber. Thus, the DC
motor internal to the EHA converts electrical power into mechanical power. The collocated
pump converts this mechanical power into hydraulic power. The hydraulic power, acting against
the piston, is converted to mechanical power capable of moving large flight control surfaces.
* Correspondence to: C. H. Houpis, Department of Electrical and Computer Engineering, Air Force Institute of
Technology, Bldg 640, 2950 P St, Wright-Patterson AFB, OH 45433-7765, U.S.A.
CCC 1049-8923/97/060591—18$17.50
( 1997 by John Wiley & Sons, Ltd. This paper was prepared under the auspices of the US
Government and is therefore not subject to copyright.
592 M. PACHTER E¹ A¸.
A linear mathematical model of the EHA is derived in Section 2. The models for the individual
EHA components: motor, pump, piston, flight control surface (inertia) and hinge moment are
developed. The EHA model is rearranged to form a two loop feedback control structure. The
controller design process is described in Section 3. First, the inner loop controller is designed to
increase the outer loop robustness. The outer loop controller and prefilter are designed to enforce
the robust tracking of commands. The two controller designs are validated by simulations. In
Section 4, the sensitivity to sensor noise of the EHA controller, and of an alternative high gain
controller, are analysed. The multiple-input single-output (MISO) quantitative feedback theory
(QFT) robust control design2,3 is compared against a standard high gain design for sensor noise
rejection and bias handling characteristics. In Section 5, the EHA is incorporated in a flight
control system and its performance is compared against the performance of a flight control
system with first-order and fourth-order conventional hydraulic actuator models. Although the
rationale for using the EHA is reliability and maintainability driven, the flight control system
incorporating the QFT controller endowed EHA also yields improved performance.
2. ACTUATOR MODELLING
2.1. Motor
Electrical motors with rotor moment of inertia J (in-lb-sec2) and electro-mechanical damping
m
B (in-lb-sec) (see Figure 1), are subject to variations in output torque and subsequent fluctu-
m
ations in rotor speed u . Since the torque due to the load counteracts the torque generated by the
m
motor, perturbations may also be caused by variations of the load torque. This relationship is
expressed as
q (s)"q (s)!q (s) (1)

e #.$ -0!$
Figure 1. Motor model
Int. J. Robust Nonlinear Control, Vol. 7, 591—608 (1997) ( 1997 by John Wiley & Sons, Ltd.
MODELLING AND CONTROL OF AN ELECTRO-HYDROSTATIC ACTUATOR 593
where q is the load torque due to the differential pressure of the fluid in the pump. This results
-0!$
in a first-order transfer function
u (s) 1
m " (2)
q (s) J s#B
e m m
2.2. Pump and fluid

The flow rate Q generated by the pump is proportional to the motor speed
m
D
Q " mu (3)
m 2n m
where D (in3/rev) represents the pump displacement constant. The flow rate of the hydraulic
m
fluid is primarily dependent on two factors: change in chamber volume and change in pressure
due to the compressibility of the fluid effect.1 The chamber volume changes as the piston moves
through the chamber at speed XQ . The flow rate due the changes in chamber volume is then
p
expressed as $AX . Secondary fluid dynamics effects include fluid compressibility, internal
p
leakage flow and external leakage flow. These secondary effects are modelled with a first-order
differential equation
dQ(s)"Q (s)!AsX "K sP(s)#C P(s) (4)
m p s t
which results in the transfer function
P(s) 1
" (5)
dQ(s) K s#C
s t
2.3. Piston and flight control surface

The pressure developed by the pump and fluid acts on the piston (see, e.g., Figure 2), causing the
ram to extend or retract. This force then generates a torque through a hinge to deflect the control
surface. This torque has to overcome two load components: control surface inertia and aerody-
namic load. The aerodynamic load only occurs in flight, when the air pressure over the control
surface applies aerodynamic reaction forces to it. The aerodynamic load is determined by three
factors: the area of the flight control surface, dynamic pressure which varies with altitude and
airspeed, and the surface’s angle to the relative wind. The surface’s angle to the wind is determined
by the surface deflection angle and the aircraft’s angle of attack.
Figure 2. Simplified actuator control system (not drawn to scale)
( 1997 by John Wiley & Sons, Ltd. Int. J. Robust Nonlinear Control, Vol. 7, 591—608 (1997)
2.4. Flexible hinge joint model

The magnitude of F acting on the piston is equal to PA, where P is the differential pressure
A
developed by the pump and A (in2) is the surface area of the piston. Thus, the force created by the
pump and applied by the fluid can be expressed as
F "PA (6)
A
The piston dynamics, with the piston mass M (lb -sec2/in) and piston damping B (lb -sec/in), can
p & p &
be described by the second-order dynamics
F !F "(M s2#B s)X (7)
A P p p p
The resulting torque acting on the flight control surface due to this force imbalance is described
by
K
q " h (X !X ) (8)
R R p L
h
where K (in-lb/rad) is the hinge stiffness constant and R , (in) is the horn’s length. The stabilator’s
h h
inertia acts against the torque generated by the actuator, such that
q !q "(J s2#B s)h (9)
R L L L L
where q is the torque created by the aerodynamic load and stabilator inertia. The variables J
L L
(in-lb-sec2) and B (in-lb-sec) represent the mass properties of the flight control surface.
L
2.5. Stiff hinge joint model
Equations (7)—(9) represent a rather complex model of the load dynamics. The complexity of
the model can be reduced if the linkage between the actuator and flight control surface is
considered as being rigid. This is a valid assumption, since the natural frequency of the hinge for
a well designed system is much greater than the actuator’s bandwidth of interest. The assumption
of rigidity could break down at high frequencies, but can safely be ignored during the controller
design process since in the case at hand it is well above its bandwidth frequency. Hence, the piston
and load dynamics are still expressed as
F !F "(M s2#B s)X (10)
A P p p p
and
q !q "(J s2#B s)h (11)
R L L L L
Now divide equation (11) by R to obtain
h
q (J s2#B s)h
F ! L" L L L (12)
R R R
h h
Assuming rigidity of the hinge assembly, F +F . Hence, adding equations (10) and (12) results
R P
in,
q (J s2#B s)h
F ! L "(M s2#B s)X # L L L (13)
A R p p p R
h h
where q is defined by
L
q "(Load )h (14)
L !%30 L
Since rigidity implies that X "X and h "X /R , equation (13) is further reduced to
p L L p h
CA B A BD
q J B
F "F ! L " M # L s2# B # L s X (15)
e A R p R2 p R2 p
h h h
or, expressed in transfer function form
X (s) 1
p " (16)
CA B A BD
F (s) J B
e M#
s L s# B # L
p R2 p R2
h h
Equation (16) yields a simplified model of the load dynamics. A complete simplified design model
of the bare EHA, without the controller, is shown in Figure 3 using the individual component
models.
2.6. Hinge moments

The aerodynamic loads mentioned in Section 2.3 describe the torque opposing the piston’s
motion. Thus, the load generated by the flow of air above and below the flight control surface
applies a torque to the hinge assembly, which in turn adds back-pressure to the piston. This hinge
moment is modelled as follows:
C D
a(s)
q "qN S R Ch #Ch h (17)
L t h a h (s) d L
L
where S denotes the area of the control surface, qN denotes the dynamic pressure, and Ch and Ch
t a d
denote the hinge moment coefficients of the control surface with respect to angle of attack a and
surface deflection d.
The aerodynamic load’s dynamic pressure qN is a function of airspeed º and air density o, as
0
shown in equation (18).4 The air density in a standard atmosphere drops exponentially with
increasing altitude. Hence, the dynamic pressure qN increases as the Mach number gets higher and
the altitude decreases.
1
qN " oº2 (18)
2 0
The hinge moment for modern fighters with stabilators, where Ch +Ch , is modelled by
a d
C D
a(s)
q "qN S R Ch #1 h (19)
L t h d h (s) L
L
Figure 3. Bare EHA block diagram
Figure 4. Comparison of full and short period approximation for the longitudinal equations of motion
2.7. Aircraft short period approximation

If the aircraft’s forward speed is assumed constant (i.e., the speed perturbations u+0), the
X force equation can be neglected since it does not significantly contribute to the short period
oscillation.5 Thereby, the short period approximation of the aircraft’s longitudinal channel is
extracted and is written as
CD C DC D C D
a5 z z a z
" a q # d d (20)
qR m m q M e
a q d
This yields a second-order minimum-phase transfer function of the form:
a (s) !K(s#a)
" (21)
d (s) (s#b) (s#c)
e
For the frequencies of interest, the short period approximation closely resembles the full state
model. As seen in Figure 4, where the solid lines represent the full state model frequency response
and the ‘#’ lines represent the short period model frequency response the phase and attenuation
characteristics are closely matched at high frequency. The approximation, as seen in the figure, is
not valid for frequencies below 0·5 rad/sec due to the effects from the slow phugoid mode.
3. COMPENSATOR DESIGN
In order to proceed with the QFT design, the block diagram as shown in Figure 3 is transformed
to the standard inner and outer QFT loop structures shown in Figure 5.6 For this design problem
the external disturbance d (t) can be ignored, i.e., d (t)"0, whereas the d (t) disturbance models
2 2 1
a possible measurement bias. The inner loop controls the angular rate of the motor while the
outer loop controls the surface deflection. Since tracking in the inner loop is less important, for its
function is to reduce the uncertainty level of the outer loop, the inner loop prefilter can be set to
unity and the tracking performance enforcement is relegated to the outer loop’s prefilter F .
h
3.1. Inner loop control
The inner loop controls the angular rate of the electric motor. The direction of the motor
rotation determines the direction of flow of the hydraulic fluid, and ultimately the piston
direction. The rise time is the dominant design constraint in the inner loop. The purpose of the
inner loop design is to facilitate the design of a robust outer loop, without the problems associated
with excessive gains. The inner loop’s robustness effectively shrinks the region of uncertainty (i.e.,
the QFT template size) of the outer loop, while also increasing the outer loop gain margin.
Figure 6 shows the flight conditions considered. Both subsonic and supersonic operating
points are considered in order to determine the maximum QFT template width. While the
Figure 5. Two feedback loops in QFT two degrees-of-freedom structure
Figure 6. A typical modern fighter flight envelope and the flight conditions considered for a(s)/d (s)
e
dynamic pressure parameter qN plays the most significant role in enlarging the templates’ size, the
aircraft damping and natural frequencies for given altitude and airspeed also play a significant
role. Two flight conditions ac41 (1 000 ft altitude and Mach 1·1), and ac28 (50 000 ft altitude and
Mach 0·78), determine the perimeter of the plant templates. Since the flight conditions ac28 and
ac41 result in the largest template size, the EHA parameters are varied to enlarge the templates
described by these flight conditions. The largest template is at the frequency of 5 rad/sec, about
22 degrees in width and 6 dB in height.
The inner loop compensator should be of low order. A unity forward gain is sufficient to satisfy
the inner loop optimal bounds, but it also requires the outer loop gain to be high. Through several
design iteractions, a gain of 10 is used in the inner loop which reduces the gain required in the
outer loop. A pole at the origin is introduced to create a type 1 system, to ensure tracking of a step
input. This pole also allows the nominal loop to fall in the troughs of the optimal bounds, as can
be seen in Figure 7. The nominal plant transmission at 1 rad/sec can be situated as low as 48 dB
by using the phase and magnitude information, instead of 74 dB, by using the magnitude
information only. This equates to a gain reduction of about 20! The inner loop compensator’s
transfer function is
1 000(s#15)
G (s)" (22)
m s(s#1 500)
3.2. Outer loop control

A common model of conventional aircraft actuators is of fourth-order, with the dominant
pole located about !20 rad/sec.7 This point of reference is used to build the upper and lower
tracking bounds. Thus, tracking bounds that satisfy both the time and frequency specifications
Figure 7. Inner loop compensator design—the loop transmission is steered through the trough, as required
are defined as
270(s#50)(s#400)
¹ (s)" (23)
ru (s#45)(s#60)(s#2 000)
and
200
¹ (s)" (24)
rl (s#10)(s#20)
It is reasonable to conclude that the actuator is quite stiff from the load end; hence the external
disturbance can be ignored (i.e., d (t)"0). The EHA should provide adequate measurement
2
disturbance (i.e., d (t)"u (t)) rejection by attenuating the disturbance input by !20 dB or
1 ~1
more. Hence, the disturbance bound specification is defined as
5
¹ (s)" (25)
D s#50
The flight conditions that define the perimeter of the plant templates are ac28 and ac41, at the
opposite end of the envelope. A DC gain of 60 is required to bring the nominal open loop
transmission above the respective optimal bounds shown in Figure 8. A pole at 8·5 rad/sec and
a zero at 14·5 rad/sec are added to meet the optimal and stability bounds. A pair of complex poles
are added at !200$j200 to swing the loop transmission line across the !180 degree phase
line. Figure 8 shows that the nominal loop transmission satisfies all the relevant bounds. The
resulting compensator’s transfer function is shown in equation (26).
2·813793]106(s#14·5)
G (s)" (26)
h (s#8·5)(s2#400s#80 000)
A minimum-phase prefilter is designed so that its Bode plot lies between the upper and lower
performance bounds (in dashes) shown in Figure 9. The resulting filter contains a zero and two
poles at high frequency and which, at the high frequency end, are required to satisfy the tracking
Figure 8. Outer loop compensator design
specifications. The resulting prefilter transfer function is:
(s#3·6)(s#8)(s#4 444·4)
F (s)" (27)
h (s#3·2)(s#200)(s#200)
The time and frequency responses are simulated using the MATLAB Simulab block models.
This simulation included the reduced-order transfer functions used in the controller design
process and, in addition, nonlinearities and elastic modes were included to enhance the realism of
the real world plant. Figures 10 to 12 show the system response in both the time and frequency
Figure 9. Outer loop prefilter design
Figure 10. Outer loop time domain unit step disturbance response
Figure 11. Outer loop time domain unit step input response
Figure 12. Outer loop small-signal frequency domain sinusoidal input response
domains. As expected, Figure 10 shows the disturbance input is rejected at levels well below the
specification. Figure 11 shows the time domain response of the system to a unit step input. The
system shows a slight oscillation, primarily due to the pair of complex poles near the origin from
the aircraft short period mode. However, the rise and settling times are quite fast. All plants settle
to 98% of the final value within 0·3 seconds. The oscillations and slight violation of the bounds
are deemed acceptable, since the magnitudes of the violations are small.
The linear, small-signal, frequency responses are obtained by commanding the individual plants
with a sinusoidal input having a magnitude of 0·5% of the maximum piston travel and frequencies
varying from 0·5 to 200 rad/sec. The peaks of both the reference and the output signal are first
detected, then the attenuation and phase between the peaks are computed. By repeating the
process over a predefined set of frequencies, a set of data is obtained for Bode plots. The dashed
lines in both the magnitude and the phase plots are the minimum values allowed in the EHA
specifications. As can be seen in Figure 12, all plants stay well below the specification boundaries.
4. SENSOR NOISE EFFECTS

In this design, two observable states, the pump motor’s angular rate and the piston position, are
measured and used for feedback control. A tachometer is used to sense the pump’s rotational
velocity, while a linear variable differential transformer (LVDT), embedded in the actuator’s
piston chamber, senses the ram movement.1 All sensors introduce some degree of inaccuracy.
There are two basic types of measurement errors that are dealt with in this paper: random and
bias. With random errors, the signals can take on any values within a band of finite width from
the nominal values. Sensor bias errors entail offsets from the nominal value by a constant
amount. The sensor errors can either be introduced gradually over the lifetime of the sensor (i.e.,
wear and tear) or introduced abruptly due to component failures. The degree of bias errors or
random noise determines the width of the error band.
It is well-known that a high system gain tends to amplify the effects of sensor noise. Hence, the
most effective countermeasure to sense noise is to decrease the control system’s loop gain as much
as possible, while still meeting specifications, e.g., tracking and disturbance rejection.4 QFT is an
ideal design method to strike this balance economically. The troughs in the composite (tracking
and ‘internal’ disturbance rejection) bounds on the Nichols chart are exploited during loop
shaping, and they allow for substantial gain reductions.
4.1. Random sensor noise error

An additional advantage of the QFT design can be seen in Figures 14 to 16, which simulate the
effects of LVDT sensor noise (random error) in the control system as shown in Figure 13. The
maximum error toleration level of typical aircraft quality actuator LVDTs is about one per cent.8
This uncertain feedback quantity can then cause chattering, or noise, on the output as the
feedback control system tries to zero-out the error. Unit step responses with 1% sensor
uncertainties of both the high forward gain and QFT designs are shown in Figure 14. In
Figures 14 and 15, (a) the traces are the high forward loop gain design responses, while (b) the
Figure 13. Addition of LVDT sensor noise in QFT feedback structure
traces are the QFT design responses. The zero-noise responses for the respective QFT designs are
shown as dashed lines in all the figures. As can be seen from the figures, the QFT design
substantially outperforms the high gain design. The effect of sensor noise is almost negligible in
the QFT design, while noise effects are quite visible in the high gain design.
Figure 14. Unit step response with LVDT signal to noise ratio of 100
Figure 15. Unit step response with LVDT signal to noise ratio of 1
The difference is even more pronounced as the signal to noise ratio is decreased from 100 to 1.
Even with the sensor noise level equal to that of the commanded input, the QFT design nominally
tracks the unit step command. But with the high gain design, the output is fairly chaotic, with the
output amplified to about six times the commanded signal. A noise error function of the IAE
type,3 defined in equation (28), is used to quantitatively compare the controllers’ performance:
1
E"
P0 abs( y/0.*/!- (t)!y/0*4% (t))dt (28)
This noise error function measures the deviation of the noise corrupted response from the
nominal (i.e., noise free) response. The comparison of the two designs, based on equation (28), is
shown in Figure 16.
Figure 16. Noise error function due to the changes in signal to noise ratio
Figure 17. Unit step response with sensor bias error of $1%
4.2. Sensor bias

The QFT design handles sensor bias errors better than the high gain design. If the feedback
signal is offset from the actual output level due to sensor bias, control systems will chase after
an offset steady state value rather than the commanded value. Even with just 1% sensor bias
level, the output steady state error of the high gain design, shown in Figure 17(a) as
about $15% is significantly more than the error for the QFT design, shown in Figure 17(b), of
about $1%. The problem worsens in Figure 18, where a sensor bias error of about $10% is
applied. As seen in Figure 18(b), the QFT design limits the error to about $10%, while the high
gain design, as shown in Figure 18(a), amplifies the error to about $100%. While the QFT
design limits the output error to the level of the bias, the high gain design amplifies the effect of the
bias by 10 to 15 times. Again, the QFT design is better at handling sensor bias than the high gain
design.
5. FLIGHT CONTROL SYSTEM INTEGRATION

5.1. Pitch control
The true test of the actuator design entails the evaluation of its performance as an integral
component of a pitch flight control system (see Figure 19). The flight control system is designed
initially with the commonly used first-order actuator transfer function of 20/(s#20).1 The results
are then compared against the current fourth-order F-16 actuator and EHA models. The
Figure 18. Unit step response with sensor bias error of $10%
Figure 19. A pitch control flight control system
fourth-order conventional actuator transfer function is given by

d (20·2)(71·4)2(144·8) N
e " " !#5 (29)
d (s#20·2)(s2#105·1s#71.42)(s#144·8) D
#.$ !#5
The simplified first-order actuator model is frequently used in text book settings. However, the
first-order model does not correctly represent the high frequency phase lag of the actual physical
system.7 A fourth-order actuator model, such as equation (29) which accurately describes the high
frequency phase characteristics of the actual physical system, should be used to design robust
flight control systems. However, conventional actuator control systems have an undesirable
amount of phase lag at the upper end of their bandwidth. Thus, the EHA control system is
designed to also reduce the high frequency phase lag. The reduced phase lag in the actuator
facilitates the maintenance of the phase margin angle and, most important, can minimize the
chances of actuator rate saturation. Indeed, the EHA model accurately predicts the performance
of the physical system and, moreover, the EHA behaves much like a first-order model. This will
greatly facilitate the design of the overall flight control system.
5.2. Simulation
The time domain simulations of the control system using different actuator models are shown
in Figure 20. The dotted, dashed and solid lines correspond to the results for the first-order
conventional, the fourth-order conventional and the EHA models, respectively. Time domain
responses of the three actuator models are similar and the responses are slightly underdamped.
The figures of merit are shown in Table 1.
The conventional fourth-order actuator model displays more realistic phase lag behaviour
than the first-order model. The phase lag difference between the first-order and fourth-order
models at the high end of the aircraft bandwidth can be as large as 50 degrees.7 Frequency
responses of the pitch flight control systems are shown in Figure 21. The dash-dotted, dashed and
solid lines correspond to the results from the first-order conventional, the fourth-order conven-
tional and the EHA models respectively. The first-order and EHA models behave in a similar
Figure 20. Time domain simulations of the pitch control system at Mach 0·7 and 30 000 ft altitude
Table 1. The pitch flight control system figures of merit
t t t
r p s
Actuator type (sec) M (sec) (sec)
p
Ist conventional 0·38 1·032 0·85 1·10
EHA 0·36 1·044 0·75 0·95
4th conventional 0·31 1·030 0·68 0·78
Figure 21. Frequency domain simulations of the pitch control system at Mach 0·7 and 30 000 ft altitude
manner, but the fourth-order model exhibits much more lag at the high bandwidth frequency.
Some lead is added to meet the given frequency domain specifications; specifications which the
conventional fourth-order actuator (model) could not meet. Indeed the conventional fourth-order
actuator model suffers from excessive phase lag at high frequency. The phase lead is achieved by
low frequency zeros, reducing the phase lag at high frequencies to be within specifications. The
main shortfall of using the first-order model is that it does not accurately describe the frequency
domain behaviour of the actuator. Thus, the novel EHA actuator control system which consists
of the actuator and compensator components, behaves like the first-order actuator model.
6. CONCLUSIONS
The novel electro-hydrostatic actuator investigated in this paper, and for which a control system
has been designed, offers easier maintainability and a higher degree of combat survivability over
conventional actuators. In addition, and as shown in the paper, improved performance over
conventional actuator designs is demonstrated.
A mathematical model of an advanced EHA is derived from basic device principles. The
subcomponent dynamic models are used to construct the required EHA model, resulting in
a complex feedback structure shown in Figure 3. The use of QFT for the control of the actuator
‘plant’ establishes a rigorous design procedure and facilitates other actuator and EHA control
system designs in the future. The QFT controller design is compared against a standard high gain
design for sensor noise rejection characteristics. The effects of sensor noise are almost negligible in
the QFT control system response while the effects of sensor noise dominate the more conven-
tional high gain control system response. The QFT controller nominally tracks the unit step
command even with a signal to noise ratio of one, while the high gain design fails to track the
command.
The EHA model was integrated in a simple flight control system and its performance was
compared against the same flight control system with first-order and fourth-order conventional
hydraulic actuator models. Some lead is added to the EHA control system to meet the given flight
control system frequency domain specifications; specifications which a conventional fourth-order
actuator model cannot meet. Thus, the flight control system with the EHA is more stable with
respect to gain variations than the flight control system with a fourth-order conventional
actuator model. Furthermore, the amount of lag at the upper limit of its bandwidth is reduced
with the EHA model, which facilitates the design of flight control systems for which a phase
margin angle requirement must be met. In addition, this reduces the chances for actuator rate
saturation. Hence, the QFT controlled EHA behaves like a first-order actuator model.
REFERENCES
1. de Silva, C. W., Control Sensors and Actuators, Prentice-Hall, Inc., Englewood Cliffs, NJ, 1989.
2. Houpis, C. H., et al., ‘Quantitative feedback theory for the engineer’, Wright Laboratory Technical Report, WL-TR-95-
3061, Wright Laboratory, Wright-Patterson AFB, OH, 1995. Available from the National Technical Information
Service, 5285 Port Royal Road, Springfield, VA 22151, U.S.A., document number AD-A297574.
3. D’Azzo, J. J. and C. H. Houpis, ¸inear Control System Analysis and Design—Conventional and Modern, Third edition,
McGraw-Hill Inc., NY, 1988.
4. Pachter, M., ‘Class notes: EENG 640 automatic flight control I’, Air Force Institute of Technology, OH, 1993.
5. Blackelock, J. H., Automatic Control of Aircraft and Missiles, John Wiley & Sons, Inc., NY, 1991.
6. Kang, K. H., ‘Electro-hydrostatic actuator controller design using quantitative feedback theory’, MS Thesis
AFIT/GE/ENG/94D-18, Graduate School of Engineering, Air Force Institute of Technology, Wright-Patterson AFB,
OH, 1994.
7. Reynolds, O. R., ‘Design of a subsonic envelope flight control system for the VISTA F-16 using quantitative feedback
theory’, MS Thesis, AFIT/GE/ENG/93D-34, Graduate School of Engineering, Air Force Institute of Technology,
Wright-Patterson AFB, OH, 1993.
8. ‘Interface control sheets, servo cylinder, stabilator/canard’, GE Aerospace Technical Report, Johnson City, NY, 1987.

1997 pachterQFT

Uploaded by

Copyright:

Available Formats

1997 pachterQFT

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

1997 pachterQFT

Uploaded by

Copyright:

Available Formats

INTERNATIONAL JOURNAL OF ROBUST AND NONLINEAR CONTROL, VOL.

MODELLING AND CONTROL OF AN

M. PACHTER, C. H. HOUPIS * AND K. KANG

q (s)"q (s)!q (s) (1)

Figure 1. Motor model

2.2. Pump and fluid

2.3. Piston and flight control surface

Figure 2. Simplified actuator control system (not drawn to scale)

2.4. Flexible hinge joint model

2.6. Hinge moments

Figure 3. Bare EHA block diagram

2.7. Aircraft short period approximation

Figure 5. Two feedback loops in QFT two degrees-of-freedom structure

3.2. Outer loop control

Figure 8. Outer loop compensator design

specifications. The resulting prefilter transfer function is:

Figure 9. Outer loop prefilter design

4. SENSOR NOISE EFFECTS

4.1. Random sensor noise error

Figure 13. Addition of LVDT sensor noise in QFT feedback structure

4.2. Sensor bias

5. FLIGHT CONTROL SYSTEM INTEGRATION

Figure 19. A pitch control flight control system

fourth-order conventional actuator transfer function is given by

Table 1. The pitch flight control system figures of merit

You might also like