View
0
Download
0
Category
Preview:
Citation preview
Design of Optimally Convex Controller for Pitch Control ofan Aircraft
P. S. Khuntia, A. S. Chaudhuri
Abstract
The convex optimization problems for development of linear time-
invariant controllers are more prevalent in practice than was previously
thought. Since 1990 many applications have been discovered in areas of
automatic control systems of aircraft. The solution methods are reliable
enough to be embedded in a computer-aided design or analysis tool, or
even a real-time automatic control system. There are also theoretical or
conceptual advantages of formulating a problem as a convex optimization
problem. Our main goal is to develop a working knowledge of convex
optimization, i.e., to develop the skills and background needed to
recognize, formulate, and solve convex optimization problems pertaining
to aircraft control. In this paper a design method is proposed to solve
control system design problems in which a set of multiple closed loop
performance specification must be simultaneously satisfied. To utilize
this approach all close loop performance specification considered must
have the property that they are convex with respect to closed loop
system transfer matrix. For close loop performance specification a close
loop controller chosen from a set of all linear controllers determined
by trial and error such that the specification is satisfied. The
transfer matrix of the final system is determined through the convex
62
combination of the transfer matrices of the plant and the controllers.
With the close loop transfer matrices given the closed loop controller
structures and its gains are solved algebraically. In this paper we
established conditions for the existence of a designed parameter
inherent in convexity. The experimental verification deals with a
problem of pitch control of flight dynamics of a rigid body aircraft.
Key words: convex controller, parameter optimization, transfermatrices, integral square error(ISE), pitch control,
Introduction
Convex controller design is an approach which is increasingly used to
solve close loop system design problems of robotics, mechatronics, high
performance aircraft and flexible space structures[1-2] .such problems
typically required that a set of designed parameters and control gains
be adjusted simultaneously so that a prescribed close loop system
performance is achieved. This system design is termed as convex
controller design in the relevant literature [3-5]. Optimization methods
are extensively used to solve these design problems. To apply the convex
controller methods all performance specifications must be convex with
respect to the close loop transfer matrix. The close loop transfer
matrices of the systems are combined in a convex combination to form a
single transfer matrix, which satisfies that close loop performance
specification. Boyd et al [ 6 ] first pointed out that many commonly used
performance specifications, such as percentage overshot, control
efforts, robust stability are convex with respect to the close loop
63
transfer matrix. Controller design can be greatly simplified by taking
advantages of this convex property. Based on
this convex property the present paper proposes linear controller
methods for design pitch control of an aircraft.
This paper is meant for the researcher, scientist, or engineer who uses
mathematical optimization, or more generally, computational mathematics.
This includes, naturally, those working directly in optimization and
operations research, and also many others who use optimization, in
fields like computer science, economics, finance, statistics, data
mining, and many fields of science and engineering. The primary focus is
on the latter group, the potential users of convex optimization.
The goal of control engineering is to improve, or in some cases enable,
the performance of a system by the addition of sensors, which measure
various signals in the system and external command signals, control
processors, which process the measured signals to drive actuators, which
affect the behavior of the system. A schematic diagram of a general
control system is shown in Fig. 1.
64
Fig. 1. A schematic diagram of ageneral control system.
The use of the sensed response of the system (and not just the
command signals) in the computation of the actuator signals is called
the feedback control, an old idea which has been developed and applied
with great success in this century .
65
Other signals which affect system (disturbances)
System to be controlled
sensors
Sensed signals
Command signals(operator inputs)
Actuator signals
Operator display warning indicators
ControlProcessor(s)
actuators
The purpose of this paper is to describe how the fundamental problem of
controller design can be solved for a restricted set of systems and a
restricted set of design specifications, by combining a recent theo-
retical result with numerical convex optimization techniques [6-8]. The
restriction of the systems is that they must be linear and time-
invariant (LTI). If the specifications are achievable, it is possible
to find a controller which meets the specifications, although the
controller found may be complex and higher order. It may be possible
to find a simpler controller which meets the specifications, for
example by some model reduction technique [9].
Figure 2:classical feedback control setup
66
K
P+_
er ye
Classical Synthetic Open-Loop DesignClassical synthetic open-loop design methods have their origin in the
work of Bode . These methods are extremely widely studied and applied,
and are described in many current introductory control texts, such as
[10-13]. Examples of the numerous works developing and extending syn-
thetic open-loop techniques are [14-17]. In this section we will briefly
describe and comment on this kind of approach to control design.
Consider the classical feedback control setup shown in Fig. 2. Given
the plant, P, the designer must find a controller K such that results
should give satisfactory closed-loop performance. Classical open-loop
methods concentrate on designing the loop gain, L = PK, the closed-loop
transfer function from r to y, PK / (1 + PK).
Parameter optimization methods
Decomposition of the plant inputs and outputs are shown in figure-3.
The inputs to the model are divided into two vector signals. The
actuator or control signal vector, denoted by u, will consist of those
inputs to the model that can be manipulated by the controller. The
actuator signal u is the signal generated by the controller, other
input signals to the model will be lumped into vector signal w, called
the exogenous input. The sensor or measured signal vector, denoted by
y, will consist of those output signals that are accessible to
controller. The sensor signal y will be the input signal to the
controller. The output signals from the model will be lumped into a
vector signal z, called the regulated variables.
67
Figure-3 :Decomposition of the
plants inputs and outputs
Algebraic formulation of the decomposed plant
The plant as shown in the figure-4 can be described by the set of
transfer functions from each of its inputs (the components of the
vectors w and u) to each of its outputs (the components of z and y),
organized into a matrix, the transfer matrix. P and K denote the
transfer matrices of the plant and controller, respectively. Plant
transfer matrix P is partitioned as
z regulated outputs
y sensed output
exogenous inputs w
actuator inputs u
P
68
where Pzw is the transfer matrix from w to z, Pzu is the transfer matrix
from u to z, Pyw is the transfer matrix from w to y, and Pyu is the
transfer matrix from u to y. This decomposition is shown in Fig. 4.Now
suppose the controller is operating, as shown in Fig. 5. We can solve
for the closed-loop transfer matrix from w to z, which we denote
…(1)
w
u
Pzw
Pyw
Pzu
Pyu
K
∑
∑ z
y
+
+
+
+
Figure 4. The decomposed plant
69
The entries of the transfer matrix H are the closed-loop transfer
functions from each exogenous input to each regulated variable.
Various entries might represent, for example, closed-loop transfer
functions from some disturbance to some actuator, some sensor to some
internal variable, and some command signal to some actuator signal. The
equation (1) above shows exactly how each of these closed-loop transfer
functions depends on the controller K. The central theme of this
paper is that H should contain every closed-loop transfer function of
interest to us. The close loop matrix from w to z is
Design of Closed-Loop Convex Controllers
If two controllers K and each stabilize P and yield closed-loop
transfer matrices H and, respectively, then for each , there is
P
K
+ ++
+
w
v1
u
z
v2
y
Figure 5: The closed loop regulator
70
some controller that stabilizes P and yields closed-loop transfer
matrix .
Thus we can find an entire family of controllers that stabilize P, and
the corresponding closed-loop transfer matrices will lie on a line in
..A very important point is that the controller that yields
closed-loop transfer matrix is generally not .
Straightforward but tedious algebra yields
Where
may be defined as the coefficient of affine criterion of convexity .
An infinite no of such stabilizing controllers may be obtained by
varying a single parameter between 0 and 1. Out of these infinite
combination ,to select the optimal setting of convex controller we
impose certain design specifications. In the present system, the
specification chosen are internal stability, time domain and frequency
domain performance specifications and optimization of a time domain
performance index (PI) for the close loop system.
Time domain and frequency domain constrains
71
Performance specifications can be considered as convex constrains.
Primary purpose of feed back system is to ensure that one for the
regulated outputs be nearly tracking the command input. This is called
tracking performances. The following specifications are required
Steady state error 0.001 ,phase margin 450, gain margin 3 dB
Figure 6: bode plot of close loop
transfer functions
gain margin and phase margin satisfy our desired specifications of
the example discussed below[18].
Optimization value of
72
The convex optimization can be obtained for all the set of closed loop
transfer matrices given by the above equations. The set of transfer
matrices all such specifications are thus given by
H spec= H stab H t1 Ht2 H f1 H f2
As the design specification becomes stronger and stronger .the range of
parameter narrows down within limit 0 and 1. Still an infinite number
of controllers are obtained which satisfy all the specifications. as the
time domain specification are closed loop convex and H H , the optimal
value of will also produce a convex controller. The parameter is
optimized by optimizing time domain index (integral squared error, ISE
for continuous time system or sum of the squared errors SSE for discrete
time system) thus producing the optimal set of transfer matrices
H opt= H spec H SSE
Example: Modelling a Pitch Controller
Physical setup and system equations
The equations[18] governing the motion of an aircraft are a very
complicated set of six non-linear coupled differential equations.
However, under certain assumptions, they can be decoupled and
liberalized into the longitudinal and lateral equations. Pitch control
is a longitudinal problem, and in this example, we will design a convex
controller that controls the pitch of an aircraft.
73
The basic coordinate axes and forces acting on an aircraft are shown in
the figure below:
Assume that the aircraft is in steady-cruise at constant altitude and
velocity; thus, the thrust and drag cancel out and the lift and weight
balance out each other. Also, it may be assumed that change in pitch
angle does not change the speed of an aircraft under any circumstance
(unrealistic but simplifies the problem a bit). Under these assumptions,
the longitudinal equations of motion of an aircraft can be written as:
…(2)For this system, the input will be the elevator deflection angle, andthe output will be the pitch angle.
Design requirements
The next step is to set some design criteria. It is desired to design a
feedback controller so that the output has an overshoot of less than
10%, rise time of less than 2 seconds, settling time of less than 10
74
seconds, and steady-state error of less than 2%. For example, if the
input is 0.2 rad (11 degrees), then the pitch angle will not exceed 0.22
rad, reaches 0.2 rad within 2 seconds, settles to 2% of the steady-state
within 10 seconds, and stays within 0.196 to 0.204 rad at the steady-
state.
Overshoot: Less than 10%
Rise time: Less than 2 seconds
Settling time: Less than 10 seconds
Steady-state error: Less than 2%
Transfer function and the state-space
Before finding transfer function and the state-space model, let's plug
in some numerical values to simplify the modeling equations (2) shown
above.
… (3) These values are taken from the data from one of Boeing's commercialaircraft [8].
Transfer functionTo find the transfer function of the above system, we need to take the
Laplace transform of the above modeling equations (3). The Laplace
transform of the above equations are shown below.
After few steps of algebra, the following transfer function is obtained.
75
State-spaceKnowing the fact that the modeling equations (3) are already in the
state-variable form, we can rewrite them into the state-space model.
Since our output is the pitch angle, the output equation is:
MATLAB representation and open-loop response
MATLAB is used to observe the system characteristics. First an open-loop
system to a step input is obtained and if it is necessary system
characteristics need to be improved. Let the input (delta e) be 0.2 rad
(11 degrees).
.
76
From the plot, we see that the open-loop response does not satisfy the
design criteria at all. In fact the open-loop response is unstable.
Closed-loop transfer functionTo solve this problem, a feedback controller will be added to improve
the system performance. The figure shown below is the block diagram of a
typical unity feedback system.
Figure 7: Closed-loop transfer function
A convex controller needs to be designed so that the step responsesatisfies all design requirements.
Model transfer function of a pitch controller
77
The general equation of a PID controller. K(s+a)2/s [13].The
proportional–integral-derivative (PID) controller is the widely used
controller structure in control system application. Its structural
simplicity and sufficient ability many practical control problems to its
wide acceptance[190].
Possible values ok K, a, m(maximum overshot) are respectively:
3.0000 0.5000 1.0262
2.8000 0.5000 1.0273
2.6000 0.5000 1.0283
2.4000 0.5000 1.0294
2.2000 0.5000 1.0304
2.0000 0.5000 1.0313
2.4000 0.7000 1.0324
2.2000 0.7000 1.0326
2.0000 0.7000 1.0353
2.6000 0.7000 1.0353
2.8000 0.7000 1.0376
3.0000 0.7000 1.0393
3.0000 0.9000 1.0952
2.0000 0.9000 1.0960
2.8000 0.9000 1.0961
78
2.6000 0.9000 1.0967
2.2000 0.9000 1.0968
2.4000 0.9000 1.0970
Two best values are taken with respect to the minimum overshoot, K =2.4000,a = 0.9000 and K = 2.8000,a = 0.7000
Transfer function of controllers:
79
Figure 8 :plot fortwo best values
Close loop transfer function (H ):
Close loop transfer function ( )
80
Figure 9: plotfor various values of lambda
In the present paper it has been shown the closed loop system which
satisfies all the performance specification is the convex combination
of the plant and the controller. Using an algebraic procedure design
parameter, controller structures and controller gains are
simultaneously determined which successfully solves the problems of
system design with multiple simultaneous closed loop performance
specifications.
81
ConclusionThis makes a sensible formulation of the controller design problem
by considering simultaneously all the closed-loop transfer functions
of interest. The paper stresses that the closed-loop transfer matrix
H should include every closed loop transfer function necessary to
evaluate a candidate controller In the convex controller design each
design specification has been associated with a set of transfer
matrices. Most of the design specifications for pitch control of an
aircraft are closed loop convex. There might be infinite no of
stable controllers for any real value of the λ in the range 0 to
1.The overshoot and the settling time of a pitch maneuvering
dynamics would be brought down to an appreciable limit by a properly
designing a convex controller. In the paper the integral squared
error (ISE) method of optimizing the parameter λ to produce optimal
performance specification in all possible combination of pitch
dynamics has been suggested. The combine system with plant and
convex controller is stable and it satisfies the design
specification.
Reference:
[1] Iihan Tuzuc,Lleonard Meirovitch ,“Effects of flexibility on the
stability of flying aircraft” Journal of Dynamic Systems measurement,
and Control, March 2005, Vol. 127,pp 41-49
[2]Ke Fu ,James K Mills, “A convex Approach Solving Simultaneous
Mechanical Structure and Control System Design Problems With Multiple
82
Closed-loop Performance Specifications” Journal of Dynamic Systems
measurement, and Control, March 2005 Vol. 127,pp 57-68.
[3]Boyd S P, Barratt C H 1991, “ Linear Controller Design: Limits of
Performance, Prentice Hall. Englewood Cliffs, NJ.
[4] Ke Fu ,James K Mills “Convex Integrated Design (CID)Methods and its
Application Design To Linear
Positioning” IEEE Transaction Control System Technology, 2005
[5] Barratt C H and Boyd S P 1989,”Examples of Exact Trade offs in
Liner Controller Design, “IEEE Control System Magazine,” 9(1) pp 46-
52
[6] Stephen Boyd, Craig Barratt, and Stephen, Norman, “ Linear
Controller Design: Limits of Performance Via Convex Optimization” Proceedings
of IEEE, Vol. 78, No. 3. March 1990
[7]A W Roberts and D E Varberg. Convex Functions; Academic Press,1973
[8]G Strang. “Linear Algebra and its Application; Academic Press,NY,1980
[9]B D O. Anderson “Controller Reduction: Concepts And Approach” IEEE
Transaction on Automatic Control, Vol 34,No 8,August 1989,pp-802
[10] R. C. Dorf, Modern Control Systems. Reading, MA: Addi-son-Wesley,
2nd ed., 1980.
[11] G. F. Franklin,). D. Powell, and A. Emani-Naeni, Feedback Control
of Dynamic Systems
Reading, MA: Addison-Wes-ley, 1986.
[12] B. C. Kuo, Automatic Control Systems. Englewood Cliffs,NJ: Prentice-
Hall, 4th ed., 1982
83
[13] K. Ogata, Modern Control Engineering. Englewood Cliffs,NJ: Prentice-
Hall, 1970.
[14] I. M. Horowitz, Synthesis of Feedback Systems. New York,NY: Academic
Press, 1963.
[15] I. Horowitz, "Quantitative feedback theory," IFE Proc, vol.129-D,
pp. 215-226, Nov. 1982.
[16] A.G.J.MacFarlaneandB. Kouvaritakis, "A design technique for
linear multivariable feedback systems," Int. j. Control, vol. 25, no. 6
pp. 875-883, 1977.
[17] I. Postlethwaite and A. MacFarlane, A Complex Variable
Approach to the Analysis of Linear Multivariable Feedback Systems.
Berlin, West Germany: Springer-Verlag, 1979
[18] http://www.engin.umich.edu/group/ctm/examples/pitch/Mpitch.html
[19] Ming-Tzu Ho and China-Yi Lin, “PID Controller Design for robust
performances”IEEE Transaction on Automatic control,vol 48,No 8,August
2003,pp-1404
84
Recommended