الفهرس 
IntroductionOnly 14 pages are availabe for public view
 |  | from | 32 |  |  |
| | | from | 32 | | |
Abstract
allll or this thesis is (0 present advanced techniques filr designing nonlinear controllers to control the output of nonlinear systcms to reach a required set point change and
also, 10 track a required trajeclory_
Uilkrent ml’lhods li,.- controlling the Sl’Il’0inl change 0lnonli1l\:ar dynamical syslems
are applied. A multilayer feedf(Jrward artificial neural network (ANN) trained by backpropagation algorithm is used to control this system. Proportional-integral-derivative (PID) controller based on identification of the nonlinear system alter linearizing the system model around the required set point is also designed. Hybrid controller (ANN and PID) is designed. The essence of the scheme is to devide the control into two different stages. For coarse control ANN is used, which drives the system output into a pre-delined neighborhood of the sel point. The controller Ihen switches 10 the line conlrol stage at which a linearization of the system model is ilknt ilied aroulld the st’!- point. ami is coni rolled with an appropriale pID controller.
The propo~a’d :Ipproach d( ••. ~;n t reqllire pnTi~a’ k nowlnl,’_e oilite nOlllinear sYSIeIIl. as
the system’s test data (inpul- outpul pairs) provide cnough inliHlllation lill-this approach. Simulations show that the hybrid controller is the best one especially whose PID controller is tuned by fast Fourier transfiHIll (FI’T) and Zeigler Nichols (ZN) techniques. The effectiveness and feasibility of the proposed techniques are illustrated by considering a simulation study on two examples for nonlinear systems.
Different methods lilr tracking nonlinear system output to track a given reference trajectory are designed. These mcthods are; Adaptive plD controller oased on pieccwise linearization fill- the givcn nonlincar systems, kedfilrwan.l mullilayer ANN hased tracking controller. and adaptive pulse amplitude modulation control law fi)f controlling nonlinear systems oased on z-from numerical calculus method. Simulations are perfimlled to
demonstralc Ihc I\:a~;ihilily ami clkl’livcllc~;s 01 Ihe I’roposl’d coni rollers alld the oulput of nonlinear system tracks the output of second order linear reference model perfectly.
This thesis comprises seven chapters. The contents of each chapter is summarized as follows:
Chapter one contains an introduction to the contents of the thesis.
Chapter two gives a survey on hybrid control of nonlinear dynamical systems using ncural ncts and convcntional control schcmc, ncural ndwork- bascd tmcking control systcm,
and 1’\1) conlrol w;ing /.ieglL’r Nichols (IN) tuning.
Chaptn thn’l’ introduces thl~ hasie principles 01 ANN and Iheirrclation wilh thc
neural systcm 01” the human brain, a gencral basic knowlcdge of their structurcs, activation functions, learning rules, capabilities and advantages presented. The most common ANNs
allli kaminI’. rllk~; :Ill’ d(’~;nil1l’d alld da~;~:ili,’(1. J)niv;,lion of II”, ”1101 h;ll’lq”oP:Wlllioli
algorithm, which is thc basic algorithm in traiuiug ncural network, is considered in the proposed method. Major advantagcs of this type of network arc presented.
Chapter fOil I’ deals with the piecewise linearil.ation 1(lr nonlim:ar systems,
identification for first order linear systems using recursive least square (RLS) algorithm, tuning PID controller by pole placement and fast Fourier transform (FFT) techniques as fine stage in the hybrid controller and filr tracking a given trajectory. Also, the I.-form numerical calculus method is introduced Iin obtaining the time response of nonlinear systems.
Chapter riVl’ discusses dilkrent mL’lhods lilr controlling the set-point change 01” nonlincar dynamical systems. A Icedlllrward neural network trained by backpropagation algorilhm is used to control this system. I’roporlional-inlcgral- derivative (I’ID) controller hased oil i,knliliralioll ofthl’ ”,,”li”cal Sy~;klll alln lillcali/.illglhc Systelllll1
and PID obstacles.
Chapter six discusses different methods for controlling output of nonlinear systems
to track a given reference tr~iectory of the output of second order reference linear model. These methods are; adaptive PI\) controller based on piecewise linearization fi)r the given nonlincar system, a IcedllJrward multilayer t\NNhased tracking controller, and adaptive pulse amplitude modulation control law Illr controlling nonlinear systcms based on z-li:Jrm numerical calculus method as piecewise linearization.
Chapter sevelJ gives the thesis conclusion and the suggesterl work for the future
research.