Identification And Speed Control Of Ultrasonic Motors Based On Modified Immune Algorithm And Elman Neural Networks

An ultrasonic motor (USM) is a newly developed motor in servo system,which has some excellent characteristics. So USM has been used in manypractical applications. The operation of ultrasonic motor is influenced by manyfactors, strongly nonlinear characteristics could be caused by the increase oftemperature, the changes of load, driving frequency and voltage and manyother factors. It is therefore difficult to construct a precise model of the USM.Recently the main approaches to perform USM control are the artificialintelligent methods based on neural networks and fuzzy systems, which haveachieved good control effectiveness for various kinds of ultrasonic motors tosome extent. However, the existing intelligent methods for the USM controlhave some shortcomings, such as complex structures, slower convergentspeeds and lower convergent precision. So it is very important to carry outfurther research on the intelligent control method of the USM by combiningthe newest control theory and computer technique.Aiming at the mentioned problems above, this thesis proposes a noveldynamical threshold-based Elman neural network (DTENN) for ultrasonicmotors identification and control. The major contents are summarized asfollows:(1) There is an introduction to USM including its development,characteristics, control strategy and the present study. The principles of neuralnetwork and Elman neural network are also introduced.(2) In this thesis, a Dynamic Threshold Artificial Immune Algorithm(DTAIA) is proposed.The reason that genetic algorithm can easily fall into local optimum valueand get premature is its diversity is insufficient during the evolution. While theimmune algorithm do not have that problem since it has the calculations ofaffinity and density which increase the diversity and help it jump out of theoptimum value.As well as we know, in the initial period of evolution, the algorithm haslittle possibility to fall into the local optimum value because of the highdiversity. With an increase of the evolution generations, there will be more andmore antibodies with high fitness values. If threshold is a constant, thealgorithm can easily become premature and get into the local optimum sincethe diversity is getting lower and lower. If threshold is an increasing functionof evolution generations, the antibody's density will be increased efficientlywith the increase of the evolution generations and that the suppression will bemore powerful to preserve high diversity. So the algorithm would have strongability to control the reproducing process.This thesis redefined the calculation of affinity by considering thedifference between fitness and made the threshold dynamic through theevolution generation. Numerical results show that the proposed DTAIA hashigh precision and better convergence speed and strong ability to avoid thelocal optimum value.(3) This thesis proposed a DTAIA-Based Elman Identifier (DTBEI).Multilayer static networks transform the dynamic time-molding probleminto a static space mold problem. That will definitely bring many problems.But the dynamic recurrent multilayer network which introduces dynamic linksto memorize feedback information of the history influence does not have thatkind of problems. It has great developmental potential in the fields of systemmodeling, identification and control. The Elman network is one of the simplesttypes among the available recurrent networks.In this thesis, a modified Elman network is employed to identify an USM,and a novel learning algorithm based on an improved artificial immunealgorithm is proposed for training the Elman network. Numerical results showthat the proposed DTBEI can approximate the nonlinear input-outputmappings of the ultrasonic motor quite well, and its identification precision issuperior to that trained by the Elman network with the feedback algorithm asthe study algorithm.(4) This thesis proposed a DTAIA-based Elman controller (DTBEC)The neural network is a kind of information processing system. It canconstitute the highly nonlinear dynamics system, and has some characteristics,such as the ability of large-scale parallel processing, self-adaptive,self-organize, self-study, the distribute storage, and so on. So the controlsystem using the neural network method has the stronger adaptive ability andis more robust.In this thesis, a novel controller is specially designed to controlnon-linear systems using the DTAIA based Elman controller, which is calledDTBEC. The USM is still considered as an example of a highly nonlinearsystem to test the performance of the controller DTBEC. Numerical resultsshow that good control effectiveness is obtained when applying the proposedcontrol scheme to various reference speeds. The precision using the proposedmethod are obviously superior to those obtained using feedback algorithmbased Elman neural network controllers. It suggests that the controllerpresented here exhibits very good robustness.The DTAIA based Elman identification and control method possessesgood effects and the control process of USM is accurate and well adaptive.