Study On Iterative Learning Control For Harnessing The Nonrepetitive Disturbance Based-on Wavelet Tranform

Iterative learning control(ILC) is a new intelligent control method developing in recent 20 years.According to some repeatable characteristic in many complex industrial processe,current control information is revised in ILC by using the former control experience and the output errors of the system.The system output is convergent as far as possible to the expect value.The nonrepetitive disturbance always existed in the actual system that traditional iterative learning control algorithm can't harness it well. According to the nonrepetitive disturbance in ILC,a Wavelet transform-based method is proposed to harness the nonrepetitive disturbance.As a time-frequency analysis tool, Wavelet transform is employed to decompose the error at different frequency.The cutoff frequency of the filter can be identified by the maximal error component.Using the filter with time varying cutoff frequency the high frequency is eliminated and the nonrepetitive disturbance is filtered.The main contents are summarized as following:First,the basic principle of the iterative learning control is introduced and the development process,the development situation at home and abroad and the existing problems of iterative learning control are analyzed.Second,Wavelet theory about Wavelet basic concept,continue Wavelet transform and discrete Wavelet frame is discussed.And combining with the latest forefront study,application prospect of the Wavelet is proposed.Third,the ILC control system is analyzed and the error is divided.According to the simulation,the influence of differenent disturbance in ILC is summarized.Forth,the tracking error is analyzed by Wavelet transform-based mothod and the cutoff frequency is obtained.The filter is designed with the given window function and the cutoff frequency obtained before. According to the ILC feedback control system with the filter which is designed before,the update law is proposed.At last,above iterative learning control algorithms are established by Matlab and simulation is presented to illustrate the effectiveness of the proposed method.Dissertation is summarized and some open problems in theory are suggested.