Design And Research On Wavelet Neural Network Controller Based On IGA

With the development of modern industry and science and technology, manyhigh orders, strong coupling, uncertainty, unknown objects appeared in largenumbers,so traditional control methods can not meet the control requirement. Therapid progress of the modern artificial intelligence technology provides a strongtheoretical basis to solve this kind of problem. Wavelet neural network solves theproblem that the traditional engineering analysis function---Fourier function can notprovide the signals of the time-frequency localization and function characteristic forus. All of these situations determine the wavelet neural network must step into thefield of artificial intelligence.The main contents of this paper include:Firstly,only when Wavelet neural network need to optimize the training of itsvarious parameters in dealing with specific problems, can it become a network todeal with specific problems----wavelet neural network, so how to train its variousparameters naturally becomes the research topic of wavelet neural networkapplication. This paper carried on the exploration,proposing to use immune geneticalgorithm (IGA) to solve the problem of training parameters in wavelet neuralnetwork.Secondly, IGA search algorithm calculation method overcomes theshortcomings of GA individual diversity, speeds up the convergence efficiency andaccuracy of the algorithm, so the calculation accuracy is improved.At the same time;I improve some parts of Mutation operation which belongs to genetic operation in theIGA algorithm and combine Linear decreasing strategy in order to to improve theIGAalgorithm.Finally, the controller module of wavelet neural network which has passedthrough immune genetic algorithm optimization that will experiment on the two level inverted pendulum control experiment, through the simulation analysis, it will provethe validity of the controller and the anti-interference ability is strong.