Font Size: a A A

Design Of Modify Internal Model Controller(IMC) Based On Neural Network

Posted on:2017-05-26Degree:MasterType:Thesis
Country:ChinaCandidate:H W LiuFull Text:PDF
GTID:2428330623954447Subject:Aeronautical and Astronautical Science and Technology
Abstract/Summary:PDF Full Text Request
In the process of space manipulator control,the traditional PID control method is difficult to guarantee the control precision and even difficult to control because of the nonlinearity,the time delay and the uncertainty of the model.Therefore,it is necessary to have a simple structure and strong nonlinearity control performance of the controller.Therefore,this paper presents an internal model controller design method based on the NARX network to complete the nonlinear and unknown mathematical model of the controller design.In this paper,the research work mainly includes:(1)In the real performing process,the main problem of IMC is the sensibility for saturation control,which will lead to windup.In this paper,the design method of the modified IMC is proposed,and the output saturation can be improved by feeding the output directly into the proportional controller.(2)In this paper,NARX network model is used for system identification and inverse model identification.The simplified process of the graph model of neural network and the efficient implementation of automatic differential are given.According to different tasks,an optional network identification structure is given to complete the network training process and testing process.(3)In this paper,the Lipschitz coefficient is used to identify the order of the system.It only needs to input the output data to complete the system order identification.This method can avoid the traditional system identification process as a priori knowledge,but in practice,the system order is difficult to determine intuitively.(4)Because NARX network identification process is different from traditional identification,its recognition effect depends on data set,so it is a very important step to select reasonable excitation signal.In this paper,we discuss two common excitation signals and propose a hybrid signal,and compare the identification results.(5)In the end,this paper gives a concrete implementation process.Through the joint simulation of CNTK open source toolbox and Simulink,the control effect of modified internal model controller and standard internal model controller is compared.
Keywords/Search Tags:Internal Model Control, NARX Network, System Identification, Order Determination scheme
PDF Full Text Request
Related items