The Study Of Adaptive Inverse Control By Neural Network

Adaptive inverse control, which was first presented by professor Widrow,has got large development in the past decade, many results have been achievedin the study of adaptive inverse control. However, up to now, the study ofadaptive inverse control focuses on linear systems. Relatively, study and resultsof nonlinear systems are rare. Thus, it's necessary for the researchers to dosome further work in the study of nonlinear adaptive inverse control.The advances in neural networks and fuzzy systems provide theresearchers with powerful tools for nonlinear systems. This thesis aims toachieve nonlinear adaptive inverse control with these tools. I did the followingwork:First of all, because slowly convergence and easily stop at the localminimum of the inverse controller constructed by multi-layer neural network,the paper construct an adaptive inverse controller by orthogonal neural network.Through modify the parameters of S-function in the hidden layer, whichaccelerate the speed of the learning. And then design the direct inverse controlsystem for BDCM, it easily get the high performance in speed control.Secondly, the paper studies the modeling and inverse modeling by fuzzyRBF neural networks;And design model reference adaptive inverse controlsystem for thermo technical automatic control, the controller performs well andcan be easily accomplished on-line.Thirdly, because the advantage of dynamic recurrent NN, the paperdiscussed the feasibility for OIF Elman neural network used in adaptive controlsystem.. The Elman NN containing radial basis function in its hidden layer issuggested, and then it is used in adaptive inverse control for nonlinear systemcontrol and disturbance cancel.Finally, a method of nonlinear PID parameters tuning by geneticalgorithms is given, adjusting the respective nonlinear function can tune eachpart of the PID controller indirectly, more optimum parameters can be obtained,the method is used on the model of controllable pitch propeller system(CPP), itimprove the ability of anti-jamming with the control system.