时变过程的神经网络鲁棒自适应控制

目前在国际上ＢＥＮＣＨＭＡＲＫ控制研究已成为热点。本文在常规逆动态控制器的输入端增加控制偏差，并对时变系统神经网络的同一输入参数先集中在１个节点上以产生该类参数的综合作用，

ROBUST AND ADAPTIVE CONTROL FOR TIME-VARYING PROCESS USING NEURAL NETWORK

Over the past years, a number of benchmark problems have been circulating in academia to test various design methods. In this paper, a benchmark example is addressed to demonstrate the performance of the neural-net(NN) control system, in which the control deviations are added in the input of normal inverse dynamic controller and the parameters having the same property are integrated at a node in the input layer so as to output a synthetic action in the time-varying neural network. The control system combined NN estimator with NN controller is applied in the benchmark plant. The results show that the NN control system achieves the requested specifications for three stress levels and it can be concluded that the concepts mentioned in this paper are readily extendable to industrial problems.