Model Free Adaptive Control System Based On Neural Networks

As to the modern control theory, it is very important to know the precise model of the controlled object. However, most industrial proceses are featured with no knowledge or very rough knowledge on mathematical model of the process due to the related physical or chemical complexity of it.In this paper, I present a new model free adaptive control system based on neural networks.The process is tackled as a blackbox, and is estimated and controlled with a time-delay multilayer perceptron(MLP), respectively.The paper first expatiate on the structure and algorithm of the new control system, then, by comparing with other analogous systems, points out its advantages. Finally, to verify the feasibility of the new system, the paper gives a set of experiments.The experimental results show that the new control system can perform well on first-order and second-order system, nolinear system, time-variant system, the system is very robust and have a degree of noise resistance.