对传感器失效具有完整性的神经网络控制器设计

本文研究了对一些传感器失效控制系统仍保持稳定性的问题，将多层前传神经网络用于这一问题，通过对神经网络的训练，使训练后的神经网络构成的控制器对传感器故障具有完整性。该方法对达到完整性设计的要求条件较低，且不受线性系统方法的限制。文末的仿真实例说明了所给方法的有效性。

DESIGN OF NEURAL NETWORKS CONTROLLER POSSESSING INTEGRITY AGAINST SENSOR FAILURES

This paper is concerned with the design of neural networks controller which possesses integrity against some sensor failure Neural networks are introduced to solve this problem The `well trained' neural networks can replace those feedback control laws designed for normal and some sensor failures,and make the closed loop system possess integrity A little condition for the existence of the neural networks controller is required The performance of neural networks control system is very good The simulation results show that the proposed method is effective