基于神经网络非线性时滞系统的鲁棒可靠控制器设计

Reliable Robust Controller Approach for Nonlinear State-Delayed Systems Based on Neural Network

将一类非线性时滞控制系统中的非线性部分 ,用一个单隐层神经网络来近似代替 ,采用线性微分包含(LDI—— linear differential inclusion)的方法来线性化该非线性环节 ,对于线性化所产生的近似误差、时滞和执行机构故障作为系统的一部分设计可靠鲁棒控制器 ,相关的定理也一并给出。高阶微分和偏微分方程一般是用来解决这类非线性系统的主要方法 ,文章中提出的可靠鲁棒控制器设计方法克服了以上这些方法求解困难的缺点 ,仿真示例用设计好的可靠鲁棒控制器与常规极点配置法进行了比较 ,从而表明了这种方法的有效性。

The neural network is introduced in a class of nonlinear state delayed systems, and the nonlinear term is approximated by one neural network with a hidden layer. This paper adopts linear differential inclusion (LDI) method to linearize the nonlinearized item, the linearized approximation error, state delayed and the actor failure are considered as a part of the systems, and the reliable robust controller is designed and the related theorems are given. Comparing with the general design approach for reliable robust nonlinear control systems based on high order differential equations and partial differential equation, the method based on the reliable robust controller design avoids the disadvantage of their resolution for the above methods to be difficult. A numerical example is given to illustrate the contrast between the designed reliable robust controller and the routine pole assignment method. And the design procedure is proved to be effective.