基于动态神经网络的非线性系统鲁棒故障检测

Robust Fault Detection Based on Dynamic Neural Network for Nonlinear Systems

针对一类具有不确定性扰动的非线性系统,将设计的系统线性观测器产生的误差信号作为残差,采用一种具有高斯型激励函数的动态神经网络(DNN)对残差信号进行分析处理,得到了系统的鲁棒故障检测方法.文中分析了该方法的稳定性和故障检测的鲁棒性,并通过算例验证了该方法的有效性.

A robust fault detection strategy based on dynamic neural network （DNN） is proposed for a class of nonlinear systems with uncertain disturbances. The strategy is realized by using DNN with Gauss active functions to approximate the residual signal of linear observer. The stable convergence of the DNN and the robustness of fault detection are demonstrated. Simulation results showed the effectiveness of the proposed fault detection approach.