用于非线性系统的合成故障诊断方法

鉴于对线性系统的故障诊断方法比较成熟,而对非线性系统的故障诊断还有很大不足的状况,将非线性系统在多个工作点进行分段线性化.在非线性系统中借鉴线性故障诊断方法,对每个工作点的模型进行线性故障诊断设计,取得工作点密度和故障诊断性能的均衡,同时使用动态神经网络对各个工作点的工作参数进行拟合.该方法将一组改进的鲁棒观测器与一个为其确定参数的动态神经网络相结合,对非线性系统进行故障诊断.用典型的3水箱模型验证了这个合成方法,验证结果显示,该方法有很好的全局工作点适应性和对干扰的鲁棒性.

Compound approach for fault diagnosis on nonlinear system

For increasing demands on safety and reliability of dynamic systems, fault diagnosis has received more and more research interests in recent years. Unlike mature approaches in linear fault diagnosis,nonlinear approaches have limitations. In order to share mature linear approaches on nonlinear fault diagnosis, nonlinear system can be linearized piecewise. Linear diagnosis can be designed on respective working points. Optimization is applied to balance performance and density of working points. A compound approach with a set of improved robust observers on fault diagnosis on nonlinear system, with a dynamic neural network deciding parameters, is illustrated with an instance of 3-tank model. Advantages of this compound scheme such as global adaptability on working points, robust to disturbance on model,are tested to be effective.