一种基于模糊神经网络的故障分类器及其在多传感器故障诊断中的应用  

An Fault Classifier Based on Neuro-Fuzzy Networks with Extension to Multi-sensor Fault Diagnosis

作　　者：刘宜平[1] 沈毅[1] 刘志言[1] 

机构地区：[1]哈尔滨工业大学控制工程系,327信箱哈尔滨150001

出　　处：《传感技术学报》2000年第1期38-43,共6页Chinese Journal of Sensors and Actuators

摘　　要：提出了一种应用模糊神经网络进行故障诊断新方法．采用模糊神经网络作为故障分类器，离线地自适应从学习样本数据中提取各个用以描述故障状态的模糊参考模型．在诊断时，此模糊神经网络在线地得到当前系统的模糊模型描述，并将与各个参考模型相匹配，从而得出正确的诊断结果．它适用范围广泛，如用于控制系统的过程对象以及传感器、执行器故障的检测与诊断．通过对燃汽轮机控制系统多传感器故障诊断的仿真证明了此法的有效性和优越性．A new approach of fault diagnosis using neuro-fuzzy network is proposed. The neu- ro-fuzzy network is adopt as fault classifier, and it can abstract each fuzzy reference model which describes the states of control system by adaptively learning from the examples data offline. When faults happened, it obtains current fuazzy model online and then gets the correct result by contrasting to each fuzzy reference model. This approach is not only suitable to diag- nose the fault of control system, but also it is effective for the faults of sensors or actuator. The simulation results can be used to verify the validity of this approach.

关 键 词：模糊神经网络 故障诊断 故障分类器 多传感器 

分 类 号：TP212.063[自动化与计算机技术—检测技术与自动化装置] TP18[自动化与计算机技术—控制科学与工程]