基于滑模观测器和神经网络的传感器故障诊断方法比较研究

针对航空发动机传感器故障诊断中各种方法的优势和劣势,选择滑模观测器和神经网络这两种故障诊断方法分别对航空发动机转速传感器进行故障诊断研究,采用实验室搭建的发动机实验台DGEN380的实验数据,选择对航空发动机控制系统影响较大的偏置故障、漂移故障、脉冲故障、周期性干扰故障这四类传感器故障进行诊断。研究结果表明,滑模观测器和IPSO-BP神经网络都能实现航空发动机传感器的故障诊断;滑模观测器方法可以诊断出偏置故障、脉冲故障和周期性干扰故障,但不能诊断出传感器发生的漂移故障; IPSO-BP神经网络方法可以诊断出偏置故障、漂移故障、脉冲故障和周期性干扰故障。因此,滑模观测器在故障诊断中可能会出现漏诊的现象,IPSO-BP神经网络相对滑模观测器而言不会出现漏诊的现象。

Comparative Study of Sensor Fault Diagnosis Methods Based on Sliding Mode Observer and Neural Network

In view of the advantages and disadvantages of various methods in the fault diagnosis of aeroengine sensor,the two fault diagnosis methods of sliding mode observer and neural network are selected for fault diagnosis of the aeroengine speed sensor respectively.Four kinds of sensor faults,including bias fault,drift fault,pulse fault and periodic interference fault,which have great influence on the aeroengine control system,were selected for diagnosis by using the experimental data of the engine test-bed DGEN380 built in the laboratory.The results show that both the sliding mode observer and the IPSO-BP neural network can realize the fault diagnosis of aeroengine sensors.The sliding mode observer method can diagnose the bias fault,pulse fault and periodic interference fault,but can not diagnose the drift fault of sensors.IPSO-BP neural network method can diagnose the bias fault,drift fault,pulse fault and periodic interference fault.Therefore,the sliding mode observer may miss diagnosis in fault diagnosis,and the IPSO-BP neural network will not miss diagnosis compared to the sliding mode observer.