模型数据相结合的电液伺服系统传感器早期故障检测与辨识

A Combined Model-Based and Data-Driven Method for Early Sensor Fault Detection and Identification of an Electro-Hydraulic Servo System

电液伺服系统中存在各类非线性因素和不确定干扰,传感器早期故障引起的系统状态量变化往往小于干扰引起的变化,因此很难发现。本文提出了一种模型和数据相结合的电液伺服系统传感器早期故障诊断方法,通过基于模型的方法对系统中难以准确测量的外干扰力进行有效解耦,通过建立观测器组对系统中常见的传感器故障进行辨识,针对系统中的建模误差、测量噪声等剩余不确定干扰,采用基于神经网络的方法进行模型补偿,通过建立补偿规则,可进一步提高对早期故障的敏感性。仿真和实验结果验证了所提出的混合式故障诊断方法的有效性。

There are various nonlinear factors and uncertain interference in the electro-hydraulic servo system.The changes of the system state caused by the early sensor faults are often less than the change caused by the disturbance,so it is difficult to be detected.This paper presents a combined model-based and data-driven method to diagnose early sensor faults in an electro-hydraulic servo system.The model-based method is used to decouple external disturbance forces that are difficult to measure accurately in the system,and a batch of observers are designed to identify common sensor faults.Then a neural network based compensation method is proposed to further reduce the influence of the remaining uncertainties such as modeling error and measurement noise in the system.By establishing possible compensation rules the diagnosis sensitivity to early faults can be further increased.Finally,simulation and experimental results verified the effectiveness of the proposed hybrid fault diagnosis method.