基于时序分析与神经网络的气阀机构故障诊断

Fault Diagnosis of Valve Train by Using Time-series Analysis and Neural Networks

通过模拟柴油机气阀机构的两种主要故障 :气门漏气和气门间隙异常进行实验 ,采集缸盖表面的振动信号。利用时间序列分析方法对振动信号建立AR和ARMA模型 ,利用其参数及残差等指标作为特征参数 ,提取时域的均方根等指标。最后利用人工神经网络进行故障模式识别。结果表明方法是可行的 。

By simulating the two main faults of the valve train:gas leakage and abnormal lash,the vibration signals of cylinder head had been measured.Based on the time series analysis method,the AR and ARMA models of cylinder head vibration signals were set up.The model parameters and vaiance were used as the characteristic paramenters to extract the mean square root of time domain.At last,neural network was used to diagnose the valve train faults.The results show that it is feasible and effective to diagnose the valve train faults by using Time Series analysis method and NN.