小波除噪、神经网络与发动机故障诊断

Wavelet Noise Filtering, Neural Network and Engine Fault Diagnosis

给出了利用起动电压波形来检测气密性这一简单易行的实验原理与方法。并对代表性波形进行小波除噪处理,在此基础上,提出了6个故障信息特征参数,通过对径向基神经网络的训练和检测,证明该神经网络能够成功的进行故障模式的辨识。从而为发动机压缩性故障诊断提供一个较好的方法。

In this paper an experiment method, of which main purpose was to analyze the cylinder gastightness from starting voltage waveforms was discussed. The noise was filtered from typical waveforms, and 6 fault symptom parameters were put forward. The Radial Basis Function Netword (RBFN) was trained by putting these symptom parameters in. This network can distinguish the fault modes preferably; therefore it will be favourable to diagnose the cylinder compression ratio fault.