基于支持向量机的柴油机气阀漏气的声发射诊断研究

Research on Acoustic Emission Diagnosis of Valve Leakage for Diesel Engine Based on Support Vector Machine

通过分析支持向量机用于柴油机小样本诊断的特点,进行不同程度的气阀漏气故障模拟试验,并结合参数分析法和小波包分析法提取了故障特征参数,建立了气阀漏气故障的诊断模型。研究结果表明:高通滤波后缸盖声发射信号的工作循环有效值、排气阀关闭段有效值和燃烧段有效值,以及经三层Daubechies8小波包分解得到的三个频带(S2、S3、S4)的能量比可有效反映气阀漏气故障,组成6维特征向量建立的诊断模型能有效实现对不同程度气阀漏气故障的准确诊断,且分类性能稳定。

Features of applying support vector machine （SVM） to diesel engine fault diagnosis with small samples were discussed. The different levels of exhaust valve leakage in diesel engine were simulated in test laboratory. By combining parameter analysis method and wavelet packet analysis method to extract the sensitive fault characteristic parameters, the diagnosis model was established based on SVM. It is confirmed that the effective values in whose cycle, exhaust closed section and combustion section of the high-pass filtered cylinder head acoustic emission signals and the energy ratios of 3 frequency bands i. e. $2,$3 and $4, obtained by 3-level Daubechies8 wavelet packet decomposition can effectively indicate valve leakage faults, and the diagnosis model based on the six fault characteristic parameters can be used effectively to diagnose different level valve leakages and its classification performance is stable.