基于时频奇异谱和RVM的柴油机故障诊断研究

STUDY ON DIESEL ENGINE FAULTS DIAGNOSIS BASED ON TIMEFREQUENCY SINGULAR VALUE SPECTRUM AND RVM

提出一种基于双树复小波包时频奇异谱和关联向量机的柴油机故障诊断方法,针对连续小波时频分布计算量大,分析速度慢,运用双数复小波包分解提取柴油机缸盖振动信号中的时频分布特征,并进行奇异值分解,结合柴油机运行的时域特点,通过特征优选,形成组合特征集,输入关联向量机多类分类器,从而实现柴油机的故障诊断。试验结果表明,该方法分析速度快,故障识别效果好。

A faults diagnosis method of diesel engine based on timefrequency distribution singular value spectrum and RVM（relevance vector machine） was proposed.In order to gain computation efficiency,the discrete DT-CWPT（dual-tree complex wavelet packet transform） transform method was used to abstract the time-frequency distribution of the diesel engine cylinder head vibration signals,and the feature was extracted by singular value decomposition based on the time domain characteristics of engine,and then the features after optimum seeking were used as the input parameters of the RVM to realize the identification of diesel engine faults.Simulation results indicate that the method has high computation efficiency and good performance of accuracy.