基于CPFs的齿轮箱复合故障特征提取

Composite Fault Feature Extraction of Gearbox based on CPFs

由于方法选择不当,齿轮箱中复合故障的特征提取会出现漏诊断或误诊断现象,LMD(Local mean deconvolution)对信号分解时由于噪声影响,会出现EMD(Empirical mode decomposition)相似的模态混叠现象,常导致能量泄漏或误诊现象。提出了一种CPFs-MOMEDA(Combined physical functions-Multipoint optimal minimum entropy deconvolution adjusted)的齿轮箱复合故障诊断方法。首先通过LMD对原信号降噪,得到一系列的PFs,通过相关系数法剔除虚假分量和残余成分;计算每层PF(Production function)的多点峭度,提取故障特征周期,将不含周期性冲击的PFs二次剔除,为了保持原信号的完整性,通过组合乘积函数方法重新组合具有相同周期的PF;最后设定不同的周期区间,通过MOMEDA对组合后的信号降噪,进一步提取故障特征。并将此方法应用在齿轮箱复合故障特征提取中,验证了此方法的可行性。

Due to the improper selection of the method,the leakage diagnosis or misdiagnosis usually occurs when the composite fault features in the gearbox are extracted.As a result of noise,there will be the phenomena of mode aliasing after LMD decomposition.In view of this,a hybrid fault diagnosis method of the gearbox based on CPFs-MOMEDA is proposed.First,the original signal is denoised by LMD,and a series of PFs are obtained.And the false component and the residual component are removed by the correlation coefficient method.Then,the definition of multi-point kurtosis is introduced,and the fault feature period is extracted by calculating the multi-point kurtosis spectrum of each layer of PF,and the PF without periodic impact is eliminated.In order to maintain the integrity of the original signal,PFs with the same period are recombined by the combined product function method.Finally,the recombinant signal is denoised by MOMEDA by setting different period intervals to further extract the fault characteristics.And the method is applied to the feature extraction of gearbox composite faults to verify the feasibility.