加权firm阈值奇异值分解及其旋转机械故障诊断

Weighted Firm Threshold Singular Value Decomposition andRotating Machinery Fault Diagnosis

奇异值分解(Singular value decomposition,SVD)作为一种有效的信号降噪方法广泛应用于旋转机械振动信号周期性瞬态冲击提取中。传统SVD以能量为导向,无法提取出能量较弱但含故障信息丰富的奇异分量(Singular Component,SC)。为此,提出加权firm阈值奇异值分解(Weighted Firm Singular Value Decomposition,WFSVD)方法。该方法首先引入平方包络谱峭度(Squared Envelope Spectrum Kurtosis,SESK)作为量化故障敏感度的指标,以评估各个SC所含故障信息的丰富程度;其次,将SESK作为权重因子引入到基于firm阈值的SC去噪中,设计基于SESK的加权firm阈值SC去噪策略;最后,重构信号,实现信号降噪并有效提取故障特征。对于仿真信号与试验数据的分析验证了所提方法在周期性微弱瞬态冲击提取及旋转机械故障诊断中的有效性。

Singular value decomposition(SVD),as an effective signal denoising method,has been widely used in periodic transient impulses extraction of vibration signals generated by rotating machinery.Since traditional SVD is energyoriented,the singular components(SC)with valid information but weak energy cannot be extracted.Therefore,the weighted firm singular value decomposition(WFSVD)method is proposed in this study.In this method,the squared envelope spectrum kurtosis(SESK)is introduced as the indicator of fault sensitivity quantification to evaluate the richness of fault information contained in each SC.Then,the SESK is introduced as the weight factor into SC denoising based on the firm threshold.Finally,the reconstruction signal is obtained by designing a weighted firm threshold denoising strategy based on SESK.The analysis of simulation signal and experimental signal verifies that the proposed method can extract weak periodic transient impulses effectively in rotating machinery fault diagnosis.