基于自学习字典的盲提取方法在滚动轴承多故障诊断中的应用

A Blind Source Extraction Method based on Self-learned Dictionary and Its Application in Fault Diagnosis of Bearing Multi-type Fault

当齿轮箱中的多个轴承同时发生故障时,由于各故障源之间的相互耦合效应,常规盲提取方法难以对其进行有效特征提取。为解决上述问题,提出一种基于稀疏表征自学习字典理论的盲提取方法。首先,将稀疏表征自学习字典方法用于滚动轴承多故障信号分析,得到一系列自学习字典集;然后,利用学习到的字典集重构滚动轴承多故障信号以消除噪声及干扰信号;最后,将盲提取方法用于重构后的滚动轴承复合故障信号,抽取出滚动轴承各单一故障信号,再逐一对单一故障信号进行包络解调分析,以获取相应的故障特征。通过实验,对所述方法的可行性及有效性进行了验证。

When multiple bearings in a gearbox failure simultaneously,conventional blind source extraction(BSE)on the vibration signals of bearing multi-type faults would not be ideal due to the mutual coupling effect among each of the fault sources.A BSE based on sparse representation self-learned dictionary method is proposed to solve the above problem.Firstly,apply the self-learned sparse dictionary originating from sparse representation on the multi-type faults vibration signals directly and a set of self-learning dictionaries are obtained.Then,the multi-type faults vibration signals are re-constructed basing on the obtained learned dictionary to eliminate noise and interference signals.Finally,apply the BSE method on compound fault signals of reconstructed rolling bearings,each single fault signal of rolling bearing is extracted,and then the envelope demodulation analysis is carried out one by one to obtain the corresponding fault characteristics.Feasibility and effectiveness of the proposed method are verified through experiment.