基于张量分解的滚动轴承复合故障多通道信号降噪方法研究

Research on Multi-channel Signal Denoising Method for Multiple Faults Diagnosis of Rolling Element Bearings Based on Tensor Factorization

针对滚动轴承多通道信号同时滤波的问题,提出一种新的基于张量分解的多维降噪技术。该方法在高维空间中将时域信号、频率和通道建模为一个三阶张量模型,首先通过Tucker3分解张量模型,然后通过L曲线准则来选取截断高阶奇异值分解(Truncated HOSVD)的截断参数,根据截断参数求解张量模型,最后根据原张量的组建方式通过逆变换得到新的目标张量。通过滚动轴承复合故障仿真分析对所提出的滚动轴承复合故障检测技术的性能进行了评价,然后应用该方法对轴承试验台采集的振动信号进行降噪分析,试验结果表明该方法在滚动轴承复合故障多维降噪以及特征提取中的有效性和可行性。基于张量分析的多维信号滤波技术将拓宽大数据时代处理异构、多维数据的视野。

Aiming at simultaneously multi-channel signal filtering and multiple faults diagnosis of the roller bearings, a novel multi-channel signal denoising technique is proposed based on tensor factorization. The original vibration signals are first formulated as a 3-way tensor via temporal signal, frequency and channel in a high dimensional space. The tensor model is factorized using Tucker-3 decomposition. Then, key parameters used in the truncated HOSVD are obtained according to the L-curve criterion. Finally, tensor model is solved by those selected truncated parameters, while target tensor is reconstructed by inverse transformation. The performance of the developed technique in detecting faults of roller bearings has been thoroughly evaluated through simulation signals. The presented approach is then applied to reduce noise in vibration signal collected on bearing test-rigs. Experimental results showed that the feasibility and validity of the proposed method in multiple faults detection of the bearings. Multidimensional signal filtering based on tensor will broaden the view in dealing with heterogeneous and multiaspect data in an age of big data.