基于奇异值分解和独立分量分析的滚动轴承故障诊断方法

Rolling Bearing Fault Diagnosis Method Based on Singular Value Decomposition and Independent Component Analysis

针对强背景噪声下难以提取滚动轴承早期故障信号中故障特征频率的问题,提出奇异值分解和独立分量分析相结合的滚动轴承故障诊断方法。该方法首先利用相空间重构将一维时域矩阵拓展到高维矩阵,得到吸引子轨迹矩阵;然后对轨迹矩阵进行奇异值分解降噪,依据奇异值差分谱阈值原则选取相应阶次分量进行重组构造虚拟噪声通道;接着将重组信号和观测信号进行独立分量分析分离;最后利用能量算子解调方法提取出有效的故障特征分量,进而识别故障类型。滚动轴承故障诊断实验和仿真结果表明该方法有效可行。

To solve the problem that it is difficult to extract the characteristic frequency of the fault in the early fault signal of the rolling bearing under strong background noise, the signal analysis method of singular value decomposition-independent component analysis was proposed. At first, phase space reconstruction was used to extend the one-dimensional time-domain signal to higher dimensions, and obtain the attractor trajectory matrix. Then singular value decomposition was performed on the trajectory matrix to reduce noise. According to the singular value difference spectrum threshold principle, the corresponding order components were selected for recombination to construct the virtual noise channel. Then the recombined signal and the observation signal were separated by ICA. Finally the energy operator demodulation method was used to extract the effective fault feature components to identify the fault type. The fault diagnosis experiment and simulation results of rolling bearing showed that the method is effective and feasible.