基于相空间重构与独立分量分析的局部独立投影降噪算法

Local independent projection de-noising algorithm based on phase-space reconstruction technique and independent component analysis

针对故障分析信号中存在噪声问题,提出一种将相空间重构与独立分量分析相结合的局部独立投影降噪算法。其中相空间重构的目的在于从高维相空间中恢复混沌吸引子,独立分量分析能够找到信号的主流形,选择邻域是为了将特征相近的相点结合在一起。使用该方法对正弦仿真信号和Lorenz仿真信号进行降噪处理,结果表明局部独立投影降噪算法的降噪效果与局部独立分量分析算法降噪效果接近,但优于全局投影降噪算法。运用该方法对低速重载轴承振动信号进行分析,准确判断出轴承故障。

A noise problem exists in failure analysis of signals. Local independent projection de-noising algorithm combining phase-space reconstruction and independent component analysis was proposed. Phase-space reconstruction could restore chaotic attractors in a high-dimensional phase space. Independent component analysis could find mainstreams of a signal shape. Neighborhood selection could combine the phase points with similar features. This method was used to de-noise a sinusoidal simulation signal and a Lorenz simulation signal. The result showed that the noise reduction effect of local independent projection analysis is similar to that of local independent component analysis, but better than that of golbal projection algorithm. Vibration signals of a bearing with low rotating speed and heavy load were analyzed by using this proposed method and the bearing faults were identified successfully.