基于奇异谱相对熵与灰色绝对关联度的监测数据特征分析

提出将奇异值分解(SVD)与相对熵、灰色绝对关联度算法相结合,对武广线车轮踏面磨损4工况监测数据提取特征进行分析,旨在充分认识踏面工作状态经历正常-轻微磨损-中度磨损-严重磨损接近镟修这一过程中的信号变化规律,并开展基于奇异谱熵特征提取的对照实验。后续仿真结果表明:踏面性能退化越深,其监测信号与正常状态的相似性就越小,所得奇异谱相对熵特征值越大,灰色绝对关联度特征值就越小,即此二维两个特征是衡量车轮踏面性能退化过程的有效指标;其次,奇异谱相对熵的特征分析结果明显优于对照实验中的奇异谱熵。

Monitoring Data Feature Analysis Based on Singular Spectrum Relative Entropy and Grey Absolute Relational Grade

To recognize the changing rules of vibration signals, during the whole process of wheel tread working conditions changes from normal state to slight wore, and then medium wore, at last heavily wore state right before lathing. A new feature extraction model, which based on the singular value decomposition (SVD) combined relative entropy as well as grey absolute relational grade algorism, was proposed in this paper to analyze the Wuhan-Guangzhou PDL GPS measured data of wheel tread wears under 4 operating conditions. Meanwhile, the singular spectrum entropy based controlled experiment was conducted in this paper. The follow-up simulation results proved that when the wheel tread got heavily degraded, the similarity between normal condition signal and heavy-degraded state signal got smaller. As a result, the relative entropy value got larger, whereas the grey relational grade value got smaller. Namely, this two features can effectively describe the performance degradation process of wheel tread. What’s more, the relative entropy was more preferable than the singular spectrum entropy in measuring the wheel tread wearing degrees.