基于熵值比-DTW度量奇异值相似度指标的滚动轴承退化历程辨识

Identification of Degradation Processes for Rolling Bearings by Singular Value Similarity Index Based on DTW Optimized by Entropy Ratio

针对滚动轴承健康状态监测现有退化指标单调性及鲁棒性差,且数据波动造成不同退化状态区分度低等问题,提出了一种基于熵值比-动态时间规整(DTW)度量的奇异值相似度指标并用于辨识滚动轴承不同退化历程。首先,利用奇异值分解(SVD)算法对不同时刻采集的轴承信号矩阵进行分解,将奇异值作为轴承退化特征;其次,基于DTW算法计算轴承连续退化奇异值时间序列的相似度,表征轴承的全寿命周期历程;最后,考虑到轴承不同退化状态的差异性,将熵值比作为权值对相似度指标进行优化,提高相似度退化指标的单调性及对早期异常点的敏感性。对IMS轴承全寿命周期数据的研究结果表明:奇异值相似度指标的单调性、鲁棒性及敏感性较好,可有效避免数据波动对轴承健康状态带来的干扰,能更准确地反映轴承全寿命退化历程。

Aimed at the problems of poor monotonicity and robustness of existing degradation indexes in health state monitoring of rolling bearings and low degree of discrimination for different degradation states caused by data fluctuation,a singular value similarity index based on dynamic time warping(DTW)optimized by entropy ratio is proposed and used to identify different degradation processes of the bearings.Firstly,the singular value decomposition(SVD)algorithm is used to decompose the bearing signal matrix collected at different times,and the singular value is taken as bearing degradation feature;Secondly,the similarity of singular value time series of continuous degradation of the bearings is calculated based on DTW algorithm to describe the full life cycle of the bearings;Finally,considering the differences of different degradation states of the bearings,the entropy ratio is used as weight to optimize the similarity index,so as to improve the monotonicity of similarity degradation index and the sensitivity to detect early abnormal point.The research results using IMS bearing full life data show that singular value similarity index has good monotonicity,robustness and sensitivity,which can effectively avoid the interference of data fluctuation on bearing health status,and can more accurately reflect the full life degradation process of the bearings.