A Degradation Condition Assessment Technique Based on online_ISSE Degradation Feature and Logistic Regression Model

Abstract

A degradation assessment technique based on an online improved symbol sequence entropy online_ISSE and a logistic regression model is proposed in this paper. Firstly, the threshold factor is introduced to retain the `coarse graining' information of direction changing and amplitude information, the `sensitivity' of improved symbol sequence entropy (SSE) to impact components is reduced and improved symbol sequence entropy (ISSE) is proposed. Then, a sliding window and Weibull distribution theory are used to effectively filter out the influence of fluctuations in the ISSE feature sequence, forming the degradation feature named online_ISSE. Finally, a logistic regression model is trained and constructed, and the health factor CV is calculated online to assess the degradation condition of the unknown signal samples. The lifetime vibration signal of the hoisting gearbox monitored from #8114 quay crane of the Shanghai Port Container Terminal is introduced for instance analysis. The results show that the proposed ISSE has a better effect in describing the complexity pattern than the SSE algorithm and that the degradation condition can be tracked and assessed accurately based on the technique proposed.