Open AccessA Novel Degradation Feature Extraction Technique Based on Improved Base-Scale Entropy
Author(s) -
Zhenyi Chen,
Chen Shao,
Xiong Hu,
Bing Wang,
Daobing Zhang,
Xiaomei Tao
Publication year - 2021
Language(s) - English
DOI - 10.20855/ijav.2020.25.11717
Subject(s) - entropy (arrow of time) , amplitude , computer science , sample entropy , algorithm , feature extraction , pattern recognition (psychology) , data mining , artificial intelligence , mathematics , physics , thermodynamics , quantum mechanics
In order to track the performance degradation trend accurately, a novel degradation feature extraction technique is proposed based on improved base-scale entropy. A unified base scale is proposed and a new symbol standard is defined to overcome the disadvantages of the base-scale entropy method, so as to symbolize signal amplitude to characterize information amount under different degradation conditions quantitatively. A lifetime dataset of rolling bearing from the IMS Bearing Experiment Center is introduced. For instance, analysis and some entropy-based techniques including fuzzy entropy, approximate entropy and sample entropy are imported for comparison. The results show that the improved basic-scale technique is able to characterize information amount of the signal amplitude distribution, so that the characterizing performance degradation degree of bearing shows a proportional relationship. When comparing the entropy-based techniques, the improved base-scale entropy technique has a faster calculation speed and better algorithm stability.