Anomaly detection of rolling elements using fuzzy entropy and similarity measures

The ability of detecting faults in rotating elements is highly desired in machine condition monitoring application (MCM). On many MCM platforms, discriminating attributes based on time and/or frequency domain of the acquired vibration data are used to classify the element under monitoring into normal and abnormal conditions. However, having such diagnostic ability is still insufficient in our global goal towards predictive maintenance. To achieve true predictive maintenance, the development tool must be able to provide a certain level of real time computation capability. In this paper, the authors propose a novel method based on fuzzy entropy and similarity measure for monitoring the health conditions of ball bearings on-line. The practicalities of the effectiveness and speed of the method are verified empirically, and results are presented towards the end of this paper.