Bearing Health State Representation and Fault Identification Based on Multisource Sensor Signal Feature Information Fusion

The acoustic emission (AE) sensor exhibits significant advantages in detecting metal fatigue cracks; however, its application in the actual testing of bearing fatigue processes has not been sufficiently explored. This study validates the efficacy of AE sensors in identifying bearing fatigue faults and proposes a method for representing bearing health status and fault detection through the fusion of multisource sensor feature information. Initially, feature information from both time-domain and frequency-domain of AE and accelerometer sensors was extracted. A comprehensive index considering feature correlation and standard deviation was designed for feature selection. A multisource sensor features information fusion method based on Mahalanobis distance (MD) was proposed, using real-time variations in MD as an indicator of bearing health. Furthermore, an initial fault localization method combining MD with sensor signal envelope spectrum data was introduced. The methodology was validated through designed accelerated fatigue life tests on bearings. Results demonstrate that in practical applications, AE and accelerometer sensors exhibit strong complementarity, and their combined use can more effectively represent bearing health status and enhance fault detection capabilities.