Multiscale Sparsity Measure Fusion for Bearing Performance Degradation Assessment

As the basis of condition-based maintenance, machine health monitoring aims timely to detect the incipient faults and quantitatively assess the machine performance degradation. This study proposes a multiscale sparsity measure (SM) fusion framework to enhance the performance of bearing degradation assessment with SMs. The framework consists of the following three steps. First, the multiscale SMs are constructed by combining the multiscale analysis and adaptive weighted signal preprocessing technique (AWSPT) for enriching the bearing status features. Second, a fusion scheme including the offline training and online testing stages is presented at the feature level, where the multiscale SMs are rapidly fused into a diverse feature. Third, an adaptively normalized transforming technique is designed to scale the diversity feature from 1 to 0. The transformed confidence values (CVs) are used as a monitoring indicator to realize the quantitative bearing performance degradation assessment. Two experimental cases validate that the proposed method is more effective for detecting the incipient defects and describing the different degradation phases of the tested bearing than the classical SMs and the advanced AWSPT-based SMs.