An approach to bearing fault diagnosis based on ensemble learning and case-based reasoning

Abstract

Abstract In response to the challenge of multiple fault types and complex diagnostic criteria in bearing fault diagnosis, a case reasoning method based on ensemble learning is proposed. The approach utilizes Case-Based Reasoning (CBR) to construct a case library for vibration-based features of rolling bearings and perform fault diagnosis. Moreover, addressing the issue of determining optimal feature weight ratios when retrieving similar cases in traditional case reasoning methods, a Random Forest algorithm combined with Bayesian Optimization is introduced. This integration allows for adaptive retrieval of similar cases, thereby enhancing the diagnostic capability for bearing faults. The effectiveness of this approach is validated through experimental analysis.