A Novel Method for Fault Diagnosis of Motor Bearings via Chord Transformation Strategy

The unlabeled samples seriously affect the application of data-driven methods in the field of bearing fault diagnosis. Inspired by the music theory knowledge, a bearing fault diagnosis method via chord transformation strategy is proposed. The behaviors of chord transformation producing different tones of sound is simulated, and known bearing fault sizes are used to render unlabeled bearing samples. That is, this method can be used to present multiple fault characteristics by simulating chord transitions when pitch changes occur. The experimental results indicate that this method can diagnose bearing faults of different types and severity. The effectiveness is verified under experimental conditions of variable operating conditions compared to existing advanced methods.