A physics-informed generalization framework for cross-condition bearing fault diagnosis with limited labeled data

Key Points

• This framework enhances cross-condition diagnosis of bearing faults, showing significant accuracy improvements.
• The analysis demonstrates a 25% increase in accuracy when using limited labeled data compared to traditional methods.
• Using a physics-informed model, this approach offers insights into diagnosing faults across different conditions with less data.
• The findings suggest implications for industries relying on predictive maintenance, potentially lowering operational costs.