MC-FSTG: A fine-grained spatio-temporal graph neural network based on multi-scale operating conditions for bearing temperature prediction in high-speed trains

Key Points

• Improved temperature prediction accuracy with the MC-FSTG model is significant for high-speed trains.
• The model achieved a 15% increase in prediction accuracy compared to traditional methods.
• Observational analysis across varying operating conditions demonstrated the model's robustness and adaptability.
• This finding may enable better maintenance strategies, reducing potential failures in complex transportation systems.