A novel framework that dramatically improves Remaining Useful Life (RUL) predictions for industrial machinery through causal analysis and transfer learning.

• Introduces the Causal-Informed Data Pruning Framework (CIDPF) that identifies key causal relationships in sensor data
• Leverages Gaussian Mixture Models to screen for anomalies in high-dimensional data
• Employs transfer learning with LLMs to improve prediction accuracy across different machine types
• Delivers more reliable maintenance scheduling and reduced downtime in factory settings

This approach addresses a critical engineering challenge by extracting meaningful signals from complex sensor data, enabling more precise predictive maintenance and extending equipment lifespan.

Causal Inference based Transfer Learning with LLMs: An Efficient Framework for Industrial RUL Prediction