LSTM-BASED MACHINE LEARNING FOR REMAINING USEFUL LIFE PREDICTION OF BEARING MOTORS USING MULTI-SENSOR MONITORING

This study presents an advanced Long Short-Term Memory (LSTM) machine learning framework for predicting the Remaining Useful Life (RUL) of bearing motors through multi-sensor monitoring. Critical parameters, including vibration (RMS), acoustic emission, temperature, stator current, and rotational speed (RPM), were simulated over a 1000-day operational period for three motors with varying conditions. Failure thresholds were defined to represent severe operational conditions. The LSTM model achieved RMSE values of 28.15, 30.29, and 29.21 days and R² values of 0.989, 0.9876, and 0.9877 for training, validation, and test datasets, respectively. These results demonstrate high predictive accuracy and reliability. Integrating multi-sensor data improves the model’s robustness and supports proactive maintenance planning. The study provides a foundation for future integration of LSTM-based predictive models with IoT-enabled real-time monitoring systems in industrial applications.