A Real-Time Diagnostic System Using a Long Short-Term Memory Model with Signal Reshaping Technology for Ship...

Area of Science:

• Marine Engineering
• Predictive Maintenance
• Signal Processing

Background:

• Insufficient ship maintenance capacity hinders operational efficiency.
• Accurate assessment of ship propeller bearing health is crucial for preventing failures.
• Existing diagnostic methods lack the precision and efficiency required for modern maritime operations.

Purpose of the Study:

• To develop an advanced ship propeller diagnostic system.
• To enhance operational efficiency and address limitations in ship maintenance capacity.
• To accurately predict the Remaining Useful Life (RUL) of ship propeller bearings.

Main Methods:

• Developed a Diagnosis and RUL Prediction Model utilizing Long Short-Term Memory (LSTM) networks.
• Employed synchronized signal reshaping technology for processing vibration signals.
• Integrated bearing aging experimental data and real-world vibration measurements for system development.

Main Results:

• The developed LSTM-based model achieved a Mean Squared Error (MSE) of 0.018 and Mean Absolute Error (MAE) of 0.039.
• The system accurately determines the current status and Remaining Useful Life (RUL) of propeller bearings, even in worn stages.
• Demonstrated superior prediction accuracy and computational efficiency compared to traditional Recurrent Neural Network and Support Vector Regression models.

Conclusions:

• The integrated ship propeller diagnostic system offers a practical solution for improving maintenance capacity.
• The system enhances ship operational efficiency through accurate real-time health monitoring and RUL prediction.
• This research contributes a robust tool for proactive maintenance in the maritime industry.