Machine Learning-Based Predictive Maintenance at Smart Ports Using IoT Sensor Data.

Area of Science:

• Maritime Logistics
• Mechanical Engineering
• Data Science

Background:

• Seaports are vital nodes in global containerized cargo logistics, relying on efficient container handling equipment (CHE).
• Inefficient maintenance of CHE leads to operational disruptions, supply chain delays, and increased waiting times.
• Intelligent maintenance strategies are essential for optimizing port operations and resource utilization.

Purpose of the Study:

• To develop a machine learning (ML)-based approach for predicting faults in CHE.
• To improve the reliability of port equipment and enhance overall port performance.
• To address inverter over-temperature faults caused by issues like fan failures and clogged filters.

Main Methods:

• A statistical model was developed to assess the health of the hydraulic system.
• Several ML models were trained and evaluated, including artificial neural networks (ANNs), decision trees (DTs), random forest (RF), Extreme Gradient Boosting (XGBoost), and Gaussian Naive Bayes (GNB).
• The models were used to predict inverter over-temperature faults in CHE.

Main Results:

• Artificial neural networks (ANNs) demonstrated the highest performance in fault prediction.
• ANNs achieved 98.7% accuracy and a 98.0% F1-score in predicting specific CHE faults.
• The developed ML models effectively identified potential equipment failures.

Conclusions:

• The ML-based approach significantly improves the prediction of CHE faults, enhancing equipment reliability.
• Implementing intelligent maintenance strategies at ports can optimize operational efficiency and resource management.
• The study highlights the potential of ANNs in maritime logistics for predictive maintenance.