A Multi-agent Digital Twin Framework for Predictive Maintenance Using Machine Learning and Genetic Algorithms: A Case Study Morocco, Tangier Med Port

This study proposes a digital twin–based predictive maintenance framework for crane systems in container terminals. The system integrates real-time sensor simulation, machine learning–based failure detection, and genetic algorithm–based maintenance optimization. Eight digital twin models of cranes were simulated on the Azure Digital Twins platform and monitored through a custom-designed dashboard. Using synthetic yet dynamic data, the framework forecasts failure risk and remaining useful life with a random forest model, while maintenance schedules are optimized to minimize cost and downtime. The results demonstrate the system’s novelty, scalability, and flexibility, as it unifies data-driven forecasting and evolutionary optimization within a real-time digital twin environment, establishing a foundation for next-generation predictive maintenance systems in ports.