AI-ready simulation-based digital twin framework for operational planning and optimization in container terminals

Container terminals, subject to the pressures and competitiveness of today’s logistics chain, face the challenge of optimally coordinating cranes, handling equipment, and transportation. Discrete-event simulation and digital twins are used for this purpose, but most existing approaches require real-time data that small-and medium-sized terminals typically do not have. Furthermore, few studies address how to progressively incorporate artificial intelligence. To address this gap, we present a five-phase framework for building simulation-based digital twins for operational planning in medium-sized container terminals. Each phase provides a connection point where optimization or learning components can be attached. We validate two of these components: a machine learning metamodel trained on 147 model-generated configurations, and a metaheuristic search algorithm that uses the metamodel to identify better resource allocations. Total operation time drops by 33% compared to the real baseline. We demonstrate the framework on the MSC-ADYA vessel at the Port of Málaga’s container terminal (Noatum Terminals), handling 3,016 container movements with two ship-to-shore cranes, eight trucks, and four RTG cranes. The digital twin runs in FlexTerm and produces indicators on crane utilization, handling rates, truck activity, and yard occupancy. Managers can compare configurations before committing to them. The work fits within the Industry 4.0 and Industry 5.0 agendas for transparent, human-centered decision-making and offers a practical entry point toward AI-supported planning without real-time data infrastructure.