A Probabilistic AI-Driven Digital Twin Framework for Customer-Centric Optimization of Maritime Supply Chains under Market Uncertainty

Maritime supply chains increasingly operate under volatile market conditions, demand uncertainty, and frequent disruptions, exposing the limitations of optimisation approaches based on static assumptions and cost-dominant objectives. This study develops a probabilistic, AI-driven digital twin framework to enable customer-centric decision-making in maritime logistics systems. The framework constructs a continuously updated digital representation of the physical network using multi-source data streams, including port operations, vessel movements, demand signals, and environmental factors. The proposed architecture integrates deep learning for demand forecasting, reinforcement learning for adaptive control, and Bayesian inference for uncertainty quantification within a unified modelling environment. In contrast to conventional models, customer-oriented performance indicators, including service reliability, delivery timeliness, and information visibility, are embedded directly in the optimisation and reward structures. This design enables the system to balance operational efficiency with service quality under stochastic and rapidly changing conditions. The framework is evaluated through a calibrated case study of a containerised maritime logistics network under multiple demand and disruption scenarios. Compared with a deterministic baseline, the proposed approach improves service-level performance by up to ۲۳%, reduces operational cost by up to ۸.۶%, decreases average delays by more than ۳۰%, and shortens recovery time following severe disruptions by approximately one-third. Sensitivity and robustness analyses indicate stable performance across a wide range of demand variability and network disturbances. The study contributes by integrating digital twin technology, artificial intelligence, and probabilistic modelling into a single decision architecture while explicitly incorporating customer-centric objectives. The results demonstrate that such integration yields more stable, adaptive, and service-oriented supply chain performance, providing a practical foundation for next-generation intelligent maritime logistics systems.

کلیدواژه ها:

Digital Twin ، Maritime Supply Chains ، Customer-Centric Optimisation ، Artificial Intelligence ، Market Uncertainty ، Reinforcement Learning ، Bayesian Inference ، Logistics Resilience ، Demand Forecasting ، Smart Ports