Fuzzy–AI digital twins for sustainable, resilient food supply chains

This study proposes a prescriptive Fuzzy–AI Digital Twin framework for food supply chains operating under deep uncertainty. The digital twin streams real-time operational data from production, cold storage, transport, and retail into a learning optimization loop. Uncertainty in demand, lead times, capacities, spoilage, and emissions is modeled using adaptive fuzzy sets, while a deep-learning surrogate accelerates a multi-objective NSGA-II optimizer to co-design decisions across cost, CO₂e, and resilience. In a perishable dairy case, the hybrid approach reduces total cost by 12.2%, lowers emissions by 15.2%, and raises the resilience index by 22.5% relative to the Fuzzy–AI model without Digital Twin integration. The optimizer stabilizes in ∼35 generations, and under severe disruptions the network recovers to 95% capacity in ∼5 h (vs. >10 h traditionally). Conceptually, the framework operationalizes resilience via adaptability (sensing disruptions, simulating alternatives, and reallocating resources in real time) rather than costly redundancy. Managerially, it enables simulate-before-implement orchestration and transparent cost–carbon–resilience trade-offs aligned with Scope-3 reporting. The results indicate that digital twins combined with fuzzy reasoning and learning assisted optimization can deliver simultaneously lower cost, lower emissions, and higher resilience for perishable food chains.