A novel mixed-integer linear programming model for dynamic reconfiguration optimization in sustainable reconfigurable supply chains 5.0

Reconfigurable Supply Chains (RSCs) offer a flexible, resilient alternative to rigid supply chain structures by allowing rapid structural changes in response to disruptions. This paper tackles the dynamic multi-echelon RSC reconfiguration problem for the electronics sector, where customer demand follows Gaussian, cyclic, and triangular fluctuation patterns. We propose a hybrid framework that couples a Knowledge Graph (KG) implemented in Neo4j with a Mixed-Integer Linear Programming (MILP) model. The KG captures and updates the relationships between raw-material suppliers, transporters, production sites, distribution centers, and retailers, while the MILP selects, at each demand arrival, the configuration that minimizes both the total economic cost and greenhouse gas (GHG) emissions. To assess performance, we generated small-, medium-, and large-scale instances. Numerical experiments show that the model provides solutions in less than 0.2 s per period for small instances and less than 1 s per period for medium instances, while the large instance (55 entities) requires about 21 h of total computation time due to the problem's combinatorial complexity. Sensitivity analyses confirm the adaptability of the approach to demand volatility: as demand increases from about 100 to 1500 units, the number of active suppliers and distribution centers rises from one to three, and, with balanced weights, dynamic reconfiguration reduces both economic cost (up to 1.7%) and GHG emissions (up to 3.7%) compared with a static configuration without reconfiguration. To the best of our knowledge, this is the first study to directly link dynamic supply chain reconfiguration with KG-based reasoning, enabling explainable and sustainable decisions across all echelons.