ARTIFICIAL NEURAL NETWORKS AND METAHEURISTIC OPTIMIZATION TO INCREASE THE PERFORMANCE OF SUPPLY CHAINS UNDER DISRUPTIONS

Nowadays, global supply chains are subject to unforeseen disruptions which deteriorate their overall performance. In this work, we propose a new AI-based framework to run the replenishment operations of the factory node for a three-echelon SC with production capacity constraints. We focus on the factory since the delivery lead times from the supplier to the factory are stochastic and can experience sudden delays due to unforeseen disruptive events. The factory node adopts the smoothing order-up-to (S-OUT) replenishment policy and the order quantity must be subject to a minimum order quantity (MOQ) constraint. The proposed framework consists of a digital twin, which periodically interacts with a virtual model of the physical supply chain, and a machine learning engine powered by a metaheuristic algorithm that dynamically adjust the replenishment parameters to maximize the fill rate while reducing the stock level. To this end, Artificial Neural Networks (ANNs) and Particle Swarm Optimization (PSO) are employed in the dynamic selection of the S-OUT parameters. To assess the effectiveness of the proposed approach, an experimental campaign was carried out; there, the AI-based framework is compared with the static scenario, in which the same replenishment parameters are adopted over the entire time horizon. The analysis of results demonstrates the effectiveness of the proposed machine learning method to promptly face the effects of disruptions and increase the performance of SCs.