Resilient supply chain network design under super-disruption considering inter-arrival time dependency: a new data-driven stochastic optimization approach

During large-scale disruptions, particularly super-disruptions such as global pandemics or large-scale natural disasters, supply chains are exposed to significant adverse impacts. This paper addresses the resilience in a supply chain network design problem under disruption risk by explicitly modeling the dependency between the inter-arrival times of disruptive events and severity of their consequences. A novel data-driven stochastic optimization framework is proposed to consider the ripple effects that typically propagate across supply chain networks following severe disruptions. Specifically, we have devised a hybrid methodology that integrates a clustering algorithm (unsupervised machine learning technique), a phase-type disruption model, and a two-stage stochastic model. To elaborate, a genetic-based clustering algorithm is used to identify the structure dependencies in the input data. Phase-type distributions and their associated theorems are then used to determine the probability distributions of disruptions. A novel mathematical model is developed to design the supply chain using the scenarios generated based on the obtained distributions, which is then solved using the Lagrangian decomposition combined with a new hyper-matheuristic algorithm. The computational efficiency and practical value of the proposed approach are demonstrated through a real-world case study. The findings highlight the effectiveness of developed methodology in designing a resilient supply chain, the proposed resilience strategies substantially improve the supply chain’s performance compared to a non-resilient approach.