Towards resilience: Primal large-scale re-optimization

Perturbations are universal in supply chains, and their appearance has become more frequent in the past few years due to global events. These perturbations affect industries and could significantly impact production, quality, cost/profitability, and consumer satisfaction. In large-scale contexts, companies rely on operations research techniques. In such a case, re-optimization can support companies in achieving resilience by enabling them to simulate several what-if scenarios and adapt to changing circumstances and challenges in real-time. In this paper, we design a generic and scalable resilience re-optimization framework. We model perturbations, recovery decisions, and the resulting re-optimization problem, which maximizes resilience. We leverage the primal information through fixing, warm-start, valid inequalities, and machine learning. We conduct extensive computational experiments on a real-world, large-scale problem. The findings highlight that local optimization is enough to recover after perturbations and demonstrate the power of our proposed framework and solution methodology.