Disruption mitigation and recovery in supply chains using portfolio approach

Abstract

An innovative portfolio approach and stochastic MIP formulations with an embedded network flow problem are developed for selection of primary and recovery suppliers and assembly plants in the presence of supply chain disruption risks. Local and regional multi-level disruptions of suppliers and assembly plants are considered. Unlike most of reported research on supply chain disruption management a disruptive event is assumed to impact both a primary supplier of parts and the buyer’s firm primary assembly plant. Then the firm may choose alternate (recovery) suppliers and move production to alternate (recovery) plants along with transshipment of parts from the impacted primary plant to the recovery plants. The resulting allocation of unfulfilled demand for parts among recovery suppliers and unfulfilled demand for products among recovery assembly plants determines recovery supply and demand portfolio, respectively. The selection of supply and demand portfolios is determined simultaneously with production scheduling in assembly plants. An integrated decision-making approach with the perfect information about the potential future disruption scenarios is compared with a hierarchical approach with no such information available ahead of time. In the integrated approach a two-stage stochastic model is applied, in which the first stage decision considers disruption scenarios to happen in the second stage so that the impact of disruption risks is mitigated. The second stage decision optimizes the supply chain recovery process. The scenario analysis indicates that for the hierarchical approach the best-case and worst-case disruption scenarios are, respectively subsets and supersets of the corresponding scenarios for the integrated approach. In addition to risk-neutral decision-making based on expected cost or expected service level optimization, an integrated risk-averse approach is developed using CVaR risk measure. The findings indicate that the developed portfolio approach leads to computationally efficient MIP models with a very strong LP relaxation.