UPW-Aware Semiconductor Supply-Chain Planning Under Disruptions: Qualification, Reclamation and Stochastic-Robust Optimization

Abstract

Semiconductor manufacturing depends on geographically concentrated, capital-intensive networks and on ultrapure water (UPW), whose availability is constrained by basin conditions, treatment requirements, reclamation capacity, and process qualification. This study develops a multi-period two-stage stochastic-robust multi-objective mixed-integer linear programming model that jointly determines capacity expansion, UPW-reclamation investment, and backup-fab qualification before disruption, while coordinating production, transportation, inventory, qualified outsourcing, freshwater feed, reclaimed feed, emergency water, and unmet demand after uncertainty is revealed. The four objectives minimize lifecycle cost, carbon emissions, scarcity-weighted freshwater footprint, and the unmet-demand ratio. A stabilized adaptive multi-cut column-and-constraint generation method, SAMC-C&CG, combines critical-pattern initialization, violation-based cut selection, duplicate and dominance screening, incumbent completion, and conditional L1 stabilization. The computational study uses 60 independent synthetic networks, mechanism and algorithmic ablations, an 18-run sensitivity analysis design, a five-level AUGMECON2 experiment, and 1,000 common-random-number out-of-sample realizations. Across the 60 networks, the proposed policy reduces lifecycle cost by 1.593% and scarcity-weighted freshwater footprint by 11.900%, while carbon emissions increase by 0.673% and mean service changes by only −0.092 percentage points. Out of sample, cost, carbon, and freshwater decrease by 1.586%, 0.950%, and 19.029%, respectively, while service decreases by 1.254 percentage points. The results support integrated qualification and reclamation planning but also demonstrate that finite scenario representations can overstate service invariance in-sample.