A novel hybrid MILP-based solution strategy to industrial-scale AWS scheduling problems

AGUIRRE, ADRIÁN MARCELO; MÉNDEZ, CARLOS ALBERTO

Barcelona

Workshop; CYTED-HAROSA Workshop & Meeting on Applied Optimization & Distributed Computing; 2011

Angel A. Juan, Daniel Riera, Joan Manuel Marquès & Alejandra Pérez-Bonilla - Universitat Oberta de Catalunya (UOC) - 2011 CYTED-HAROSA Workshop General & Local Committee Chairs

The Automated Wet-etch Station (AWS) is one of the most critical stages of a modernsemiconductor manufacturing system (SMS), which has to simultaneously deal with manycomplex constraints and limited resources. Due to its inherent complexity, real-world automatedwet-etch station scheduling problems are very difficult to solve using traditional mathematicalformulations. Thus, heuristic, meta-heuristics and simulation-based methods have been reported inliterature to provide feasible solutions with reasonable CPU times.This work presents a novel hybrid MILP-based decomposition strategy that joins the advantages of a rigorous MILP (Mixed Integer Linear Programming) continuous-time formulation with theflexibility of dynamic heuristic procedures. The schedule generated provides near-optimaldynamic solutions to challenging industrial-sized automated wet-etch stations scheduling problemswith a moderate computational cost. Also, this methodology provides more than a 10% ofimprovement in comparison with best results found in literature for the most complex probleminstances analyzed.