An Improved Imperialist Competitive Algorithm Based Rolling Horizon Strategy for Photolithography Machines Scheduling

摘要

This paper focuses on the photolithography machines scheduling problem in wafer fabrication factories with the objective of total completion time minimization. The operation constraints in photolithography machines include wafers arriving over time, dedicated machine constraints for critical layers, auxiliary resources constraints and unrelated parallel machines. As the problem is NP-hard and traditional scheduling algorithms cannot achieve a global optimization while considering all the constraints, an improved imperialist competitive algorithm based rolling horizon strategy is proposed. To balance the computation efficiency and optimization performance, a variable time interval approach is presented to decide the scheduling point. To search for global optima in every local schedule, the result of SRPT rule is incorporated into the local objective function. To avoid converging to local optima, an adaptive assimilation operation and a socio-political theory based competition operation are adopted in the imperialist competitive algorithm. To verify the proposed algorithm, numerical tests are conducted using historical data from a semiconductor manufacturing factory in Shanghai. Results show that the average cycle time of photo can be decreased by 14.0%.