Semiconductor wafer fabrication production planning using multi-fidelity simulation optimisation

Abstract

Semiconductor wafer fabrication is a complicated and time-consuming production process in the semiconductor manufacturing industry. It is very important for the manufacturer to come up with production plans that can most efficiently utilise the manufacturing equipment and fulfil customer orders placed in a planning horizon. Because of the complexity of the manufacturing processes, it is necessary to use high-fidelity discrete-event simulations to provide accurate estimates of delivery lead time for any given production plan. However, high-fidelity simulations are time-consuming, and thus decision-makers may only evaluate a small number of production plans once customer orders are received. In this paper, we propose the use of a multi-fidelity simulation optimisation approach to efficiently evaluate and select the best production plan from a large set of alternative plans under consideration. We develop an open queue approximation model for a wafer fabrication system and then use the low-fidelity estimates of lead times obtained from the approximation model in a recently developed multi-fidelity simulation optimisation method. Simulation experiment results show that the multi-fidelity approach significantly improves the computational efficiency of simulation-based production planning.