Abstract: In order to improve the solution accuracy and convergence effect of traditional simulation optimization methods, a two-stage simulation optimization method based on segmented linearly allocated clearing functions (ACF) is proposed to solve the wafer release planning, which is used to decide the optimal release rates for different types of products in each planning cycle. The method first constructs a segmented linear fitting model to fit ACF. A mathematical release model based on ACF is established and solved with the objective of minimizing production costs to quickly obtain an approximate global optimal solution as a high quality initial solution. Then, a discrete-event simulation model is developed combining with simulated annealing algorithm to form a simulation optimization framework. A satisfactory solution that meets the production reality is obtained by performing a neighbourhood search based on the initial solution. Experimental results show that the two-stage simulation optimization method based on segmented linearly ACF saves more than 27% of production cost compared to the comparison methods and the current enterprise methods.

Key words: allocated clearing functions, two-stage simulation optimization, wafer fabs, production release planning