Simulation and analysis of scheduling rules for semiconductor manufacturing line

Semiconductor manufacturing has the characteristics of great size, reentrance, uncertainty and multi-objective optimization. A pheromone based dynamic scheduling algorithm, called PB, can optimize multiple performances, such as on-time delivery, cycle time, output, wafer in process and movement. Three models were set up according to Intel Minifab, Wein 24 machines and a real semiconductor fabrication line. With two release strategies, which are deterministic input and CONWIP input, first we discuss how the properties of the whole system change with different parameters alpha , beta , then we can conclude a rule to match alpha , beta in order to achieve a good performance. Finally, PB is compared with other four heuristic algorithms, which are FIFO, SRPT, EDD and CR. The simulation results show that the PB algorithm can effectively improve the cycle time and on-time delivery, and it can also be at least second optimal in all the other properties and optimize multi-objective performances.