Multiproduct Lot Merging Methods For Semiconductor Wafer Fabrication

Semiconductor wafer fabrication (SWF) is one of the most complex systems in industry manufacturing. With the vigorous development of domestic information industry and electronics technology, the demand for semiconductor manufacturing upgrade is accelerating. Under the background of consumers’ diversified demand, and orders for small quantities of varied products, multiproduct lot clustering and merging is becoming one of the hotspots and focus of research in semiconductor scheduling optimization. It mainly solve the problem of how to sort, cluster, dispatch the lots waiting in the buffer, and how to improve the efficiency of SWF. It also poses a novel way to production optimization of complex manufacturing systems research. The main content of this paper includes:(1) Briefly introduces semiconductor manufacturing lot merging theory, and constraints in actual production line. Then we give data pre-processing method for different types of data from semiconductor production line, and the data clustering algorithm used for lot merging. These provide the theoretical basis for future research.(2) Considering large order quantities, job varieties, due date differences and other characteristics in semiconductor wafer fabrication system, we propose a semiconductor multiproduct lot merging method. A compound lot dispatching rules are applied for lots waiting in buffers, and an iterative improvement procedure is used for optimizing previous merging strategy. To solve the problem caused by uncertain events, a dynamic lot weighting factor revising scheme is proposed based on closed loop control strategy.(3) With the objective of minimizing total cycle time and tardiness, we propose a lot clustering algorithm based on extracting lot characteristic properties from semiconductor production line. The output lot set of fuzzy C means clustering algorithm is to further improve the original lot merging method, and improve the performance of the production line.