Researches On Scheduling Problems Of Semiconductor Wafer Fabrication Based On Batching Decision

Semiconductor manufacturing is one of the most complicated manufacturing systems nowadays. Semiconductor manufacturing scheduling is an important problem of semiconductor manufacturing system and it is also one of difficult topics of the scheduling theory study. Semiconductor manufacturing has the characters of complex technological process, re-entrant processing flow, simultaneous processing of various products, and mixed processing model etc.. The technological process is probably complicated due to coexistence of multi-order and multi-product. In semiconductor production process, there are two kinds of the machines:discrete processing machine and batching processing machine, these machines have various processing methods. Lot is the minimum unit of semiconductor manufacturing; in general, there are dozens of wafers in a lot. Discrete processing machine processes lots one by one, while batch processing machine processes a number of lots simultaneously, but it must do batching decision before processing, batching means that some lots have to be grouped into a batch.Firstly, the thesis introduces semiconductor manufacturing system and the characteristics of semiconductor wafer fabrication. After the importance and literature of semiconductor wafer fabrication scheduling problems based on batching decision are discussed, the difficult points and deficiency of the scheduling problems are analyzed. The scheduling problems based on batching decision are also classified.Secondly, the thesis considers multiple orders per job scheduling problem to minimize maximum tardiness. Two sub-problems of grouping multiple orders into lots and scheduling these formed lots are studied. An efficient genetic algorithm is applied to the problem, in which a representation based on order and product is proposed and binary tournament selection, order crossover, and swap mutation are used to optimize the problem. The experimental results show that the algorithm has good performance by using three benchmark problems and the solutions of the algorithm with different parameters.Finally, the thesis addresses job shop scheduling problem with batching machines for minimizing makespan, and proposes a scheduling algorithm based on swarm-based neighborhood search. A batch based representation is used after batches are formed by using lots of the same product sequentially and a decoding procedure with batching decision is proposed. A swap operator is then applied to some individuals and an insertion is done on other individuals. The worst solution of the swarm is replaced with the elite solution every iterations. The computational experiments demonstrate the feasibility and efficiency of the algorithm. Machine flexibility is introduced to diminish the long waiting time of batch on batching machines. The flexibility of discrete machine is considered and flexible job shop scheduling with batching machines is investigated.