半導體成品測試製程中,需要經過一道預燒(burn-in)作業以提昇產品可靠度。然而預燒作業所需要的處理時間遠大於其它製程的加工時間,使得預燒作業成為整個成品測試流程中最關鍵之製程。 因此本研究針對具有成批處理特性的預燒作業,考量工作非同時抵達、不同工作大小(non-identical job size)、批次處理時間為同批工作中加工時間最長者、批次容量有限與每件工作皆有到期日限制之生產環境,運用模擬退火法探討以「總完工時間」最小化與「總延遲時間」最小化為績效指標的單一批次機台排程問題。並將同樣求解單一批次機台排程問題但未考慮工作大小之GRLPT、R1與R2等三種啟發式演算法加入工作大小限制,修改成GRLPT_S、R1_S與R2_S等三種啟發式演算法。以一系列的模擬實驗,分析探討各種排程方法在不同生產環境求解品質之優劣。由實驗結果顯示模擬退火法平均求解結果,於總完工時間指標會與GRLPT_S、R1_S、R2_S等三種啟發式演算法平均求解結果差異不大,但於總延遲時間指標卻能有效改善5.2%。
To improve the reliability of semiconductor products, burn-in operations is essential in the final testing processes. While the time needed at burn-in operation is much more than the other ones, it makes burn-in operations the bottleneck of the procedure in final test. This study is aimed at burn-in operations characterized batch processing. Dynamic arrival of jobs with due date, non-identical job size, batch processing time as the longest processing among all jobs in a batch, and limitation on the sum of job sizes in a batch are considered. Procedure of simulated annealing to minimize makespan and total tardiness of jobs on a single batch-processing machine is proposed in the study. GRLPT, R1, and R2 heuristic algorithms in literature, in which jobs size is assumed identical, are modified as GRLPT_S, R1_S and R2_S in this study to accommodate job sizes in a batch. Various experiments were conducted to compare the proposed method with those modified algorithms. The computational results show that the proposed algorithm performs quite well as GRLPT_S, R1_S and R2_S on the criterion of makespan, while the proposed algorithm improves the criterion of total tardiness by 5.2 percent average.