电子学报
電子學報
전자학보
Acta Electronica Sinica
2015年
10期
2082-2087
,共6页
曹政才%彭亚珍%李博%刘民
曹政纔%彭亞珍%李博%劉民
조정재%팽아진%리박%류민
半导体生产线%动态调度%紧急订单预测%鼓-缓冲-绳子
半導體生產線%動態調度%緊急訂單預測%鼓-緩遲-繩子
반도체생산선%동태조도%긴급정단예측%고-완충-승자
semiconductor wafer fabrication%dynamic scheduling%rush order prediction%drum-buffer-rope
实际半导体制造过程调度问题具有大规模、不确定、带复杂约束及多目标等综合复杂性,要确保上述生产过程优化运行,必须及时有效进行动态调度。鉴于半导体生产线具有多重入特征及紧急订单对常规订单产生影响,本文给出一种基于“鼓-缓冲-绳子”(Drum-Buffer-Rope,DBR)和自适应模糊推理系统(Adaptive Neuro-Fuzzy Inference Sys-tem,ANFIS)相融合的半导体生产线动态调度方法。首先以最大化瓶颈设备有效产能及保证生产线负荷均衡为目标,将投料控制与工件调度有机结合进行 DBR 优化算法设计;其次根据生产线运行过程中积累大量的历史数据与实时数据,利用 ANFIS 构建紧急订单相关信息预测模型;再次结合专家经验知识,利用模糊推理系统将预测结果与相应的DBR 算法相融合,使生产线提前调整投料策略,保证紧急订单到来时生产线能够有效完成其加工任务及减小紧急订单与常规订单之间的相互影响;最后通过某半导体生产线进行仿真验证,该方法能够实现生产线的多目标优化,为解决实际半导体生产调度问题提供参考。
實際半導體製造過程調度問題具有大規模、不確定、帶複雜約束及多目標等綜閤複雜性,要確保上述生產過程優化運行,必鬚及時有效進行動態調度。鑒于半導體生產線具有多重入特徵及緊急訂單對常規訂單產生影響,本文給齣一種基于“鼓-緩遲-繩子”(Drum-Buffer-Rope,DBR)和自適應模糊推理繫統(Adaptive Neuro-Fuzzy Inference Sys-tem,ANFIS)相融閤的半導體生產線動態調度方法。首先以最大化瓶頸設備有效產能及保證生產線負荷均衡為目標,將投料控製與工件調度有機結閤進行 DBR 優化算法設計;其次根據生產線運行過程中積纍大量的歷史數據與實時數據,利用 ANFIS 構建緊急訂單相關信息預測模型;再次結閤專傢經驗知識,利用模糊推理繫統將預測結果與相應的DBR 算法相融閤,使生產線提前調整投料策略,保證緊急訂單到來時生產線能夠有效完成其加工任務及減小緊急訂單與常規訂單之間的相互影響;最後通過某半導體生產線進行倣真驗證,該方法能夠實現生產線的多目標優化,為解決實際半導體生產調度問題提供參攷。
실제반도체제조과정조도문제구유대규모、불학정、대복잡약속급다목표등종합복잡성,요학보상술생산과정우화운행,필수급시유효진행동태조도。감우반도체생산선구유다중입특정급긴급정단대상규정단산생영향,본문급출일충기우“고-완충-승자”(Drum-Buffer-Rope,DBR)화자괄응모호추리계통(Adaptive Neuro-Fuzzy Inference Sys-tem,ANFIS)상융합적반도체생산선동태조도방법。수선이최대화병경설비유효산능급보증생산선부하균형위목표,장투료공제여공건조도유궤결합진행 DBR 우화산법설계;기차근거생산선운행과정중적루대량적역사수거여실시수거,이용 ANFIS 구건긴급정단상관신식예측모형;재차결합전가경험지식,이용모호추리계통장예측결과여상응적DBR 산법상융합,사생산선제전조정투료책략,보증긴급정단도래시생산선능구유효완성기가공임무급감소긴급정단여상규정단지간적상호영향;최후통과모반도체생산선진행방진험증,해방법능구실현생산선적다목표우화,위해결실제반도체생산조도문제제공삼고。
The scheduling problem of actual semiconductor production process is of large scale,uncertainty,complicated con-straints and multiple objectives,and the optimization of the process depends upon a timely and efficient dynamic scheduling.Consid-ering the multi-reentrant characteristic of semiconductor wafer fabrication and the influence of rush orders upon regular orders,a Drum-Buffer-Rope (DBR)theory and Adaptive Neuro-Fuzzy Inference System (ANFIS)integrated dynamic scheduling algorithm is proposed.Firstly,a DBR optimization algorithm based on the combination of releasing control and lots scheduling is designed, aiming at maximizing the effective capacity of bottleneck equipment and ensuring the wafer fabrication load in a balance.Secondly, using the historical and real-time data obtained in the wafer fabrication,a prediction model of rush orders is constructed on the basis of ANFIS.Thirdly,the prediction result is integrated with the DBR algorithm combined with the experts’knowledge and practice experiment,to adjust the wafer fabrication before rush orders come,which guarantees that the rush order could be completed effec-tively,and reduces the interplay between the rush order and the regular one.The proposed scheduling algorithm is simulated on a certain semiconductor wafer fabrication system,and the result shows that the method can realize the multi-objective optimization and provide reference to the practical semiconductor wafer fabrication scheduling.