THE DESIGN OF INTELLIGENT JOB SHOP SCHEDULING SYSTEMS

• Main Authors: HUANG, HUA-SHAN, 黃華山
• Other Authors: ZHAO, RONG-YAO
• Format: Others
• Language: zh-TW
• Published: 1990
• Online Access: http://ndltd.ncl.edu.tw/handle/30239503091668791102

博士 === 淡江大學 === 管理科學研究所 === 79 === This research attempts to employ the knowledge-bassde approach to solve the job shop scheduling problems.First,three job types（series,fan,and assimbly）of scheduling problems are investigated and,the reasons why many previous researches achieved different results are expolred. Three intelligent scheduling systems for solving these job shop scheduling problems are then developed. The nucleus of each of these systems consists of three parts:the frame base,the rule base and the inference engine.The frame base is constructed by using the segment representation approach.This approach not only simplifies the representatoin but also deals with the fan and assembly jobs easily.Moreover,a hierarchical control strategy is proposed to manage the transmission and maintenance of data in the frame base more efficiently. To overcome the problems that different researches have achieved different results,and to re,edy the fact that few researches treated the mord complicated assembly shop scheduling problem,the experiments to simulate the performance of the dispatching rules under various conditions are made.The results show each of the dispatching rules proposed by the previous researchers may behave best under what kinds of conditiopns,and then,on the basis of the results,rule bases are constructed.That is,the efforts of previous researchers are integrated as the knowledge sources of the rule bases.This enables three proposed systems always to syggest the dispatching rule which is most suitable for a job processing operation in any situation during the scheduling process. The inferencd engine dirdcts the entire progress of the whole system.Irs first three processes are applied to deal with normal situations in the scheduling while the fourth process deals with abnormal situations.Thus,each of these three proposed systems can not only handle the dynamlc situations in the job shop environment so as to make good scheduling decisions according to the changing situations,but also control the abnormal events in the scheduling. Finally,three types of experimental job shop environments are rested by using the LISP-based software tool--EJAUNDICE.The results reveal that all of the proposed systems can respectively construct better schedules than those performed by the conventional dispatching methods,such as SPT,COVERT,DDATE,SLK/OPN,LWKR and NUJOB-SPT. 本研究嘗試以知識為基礎的方法（Knowledge-based approach）來解決訂貨生產排程的問題。首先，本研究對三種不同型態的訂貨（線型 ，扇型和裝配型）生產排程的問題加以探討，並找出不同學者獲得不同研究結果的原 因，然後提出三種智慧型排程系統。 此三種系統中的主要部份為：框架庫（Frame base ），規則庫（Rule base ）和推論機（Inference engine ）。本研究使用分支（Segment） 表示法來建立框架庫，以簡化表示的方式，同時一 個階層式的管理策略，來管理和加速資料在框架庫中的傳輸。 為了解開何以不同學者會得到不同的研究結果，以及彌補很少有關裝配型訂貨生產排 程方面研究的缺憾，本研究乃進行多項實驗來模擬以前學者所提之各種派工法在各種 不同情況下的表現（Performance）。 實驗結果顯示了各種派工法在何種情況下可 最佳，接著根據此項結果來建立規則庫；亦即，本研究結合了以前學者的研究結果以 做為規則庫中的知識來源。這使得本研究所提之三系統在任何情況下，皆可選用最適 合該情況的派工法做為排程的依據。 本研究所提之推論機含有四個處理步驟，前三個步驟用來處理工廠正常運作的情況， 而最後一個步驟則用來處理特殊的狀況，以使得所提的三個系統可同時處理正常和特 殊的狀況，而符會工廠的實際需要。 最後，我們用以LISP寫成的軟體EJAYNDICE 來建立此三系統。根據實驗結果顯示，本 研究所提之系統所產生的排程皆優於任何一種傳統的派工法（例如SPT，COVERT，DDA TE，SLK/OPN，LWKR 和NUJOB-SPT）所產生的排程。