應用自主性代理人機制於多階非相關平行機臺之研究

Applying Autonomous Agent in the Multi-stage Unrelated Parallel Machines Scheduling Problem

10.6346/NPUST.2014.00126

許國屏

代理人技術 ； 多階非相關平行機臺 ； RMM模式 ； 派工法則 ； Agent-oriented technique ； Multi-stage Unrelated Parallel Machines ； Recursive Modeling Method ； Dispatching Rules

屏東科技大學工業管理系所學位論文

2014年

碩士

黃祥熙

繁體中文

隨著電子與半導體產業的工廠規模擴大與產能不斷擴張，零件都需經過多道加工程序，形成多階段平行機臺之加工問題。本研究應用多代理人協商於多階非相關平行機臺之模式建構，機臺將具有加工兩種以上不同零件能力，零件可於不同機臺進行加工，而不同機臺對相同零件則有不同加工效率，本研究運用報酬矩陣配合循環模式法之決策機制作為機臺間溝通的模式並尋找最佳加工排程。系統模式將透過Em-Plant模擬系統建構，針對代理人協商機制與傳統SPT及FCFS派工法則進行比較。結果顯示，透過代理人模式進行溝通能選擇價值較高的產品優先加工，使產品的總產出價值達到最高，並降低在製品數量與平均加工流程時間，使緩衝區變動量降低，達到工件派遣擁有更優良的選擇依據。

With the expansion of scale of capacities and plant sizes for electronic and semiconductor industry, every part needs to go through multi-stage processing procedures to be a finished product. This research is the so called multi-agent-based architecture for the multi-stage unrelated parallel machines scheduling problem. In this production system, machines are able to operate two or more different parts and each part can be processed on different machines, whereas different machines have different working abilities to process the same job. In this research, the pay-off matrix and Recursive Modeling Method for decision-making mechanism are applied, in which machines can communicate with each other and obtain the feasible solution. An Em-Plant simulation model was developed with the factors, including the product interval time, value of products, buffer capacity and process time. These indicators are used as the objectives to find out the most appropriate production sequence. In this research, the agent negotiation mechanism will be compared with SPT, FCFS dispatching rules. The result of this thesis shows that the performance of agent negotiation mechanism is working as balancing different performance indices. Furthermore, the proposed method can maintain a stable system in the different situations.

1. 張彥文(2017)。以即時資料為基礎的作業現場製程規劃與彈性管控系統。中原大學資訊管理學系學位論文。2017。1-79。