自動物料搬運系統分散式優化

Decentralized Control for Automated Material Handling Systems

呂思達(Starr Lyu)

指導教授 ： 吳政鴻

國立臺灣大學/工學院/工業工程學研究所/碩士(2018年)

https://doi.org/10.6342/NTU201802769

摘要

現代工業普遍採用最近工作優先的集中式控制方式。這種集中式的控制都是通過中央處理器根據相關派遣規則對OHT進行派遣的控制，這就無法及時保證在當下計算所得派遣規則是最優的。而且，這種集中式派遣規則的運算複雜度是隨著運輸OHT的數量增長而增長，在一定的計算機運算能力限制下，中央處理器無法進行有效的任務派遣。OHT在實際運行過程中，使用最近工作請求優先的集中式派遣規則會導致OHT堵塞以及加劇瓶頸站點排隊情況。對自動物料搬運系統進行集中式控制，會使派遣面臨一個啟發式的非最佳解以及多餘的時間延遲，這些問題都可能造成OHT在軌道上的堵塞。因此，為解決上述問題，本研究換了一種思路採取對一種分散式的控制方式對OHT進行派遣。一方面本研究採取機台與OHT的通訊，根據OHT的位置等其他信息對OHT進行搬運任務的最佳派遣；另一方面採用相關物聯網技術使OHT與OHT之間有能力進行兩兩通訊，根據兩輛OHT的相關信息改變OHT的派遣。
本研究基於V2V、V2M的理念對半導體製造工廠中的自動物料搬運系統進行離散式控制，在V2V與V2M的基礎上提出基於車與車通訊-未被派遣任務OHT與已被派任務OHT的交換（V2V-AAE）、基於車與車通訊-未被派遣任務OHT與已被派任務OHT的交換（V2M-PUA）與基於車與車通訊-未被派遣任務OHT與已被派任務OHT的交換（V2V-ITC）控制方法，以國際半導體技術聯盟（International SEMATECH）在300mm晶圓製造廠與物料搬運系統建模報告的中的某個產品的加工流程資料為例進行分析，建立自動物料搬運系統的FlexSim模擬系統，檢驗本文提出之方法的優劣，並可進一步分析半導體製造工廠的自動物料搬運系統的相關特征。
本研究經過範例實驗分析與模擬得到以下觀點：AMHS的分散式控制相比于原先的集中式控制在總產出、產品週期時間、搬運時間等指標上顯示出了明顯的改進。一方面是分散式控制相較於集中式控制節省了與中央控制系統信息傳輸的時間，另一方面通過V2V、V2M的通訊使OHT隨時根據搬運請求與軌道路徑狀況作出相應調整，可以節約等待時間、搬運時間或者堵塞時間。AMHS的分散式控制在2×2網格型佈局、2×1網格型佈局與單一環形佈局中都顯示出了明顯的改進效果。增加OHT的數量可以明顯改進AMHS的表現，但是改進效果的邊際效應隨著OHT數量增加遞減，同時分散式控制從總產出上看改進效果存在隨著OHT數量增加改邊際效應遞減，而從產品週期時間上與等待時間上看存在邊際效應遞增的效果。隨著系統利用率的增加，產品的總產出、產品週期時間、等待時間、搬運時間以及堵塞次數與OHT利用率都在增長，並且存在一個邊際效應遞減的趨勢。分散式控制在相對集中式在總產出改善量隨著系統利用率的增加而增加。當機台加工時間服從均勻分配與常數分配時，AMHS表現差異不大，且分散式控制方法相比較集中控制方法改進程度持平；但在指數分配下，AMHS表現與之前兩種分配有差異。

關鍵字

AMHS ； V2V ； V2M ； 分散式控制 ； FlexSim

並列摘要

Modern industry generally adopts centralized control methods like nearest job first. This centralized control is controlled by the central processor to dispatch OHT according to the relevant dispatch rules, which cannot guarantee in time that the dispatch rules currently calculated are optimal. Moreover, the computational complexity of this centralized dispatching rule increases as the number of transported OHT increases, and the central processor cannot perform effective task dispatching due to certain computer computing power limitations. During the actual operation of OHT, the use of centralized dispatching rules with the most recent work request priority will lead to OHT congestion and increased bottleneck. Therefore, in order to solve the above problems, this study has replaced the idea of adopting a decentralized control method to dispatch OHT. On the one hand, this study adopts the communication between the machine and the OHT, and the best dispatch of the OHT to carry the task according to other information such as the location of the OHT; on the other hand, the related Internet of Things technology enables the OHT and OHT to communicate with each other. Change the dispatch of OHT based on information about the two OHTs.
Based on the concept of V2V and V2M, this study discretely controls the automatic material handling system in semiconductor manufacturing plants. Based on V2V and V2M, V2V-AAE, V2M-PUA and V2V-ITC control methods are proposed to the International Semiconductor Technology Alliance. (International SEMATECH) analyzes the processing flow data of a product in the 300mm wafer fabrication plant and material handling system modeling report, and establishes the FlexSim simulation system of the automatic material handling system to test the advantages and disadvantages of the method proposed in this paper. The relevant features of the automated material handling system of semiconductor manufacturing plants can be further analyzed.
This study has obtained the following observations through sample experimental analysis and simulation: the decentralized control of AMHS shows a significant improvement in the total output, product cycle time, and transport time compared to the original centralized control. On the one hand, the decentralized control saves the time of information transmission with the central control system compared to the centralized control. On the other hand, the communication between V2V and V2M enables the OHT to adjust accordingly according to the handling request and the track path condition, which can save waiting time transport time or blocking time. The decentralized control of AMHS shows significant improvement in 2×2 grid layout, 2×1 grid layout and single ring layout. Increasing the number of OHT can significantly improve the performance of AMHS, but the marginal effect of the improvement effect decreases with the increase of the number of OHT. At the same time, the decentralized control shows that the improvement effect from the total output has a diminishing marginal effect as the number of OHT increases. There is an effect of increasing marginal effect on cycle time and waiting time. As system utilization increases, total product output, product cycle time, waiting time, handling time, and number of clogging and OHT utilization are increasing, and there is a trend of diminishing marginal effects. Decentralized control increases in total output as the total output improvement increases with system utilization. When the processing time of the machine obeys the uniform distribution and constant distribution, the performance of AMHS is not much different, and the degree of improvement of the distributed control method is flat compared with the centralized control method. However, under the exponential distribution, the performance of AMHS is different from the previous two.

並列關鍵字

AMHS ； V2V ； V2M ； Decentralized Control ； FlexSim

參考文獻

[1] Agrawal, G. K. and S. S. Heragu (2006). "A survey of automated material handling systems in 300-mm SemiconductorFabs." IEEE Transactions on Semiconductor Manufacturing 19(1): 112-120.

[2] Applegate, D. L., et al. (2006). The traveling salesman problem: a computational study, Princeton university press.

[3] Babiceanu, R. F. and F. F. Chen (2005). Performance evaluation of agent-based material handling systems using simulation techniques. Proceedings of the 37th conference on Winter simulation, Winter Simulation Conference.

[4] Bako, B. and P. Božek (2016). "Trends in simulation and planning of manufacturing companies." Procedia Engineering 149: 571-575.

[5] BARTHOLDI III, J. J. and L. K. Platzman (1989). "Decentralized control of automated guided vehicles on a simple loop." IIE transactions 21(1): 76-81.