一般性流程生產線緩衝區配置之研究

Buffer allocation in production lines

陳少開(Shao-Kai Chen)

指導教授 ： 張舜德

共同指導教授 ： 李緒東

國立臺北大學/商學院/企業管理學系/碩士(2008年)

摘要

在生產線上選擇適合的緩衝區大小以提昇產能，是一個具有實際需求的問題。在流程生產線中，眾多工作站的生產率及損壞率皆各自獨立，這個因素讓緩衝區配置的工作，變得更加困難。生產線若面臨頻繁的產品調整或更改，則緩衝區的配置也必須跟著變動。過去有許多緩衝區的配置方法，多半不適用於緩衝區動態配置。如使用正解法來做緩衝區配置，將涉及複雜的數學運算及假設，且配置結果與生產線型態有關，因此不適合動態配置；若使用傳統啓發式方法來做配置，這些方法往往基於某些特定假設，而不適於所有生產環境，並且容易陷入區域最佳解，而無法找到全域最佳解，所以也不適合。
　　人工智慧的運用，能夠廣泛的解決許多問題，本研究使用人工智慧的技術，來解決動態配置的問題。首先使用基因演算法和模擬軟體搭配，試圖快速找出一系列的最佳解，接著再將最佳解套入類神經網路做訓練，所產生的類神經網路模型可即時找出緩衝區最佳配置之情形。
　　結果證實，經完整訓練的類神網路模型，在多樣參數條件下，可成功找出最佳緩衝區配置，本研究使用基因演算法搭配類神經網路，能夠有效解決動態緩衝配置的問題。

關鍵字

緩衝區配置 ； 不平衡生產線 ； 不可靠生產線 ； 基因演算法 ； 類神經網路

並列摘要

Choosing suitable buffer setups for transfer lines to augment throughputs is a pragmatic issue. To find out a preferable buffer arrangement is a complex problem due to the interaction between the production rates and breakdown rates of workstations. Prior studies mainly concentrated on buffer allocation problem with static layouts. Two major approaches are used: 1) mathematical method, whose results are in associated to production line types and involves complicated mathematical calculation; and 2) heuristic method, which may not suitable to all the practical of production lines and it may easily trapped in local optimization of solutions. When transfer lines with frequent product adjustments and setup changes, the buffer allocations will alter accordingly, and it becomes more difficult to balance between effectiveness (by using mathematical method) and efficiency (by using heuristic method).
The artificial intelligence (AI) is widely applied to accommodate many problems with appropriate solutions. In this study, an AI based method is proposed to solve the buffer allocation of unreliable / unbalance transfer lines. Firstly, Genetic Algorithm (GA) combined with simulation method is used in attempting to rapidly figure out the best solutions, and then the best solutions are feed into Artificial Neural Network (ANN) for predicting preferable buffer layouts.
Through well trained ANN, it may successfully find out the best buffer allocation. This study uses GA and ANN to solve the problems of dynamic buffer allocation and it shows a possible approach to balance the effectiveness and efficiency for this practical problem.

並列關鍵字

buffer allocation ； unbalance production line ； unreliable production line ； genetic algorithm ； artificial neural network

參考文獻

1. 吳泰熙，吳奕樺，張欽智，2006年「以基因演算法求解單原片方形物件排列問題」，科學與工程技術期刊（第二卷 第三期）。

2. 林豐澤，2005年，「演化式計算下篇：基因演算法以及三種應用實例」，智慧科技與應用統計學報（第三卷 第一期）。

5. Altiok, T. and Stidham S. (1983), “The allocation of interstage buffer capacities in production lines,” IIE Transaction, 15, 292-299.

6. Altiparmak, F., Dengiz, B. and Bulgak, A.A. (2007), “Buffer allocation and performance modeling in asynchronous assembly system operations: An artificial neural network metamodeling approach,” Applied Soft Computing, 7, 946-956.

8.　Bulgak, A.A., Diwan, P.D. and Inozu, B. (1995), “Buffer size optimization in asynchronous assembly systems using genetic algorithms,” Computers ind. Engng., 28, 309-322.