連續批量排程機制之構建--以薄膜液晶顯示器組立製程為例

Abstract

薄膜液晶顯示器包含陣列、組立與模組三大主要製程，其中組立製程之中段係由數個批量工作站所組成。多數的學者皆僅考量單階多機或二階單機之批量排程問題，尚無針對多階多機之批量排程問題作一探討。另外，組立製程考量到品質的因素，增加了產品等待加工時間不可過長之限制，更加深了排程之複雜度。因此，本文在符合等候時間限制且最小化其瓶頸工作站換線次數的條件下，針對組立廠之連續批量工作站，發展一混合整數規劃與一啟發式法則，以解決其排程問題。 本文首先透過「需求規劃模組」估計各產品到臨連續批量工作站之時間點，並設定規劃時格以降低規劃複雜度；接著，透過「產能粗估模組」推估各工作站在各規劃週期之最大可用產能，並且考慮機台之整備時間，以各工作站最大可換線次數來定義連續批量工作站之瓶頸所在。在得知瓶頸工作站以及相關生產資訊後，「數學規劃解法」考量各批量工作站之最大加工批量數與產品等候時間限制，期望在達到產出目標之前提下，排訂連續工作站之詳細排程，並且儘量減少不必要之機台設置時間，以避免突發狀況之發生。而為了解決真實世界之問題，本文發展一「啟發式法則」快速求解連續批量工作站之排程問題，其主要利用限制理論之精神，先排定瓶頸批量工作站之排程，並且以最大加工批量排訂以達到最小化設置時間之目標。 實驗結果顯示，設定規劃時格可大幅降低排程複雜度，以利數學規劃解法與啟發式法則之計算；而產能粗估模組可明確定義出批量工作站之瓶頸所在。吾人所發展之數學規劃解法，可求得最佳解，使瓶頸工作站設置次數最小化，且符合等候時間之限制，而啟發式法則則可於數十秒鐘即可找到一合理解，增加了實際運作之可行性。但針對節省總換線次數而言，數學規劃解法之績效大多優於啟發式法則。

The three main manufacturing process of Thin Film Transistor – Liquid Crystal Display (TFT-LCD) are TFT Array process, Cell Assembly process and Module Assembly process. The Cell Assembly process includes several batch workstations; and the panel cannot wait too long without being processed after leaving the previous batch workstation. Previous researches and papers only consider the scheduling problems for single-stage with multiple-machines or for two-stages each with single-machine. Seldom scholars take into account the multiple-stage multiple-machine scheduling problem that exists in the cell assembly process. In view of this complex problem, we build the scheduling mechanism for contiguous batching operations by developing a mixed integer programming model and a heuristic rule which considers the waiting time constraint between workstations and the setup time minimization of bottleneck workstation. The proposed scheduling mechanism contains three modules: demand planning module, mixed integer programming (MIP) module, and heuristic rule module. First of all, the demand planning module approximates the arriving time of jobs to the first batch workstation. We set a new planning time unit to reduce the problem solving time. Secondly, we calculate the maximum available capacity during each planning period in the capacity evaluation module. In this module, the maximum available setup times is calculated for each workstation in order to recognize the bottleneck batch workstation. Then the mixed integer programming (MIP) module and the heuristic rule are built to set the detail schedule for contiguous batching operations considering the maximum lot sizing of each batch workstation and the waiting time constraint between two consecutive operations. The objective of this MIP model is to satisfy the throughput target and to minimize the total setup times of bottleneck workstation. For the sake of solving the real world cases rapidly, the heuristic rule is developed based on the concept of Theory of Constraint. The heuristic rule adopts full batch size policy to set the bottleneck workstation’s schedule so as to minimize the total setup time. Experimental result shows that the complexity of the scheduling problem is reduced by the new time unit, thus the time for solution deriving is shorten. Also, the MIP model can derive an optimal solution satisfying the waiting time constraint between workstations and the total setup time minimization. Finally, the heuristic rule can search out a feasible solution only in several seconds, and it enlarges the application range in this way. However, it may derive a schedule with more total setup times than the solution of MIP model.