具擴充性之多機台動態派工與預防保養方法

Abstract

本研究利用線性規劃進行產能配置，降低動態派工與預防保養策略時的複雜度(Decomposition)，應用於大型的多產品多機台的生產系統中。其中機台的健康狀態與產品的派工以及機台預防保養策略相關，在此生產環境下，若無考量妥善的派工與保養決策將造成產品週期增加，進而系統生產成本上升，因此凸顯適當的派工與保養策略之重要性。 然而利用動態規劃(Dynamic programming)建立多產品多機台的動態派工與預防保養策略模型(Dynamic Dispatching with Preventive Maintenance Model, DDPM)時，考慮所有健康狀態下的機台數量導致當系統機台數目增加，維度越大以致於無法在合理時間內得到適當的動態派工與保養之決策，故本研究提出LPD (Linear Programming Decomposition)方法將考量的健康狀態化簡為單一機台，目標為最小化等候時間，模擬在同質機台系統下以兩產品三機台與兩產品四機台生產系統為例，以及異質機台系統下以兩產品兩異質機台系統進行模擬分析，並與DDPM以及其他派工方法比較。結果顯示LPD方法的表現接近於DDPM，且優於其他方法能夠有效提升生產效率且降低產品等候成本，顯示LPD能夠在合理時間內獲得妥善的動態派工與預防保養策略。

This research decomposes the dynamic dispatching and preventive maintenance strategy for large-scale production system by capacity allocation with linear programming model. The health condition of machines is dispatching and maintenance dependent. Therefore, an improper dispatching and maintenance rules will lead to an increase in the cycle time of product, and thus stress the importance of appropriate dynamic dispatching and preventive maintenance policy. However, when the dynamic programming is used to develop the Dynamic Dispatching with Preventive Maintenance (DDPM) model, the consideration of the number of machines in all health conditions will result in the fact that an appropriate policy cannot be obtained within a reasonable time as the scale of production system becomes larger. Hence, we propose Linear Programming Decomposition (LPD) as a method to divide multiple machines system into several single machine problems so as to consider only one machine’s health condition. The simulation results show that the performance of LPD is close to DDPM and superior to other dispatching rules under both homogeneous machine system and heterogeneous machine system, which verifies that an approximately optimal dispatching with preventive maintenance can be obtained by LPD within a reasonable time.