捷運系統成為大都會區交通運輸軌道建設之一，捷運班次多、準點及安全等各項服務，便於民眾短程往返。其中，月台門設備扮演捷運安全的重要設備之一，基於安全上的考量，維修人員必須於夜間進行月台門設備維護保養作業，決策者依照勞基法及公司設備標準作業程序，安排維修人員出勤任務。有鑑於決策者依照設備位置及維修人力直接進行設備維護排程作業，並非有效利用科學方式進行分析。因此，為有利於決策者憑藉科學數據，如何有效指派維修人力執行任務，成為捷運營運事業單位重要的問題。 本研究以捷運營運事業單位月台門設備維護保養作業為例，利用數學模式建構最佳化維修人力指派作業，應用電腦套裝軟體工具LINGO 17.0版本求解，再搭配Microsoft Excel彙整數據圖表化，經與人工排班對照，可提高人力資源利用分配，且有效進行月台門設備維護保養排程，藉由本研究模式參數作深入專案分析，使決策者參照利用。 ;The MRT system has become one of the transportation rail constructions in metropolitan area. The MRT has a wide range of services, such as frequency, punctuality and safety on the MRT schedule, so as to make it convenient for people’s short-range journey, in which the platform gate equipment plays an important role in the MRT security. For safety reasons, maintenance personnel must carry out maintenance operations on the platform gate equipment at night. The decision maker shall arrange the maintenance personnel to be on duty according to the Labor Standards Act and the company′s equipment SOP. In view of the fact that the decision makers directly carry out the equipment maintenance scheduling in accordance with the equipment location and maintenance manpower, it is not an effective use of scientific method for analysis. Therefore, in order to facilitate decision makers to rely on scientific data, how to effectively assign maintenance personnel to perform tasks has become an important issue for MRT business entity. This study takes the maintenance of the platform gate equipment for the MRT business entity as an example, the mathematical model is used to construct the optimal maintenance manpower assignment, and the LINGO 17.0 software package to solve the problem. And then, diagramming from data compiled by Microsoft Excel is matched in comparison with manual shift scheduling to improve the utilization and allocation of human resources, and effectively carry out the maintenance schedule of equipment on the platform gate. The decision-makers can make reference and use through in-depth case analysis of the parameters in this research model.