| 摘要: | 台灣的大眾捷運系統（Mass Rapid Transit, MRT）近年來大幅擴展，逐漸成為城市地區交通的主要形式。為了確保大眾捷運系統的穩定和正常運行，必須利用有限的資源安排良好的維護計劃，以有效且高效率地執行設施維護工作。實務上，目前主要是維修人員根據自身過往的經驗進行規劃捷運設施維護排程，然而，使用試誤法的人工調整方法既不系統化且效率不佳，此外，在正常營運中出現設施故障時，必須派出維修人員對設施進行檢查、故障排除和維修，以盡量減少對捷運正常運營的影響，此時通常需要指派維護人員進行檢修，並且可能需要對先前計劃的維護排程進行調整。本論文的第一部分利用時空網路流動技巧，為捷運設施建立一系統性且高效率的周維修排程模式，該模式考慮了預估的故障檢修需求、定期的預防性維護需求，現有的人力限制包括的維護人員授證等要求，並且為了提高維護效率及降低成本，我們制定了一項策略，要求屬於同一維修廠的維護人員進行緊急維護工作，目的是盡量減少總體維護成本。
在目前的實務中，每個維修股的負責人或代理人負責規劃該股的每日維修排程， 這包括基於定期且重複的已知預防性維修項目，並考慮在計劃日需要處理的故障設施檢修項目，以及可用的自有維護人力、來自外包支援廠商的資源以及其他相關因素。在本文的第二部分中，我們應用第一部分中開發的網路架構來開發日維修重排程模式，該模式考慮了在日常維護計劃中須執行的故障檢修之確定性需求，並以最大限度地減少隔日的維護成本及降低排程變動為目標。為了更好地滿足實際需求，日維修重排程模式包含了在同一維修課內的不同維護團隊之間互相支援故障檢修工作的策略，以及允許維護團隊在同一場站同時工作數量之限制。
在本論文中，我們開發了兩種模型：周維修排程模式和日維修重排程模式，並應用網絡流動概念來呈現捷運維修工作隊的移動。 此外，為了能夠解決第一部分之周維修排程龐大而複雜問題，發展一以維護日為分割基礎的分治法求解演算法以協助高效率的求解問題；相對地，日維修重排程模式的問題規模較小，可以直接用數學規劃軟體進行求解，這些模式和開發的演算法經過案例測試驗證後，確認可有效協助大眾捷運決策人員解決維修排程問題。
;Recently in Taiwan, the Mass Rapid Transit (MRT) system has been greatly expanded, gradually becoming the main form of urban transportation in the region. To ensure the stable and normal operation of the MRT system, it is necessary to take advantage of limited resources to conduct a well-scheduled maintenance plan that can effectively and efficiently carry out facility maintenance work. Currently, MRT facility maintenance scheduling is mainly planned by maintenance staff, rely on past experience. However, the manual adjustment method, which uses a trial-and-error approach, is neither systematic nor efficient. Furthermore, when facility failure occurs during regular operations, maintenance personnel must be dispatched urgently to troubleshoot, repair, and inspect the facility to minimize the impact on regular MRT operations. Repair work often requires assigning maintenance staff to carry out repairs and can require significant adjustments to the previously planned maintenance schedule. In the first part of this thesis, the time and space network flow technique is used to create a systematic and efficient weekly maintenance schedule for MRT facilities. This model takes into account the estimated demands for repair works, regular preventive maintenance needs, and existing manpower constraints, including the certified maintenance staff requirements. Furthermore, to improve the maintenance performance and reduce the cost of maintenance, a strategy has been developed that involves requiring maintenance staff belonging to the same plant to carry out emergency maintenance work to minimize the overall maintenance cost.
In current practice, the head of each maintenance subsection or deputy is responsible for planning the daily maintenance schedule for their group. This includes preventive maintenance based on known regular recurring maintenance items, as well as troubleshooting to plan repairs for faulty facilities that require attention on the day the schedule is made. They also have to take into account the available self-owned maintenance manpower, supporting resources from outsourced contractors, and other relevant factors. In the second part of this thesis, we apply the network architecture developed in the first part to develop a daily maintenance rescheduling model. This model considers the needs identified from repair works in the daily maintenance schedule and has the objective of minimizing maintenance costs and reducing the variance of scheduled works for the next day. To better align with practical needs, a strategy for sharing repair work across different maintenance teams within the same section has been developed. In addition, the model takes into account quantity restrictions, which allow maintenance teams to work simultaneously at the same site.
In this thesis, we develop two models: a weekly maintenance scheduling model and a daily maintenance rescheduling model and apply the network flow concept to represent the movement of MRT maintenance team. Additionally, to efficiently solve the large and complex problem of weekly maintenance scheduling in the first part, we have developed a heuristic algorithm that applies a divide-and-conquer approach based on the maintenance day. In contrast, the scope of the daily maintenance rescheduling model is smaller and can thus be solved directly using mathematical programming software. After testing and verification through case studies, these models and the developed algorithm are useful for decision-makers in the MRT system in solving the problem of maintenance scheduling. |
