運輸設施的狀態對於一個國家來說相當的重要，狀態的好壞可以直接影響到經濟的發展以及國民的生活，要讓運輸設施都保持在最佳的狀態就必須用生命週期管理的角度來做維護，並使用最佳化的方式來做規劃，但使用最佳化方式產生出來的維修策略往往都與實務單位的工作流程不相容，一般實務的工作流程為經由儀器或者人工的方式測量而得設施的狀態值，決策者再參考測量數據做維修方式的決策，因此很難理解以最佳化方式所規劃出來的預擬維修計畫，所以最佳化的方式很難讓實務單位所採納，本研究所提出的維修門檻值策略跟實務單位的工作流程較相似，同時提出門檻值調整的方法，探討維修門檻值策略藉由門檻值的調整所得到效益，並且將維修門檻值策略與預擬計畫策略(也就是一般傳統最佳化的決策方式)來做分析與比較。在過去的研究中，對於運輸設施最佳化所提出的相關數學模式都不是相當完備，然而使用不完備的模式求解出來的結果往往會與實際情況有落差，因此本研究使用動態混合模式的架構來建立模型，動態混合模型是由事件產生器、有限狀態機、模式選擇器及切換系統所組合而成的，使用此種架構所建立的模型可以同時考量設施不同的劣化方式以及不同的維修效果，使求解出來的結果可以與真實情況更接近。預擬計畫策略與維修門檻值策略在大型問題上的求解都有一定的困難度，因此本研究使用拉式鬆弛演算法來求解預擬計畫的問題。最後以桃園縣大溪鎮的路網作為實例驗證的目標，使用不同的預算來探討預擬計畫策略與維修門檻值策略的效果差異，藉由測試的結果整理出一些建議，以利實務單位在做決策的時候可以當作參考的依據。The condition of transportation infrastructure is very important for a country. The infrastructure has a direct impact on the economic development and the quality of life for people. In order to improve the functionality of transportation infrastructure, the concept of life-cycle management has been introduced to the infrastructure maintenance decision-making process. In practice, transportation agencies measure the conditions of transportation facilities by equipment or manual inspection, and then make maintenance decisions based on the measurements and maintenance thresholds. However, most of the optimization methods generate predetermined maintenance plans (a detailed future plan for the types and timings of maintenance actions), which are incompatible with the workflow of the transportation management agencies and difficult to interpret. As a result, the pre-determined plans are not well accepted by transportation agencies in practice. The past mathematical models for optimizing the transportation infrastructure are not realistic in many ways. Therefore, the solutions obtained by these models are not optimal in the actual situations. To address the problem, a method of finding the optimal maintenance thresholds is proposed in this research. This study uses hybrid dynamic mode (HDM) to describe the transportation infrastructure. The dynamic hybrid mode includes event generator, finite state machine, mode selector and switched systems. This model considers different deterioration and maintenance effects simultaneously. As a result, the solutions are expected to be superior in the actual situations. However, HDM is difficult to solve for large problems. A Lagrangian relaxation algorithm is developed to find the solutions for the pre-determined plan strategies efficiently.The road network of Dasi Township, Taoyuan County, is used as the numerical example to test the proposed methodology. Various levels of budget are tested to analyze the differences between predetermined plan strategy and maintenance threshold strategy. Finally, the conclusions and future directions for research are provided.
