| 摘要: | 現有的運輸設施生命週期管理之模式有三項限制：第一，實務單位習慣於利用門檻值(threshold) 設立維修條件來進行設施管理，但現有最佳化模式產生之維修決策過於詳細且缺乏彈性，與施工單位之工作流程不相容，實務界接受度甚低。第二，目前之生命週期管理多假設設施之劣化模式為唯一，但許多研究指出經過維修之設施其劣化行為與維修前有所不同。第三，當考量路網層次之多設施問題時，設施之間的相關性常被忽略，導致施行後之結果與預期不符。基於以上理由，本計畫擬利用動態混合系統(dynamic hybrid system)，同時考量運輸設施之離散行為(門檻值條件、不同維修策略下之劣化模式切換…等) 與連續之狀態變數，並將最佳化納入維修門檻值之訂定，最後，再以雙層規劃(bi-level programming) 分別描述維修決策與交通量指派，以完整考量維修決策對車流分佈之衝擊。以上改變預期可提昇網路層級的維修決策正確性，以增進生命週期管理之效益。經過單一設施之初步測試，動態混合系統可有效應用於運輸設施之生命週期管理，亦可透過離散狀態變數考量不同維修策略造成設施劣化機制的切換，再者，利用門檻值之維修策略之效果與最適控制接近，且與實務單位之工作習慣相近，可簡化維修策略之施行以及增加接受度，有極高之研究潛力以及實務價值。後續將延伸至多設施問題，提出求解方法，並以實際路網進行測試，以驗證本研究計劃之實務可行性。There are three major limitations in the methodologies of transportation life-cycle management. First, the solutions generated by optimization frameworks such as optimal control contain detailed maintenance policy for every facility at every time period. However, they are incompatible with the work flow of transportation agencies, which traditionally adopt threshold-based rules for maitnenance decision-making. That is, a maintenance action is applied when the facility state is below a certain threshold. Second, the deterioration mechanism is often assumed to be homogeneous regardless of the maintenance action being carried out. However, many studies have shown that the deterioration mechanism before and after maintenance could be different. Third, the interdependency between facilities is sometimes ignored or not captured properly and, as a result, the maintenance policies created for network-level maintenance problems are erroneous. Based on the above reasons, this study plans to use dynamic hybrid systems to include both discrete behaviors (e.g. maitenance thresholds, deterioration mode switches due to action changes) and continuous variables of facility states. To be consistent with the methods adopted by the transportation agencies, the study will consider threshold-based rules in the maintenance decision-making process. Moreover, the determination of the values for the thresholds will be conducted with an optimization model, which will improve the current practice of maintenance procedures adopted by transportation agencies. The study will also formulate the problem with bi-level programming to include maintenance decision-making and traffic assignment. By considering the impact of the capacity reduction and route choice change caused by the maintenance actions, the model is expected to create better outcomes and the benefit of life-cycle management will be increased. After some preliminary studies, dynamic hybrid systems are shown useful for transportation life-cycle management. It is possible to include multiple maintenance actions and take in account the deterioration mode switches caused by action changes. Moreover, the performance of optimal threshold approach is close to that of optimal policies and it is simple to apply and easy to understand by the transportation agencies. Therefore, the proposed approach is adequate and has great potential. The remaining tasks for the study are the following. Multiple facilities and their interdependency will be studied and solution methods will be designed. The proposed methodology will also be validated with an actual road network to understand its applicability for real-world problems. 研究期間：10008 ~ 10107 |
