| Summary: | 碩士 === 國立交通大學 === 管理學院運輸物流學程 === 105 === Level crossing is usually regarded as one of the most vulnerable locations to train accidents of a railway system. There are a total of 435 level crossings of Taiwan Railway Administration (TRA), implying the importance to propose effective prevention and relief operation to level-crossing accidents. Once a level-crossing accident occurs, a dispatcher has to make a decision to relive the potential impact caused by the accident immediately. In practices, such a decision is usually made according to the experiences and professional judgement of the dispatcher. However, it is not an easy job to make a high-quality decision under complex and uncertain intertwined relationships among accident evolution, delay propagation of neighboring trains, and relief operations.
With well-informed situation regarding the potential consequences of various operation adjustment strategies right after the occurrence of a level-crossing accident, dispatchers can surely improve their decision quality and effectively curtail possible impact. To this end, this study aims to develop a decision support model for operation adjustments in response to level-crossing accidents.
To do so, this study firstly estimated an accident duration model based on historical level-crossing accidents by regressing accident duration time on contributing factors including train damage, railway damage, fatality, injury, and distance to police station. With predicted accident duration, blockage of single line or double line, and frequency density of neighboring trains, a delay propagation simulation model is developed to estimate the total passenger delays under various operation adjustment strategies, including Do nothing, Passenger-transfer, Single-line operation. In terms of passenger total delay, Single-line operation strategy is the best response choice in most of cases. However, in terms of total sum of squared passenger delay, if the accident duration is long, passenger-transfer strategy become favorable.
|
|---|
