鐵路運輸是在陸上運輸系統中最可靠的系統，但一發生事故就會造成嚴重影響。根據台鐵過去的統計資料，機務事故當中與旅客息息最切身相關的車輛事故每年占了整體事故超過4成，高居台鐵運務、工務、機務、電務四大處之首。因此本研究目的在建立一套車輛故障的分析模式以用來預測未來可能發生的肇事情況，作為改善參考。本研究蒐集台鐵機務部門2010 年至2014 年五年間之事故資料，利用一般線性模式中的卜瓦松迴歸及負二項迴歸，針對鐵路行車特性及軌道幾何等等不同特性，構建鐵路事故次數與故障次數模式預測。 研究結果發現，事故次數模式與故障次數模式適合以卜瓦松迴歸預測事故，同時尖離峰跟平假日不同的乘載條件下對事故有顯著影響。 ;Railway transport is the most reliable system in the land transportation system, but it may cause serious effect while the accidents happen. According to the Taiwan Railway Administration’s former statistic data, rolling stock accident accounted for above 40% of the overall rail accidents which closely related to the passenger. Therefore, the purpose of the study aims to establish rolling stock breakdowns analysis model to predict the future accidents. The study collecting the Taiwan Railway Administration five years of accident data from 2010 to 2014, using general linear models of Poisson and Negative Binomial Regression, in relation to railway travel characteristics, track geometry and so forth. Constructing the railway accident frequency model and failed mode to predict accident in the future. . The result indicated that railway accidents’ frequency model and failed model are suitable to using Poisson Regression is better than Binomial Regression to predict accidents in the future. The peak/off-peak hours and rush /general hours under different loading conditions have significant influence on accident.
