中山高速公路肇事分析模式之研究

Abstract

在國人生命價值及時間價值不斷提高之今日，如何改善中山高速公路 肇事之課題益顯重要；本研究旨在建立一適當之肇事分析模式，期能有效 改善中山高速公路肇事情形，增進安全。以民國80年至84年之列管肇事為 研究對象，為使模式得以掌握實際肇事特性，本研究首先針對中山高速公 路肇事變動趨勢、肇事發生分配型態及肇事影響因素進行分析，以作為模 式構建之基礎；接著採用一般化線性模式(GLM)中之卜瓦松迴歸及負二項 迴歸技術，考量公路供給特性、交通特性及管制措施三大類肇事影響因素 ，分別構建高速公路三種路段--主線、交流道及收費站之肇事分析模式。 研究結果發現三種路段之肇事分析模式有所差異，除模式解釋變數不 盡相同外，亦各有不同之適用模式--收費站路段適用卜瓦松模式，而主線 及交流道路段則以負二項模式較佳；顯示主線及交流道路段之肇事發生存 在過度離散(即肇事發生次數之變異數大於平均數)情形。經由肇事因果分 析，可獲致以下結論：1.公路幾何特性對於肇事之發生有顯著影響，不僅 受路段本身幾何特性影響，亦受到毗鄰路段之影響；避免陡下坡及曲度差 距過大之設計、減少交流道路段之衝突點數及廢除戰備跑道均有助於減少 肇事之發生。2.肇事次數隨每車道車公里數及重車比例之增加而增加。3. 提高速限有助於減少主線路段之肇事發生。本研究所構建之模式除可供作 肇事因果分析外，尚可進一步應用於肇事預測、危險路段判別及交通安全 相關策略之效果評估。 The modeling of Sun Yat-Sen National Freeway accidents continues to be of interest because of the frequency and severity of the accidents and thecongestion associated with them. This study is to build appropriate accidentanalysis models which help understand the cause-and-effect of accidents onfreeway. Fatal and severe injury freeway accidents from January 1991 toDecember 1995 are collected. After analyzing the characteristics of accidents,this study uses poisson and negative binomial regression techniques toestimate the effects of road geometrics, traffic characteristics and speedregulation on accident frequency for three road sections -- mainline, inter-change, and toll station. The results show that there exist differences amongthree accident analysis models on different road sections. The variablesinvolved in each model as well as the appropriate regression techniques fordifferent models are not all the same.Poisson regression model can welldescribe the toll station sections, while negative binomial regression modelappropriately depicts the mainline and interchange sections, which representsa high variance-mean ratio of accident frequency on both mainline and inter-change sections. The cause and effect analysis concludes that (1)avoiding roadgeometric design with steep downgrade and sharp curvature, reducing the numberof conflict points on interchange sections, modifying the "warlike preparationrunways" are effective in reducing accidents; (2)high vehicle kilometers perlane and heavy vehicle ratio will increase the accident frequency; (3)raisingthe upper speed limit on mainline sections will reduce the accident frequency.In addition to conduct accident cause-effect analysis, these models can alsobe used in accident prediction, dangerous sections diagnosis, and effective-ness evaluation for traffic management strategies.