標題: 國道隧道速限影響交通事故之研究
Effects of Speed-limits on Crash Frequency in Freeway Tunnels
作者: 許書銘
吳宗修
Hsu, Shu-Ming
Woo, Hugh
運輸與物流管理學系
關鍵字: 事故調整因子;安全績效函數;隧道;負二項;crash modification factors;safety performance function;tunnel;negative binomial distribution
公開日期: 2017
摘要: 隧道封閉的環境對於事故頻次的影響尚未經實證研究或具有共識的模式,且國內少有關於隧道路段研究,多以整體路段以隧道為類別變數進行分析;本研究針對國道高速公路隧道進行研究,衡量隧道的幾何設計與交通特性對於事故頻次的影響,使用不同變數轉換的方式建立安全績效函數,選擇表現較好的模式後,再透過事故調整因子(crash modification factors, CMF)衡量各個因素在不同範圍下對於事故頻次的影響變化,有別於針對事前事後的事故調整因子,本研究採用的是選取反應變數的基準狀態,並比較基準狀態以外對於事故發生的影響。在建立安全績效函數之前,本研究參考其他文獻對於變數處理的方式建立五種不同變數轉換型式的模式,其中包括不使用任何變數轉換、對日交通量取對數、對日交通量與隧道長度取對數、對日交通量取自然對數、對日交通量與隧道長度取自然對數等五種,並使用概似比指標與貝氏資訊準則選擇表現較好的變數轉換形式,其中對日交通量與隧道長度取自然對數的變數轉換型式表現最佳。本研究並使用實證貝氏分析法建立事故調整因子來估算不同速限之間的關係以及不同車道數之間的關係,透過設定速限90公里/小時為基準狀態,比較其他速限相對於基準狀態下對於事故數的影響,結果顯示速限80公里/小時與100公里/小時以上均相較於90公里/小時會較少事故發生;設定三車道隧道為基準狀態,結果顯示雙車道隧道與四車道隧道相對於三車道隧道較少事故發生。
The effect of tunnels’ environment on crash frequency has not been proved by empirical research or consensus model, and there are few crash researches about tunnels. Most researches treat tunnels as category variables. The focus of this research is on the effect measurement of crash frequency by geometric design and traffic characteristics. Different methods using transformed variables were used to establish safety performance functions. After choosing a better model, crash modification factors were applied to measure its effect to crash frequency within different ranges of variables. Different from before-after crash modification factors, this study adopts variables’ baseline condition and measures relative effects in other conditions. Before establishing safety performance function, five models with different variable transformation, including original variables, logarithm of annual average daily traffic, logarithm of annual average daily traffic and tunnel lengths, natural logarithm of annual average daily traffic, and natural logarithm of annual average daily traffic and tunnel lengths. Likelihood ratio index and Bayesian Information Criterion were then applied to choose the model with better explanatory variable transformation. The results show that the model with natural logarithm of annual average daily traffic and tunnel length is more better. Empirical Bayesian approach was applied to estimate the relationship between each speed-limit and each lane numbers. By setting 90 km/h as the baseline condition, it appears that there are less crashes in 80 km/h and 100 km/h sections. By setting three-lane as the baseline condition, it appears that there are less crashes in two-lane and four-lane sections.
URI: http://etd.lib.nctu.edu.tw/cdrfb3/record/nctu/#GT070453662
http://hdl.handle.net/11536/142966
Appears in Collections:Thesis