Models for ranking railway crossings for safety improvement

• Published in: Operational Research in Engineering Sciences: Theory and Applications 2020-12, Vol.3 (3), p.85
• Main Authors: Kasalica, Sandra, Obradović, Marko, Blagojević, Aleksandar, Jeremić, Dušan, Vuković, Milivoje
• Format: Article
• Language: English
• Subjects: railway crossings, high-risk locations, accidents, regression models
• ISSN: 2620-1607; 2620-1747
• DOI: 10.31181/oresta20303085k

Analysis of high-risk locations, accident frequency and severity for railway crossing is necessary in order to improve the safety and consequently diminish the number of accidents and their severity. In order to extract the necessary parameters that quantify the risk associated with railway crossings in Serbia, we have carefully analyzed available statistical models commonly used in this kind of studies. A zero-inflated Poisson model and a multinomial logistic model were used for the assessment of accident frequency and accident severity respectively. In order to quantitatively evaluate the risk, a well known measure – total risk was modified and a new measure for risk – empirical risk was introduced. The road sign warning device (p=2.76∙10^(-9)) , exposure to traffic (p=4.3∙10^(-7)), and maximum train speed at a given crossing were (p=1.36∙10^(-5)) significantly associated with probability of accident frequency and significantly influenced the expected total number of fatalities or injuries caused by traffic accidents.