عنوان مقاله [English]
Interfaces of two different transportation systems of rail and road are among the most hazardous points for traffic safety annually. Accidents not only bring about casualties of road and rail users but also cause stops in road and rail services and ruin equipment. Among the variety of accidents, train-vehicle crashes are some of the most severe types. Identification of effective factors in accident severity is vital for reduction programs. Usage of statistical models is a determining method for identifying black spot crossings. Such models are developed based on the relationship between accidents, on the one hand, and geometric design, control devices and traffic attributes on the other, which help to compute the amount and consequences of damage in particular places. In Iran, no model has been prepared for predicting accident severity so far. In this study, with the aid of grade crossing characteristics and accident histories from 1381-1385, such a predictive model has been developed using generalized linear regression (Poisson and Negative Binomial) methods. Modeling is performed with SAS 9.1 software. Model coefficients in generalized linear regression methods are estimated via maximum likelihood (ML) methods. In analysis, the confidence levels are set at the 90th percentile. In the provided severity prediction model, six important factors are distinguished that are similar to the other prediction models, which are compatible with engineering presumptions. Road width, type of road, train speed, presence of road curves in crossings, sight distances and the presence of humps are significant and are introduced in the negative binomial model. Considering the estimated coefficient for each factor, and expected changes in the future, a new outlook for the safety situation of grade crossings and the severity of accidents can be imagined. Based on the produced model, using humps and an improvement in sight distance, significant impact on accident severity has occurred. For instance, the use of hump reduces about 38% of accident severity at crossings. Use of this method is suggested for prioritization of grade crossing security, and prediction of future crossing situations, by improving characteristics, resource allocation, etc.in Iranian railway systems.