Improving Railroad Grade Crossing Safety: Accident Prediction Models Using Macro- and Micro-Scale Analysis
Rahim Benekohal, Professor, N/A
This project is working to 1) develop a methodology for analyzing rail crossing crashes at a micro level to discover trends at a single crossing or a series of crossings along a corridor or a region and 2) improve the accuracy of crash predictions by incorporating findings from the microscopic analyses and studying the regional trends that emerge in this analysis but are not observed at a national level. This new micro approach has resulted in the identification and valuation of the contribution of new variables not previously considered. For example, findings indicate that the distance to the nearest highway-highway intersection is an important factor in gated-crossing crash prediction. It was also found that the angle between railroad and highway is an important factor in crash prediction for crossings with flashing lights. Improvements in crash prediction can result in a more reliable ranking and selection of crossings for safety improvements as well as more accurate cost-benefit estimations, which can help optimize resource allocation procedures and therefore reduce crash frequencies to a greater extent than current practices. The addition of information resulting from the micro analysis into macro models is expected to further enhance predictions and to provide a multi-scale perspective not previously studied in this context. To complement dynamic tree analysis for finding contributing factors to crash frequency, a "crossing cluster" is calculated for each crossing. The crossing cluster represents the contribution of a variable to crash frequency at that crossing, which is often larger than the contribution found in the dynamic tree.