Analysis Of Injury Severity In Accidents At Highway-Rail Grade Crossings Based On Improved Logit Model

As China’s economy continues to progress,the road and railway transportation system is becoming more and more prosperous.As the intersection of railway and roadway,highway-rail grade crossings(HRGCs)present the possibility of a collision between vehicles and trains.Given the substantial difference in mass between the two,such accidents have the potential to result in severe casualties and costly property damage.In addition,due to its long braking distance,large load capacity,and large inertia,the severity of accidents on trucks is often greater than that of other types of vehicles.Therefore,on the basis of the Multinomial Logit model,this paper establishes a Mixed Logit model considering the heterogeneity of the mean and variance,and constructs prediction models for the injury severity of highway-rail grade crossings accidents for auto-train and truck-trian,respectively,to capture the difference of auto-trian accidents and truck-trian accidents.Considering the heterogeneity of accident data,explore the influencing factors of injury severity and the relationship between factors.First,a total of 14,415 accident data from 2012 to 2021 were selected from the Federal Railway Administration(FRA),and a total of 38 independent variables were selected from four aspects: driver characteristics,vehicle characteristics,environmental characteristics,and crossing characteristics.After data cleaning,filling and feature engineering,a total of 56 independent variables were obtained.Subsequently,by establishing a Multinomial Logit model for the injury severity of the accident,and through the likelihood ratio test,it was verified that there is a significant difference between the accident data of autos and trucks,and the modeling by vehicle type is meaningful.Furthermore,in order to capture the heterogeneity of different causal factors on injury severity,this study employed a Multinomial Logit Model and introduced random parameters to construct the Mixed Logit model;then,on the basis of the Mixed Logit model,the heterogeneity of the influence of factors on injury severity was further revealed by considering the heterogeneity of the mean and variance of random parameters.The Mixed Logit model considering the heterogeneity of mean and variance was constructed.The results show that the Mixed Logit model considering the heterogeneity of mean and variance has the best goodness of fit through three evaluation indexes.Finally,based on the Mixed Logit model considering the mean and variance heterogeneity,the impact of each influencing factor on the injury severity is described by the average marginal effect value.Based on the model results,targeted measures to improve the safety of highway-rail grade crossings are proposed.The research results show that the Mixed Logit model considering the heterogeneity of the mean and variance can further explore the heterogeneity of the influencing factors on the injury severity of the event,and has a good goodness of fit.At the same time,the factors affecting the injury severity of accidents are analyzed from the perspectives of auto and trucks,and it is found that there are significant differences in the factors affecting the accidents of the two types of vehicles,which provides a theoretical basis for relevant management departments to formulate differentiated management countermeasures.