Modeling And Prediction Of The Impact Of Geohazard Accidents On Mountain Roads

A mountainous road network with special geological and meteorological characteristics is extremely vulnerable to nonrecurring accidents,such as traffic crashes and geohazard breakdowns.These accidents significantly impact the operation of the road network.Timely and accurate prediction of the duration and spatial extent of geological hazard incidents is of significant importance to regional traffic safety management and control schemes.Due to the differences in the characteristics of road geohazard accidents and traffic accidents,the impacts caused are not the same.Existing studies on the scope of accident impacts have mainly focused on traffic accidents occurring on urban roads or highways,and fewer studies have focused on the impact of geohazard accidents on mountainous road networks.Therefore,this paper analyses the distribution pattern and the key influencing factors of the spatial and temporal extent of road geological disaster accidents based on the road geological disaster accident data collected in Yunnan Province.Survival analysis theory is applied to model accident duration.A model for predicting the spatial extent of accidents was constructed using fuzzy clustering and Logit models.The study results can guide traffic managers to formulate effective control measures and accident response plans promptly,and provide reliable traffic information for travelers.The main work of the thesis is as follows.(1)The spatial and temporal variation characteristics of highway geological hazard accidents in Yunnan Province were explored,and the distribution patterns of accident duration and spatial extent were statistically analysed.The study found that geohazard accidents are concentrated in areas with active crustal movement and steep hilly terrain,and that the rainy season is the high incidence period for road geohazard accidents.There are differences in the frequency distribution and spatial and temporal impacts of different types of accidents.Mudslide accidents are the most common,but landslide accidents have the longest average duration and the largest spatial extent.The average duration and spatial extent of accidents do not vary significantly between different weeks and months.Accidents with secondary accidents or accidents that result in stranded vehicles or people have a longer and wider spatial and temporal impact,and different accident handling methods and departments can have an impact on the duration and spatial extent of accidents.(2)Analysis of factors influencing the duration of highway geohazard accidents and duration prediction modeling.The relationship between accident duration and individual factors was explored based on historical accident data using the Kaplan-Meier nonparametric model.The results showed that accident type,secondary accidents,impounded vehicles or personnel,closed roads,morning rush hour and the level of accident management had a significant effect on the duration of highway geohazard accidents.Afterward,the relationship between accident survival time and all influencing factors was analysed using a Cox proportional hazards regression model,as well as ranking the importance of each factor.Based on the results of the Cox regression model,the influencing factors were selected for combination and used as inputs to the stochastic survival forest prediction model,and the performance of the constructed model was analysed in comparison with other models.The results show that the stochastic survival forest road geohazard accident prediction model performs best in prediction performance compared to other survival analysis models and machine learning models when the inputs are accident type,secondary accident,impounded vehicle or person,closed road,morning rush hour,level of accident management,road type and day of the week,and is a ffavorabletool for road geohazard duration prediction.(3)Analysis of factors influencing the spatial extent of highway geohazard accidents and modeling of spatial extent prediction.The spatial extent of highway geological hazards was classified into three categories: small,intermediate and large by using range fuzzy C-mean clustering analysis.The association rules were then applied to explore the dependency relationships between the three dimensions of accident type,time situation and accident form and the spatial extent of accidents.The results showed that all four types of accidents occurred mostly in the small range,while the large and intermediate range accidents occurred mainly in the third quarter,and the large range accidents often had vehicles or people stranded.Finally,the ordered logit model and the multinomial logit model were constructed separately to predict the spatial extent of accidents.The results showed that the regression equations obtained from the multinomial logit model fitted better,had better prediction accuracy than the ordered logit model,and were more applicable in analyzing the spatial extent of road geological disaster accidents.