Accident Prediction And Safety Hazards Digging Of Ordinary Arterial Highway In Mountain Area Based On Accident Scenario Classification

Ordinary arterial highways in mountain areas are the main transportation arteries,which undertake a large number of passengers and freights transportation,and play an important role in economic development.However,because of the complex terrain and special geological conditions for ordinary arterial highways in mountain areas,some limit design indicators have been used in road alignment design.The traffic safety and management facilities of ordinary arterial highways in mountain areas are always lack of due to limited funds or construction conditions,which further leads to frequent traffic accidents.According to some investigation data,the accident rate of ordinary arterial highways in mountain areas is often several times that of other plains,which imposes serious negative impact on economic development and people's health.The paper takes the traffic safety of ordinary arterial highways in mountainous areas as the research object.Through excavating the influencing factors of traffic safety,analyze traffic accidents' possibility and severity.Combining traffic accidents duration and space influence scope,traffic accident scenarios have been classified.Then,the strong association rules between traffic accident scenarios and other influencing factors are extracted,and the corresponding accident probabilities under the weak association rules for different elements combinations are predicted.According to road conditions,traffic environment and other factors,the traffic accident scenarios can be inferred in advance to realize the active prevention and control for traffic safety.Based on the relationship between traffic accidents and road network,traffic scenarios have been clustered to obtain the distribution of traffic safety risky points and the risk levels.The paper studies the traffic risk identification of ordinary arterial highways in mountainous areas from four aspects: accident duration and space influencing scope analysis,traffic accident scenarios classification,accident association rules extraction and scenarios prediction,and traffic risky points identification.The research conclusions provide theoretical support and decision-making foundation for accident risk control of ordinary arterial highways in mountain areas.The main contents include:First,the special road and environmental characteristics of ordinary arterial highways in mountainous areas are analyzed,further to explore the causes of traffic accidents and clarify the key influencing factors.Corresponding to the proliferation process of traffic accidents,for the time,accident duration is divided into three phrases,including accident detection and response phase,accident handling phase and traffic flow recovery phase.The traffic wave theory is used to establish the traffic accident duration model.For the spatial influence,based on the propagation regulations of traffic accidents,an elliptical-like traffic accident spatial impact scope level model based on the improved Gaussian Cloud model is proposed,considering the two situations of traffic accidents not occupying the opposite lanes and occupying the opposite lanes.The spatial influencing scopes of the current road and intersection roads are quantified and graded.The ellipse-like traffic accident impact level model takes into account the difference in accidents' propagation speed on the road in different directions,and reflects the heterogeneity of accidents' spatial propagation.Then,the concept of accident scenario is introduced,and construct a method system for accident scenarios discrimination based on the possibility and severity.Accident possibility considers objective factors such as roads and external conditions,and the severity includes two classes' indicators of accident attributes and traffic attributes,based on which,indicator system for accident scenarios discrimination has been established.For multi-attribute decision-making problems,the CRITIC method is used to determine the weight of each attribute.The rank-sum ration method is introduced to obtain the judgment matrix of accident scenario division,through calculating the cumulative probability of the rank-sum ratio weights of the two major categories of accident probability and severity.The concept of accident scenarios introduced in this paper considers the difference of accidents' impact degree due to their unique attributes,and constructs a multi-factor accident scenario identification method,which can better reflect accidents' impact than the single index of severity.Furthermore,based on accident scenarios classification,the association rules between traffic accident scenarios and factors such as roads,traffic and external environment are extracted,and predict the accident scenario's probability under the combination of multiple rules.Because of the different impacts of various factors,this paper introduces an improved Apriori algorithm,using multi-dimensional and multi-layer AHP weighted association rules to excavate and extract strong association rules between accident scenarios and various elements,and obtain the combination rules' identification method of risk factor for ordinary arterial highways in mountainous areas.Focused on other combinations of risk elements corresponding to strong association rules,the Bayesian network method is used to infer and predict accident scenarios' probability,and the weak association rules of risk factors combinations are obtained.Accident scenarios can be obtained when element risk values are known based on strong association rules;while weak association rules can obtain the corresponding probabilities of various traffic accident scenarios.Finally,a method of identifying and classifying traffic accidents' risk points of ordinary arterial highways in mountainous areas based on network spatiotemporal kernel density estimation is proposed.Considering road network's heterogeneity and traffic accidents' spatial specificity,a kernel density estimation method based on network distance is proposed taking accident scenarios as clustering index.The zero-expansion negative binomial regression model is used to estimate the kernel density value by sections.A method to determine the grading threshold of the nuclear density value is proposed based on the statistics principle.According to the method,the kernel density value of the accident scenarios is graded,and the first,second and third grade danger spots are obtained.The identification of traffic accident risk points takes the accident scenario as the clustering index,and considers accidents' possibility and severity.The classification of accident risk points provides the basis for traffic department to implement the traffic safety management measures in different stages.