| In recent years,while the rapid growth of freeway mileage has brought convenience to people's life and production,the accompanying traffic crash and hidden safety hazards cannot be ignored.In existing freeway crash risk related research,the research objects focus on specific geographic landforms or single road section types,ignoring the impact of regional type differences on freeway crash cause analysis,and the identification of crash sign factors and assessment of traffic safety levels.As high-precision traffic flow data becomes available,the static and passive traditional freeway safety improvement methods are gradually replaced by active safety control technology based on real-time dynamic traffic data.However,in the study of the relationship between dynamic traffic flow characteristics and traffic safety,there is still the problem that regional type differences are not highly targeted.In addition,the traditional freeway traffic safety evaluation research mainly focuses on “crash intensity analysis” on the micro-section level and lacks relevant research that considers the differences in macro-regional types.Therefore,the traditional freeway traffic safety analysis method is difficult to provide accurate guidance for the safety management of different regional types of freeways.In view of this,this thesis takes the freeway under the condition of regional type difference as the research object and carries out the research according to the logic of "crash cause difference judgment-dynamic traffic flow crash risk identification-traffic safety level evaluation".The focus is on answering the following key scientific questions: whether the correlation between freeway risk factors and crashes of different regional types is the same;what kind of deep-seated self-correlation laws exist among the variables of crash dimension and cause dimension;what differences exist between the traffic flow of each regional type of freeway and in the relationship between operating status and traffic safety;how to use high-precision traffic flow data to effectively study the dynamic crash risk mechanism of freeways under different regional types and traffic conditions,and how to quantify the level of freeway traffic safety under the condition of regional type differences.The research includes four aspects:(1)Analysis of the causes of the different regional type freeway traffic crashes based on the improved WODMI-Apriori association rule mining algorithmThe study area is divided into urban,rural and mountainous freeways,and a multi-dimensional interactive improved Apriori association rule mining algorithm(Weighted Orientated Multiple Dimension Interactive-Apriori,WODMI-Apriori)considering orientation constraints and index weighting is proposed.Based on the Interval Analytic Hierarchy Process(IAHP)and the subjective and objective joint weighting model of Gray Correlation Degree,the weights of the data fields are optimized,and the improved association rule mining algorithm is used to map the three different types of freeways,using multi-dimensional interactive association rule mining and calculation based on angle,dimensional interaction angle,and crash dimensional autocorrelation angle.The results show that different regional types of freeways have different crash occurrence mechanisms,and each dimension level in them also has different association laws.The results show that the improved WODMI-Apriori algorithm can better reveal the differences in the crash causes and risks in different regional types of freeways,and its algorithm is more accurate than the traditional Apriori association rule algorithm in urban,rural,and mountainous freeway conditions,with an increased accuracy of 82.7%,88.5%,and 80.5% respectively.(2)Safety risk assessment of freeway traffic flow status under the condition of regional type characteristics differenceFirstly,the traffic state is divided into saturated flow and unsaturated flow based on the LOS theory,and a total of six units to be evaluated are established in combination with the division of the three regional types.Then,the Case-Control Pairing method is used to carry out data matching and sample structured design on traffic flow and crash data.Finally,the Conditional Logistic Regression based on Markov Chain Monte Carlo(MCMC)method is used to quantitatively evaluate the risk of freeway crashes under different regional types and traffic conditions.The results show that the flow,speed and occupancy rate are highly correlated with the freeway area type and traffic state,and the freeway area type and traffic state are all significantly correlated with traffic safety.Among them,the crash risk in the urban area/saturated flow state is the largest,and its crash risk is 29.6 times that of the crash risk in the rural area/unsaturated flow state.(3)Research on the internal mechanism of different regional freeway traffic crashes based on dynamic traffic flow characteristicsFirst,from the basic information of traffic flow,a total of twenty related traffic flow variables are selected,including vehicle fleets in traffic flow,lane changing behavior of vehicles,sudden changes in traffic flow variables in a short time,and vehicle following behaviors,which reflect the dynamic characteristics of traffic flow.Subsequently,the Random Forest algorithm is used to calculate the crash sign traffic flow variables under different area types and traffic conditions.Finally,according to the Random Forest crash sign variables,Bayesian Logistic Regression method is used for modeling,and the statistical relationship between traffic flow variables and crash risk in different regional types and traffic conditions is constructed.The results show that in different regional types,the factors that affect traffic safety are different,and the importance of the same factor in different regional types is also different,further verifying that different regional types of freeways have different crash occurrence mechanisms.In addition,the results of multiple models show that the real-time crash risk assessment method that simultaneously considers the differences in freeway area types and traffic conditions can more comprehensively and accurately capture the relationship between traffic flow dynamics and traffic safety.(4)Traffic safety evaluation method for freeway with regional type differencesFive different regional types of freeways are selected as units to be evaluated,and a total of eight evaluation indicators are considered from the four aspects of safety,efficiency,economy,and environment.A comprehensive freeway safety evaluation system is constructed,the traditional Osculating Value model is improved,and an improved Osculating Value evaluation model based on Entropy Weight is proposed,and the comprehensive traffic safety level of each regional freeway is evaluated from two perspectives: annual and quarterly.The evaluation results show that the pros and cons of each road section are ranked differently,and each index reflects different degrees of importance in the evaluation process,indicating there are significant differences in the traffic safety level of different regional types of freeways.In addition,the calculation results of the improved Osculating Value method are obviously different from the traditional osculating value method,mainly because the traditional Osculating Value method has equal weighted evaluation indicators.In order to avoid result bias,it is necessary to optimize and improve the traditional Osculating Value model.The Osculating Value method does not need to determine the subjective parameters,is quick to calculate,has high resolution of results,and can be used as an effective method in the evaluation of freeway traffic safety.There are 75 figures,48 tables,and 235 references in this thesis. |
PDF Full Text Request