Research On Prediction Of Asphalt Pavement Roughness Of Expressway In Shanxi Province

Pavement roughness is one of the important indicators for evaluating pavement surface performance.which has an important impact on vehicle driving comfort and road safety.Predicting pavement roughness is helpful to grasp the pavement surface roughness.The changing trend lays the foundation for the determination of pavement maintenance timing and maintenance plan,and is beneficial to the optimal allocation of pavement maintenance funds.The main goal of this article is to establish a highway asphalt pavement roughness prediction model suitable for Taiyuan Company and Changzhi Company.The basic road data,performance detection data,road traffic load data and road maintenance historical data of expressway network under the jurisdiction of Taiyuan and Changzhi are collected.The whole road network is split according to the rules,The sections with wrong performance detection data were corrected by interpolation method,the sections with missing performance detection data for one year were interpolated and supplemented,and the sections with missing performance detection data for two years or more were deleted.The evaluation criteria of pavement smoothness under different standards are compared.The historical status of pavement riding quality index RQI within the road network of Taiyuan and Changzhi companies is analyzed.The influencing factors of pavement roughness development are analyzed,and the sections of the whole road network are grouped according to infrastructure type,traffic load,latest maintenance type,environment and climate zoning,among which infrastructure types are divided into three categories:pavement,bridge and tunnel;Combined with BP neural network,the traffic load is divided into three categories by using mean influence value algorithm and hierarchical clustering algorithm:light traffic,medium traffic and heavy traffic;The latest maintenance type is divided into three categories:daily maintenance,preventive maintenance and rehabilitation maintenance;Environment and climate zoning can be divided into two types:warm temperate zone cold temperature heavy semi-arid climate zone and warm temperate zone cold temperature semi-humid climate zone.Investigating the research status of common pavement roughness prediction models at home and abroad,combining with the decay law of pavement roughness of Taiyuan Company and Changzhi Company,S-shaped curve model and exponential model are selected as the standard forms of prediction models.The parameters,prediction accuracy and fitting effect of each group of sections decay equation are obtained by using least square method through nonlinear regression analysis,and it is concluded that S-shaped curve model is superior to exponential model,so S-shaped curve model is selected as the standard form of pavement roughness decay equation.MAE,MSE and R~2are used to evaluate the prediction model,and finally MAE=0.191,MSE=0.0190 and R~2=0.852 are obtained for the whole road network.Finally,BP neural network model and GA-BP neural network model are designed to predict the pavement roughness.The input variables include the historical detection data of pavement roughness and the influencing factors of pavement roughness development and change,and the output variables are the predicted values of pavement roughness in the next year.The BP neural network model is compared with GA-BP neural network model,S-shaped curve model and GA-BP model,and the prediction accuracy and fitting effect of GA-BP model are the best,which shows that GA-BP model can be better applied to pavement roughness prediction.