Research On Rutting Prediction Model Of Asphalt Pavement Based On Full-scale Testing Road

Rutting is one of the main damage forms of asphalt pavement,which is of great harm to the pavement safety and driving safety.Therefore,researches on rutting prediction model,especially effective method of variable indexes selection and model establishment so as to establish accurate and effective rutting prediction models suitable for different situations,are of great significance to the design of anti-rutting pavement,the maintenance of rutting damage and the improvement of asphalt pavement service life.Based on National Key Research and Development Project of China “Research on High-throughput Response Data Mining and Long-life Road Service Performance Verification Technology of Full-scale Testing Road”(NO.2017YFC0840201),rutting prediction models using exponential empirical method,mechanical-empirical method and empirical method of artificial neural network(ANN)were established under wide pavement rigid domain by means of literature research,theoretical analysis,mathematical statistics and data mining.Moreover,the accuracy of rutting prediction models was verified and its applicability and advantages were compared and analyzed.First of all,based on the characteristics analysis of full-scale testing road,STR7,STR4,STR16 and STR19 were selected as the representative pavement structures of semi-rigid base,rigid base,flexible base and full thickness asphalt pavement respectively.The data of traffic load,environmental condition,asphalt layer thickness,pavement structure and material were analyzed.Through above analysis,accumulative equivalent single axle loads(ESALs),environmental temperature,ambient humidity and center point deflection were determined as the four main influencing factors of rutting prediction model.Secondly,the method of variable indexes selection for rutting prediction model was researched.Taking STR7,the standard test pavement structure in full-scale testing road,as an example,the common influencing factor of accumulative ESALs,environmental temperature,ambient humidity and center point deflection were analyzed based on the factor analysis method of data mining technology.Results show that environmental temperature,accumulative ESALs and center point deflection have an important impact on the rutting deformation except ambient humidity.These three factors can be used as the variable indexes of the rutting prediction model.Through the factor analysis of four main influencing factors,the method of variable indexes selection for rutting prediction model is given,which may provide guide for the variable indexes selection of rutting prediction model in other research.Thirdly,the method of establishing rutting prediction model was studied.Based on the analysis results of rutting depth and influencing factors,the exponential empirical prediction model framework was determined;based on the Burgers model,the Burgers model was modified and the mechanical-empirical rutting prediction model framework was determined;based on the characteristics of ANN algorithm structure,empirical rutting prediction model framework of ANN was determined.Based on optimization algorithm technology and ANN algorithm technology,rutting prediction model using exponential empirical method,mechanical-empirical method and ANN empirical method were established respectively under wide pavement rigid domain.Moreover,fitting effects of rutting prediction models for asphalt pavement with different rigid base were evaluated and compared.The result shows that fitting effects of these three models are satisfactory and that of the mechanical-empirical prediction model is generally better than that of the two empirical prediction models.Through screening model variable indexes,determining these three rutting prediction model frameworks,establishing three rutting prediction models under wide pavement rigid domain and evaluating their fitting effects,the method of establishing rutting prediction model is given,which may provide guide for the establishment of other rutting prediction models.Lastly,based on different characteristics of the two types of prediction models,their applicability in different situations is clarified.Based on the fitting effects of these three models,their applicability to different asphalt pavement with different rigid base is clarified.Compared with the rutting prediction model established in other studies,the data base of the rutting prediction model established in this paper is more reliable,the model is more practical and applicable,which is of great significance for the application of the rutting prediction model and the prevention and control of rutting diseases.