Research On Data-driven Multi-model Combined Prediction Of Pavement Service Life And Development Of Integrated System

Reasonable prediction of asphalt pavement service life decay trends is of great significance and socio-economic value for timely maintenance and repair measures to extend the service life of pavements.However,the decay of pavement service life is not only the deterioration of performance from one perspective,but often the decay of performance in several aspects simultaneously;therefore,there are obvious shortcomings in predicting the decay of pavement service life from only one perspective.In addition,a unified information integration management database has not been established in the field of highway engineering in China,resulting in the waste of a large amount of practical data resources.To solve the above problems,this paper firstly,based on the results of on-site disease research and disease cause analysis in the Hebei area,and facing the actual needs of the project,determines the decay law and establishes the prediction model from three perspectives of interlayer binder material,asphalt and asphalt mixture and pavement technical index,and proposes the method of multi-model combination to predict the service life of asphalt pavement;secondly,this paper introduces the concept of knowledge mapping Secondly,this paper introduces the concept of knowledge mapping into the field of highway engineering and establishes a knowledge map and search engine system for asphalt pavement information;finally,a multi-model combination prediction system for asphalt pavement service life is constructed based on the knowledge map and multi-model combination prediction method.The integrated design objectives of information storage,information retrieval,service life prediction,and auxiliary decision-making in highway engineering are realized.The main research results of this paper are as follows:(1)Through the field disease research,it is found that the proportion of cracking type disease is large,and the generation of this disease is closely related to interlayer slip and asphalt mixture fatigue cracking,and the value of pavement performance state evaluation index is proportional to the severity of the disease.Combined with the analysis of the disease research results,this paper selects three aspects from the interlayer bond state,asphalt mixture fatigue life,and pavement performance technical indexes to establish performance prediction models respectively,and adopts the multi-model combination method to predict the service life of in-service asphalt pavement comprehensively.(2)Through the experimental study of the properties of interlaminar bonded materials,a composite cohesive region intrinsic model was proposed to characterize the decay process of interlaminar bonded materials,and the reliability of the model was verified by using finite element simulation,and the simulation results showed that the composite cohesive region intrinsic model could accurately simulate the damage process of interlaminar bonded materials under the action of positive and shear stresses.Based on the performance test results,the shear fatigue life prediction equation based on the actual stress ratio was further established by the 45° oblique shear fatigue test and time-displacement relationship curve,and the change law of residual shear strength after different shear fatigue times was analyzed;finally,a nonlinear shear fatigue damage model independent of stress level was established based on the residual strength theory and the damage variables defined by residual shear strength Finally,based on the residual strength theory and the damage variables defined by the residual shear strength,a nonlinear shear fatigue damage model independent of stress level and a residual shear strength decay prediction model were established.(3)Through the study of the fatigue performance of mastic powder-modified asphalt and its mastic,the nonlinear fatigue damage performance and fatigue damage evolution characteristics were analyzed,and the corresponding fatigue life prediction model was established;further,considering the fatigue cracking resistance of asphalt mix as well as the drilling method of field pavement core samples and the convenience and accuracy of fatigue prediction model application,a fatigue life prediction model based on dynamic load semicircular bending test was proposed with The fatigue life prediction model was proposed based on dynamic load semicircular bending test,with critical strain energy release rate as the characterization parameter of asphalt mixture anti-cracking performance,and asphalt saturation,void ratio,stress level,and dynamic modulus as the basic prediction parameters,and the prediction accuracy of the model reached 84.85%.(4)Based on pavement disease detection and performance index evaluation,a multivariate nonlinear empirical regression prediction model of the pavement performance evaluation index was established.The prediction accuracy of all indexes was higher than 95%.The threshold and weight coefficients of the traditional BP neural network were further optimized using the fireworks algorithm,and the prediction model of asphalt pavement technical indexes based on the optimized BP neural network with the fireworks algorithm was established.The accuracy of its prediction results reached 98.35%,which indicates that the BP neural network model optimized based on the fireworks algorithm has a more desirable prediction accuracy,especially since it can significantly improve the data variability prediction accuracy in the region with significant variability.(5)By introducing the concept of knowledge graph into the field of highway engineering,a knowledge graph of asphalt pavement information was constructed;secondly,an asphalt pavement information search engine based on the knowledge graph was established to realize the information interaction and target information matching between the web server and the graph database;finally,based on the residual shear fatigue life prediction model of interlayer binder material,the nonlinear fatigue life prediction model of asphalt mixture and the pavement Finally,the multi-model combination prediction function of pavement service life and the weight relationship of each model are proposed based on the interlayer binder material residual shear fatigue life prediction model,asphalt mixture nonlinear fatigue life prediction model,and pavement technical index prediction model(FWA-BP),and a multi-model combination prediction system for the service life of asphalt pavement dedicated to regional road networks is constructed by integrating the knowledge graph.Based on the above research,this paper adopts various technical means to build an integrated asphalt pavement information management system,which can effectively solve the status quo of "knowledge silos" faced by the highway industry and initially realize the integrated design goals of storage,retrieval,pavement service life prediction and auxiliary decision making of asphalt pavement structure,material,and performance testing data.