Research On Multi-scale Parameter Database And Cross-scale Correlation Method Of Asphalt Mixture

Asphalt mixture is a kind of multiphase composite material composed of aggregate,asphalt,mineral powder,etc.the size and physical and mechanical properties of the composite materials are different.The structure after mixing and forming is also random.The complexity and randomness of the materials lead to the diversity of the mesostructure and the complexity of mechanical behavior.However,the mixture model in the present design method is too simplified to reflect the complexity and randomness of asphalt mixture,and then the fine structure and macro volume parameters of the mixture can not be accurately predicted by theoretical model.Therefore,it is necessary to study the relationship between asphalt mixture gradation,microstructure and macro-volume parameters based on the real meso-structure and a large number of test data,and establish a cross-scale correlation model to reveal the quantitative relationship between the three.At present,there are still some problems in the research,such as the lack of research samples,the traditional research angle and the lack of advanced research methods.Based on the above problems,this study first establish the meso-structure data pool of asphalt mixture,improves the segmentation ability of CT image based on adaptive image preprocessing and deep learning model,extracts the real meso-structure of asphalt mixture,and constructs a well differentiated and representative meso-structure index group to characterize the distribution and morphology characteristics of voids and aggregates,The above indexes are calculated for all images in the database,and the data are preprocessed and combined to complete the establishment of the meso-structure database;Secondly,based on the collection of volume parameters of asphalt mixture,the macro-volume parameter data pool of asphalt mixture is established.Through index screening,missing value filling,error data screening and outlier screening,the data quality is improved to complete the establishment of macro-volume parameter database;Furthermore,based on the stepwise regression and characteristic engineering,the gradation meso-structure model is established,and the model is tested to verify the significance of the model.An example is given to introduce the application background and method of the model;Finally,based on the ensemble learning model and Bayesian optimization,the gradation-macro-volume parameter model is established,and the evaluation and error analysis of the model are carried out.An example is given to introduce the application background and method of the model.The research results provide a theoretical basis for revealing the formation mechanism of asphalt mixture performance and optimizing the design method of asphalt mixture.