A laser texture scanner was used to measure the mean profile depth of outdoor real pavement surfaces and slabs manufactured in the laboratory and photograph to build a database, group the images based on the mean profile depth range of the same asphalt, and then machine learning was used to predict the surface macrotexture of asphalt from asphalt images. The mean texture depth and mean profile depth of slabs with different aggregate gradations and bitumen content were measured by the sand patch test and laser scanning, respectively. And study the relationship between the results of sand patch test and laser scanning and the maximum aggregate size, porosity. Try to build some models that mean texture depth and mean profile depth can predict each other by regression fitting. Try training a convolutional neural network model to identify the average profile depth range of different asphalt images, and further predict the mean texture depth.
Explore the statistical relationship between the macrotexture and void properties of the real asphalt mixture and try to establish some predictive models between the parameters through regression analysis. The project uses X-ray computed tomography (CT scan) to measure the void properties of the asphalt mixture (such as macroporosity, average void diameter, void shape characteristics, etc.) and use the two parameters of mean profile depth and wavelength to quantify the macrotexture. Based on software simulation, measure the volumetric properties of the asphalt mixture (such as the number of aggregates and mastic content). The research includes the correlation between the topological properties of air voids and voids in mineral aggregate and the correlation between void properties, volumetric characteristics, and macrotexture.
Use the data obtained from the real asphalt mixture test to compare the relevant data of the equivalent virtual mixture simulated by the verification and optimization software. To obtain relevant data closer to the real asphalt mixture through the software. The project will use the laboratory measurement results of the real mixture to verify the data corresponding to the equivalent virtual mixture simulated by the software, including the macrotexture of the asphalt surface based on laser measurement and the void properties data calculated based on CT scanning. I will try to continuously use more experimental test results of real mixtures with different gradation porosity to optimize the software numerical test results' accuracy. The software can simulate equivalent virtual specimens with the same porosity as the asphalt mixture prepared in the laboratory according to the gradation and air voids of the real asphalt mixture. The macrotexture data and gradation will be used to simulate an equivalent virtual mixture with the same surface texture after completing verification and optimization.
The surface texture of the asphalt mixture is used to predict the topological properties of the voids and the degree of compaction. Based on the above research to optimize the numerical test results of the virtual test piece simulated by the software to the extent that it is consistent with the test data of the actual test piece to simulate the surface texture according to the gradation and macrotexture data of the real asphalt mixture. Furthermore, the internal microscopic distribution is close to the equivalent virtual mixture of the actual test piece, and the test data of the air voids content and volumetric properties corresponding to the asphalt mixture of this texture are obtained. Through the three-dimensional visual simulation of the asphalt mixture, the sample's complete spatial structure and surface texture can be observed intuitively. Digital image technology is used to measure the data of the topological properties of the voids of the virtual mixture and compare it with the expected air voids content calculated based on the volumetric properties to determine the degree of compaction of the asphalt mixture.