Abstract
Asphalt mixture pavement has advantages such as flat surface, comfortable driving, low noise, and easy maintenance. However, asphalt mixture has poor temperature stability, is prone to brittle fracture, and has a high porosity, which is prone to rainwater and chemical erosion, especially chloride ion erosion, and is prone to aging. This article simulates and analyzes the crack resistance model of asphalt concrete based on CNN (Convolutional Neural Network). A CNN based prediction model for the crack resistance of asphalt concrete has been established based on the advantages of the CNN method. The process of predicting and measuring concrete cracking using different models was tested, and the research results showed that the error of the concrete cracking prediction curve based on this model is small, the target loss is minimal, and the training results obtained are optimal. Meanwhile, its predictive performance is significantly better than statistical methods and ACA (Ant Colony Algorithm) models. To evaluate the low-temperature cracking resistance of asphalt mixtures using low-temperature bending tests, elements are placed in the middle section of the asphalt overlay and rubber stress absorption layer, while two-dimensional plane strain elements are used in the remaining areas.