Research on structural modulus inversion method of asphalt pavement based on BP neural networ

Abstract

The current inversion methods based on pattern recognition method, database search method, genetic algorithm and other inversion methods are difficult to solve the absolute convergence problem of structural modulus inversion for asphalt pavements with more than three layers, while the deep learning models and methods widely used in the scenarios of image recognition, speech recognition, etc. as well as the ways to implement them have been increasingly improved, and they can be applied to solve the problem of structural modulus inversion for asphalt pavements. This study aims to solve the modulus inversion problem of multi-layer asphalt pavement structures, obtaining enough theoretical bending basin data of asphalt pavement structures as training samples through mechanical theory and programming calculations, and using the BP neural network model to train the prediction model of structural layer modulus inversion. The test results show that the BP neural network inverse asphalt pavement structural modulus model established in this paper can not only get the prediction results quickly and effectively, but also the prediction results have high accuracy, which provides an effective way for solving the modulus inversion problem of asphalt pavement structure with more than three layers by using the BP neural network to solve the pain point and bottleneck problems in the industry.