A data-driven model for phase angle behaviour of asphalt concrete mixtures based on convolutional neural network

Selection of asphalt concrete (AC) mixtures with proper knowledge of its phase angle characteristics is critical in designing flexible pavements and ensuring the maximum service life of pavements. To achieve this purpose, laborious and expensive laboratory testings are frequently performed, and the results are implied to field. To overcome this problem, models (mathematical or machine learning) are developed to predict the phase angle of AC mixtures. However, the complex and non-linear relationship of phase angle with its independent variables is hard to capture using simple mathematical (or statistical) models. As such, this study proposes a data-driven model based on Convolutional Neural Network (CNN) to capture and predict the phase angle behaviour of AC mixtures. Twenty-three AC mixtures are prepared in laboratory consisting of varying gradations, binder grades, and mix types to perform phase angle testing. The proposed modelling framework, trained using the dataset obtained from laboratory testing, captures 90% of the variance in the test data, which is a significant improvement as compared with other machine learning models as well as linear regression. The proposed model has the capability to capture the nonlinearity associated with AC mixtures and can be used by transport agencies and practitioners as a surrogate to tedious laboratory testing. (C) 2020 Elsevier Ltd. All rights reserved.

Automated summary

This study proposes a data-driven model based on Convolutional Neural Network (CNN) to predict the phase angle behavior of AC mixtures, which captures 90% of the variance in the test data. The model significantly improves upon other machine learning models and linear regression, providing a surrogate to tedious laboratory testing for transport agencies and practitioners.