Predicting the Marshall flow of asphalts concrete using Gradient Boosting model

Abstract

Dense-graded asphalt concrete is one of the primary materials used to construct road pavement structures. The design and quality assessment of dense-graded asphalt concrete mixtures require consideration of multiple criteria, among which Marshall flow (MF) is a main indicator. This study compiled a dataset consisting of 90 dense-graded asphalt concrete samples with nine input variables and one output variable, MF. Based on this dataset, three machine learning models were developed and evaluated for predicting MF, including Gradient Boosting, Light Gradient Boosting, and Decision Tree. The predictive performance of the models was compared over 20 simulation runs, leading to the identification of the Gradient Boosting model as the optimal approach, with a high predictive accuracy (R²test= 0.91). The study presents a guide for using the machine learning model to predict the MF of dense-graded asphalt concrete with high accuracy, providing significant time and cost savings. Additionally, SHAP diagram was used to predict the influence of each input variable on the MF of dense-graded asphalt concrete.