Optimizing Foamed Bitumen Bound Asphalt Mixture Design Using Neural Network

Abstract

Effective design of bituminous mixes for road pavements requires a robust understanding of their mechanical properties to ensure durability and safety. Conventional experimental methods for assessing these properties are time-consuming and costly. To address this challenge, advanced machine learning techniques have gained prominence in predicting bituminous mix behaviour. In this study, we focus on predicting Marshall Stability (MS) and Flow (MF) of foamed bitumen bound asphalt pavements using essential input parameters: Temperature, Foam Content, Expansion Ratio, and Half-Life. Leveraging a neural network model, accurate prediction equations and surface analyses were developed for optimizing pavement design. Furthermore, integration equations are also introduced to enhance the accuracy of the methodology. Sensitivity and Parametric Analyses provide insights into parameter impacts, and R-squared measures model goodness of fit. The research work presented not only streamlines pavement design but also advances the understanding of intricate input-output relationships in bituminous mixtures.