Akademik Veri Yönetim Sistemi
International Journal of Pavement Research and Technology, 2026 (ESCI, Scopus)
Ensuring the durability and impermeability of asphalt pavements requires proper compaction and controlled water absorption. Traditional quality control methods rely on destructive testing and trial sections, which are costly and time-consuming. This study introduces an artificial neural network (ANN)-based approach to predict in-situ compaction and water absorption using Marshall test parameters, addressing the nonlinear and multivariate nature of the problem. Unlike conventional black-box ANN applications, the proposed model incorporates interpretable visualization through contour plots, enabling engineers to understand parameter interactions and optimize mix design. A dataset comprising various nominal aggregate sizes and Marshall characteristics was used to train and validate the model. The optimal ANN architecture included two hidden layers with nonlinear activation functions, achieving superior prediction accuracy compared to linear models. Multi-objective optimization identified input combinations that maximize compaction while minimizing water absorption. The findings suggest that current specifications, which only set minimum compaction limits, should include a water absorption threshold of 1% to enhance pavement performance. This approach reduces reliance on trial sections, improves quality assurance, and offers a cost-effective solution for predicting critical performance indicators such as permeability, water absorption, and degree of compaction for flexible pavements.