Stiffness Data of High-Modulus Asphalt Concretes for Road Pavements: Predictive Modeling by Machine-Learning

This paper presents a study about Machine Learning approach for modeling the stiffness of different high-modulus asphalt concretes (HMAC) prepared in laboratory with harder paving grades or polymer-modified bitumen which were designed without reclaimed (RA) content. Notably, mixtures considered this are not part purposeful experimentation support modeling, but practical solutions developed actual mix design processes. Since models require careful definition network hyperparameters, Bayesian optimization process was used to identify neural topology, as well transfer function, optimal type needed. By employing performance metrics, it possible compare obtained by diversifying inputs. Using variables related composition, namely content, air voids, maximum and average bulk density, along categorical variable that distinguishes RAP percentages, successful predictions Stiffness have been obtained, determination coefficient (R2) value equal 0.9909. Nevertheless, use additional input, Marshall stability quotient, allows prediction be further improved, R2 values 0.9938 0.9922, respectively. However, cost time involved test may justify such slight improvement.