Modelling of preparation parameters of polymer and oily waste sludge modified bitumen using neural network coupled with multiobjective evolutionary algorithm

Abstract

ABSTRACT In this study, artificial neural networks coupled multi-objective evolutionary algorithm based on decomposition (ANN-MOEA/D) and non-dominated sorting genetic algorithm versionIII (ANN-NSGAIII) as the chemometric approaches were used for the first time in the pavement field to model the modification conditions of the base bitumen (PG58-22). The novel ternary system of styrene-butadiene-styrene and two inexpensive waste materials as independent factors and five responses were considered in this work. Under the ANN-MOEA/D conditions, as the better optimisation method, the amounts of polyethylene, styrene-butadiene-styrene, and oily waste sludge were 3.56%, 4%, and 4.34%, respectively. The response values of penetration and change of mass after short-term ageing were at their minimum values of 50 dmm and 0.06%, respectively. The response values of softening point, ductility at 10oC, and retained penetration after short-term ageing, were at their maximum values of 65oC, 30 cm, and 89%, respectively. The analysis of the results of BBR and DSR tests in the optimal conditions obtained from two metaheuristic algorithms indicated improvement in the rutting resistance (G*/sinδ), m-value, and stiffness parameters of the modified bitumen compared to the base bitumen. The optimal conditions of ANN-MOEA/D and ANN-NSGAIII of the modified bitumen reached the characteristics of PG70-28 and PG-64-22 bitumen, respectively.