Predicting the dynamic modulus of asphalt mixture using hybridized artificial neural network and grey wolf optimizer

ABSTRACT

The dynamic modulus ($|E^{\ast }|$) of asphalt mixture, which is one of the fundamental parameters in asphalt industry, is used to characterise the performance of asphalt mixtures at a wide range of temperatures and loading conditions. Direct laboratory-based measurement of $|E^{\ast }|$ is time-consuming and requires trained operators and expensive equipment. In this paper, a modelling framework based on the artificial neural network (ANN) algorithm is proposed for the estimation of the $|E^{\ast }|$ of asphalt mixtures. The architecture of the ANN algorithm significantly influences its prediction accuracy. To determine the optimised architecture of the ANN algorithm and to identify the effective variables, grey wolf optimizer (GWO) was used. A large dataset containing experimental results from different laboratories was used to develop the $|E^{\ast }|$ predictive models. The dataset included information on aggregate gradation and characteristics, volumetric properties, binder properties, test conditions, and reclaimed asphalt pavement (RAP) content. The results show that a hybrid ANN and GWO can successfully provide a framework for the estimation of $|E^{\ast }|$ with the Pearson correlation coefficient (Pearson’s R) above 0.98. A graphical user interface (GUI) was also developed as a user-friendly environment that can be easily used by the pavement engineers for predicting the $|E^{\ast }|$ values.

KEYWORDS:

• Asphalt mixture
• dynamic modulus
• grey wolf optimizer (GWO)
• artificial neural network (ANN)
• graphical user interface (GUI)

Disclosure statement

No potential conflict of interest was reported by the author(s).