Application of artificial neural network models for predicting the resilient modulus of recycled aggregates
Parisa Rahimzadeh Oskooei, Alireza Mohammadinia, Arul Arulrajah, Suksun Horpibulsuk
Abstract
In recent years, efforts have been made to utilise construction and demolition (C&D) wastes as an alternative material to natural quarried aggregates in the structural layers of railways and roads. Resilient modulus (Mr) is one of the crucial design parameters used for the construction of roads and railways. Undertaking resilient modulus testing is expensive, time-consuming and complex. This study employs artificial neural network (ANN) as a robust method for representing two separate models for predicting the resilient modulus of bound and unbound C&D materials. A comprehensive database has been collected from past publications for developing the models. In order to consider the essential factors on determination of resilient modulus, two representative parameters are proposed. Several statistical criteria were utilised to evaluate the precision and performance of the predictive models, and the best models were transformed into practical equations. A sensitivity analysis was undertaken to determine the impact of each input parameter in the proposed models. The results indicated the applicability and efficiency of the ANN method for predicting the resilient modulus of unbound and bound C&D aggregates.