Predicting Dynamic Modulus of Asphalt Mixture Using Data Obtained from Indirect Tension Mode of Testing

TL;DR

This paper develops an accurate finite element model combined with neural network back-calculation to predict the dynamic modulus of asphalt mixtures from indirect tension tests, aiding pavement performance prediction.

Contribution

It introduces a novel integration of ANN and FE modeling for precise dynamic modulus prediction of asphalt mixtures based on laboratory and field data.

Findings

ANN effectively back-calculates elastic modulus.

FE model accurately predicts dynamic modulus.

Model validated with field core data from Minnesota.

Abstract

Understanding stress-strain behavior of asphalt pavement under repetitive traffic loading is of critical importance to predict pavement performance and service life. For viscoelastic materials, the stress-strain relationship can be represented by the dynamic modulus. The dynamic modulus test in indirect tension mode can be used to measure the modulus of each specific layer of asphalt pavements using representative samples. Dynamic modulus is a function of material properties, loading, and environmental conditions. Developing predictive models for dynamic modulus is efficient and cost effective. This article focuses on developing an accurate Finite Element (FE) model using mixture elastic modulus and asphalt binder properties to predict dynamic modulus of asphalt mix in indirect tension mode. An Artificial Neural Network (ANN) is used to back-calculate the elastic modulus of asphalt mixtures. The developed FE model was verified against experimental results of field cores from nine different pavement sections from five districts in the State of Minnesota. It is demonstrated that the ANN modeling is a powerful tool to back-calculate the elastic modulus and FE model is capable of accurately predicting dynamic modulus.