Predicting fatigue performance of hot mix asphalt using artificial neural networks

Developing predictive models for fatigue performance is a complex process and can depend on variables including material properties, test conditions and sample geometry. Several models have been developed in this regard; some of these are regression models and are related to mechanistic properties i...

Full description

Saved in:
Bibliographic Details
Published in:Road materials and pavement design 2017-05, Vol.18 (sup2), p.141-154
Main Authors: Ahmed, Taher M., Green, Peter L., Khalid, Hussain A.
Format: Article
Language:English
Subjects:
artificial neural network
Artificial neural networks
Asphalt
Dissipation
DSR technique
fatigue performance
Fatigue tests
Granite
hot mix asphalt
Hot rolled asphalt
Limestone
Materials fatigue
Mathematical models
Neural networks
Performance prediction
Properties (attributes)
Regression models
Roads & highways
Rolled asphalt
Stiffness
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Developing predictive models for fatigue performance is a complex process and can depend on variables including material properties, test conditions and sample geometry. Several models have been developed in this regard; some of these are regression models and are related to mechanistic properties in addition to volumetric properties. In this work, a computational model, based on artificial neural networks (ANNs), is used to predict the fatigue performance of hot mix asphalt (HMA) tested in a dynamic shear rheometer (DSR) technique. Fatigue performance was evaluated according to three approaches: traditional, energy ratio and dissipated pseudo-strain energy. For predicting fatigue performance, two types of ANN models were developed: those dependent on test modes, that is, based on controlled test modes, and those independent of test modes, that is, irrespective of controlled test modes, using fundamental parameters, for example, stiffness modulus, phase angle and volumetric properties. In this work, limestone (L) and granite (G) aggregates were used with two binder grades (40/60 and 160/220) to prepare four mixtures with two different gradations: gap-graded hot rolled asphalt (HRA) and continuously graded dense bitumen macadam (DBM). The results revealed an excellent correlation between the predicted and experimental data. It was found that the prediction accuracy of the strain test mode was better than that of the stress test mode.
ISSN:1468-0629
2164-7402
DOI:10.1080/14680629.2017.1306928