Modeling the long term aging of asphalt cement by artificial neural networks


Creative Commons License

Taşdemir Y., Taşdemir F.

2nd INTERNATIONAL CONFERENCE ON CIVIL AND ENVIRONMENTAL ENGINEERING, Nevşehir, Turkey, 8 - 10 May 2017, pp.147

  • Publication Type: Conference Paper / Summary Text
  • City: Nevşehir
  • Country: Turkey
  • Page Numbers: pp.147

Abstract

Asphalt cement is one of the main component of the asphalt mixtures due to the its

viscoelastic properties. The aging in asphalt cement occurs during the mixing and

construction process which is called short term aging and during the service life in the

pavement that is known as long term aging. As a result of aging the viscoelastic

properties of asphalt cement is changed and age hardening can accelerate distresses such

as fatigue, low temperature cracking and moisture damage. The long-term aging

prediction models have been developed in research studies over the years. The most well -

known of them is the global aging system (GAS) model. GAS model is used in the

Mechanical Empirical Pavement Design Guide.

In this study, multi-layer perceptrons (MLP) which is one of the artificial neural networks

(ANN’s) techniques was used in modeling long term aging of asphalt cements. The

fastest MLP training algorithm, that is the Levenberg-Marquardt algorithm, was used for

optimization of the network weights. The different ANN structures were tried in terms

of iterations and hidden layer numbers. The ANN (3,9,1) appeared to be most optimal

topology for prediction of the aged viscosity value of asphalt cement. The ANN results

were compared to GAS model results. Experimental data obtained from three different

studies reported in literature include asphalt pavements with different ages, loading

history and environmental conditions. The ANN model gave R2 coefficient of 0.99,

which was higher than the value of 0.93 obtained using the GAS model. The root mean

square error of ANN model was lower than the GAS model, indicating that the developed

ANN model is able to predict the aged viscosity value more accurately than GAS model.