Looking to the future: the next-generation hot mix asphalt dynamic modulus prediction models

Abstract

This paper describes an innovative approach related to the development of a new hot mix asphalt (HMA) dynamic modulus (|E*|) prediction model by employing the artificial neural networks (ANNs) methodology. Many studies have been conducted over the last 50 years related to the development of HMA |E*| prediction models based on the regression analysis of laboratory measurements. The current study is an attempt to replace the regression analysis with the ANNs that have proved useful for solving certain types of problems that are too complex, poorly understood or resource intensive to tackle using more traditional numerical and statistical methods. The ANN |E*| prediction models were developed using the latest comprehensive |E*| database that is available to the researchers (from the NCHRP Report 547) containing 7400 data points from 346 HMA mixtures. The ANN model predictions were compared with the 1999 version of the Witczak |E*| prediction model, which is included in the mechanistic-empirical pavement design guide (MEPDG) and the new revised version as well. The sensitivity of input variables to the ANN model predictions were also examined and discussed. The ANN |E*| models show significantly higher prediction accuracy compared with the existing regression models and could easily be incorporated into the MEPDG. This approach may lead to more accurate characterisation of the HMA dynamic modulus resulting in better performance prediction, thereby reducing the risk of premature pavement failure. Reference to this paper should be made as follows: Ceylan, H., Gopalakrishnan, K., and Reference to this paper should be made as follows: Ceylan, H., AbstractThis paper describes an innovative approach related to the development of a new Hot-Mix Asphalt (HMA) dynamic modulus (E*) prediction model by employing the Artificial Neural Networks (ANN) methodology. Many studies have been conducted over the last 50 years related to the development of HMA E* prediction models based on regression analysis of laboratory measurements. The current study is an attempt to replace the regression analysis with ANNs which have proved useful for solving certain types of problems too complex, too poorly understood, or too resource-intensive to tackle using more-traditional numerical and statistical methods. The ANN E* prediction models were developed using the latest comprehensive E* database that is available to the researchers (from the NCHRP Report 547) containing 7,400 data points from 346 HMA mixtures. The ANN model predictions were compared with the 1999 version of the Witczak E* prediction model which is included in the Mechanistic-Empirical Pavement Design Guide (MEPDG) and the new revised version as well. The sensitivity of input variables to ANN model predictions were also examined and discussed. The ANN E* models show significantly higher prediction accuracy compared to the existing regression models and could easily be incorporated in the MEPDG. This approach may lead to more accurate characterization of HMA d...

Advanced approaches to hot-mix asphalt dynamic modulus prediction

“…Recently, researchers at Iowa State University (ISU) developed a novel approach for predicting HMA |E*| using the ANN methodology based on the input parameters of the Witczak |E*| model (Ceylan et al 2007). In this paper, research efforts related to the development of a simple approach for predicting HMA |E*| using the ANN methodology based on the input parameters of the Hirsch |E*| model are documented.…”

Prediction of Reflection Cracking in Hot-Mix Asphalt Overlays

“…This dissertation provides two ANN (Artificial Neural Network) models which developed by Dr. Halil Ceylan (89)…”

Development of Artificial Neural Network Predictive Models for Populating Dynamic Moduli of Long-Term Pavement Performance Sections

“…In total, five data sets are used in this study: (a) Witczak, (b) FHWA Mobile Trailer, (c) NCDOT, (d) WRI, and (e) Citgo. These data sets involve a total of 1,131 mixtures, 22,505 data points, and 67 asphalt binders and are believed to be the most substantial mixture and binder modulus databases available today (5)(6)(7)(8). The majority of mixes in the database are lab mixed and lab compacted.…”