New Predictive Models for Viscosity and Complex Shear Modulus of Asphalt Binders

Abstract

The main purpose of this paper is to present the development of a set of predictive models for the viscosity and complex shear modulus of asphalt binders. The model development was aimed at overcoming the limitations of current models used in the Mechanistic-Empirical Pavement Design Guide (MEPDG), which has been developed under NCHRP Project 1-37A and refined under NCHRP Project 1-40D. A comprehensive study was completed at Arizona State University to conduct a number of asphalt binder tests and to finalize a large binder characterization database containing 8,940 data points from 41 different virgin and modified binders. This database was used to develop the new models. The first of the models developed in this study is a fully revised version of the widely known ASTM Ai-VTSi viscosity model. This new, revised model takes into account the loading frequency applied on the binder while accurately predicting the viscosity at a specific temperature and loading frequency from the Ai and VTSi values of a binder. The other two new models developed are capable of accurately predicting the dynamic shear modulus (|G*b|) and associated phase angle (δb) of the binder at a specific temperature and loading frequency from the Ai and VTSi values of the binder. The models have been found to be rational, unbiased, accurate, and statistically sound. It is hypothesized that because of mathematical structures similar to those used in the MEPDG, the models developed in this study can be incorporated easily into a future revision of the guide.

Comparison of Witczak NCHRP 1-40D & Hirsh dynamic modulus models based on different binder characterization methods: a case study

“…This model is simpler in form and number of inputs compared to Witczak model. However, it was reported to lose its prediction accuracy when used to predict E* values for Witczak database [4,7]. It was also reported by several researchers that Hirsh model, similar to Witczak model, yields significantly biased E* predictions at the extreme low and high temperatures [4,7,[11][12][13].…”

“…The first model is called "NCHRP 1-37A" which follows a sigmoid function and predicts E* of the mix as a function of mix aggregate gradation, volumetric properties, binder viscosity and loading frequency [3]. The second model which is the NCHRP1-40D is a modified version of the latest model developed by Bari and Witczak [4]. Following is a comprehensive overview of both Witczak NCHRP1-40D and Hirsh models.…”

“…Gb* is also a part of the Superpave system for the binder characterization. In 2007, this model was modified under the NCHRP 1-40D project and incorporated since version 0.95 of MEPDG under the name "NCHRP 1-40D Gb*-based E* model" [4]. The model was developed based on a large database containing 7,400 E* data points from 346 mixtures [4].…”

“…In 2007, this model was modified under the NCHRP 1-40D project and incorporated since version 0.95 of MEPDG under the name "NCHRP 1-40D Gb*-based E* model" [4]. The model was developed based on a large database containing 7,400 E* data points from 346 mixtures [4]. The E* database for Witczak model contains non-aged, short-term oven aged for 4 hours at 135 o C, plant aged mixes, field aged cores as well as rubber modified mixes.…”

“…Consequently, over years, researchers have tried to establish E* predictive models using a number of laboratory E* databases. The most well-known models are the ones developed by Professor Witczak and his students over the last few decades [3][4][5]. Hirsh model is another well-known model for E* prediction [6].…”