Development of prediction models of high temperature crumb rubber modified binders

Abstract

The purpose of this research was to identify and model critical elements of the asphalt rubber matrix contributing to the crumb rubber modified (CRM) binder properties as measured by Superpave testing procedures. Current models are neither applicable (due to new testing procedures) nor practical (due to the difficulty of application or lack of accuracy). Therefore, this research investigated the development of predictive models for estimating binder properties within a specified range of accuracy.

Multiple crumb rubber sources were evaluated in order to gain an understanding of the differences between the various crumb rubbers. This analysis permitted identification of the important parameters, thus allowing for an accurate model to be developed. Analysis of the crumb rubber particles involved the determination of the glass transition temperature, chemical and visual analysis by scanning electron microscope.

Once the critical parameters were established, other test data was obtained from research projects conducted at other labs. This allowed a broader model to be developed, a model which would not be specific to one specific tester and lab facility. Ultimately a total of 17 virgin binder sources from 10 separate regions were evaluated; a further 12 crumb rubber sources were used in conjunction with the various binder sources.

The nonlinear empirical models for estimation of CRM binder properties were developed using the nonlinear least squares method. The accuracy of the various models was evaluated by identifying 95% confidence intervals for the binder property estimation. Validation of the models was performed using a fractional factorial design with previously untested CRM binders and this step confirmed the accuracy of the various models.

Findings suggest that the effect of the crumb rubber in the binder tends to enhance the effects of the base binder. This indicates that, for CRM binder, the properties of the base binder typically have a greater influence on CRM binder properties than the properties of the crumb rubber used in the matrix. It was possible to develop empirical models depicting the changes in viscosities, G*/sinδ values, and failure temperatures. The Rubber coefficient for viscosity (R_cv) and G*/sin[special characters omitted] (R_cg) were important parameters when estimating CRM binder properties. These coefficients are representative of the effects of the various crumb rubbers on the binders; generally, it was seen that ambient crumb rubbers had higher R_cv and R_cg values than cryogenic rubbers.