Comparison of different blending combinations of virgin and RAP-extracted binder: Rheological simulations and statistical analysis

Abstract

Abstract The linear viscoelastic (LVE) behaviour of several combinations of two binders, a straight run 35/50 bitumen and a RAP-extracted bitumen in various proportions (20%, 40% and 60% of RAP binder), was simulated. Two different methods were used to carry out the simulations, The first method consists of a blending rule from the 2S2P1D (2 Springs, 2 Parabolic elements, 1 Dashpot) rheological model, that was calibrated for straight run bitumen and RAP. The second approach is based on the self-consistent model proposed by Herve and Zaoui [1], considering the two binders as a layered composite sphere embedded in a homogenous medium. Both possible juxtapositions, 35/50 binder over RAP binder (“fresh over RAP”) and vice versa (“RAP over fresh”), were taken into account. All the simulations were compared with 2S2P1D simulations of experimental data obtained from DSR and Tension/Compression tests on samples of perfect blends of the binders, in the same proportions. The two methods yielded different results. Simulations obtained with the blending rule are remarkably close to 2S2P1D simulations of experimental data. “Fresh over RAP” simulations of the self consistent approach are closer than “RAP over fresh” simulations to the results obtained with the blending rule. A statistical analysis was also performed to compare all the considered combinations. Norm (|E*|) and phase angle (φ) of complex modulus at different frequencies (from 1\u202f×\u202f10−6 to 1\u202f×\u202f10100\u202fHz) and steady state viscosity (calculated from norm of complex viscosity, |η*|, at 1\u202f×\u202f10−100\u202fHz) were considered as responses. First, a block design approach was used to estimate that significant differences were obtained between all various simulations and 2S2P1D fits of experimental data. Then, the Turkey method was applied in order to compare all combinations with each other and to identify which pairs of combinations show significant differences. The results of the statistical analysis generally confirm the conclusions drawn from the visual judgment of the rheological simulations.