Evaluation of ANN-Based Dynamic Modulus Models of Asphalt Mixtures

Artificial neural network (ANN)-based dynamic modulus $|{\mathrm{E}}^{*}|$ models were evaluated on South Carolina’s asphalt mixtures, the majority of which contained recycled asphalt pavement (RAP). These ANNs contained similar input variables as the NCHRP 1-40D and Hirsch regression models and were implemented in the neural network toolbox of MATLAB version R2018b. Two previously published ANN-based $|{\mathrm{E}}^{*}|$ models were also evaluated on the same database. Most ANNs in the literature have been shown to predict $|{\mathrm{E}}^{*}|$ with good success; however, they have not been validated outside of their original studies. The results showed that (1) ANN-based $|{\mathrm{E}}^{*}|$ models performed significantly better than regression models; (2) ANNs with few input variables (either $V_a$, $V_{beff}$, and $G_b^{*}$ or VMA, VFA, and $G_b^{*}$) highly predicted $|{\mathrm{E}}^{*}|$ with ${\mathrm{R}}^2>0.99$ on testing; (3) ANNs can accurately predict $|{\mathrm{E}}^{*}|$ of recycled asphalt mixtures; (4) the validation performance of the two published ANNs on South Carolina’s asphalt mixtures was ranked fair; and (5) locally customized ANNs are more accurate in the estimation of $|{\mathrm{E}}^{*}|$ than globally calibrated ANNs or regression models.