A fundamental multiple-stage ANN-based sensitivity analysis to predict the dynamic modulus of hot mix asphalt considering the effect of confining stress

Dynamic modulus |E*|, of Hot Mix Asphalt (HMA) is a crucial parameter in the pavement design and analysis. The Witczak regression-based model adopted by the Mechanistic-Empirical Pavement Design Guide (MEPDG) could be considered as the most fundamental and widely used model to estimate the dynamic modulus of HMA. However, the effect of confining stress has not been considered in this model as an effective parameter. In this paper, attempts were undertaken to develop a new predicting model for |E*| of HMA considering the effect of confining stress. Artificial Neural Networks (ANNs) was administrated as the computational tool for this modeling using 1320 |E*| test results performed at the University of Maryland. Asphalt mix parameters, test frequency, temperature as well as the level of the confining stress were considered as model inputs. Also, intercept of temperature susceptibility relationship (A) and slope of temperature susceptibility relationship (VTS) represented the effect of binder viscosity features. The new model could predict the |E*| of HMA with high accuracy of R2 = 0.99. Fundamentally, a comprehensive multiple-stage ANN-based sensitivity analysis was developed to survey whether the high accuracy is enough for a model to be considered perfect. Or despite the high accuracy, there may be other significant factors which are not considered as model inputs. Also, this sensitivity analysis shows which of the inputs has a fundamental role in the model. Moreover, for evaluating the level of influence of selected inputs, another sensitivity analysis was performed by the r-Pearson method and the results indicate that the confining stress has the highest increasing and temperature has the highest decreasing effect on |E*| with a significant difference with other inputs. The Pearson results confirm the ANN-based sensitivity analysis results.