Predicting percent air voids content in compacted bituminous hot mixture specimens by varying the energy laboratory compaction and the bulk density assessment method

The research focuses on the calibration of air voids content prediction models in compacted bituminous hot specimens of dense-graded friction courses when bulk density is estimated using a) the saturated surface dry method and b) sealed specimens method adding paraffin wax; in particular, it was performed one model for bituminous specimens prepared in laboratory under an Impact Marshall compaction and another one for specimens subjected to the Gyratory Compactor energy. Fifty-six specimens were prepared using two mix design solutions studied by adopting the Marshall method. Before moving on to the calibration phase, each subset was filtered by adopting two filtering criteria for removing anomalous values: 3σ method and the Chauvenet criterion. Last one criterion was the most restrictive; it works under the assumption of a normal distribution of the measurements that was confirmed by the Kolmogorov-Smirnov test application for each subset. Two effective explanatory variables were recognized in the best-final structure of the models: the percentage of bitumen by weight of the aggregates and a volumetric property of the specimens. The weight of the coefficients of the explanatory variables changes into two prediction models due to different system of inner stresses induced by the Gyratory compaction energy compared to the Marshall energy. Validation procedures by preparing fifty-two others specimens reflecting the same characteristics as those used during the first step were then carried out. Residuals analysis based on some statistical parameters assessment (i.e. mean absolute residual and mean squared residual) and diagrams (i.e. cumulated squared residuals and performance diagrams), and one way ANOVA, have performed to confirm the effectiveness and reliability of the results.