Prediction of air voids of asphalt layers by intelligent algorithm

The objective of the study was to give a suitable prediction method of air voids of asphalt layers in the process of construction. Seven different methods are utilized to predict the air voids of asphalt layers: nonlinear data fitting, Back Propagation neural network (BPNN) algorithm, Radial Basis Function neural network (RBFNN) algorithm, support vector machine for regression (SVR), Gaussian process regression (GPR), regression trees, and random forest regression. The results of laboratory experiments and field tests showed that the intelligent algorithm of SVR is more accurate and suitable for estimating the air voids of asphalt layers. The compaction quality of asphalt layers can be evaluated by this proposed new prediction method of air voids.

Automated summary

The study used multiple methods to predict air voids in asphalt layers, with results showing that the SVR algorithm is more accurate for estimating air voids. The new prediction method can evaluate the compaction quality of asphalt layers.