Synthetic GPR datasets to evaluate hybridization inverse approach for pavement tack coat characterization—Geometrical and physical parametric study

The visible surface degradation of pavements is often the result of underlying subsurface defects. The quality of the bond between the wearing course and the binder course is a key factor in limiting issues such as delamination, stripping, etc. The use of non-destructive testing (NDT) methods, such as electromagnetic wave propagation combined with a hybrid data processing approach-machine learning and Full-Waveform Inversion-has recently demonstrated its effectiveness [1]. To support this research, a database was created using the open-source software gprMax [2] for various central frequencies on different two-layered pavement structures, modeled in accordance with current French standards. Variations in geometrical and physical parameters were applied to the wearing course, tack coat, and binder course. The objective is to build a representative database for numerical validation of an innovative hybridization method (machine learning method combined to full-wave form inversion), facilitating the characterization of the tack coat located between the two layers that form the surface layer. The challenge is to discriminate the echo from a thin bituminous subsurface layer using a ground-coupled GPR system (near-field), despite the presence of constructive interference.