Overview
The lifecycle of asphalt roads is strongly tied to their load bearing capacity and thus important to cost-benefit optimization of maintenance and management. Its determination and analysis require structural information such as the number, thicknesses, and extent of layers (asphalt, granular base). For the non-destructive detection of the layer geometry, IMP Bautest AG successfully deploys ground-penetrating radar (GPR) methods. The gathered information is used to analyze deflection measurements and to derive elastic moduli of individual road layers for the purposes of quality control, rehabilitation measures, inventory control, and as input for building information modeling (BIM). However, to date, the detection of layers in the recorded data is a challenging and time-intensive manual task done by experts whose knowledge and experience is key for solid results. Our idea is to apply new methods of machine learning (ML) to GPR data to (partly) automate the interpretation process and generate reliable and reproducible pavement layer models in a cost-effective way. This would enable us to obtain this critical information for the full road network of Switzerland and beyond for ASTRA, cantons and municipalities.
