Top-down cracking (TDC) is a distress affecting asphalt pavements and consists of longitudinal cracks that initiate on the pavement surface and propagate downwards. Such a distress is critical especially for thick asphalt pavements with open-graded friction courses (OGFC), which are common on motorways and high-speed roads. Nevertheless, many road agencies are not fully aware of the TDC issue yet and thus do not have adequate tools to detect TDC. Within this framework, as part of a larger project, this study proposes an automatic method for the recognition of TDC on the pavement. The tool developed is based on machine learning (ML) algorithms and allows to identify TDC from the analysis of pavement images. The main output provided by this tool is the information on the presence/absence of TDC on the pavement, with the related confidence level. The labeling and training of the algorithm were carried out on images of a significant portion of the Italian motorway network (400 km) that were subjected to a non-automatic visual analysis in a previous phase of the project. The algorithm was then validated considering a further 100 km trial section belonging to the Italian motorway network, from which several control cores were taken. The tool developed has the potential to be used in a pavement management system (PMS) to plan timely surface repairs/maintenance against TDC, especially when combined with a model able to predict TDC depth evolution over time.

Road pavement; Top-down cracking (TDC); Pavement management system (PMS); Condition assessment; Maintenance; Machine learning.