TACK-II: an AI Framework for Automated Tunnel Inspections and Assessment

This research project is a continuation of the recently finalized project TACK - Tunnel Automatic CracK Detection, which used a hybrid approach of deep learning and photogrammetry to show the feasibility of automatically detecting and measuring the width of cracks in the concrete tunnel lining. In TACK, a limited amount of data from three tunnels in Sweden were used as a small proof of concept.

The aim of the proposed project is a proof of concept on a large scale.

This includes collecting data from three to four tunnels and autonomously detecting and visualizing the location of cracks in the concrete lining.

A framework for digital inspections, particularly a method to autonomously assess the risk associated with cracks, will be developed. This framework will be used to assess the structural condition of the tunnel.

Lastly, the most important step of this project is to compare the results from digital inspection with results from human in-situ inspections.

Here, inspection accuracy, time and cost and knowledge transfer between inspectors and owners should be evaluated systematically. This is important to show the proposed methodology´s capability and take it one step closer to implementation.