The use of deep learning for a cost-effective tunnel maintenance

Every year 4.700 km of new tunnels are built with an annual growth value of 7%. This results in the fact, that the total amount of tunnels which must be inspected will increase in the future. Furthermore, their operation reliability must be guaranteed with safe and cost-efficient means. Today, tunnel assessment is mostly based on a slow and subjective human inspection process. Automatic defect detection will provide a more objective and quantifiable approach to the task of tunnel inspection. By manual inspection, it is difficult to assess anomalies objectively - especially cracks at an accuracy of 0.2 mm with random size and shape - and it is difficult to compare them to the historic state of previous assessment campaigns. Thus, it makes sense to aim at an automated detection procedure and a fully digitized workflow for tracking the defects over time. Recent developments in the field of artificial intelligence can be applied to the field of tunnel inspection and bring it to a higher degree of automatization.