21. System for prestressed concrete (PC) sleeper evaluation of health, focusing on vertical cracks

In recent years, longitudinal cracks have been observed in some aging PC sleepers. It is essential to understand how these cracks impact sleeper performance and to develop an efficient inspection method.
Therefore, we conducted a survey of the actual condition of PC sleepers and developed a system that estimates the length of longitudinal cracks from images taken by cameras mounted on maintenance vehicles and other equipment, allowing us to assess the structural health of the PC sleepers.

A survey of the actual load-bearing capacity of 200 aged PC sleepers and 5 new PC sleepers on a commercial line revealed that although longitudinal cracks have only a small effect on bending load capacity, the pull-out load-bearing capacity of sleeper plugs tends to decrease with increasing length of longitudinal cracks (Figure 1).
To perform an evaluation of health based on this relationship, we developed a method using deep learning to automatically extract longitudinal cracks from PC sleeper top surface images and estimate their length (Figure 2). Furthermore, using the relationship between the estimated crack lengths and Figure 1, an automatic system was developed to evaluate the health of the PC sleepers (Figure 3).

For example, using this system, it is possible to automatically detect longitudinal cracks and evaluate the health of a large number of PC sleepers—tasks that were previously done via manual visual inspection (Figure 4). This improves safety and streamlines maintenance and management operations.