Early detection of rail defects

We developed a method to detect rail defects in an early stage: We use ARRoW-measurements (sound measurements in close proximity of the wheel) to detect squats. This turned out to be a viable way to determine squats fast, accurately and at an early stage.

For infrastructure managers, timely correction of rail defects is a very important issue. Late detection of defects can lead to high maintenance costs or worse, failure of the track. Therefore, the railway network owner wants to detect damages as early as possible. One common defect type is the so-called squat. In cooperation with TU Delft and BAM Rail we developed a method to use sound measurements in close proximity of a wheel with our ARRoW system to detect squats.


Squats are rail defects on the running band on the rails. They are classified as A, B and C with increasing severity.  Squats are normally detected by ultrasonic measurements but this has the disadvantage that squats cannot be detected until the crack depth is after a certain point. It might then be too late to take adequate measures. Another way to detect squats is by visual inspection. This is labor-intensive and only allowed when the track is out of service. It is not feasible to use this method alone to inspect squats over a larger network of train rails.  There is thus need for a method that can detect squats fast, accurate and in time.

Sound measurements

Often you can hear a wheel run over a squat in a very early stage of the growth of the squat. This is a characteristic impact noise which differs significantly from the ordinary rolling noise.  It is a strong increase of the sound level.  We use ARRoW to record the sound level in close proximity of the wheel/rail interface and we developed an algorithm to detect the squats. The algorithm searches for rises of the overall sound level. If the value is above a certain threshold value, the location is marked as a squat.

Detection algorithm

The algorithm we developed is suitable to classify the B and C-squats. The squats with the more severe damage on the rail head. The less severe squats are more difficult to detect: they are smaller in size, so they are not always run over and if they are run over the cause less impact noise.  The results so far are very promising but we do see room for improvement.

April 2015

Actueel type:

TextureInspector was tested in the field during April 2015

This summer M+P has launched a new measurement system: FLaSH. FLaSH determines the pavement texture...