Detection of early squats by axle box acceleration
This thesis discusses a new method for detection of short track irregularities, particularly squats, with axle box acceleration (ABA) measurements.
A squat is a surface initiated short track defect, associated with high frequency vibrations of the wheel-rail system. High stresses in the contact patch at squats cause accumulation of plastic deformation of the rail and growth of cracks. Cracks growing in the subsurface can cause a rail fracture. Light squats can be treated by grinding of the rail surface; while mature squats lead to replacement of the rail section. For cost effective maintenance policy and operational safety squats should be detected at an early stage. Detection of light squats is the main aim of this study.
Till now, ultrasonic measurements have been mainly used for detection of squats. By that method the depth of cracks is measured; hence, it is applicable only to detection of severe squats with sufficiently large cracks. In the present work, ABA measurements were employed. The advantages associated with this method are that ABA measurements can be performed on standard operating vehicles travelling with usual traffic speeds and squats can be detected at their early stage.
The first goal of this study was to find a relationship between squats and ABA characteristics, such as magnitude and frequency content, and apply them for detection of squats. To this end, a three-dimensional finite element (FE) model was applied for dynamic simulations of the wheel-track interaction in the high frequency range. By parameter variation, the influence of the geometry of squats, speed of the train and location of the squats relative to the sleepers on ABA characteristics was studied. Local frequency characteristics of ABA at squats were obtained and their relation with the severity of squats was established. These frequency characteristics can be applied for detection of squats and their assessment.
The second goal was to improve the signal-to-noise ratio of ABA measurements to enable detection of light squats. Several methods to improve signal-to-noise ratio of ABA measurements were suggested. These included noise reduction techniques, reduction of disturbances from the wheel defects and signal enhancement by improvement of the measuring system by using longitudinal ABA. Owing to the improvement of the signal-to-noise ratio, the hit rate of moderate squats increased from 60% to 100% and the hit rate of light squats together with trivial defects (trivial defects are small rail surface defects which are so small that they will be worn away, and will therefore not grow into squats.) increased from 57% to 85%. Since light squats are larger than trivial defects and, therefore, easier to detect, the hit rate of light squats, which depends on the threshold that separates light squats from trivial defects, is higher.
The third goal was to develop an algorithm for automatic detection of squats, which enables continuous analysis of track. The initial results indicated that 78% of light squats and trivial defects can be detected automatically by ABA. The hit rate of severe squats was 100%.
The presented ABA method enables automatic detection of squats at their earliest stage, when preventive and early corrective actions can be taken. The employment of such method can significantly reduce life cycle costs of a track infected by squats.Full text: doi:10.4233/uuid:ee403235-2ddf-4ce3-9f6d-e0775448f428