A combined threshold and signature match method for the automatic detection of rail RCF cracks using an ACFM sensor

Abstract

Detection of surface breaking defects, such as rolling contact fatigue (RCF) cracks, is an on-going topic of research within the context of rail inspection. At present, detection and classification of RCF cracks using an ACFM sensor is based on low speed walking stick systems where negligible sensor lift-off changes result in a reasonably stable background signal and hence the defects can be automatically detected using a threshold method. However, in the case of high speed inspection systems, the inevitable lift-off variations (e.g. owing to the dynamics of the train bogie etc.) lead to a varying background ACFM signal which renders the threshold method ineffective. A novel method for the detection of isolated RCF cracks, namely combined threshold and signature match (CTSM), has previously been applied to ACFM scans over RCF cracks. The method proved to be effective for automatic detection of isolated RCF cracks, however, it was observed to perform poorly in response to areas of multiple RCF cracks, which often appear in clusters. This paper investigates the application of an enhanced CTSM algorithm to the detection of clustered RCF cracks. The algorithm has been applied to low speed scans over sections of rails removed from service containing real RCF cracks and to ACFM scans obtained at high speed (up to 48 km/h) over a spinning rail rig containing clusters of artificial cracks. Results suggest that the extended CTSM algorithm is effective in automatically detecting multiple RCF cracks with a high detection rate (> 90%). Further, in the case of widely spaced (> 5 mm) multiple cracks, the algorithm can also provide extra characterisation information about the number and position of cracks within a cluster.