Automatic segmentation of time-of-flight diffraction images using time-frequency techniques - application to rail-track defect detection

Ultrasonic time-of-flight diffraction (TOFD) is now a well-established technique alongside other ultrasonic testing techniques for accurate defect sizing. The current practice in the rail industry for the inspection of the rail welds and fishplate areas involves elaborate and painstaking manual inspection using a number of different probes at different positions around the track. TOFD, on the other hand, allows this procedure to be automated, providing detection, sizing and classification. In TOFD inspection, only a small fraction of the collected data actually represents defects, whereas the majority of the data is considered redundant. The first of the current processing stages which relies heavily on a skilled operator, involves pointing out those image areas containing defect areas and suppressing others. Consequently, this process consumes considerable amounts of time and effort, apart from the fact that the existence of the human factor at this critical stage invariably introduces inconsistency and error into the interpretation. Novel time-frequency analysis techniques have been combined with an artificial neural network to characterise TOFD signals and extract distinguishable features to be used for the detection, classification and sizing of rail-track defects. It is anticipated that, coupled with the necessary processing algorithms, TOFD can be used for a comprehensive automatic inspection of the rail-track, particularly fishplate and weld areas (Figure 1) with satisfactory levels of accuracy and reliability.