Improvements in Noncontact Ultrasonic Testing of Rails by the Discrete Wavelet Transform

摘要

Noncontact methods to generate and detect ultrasonic waves for the testing of railroad tracks are being investigated by a number of researchers. While noncontact testing offers potential advantages over conventional contact testing, it generally suffers from a poor signal to noise ratio of the measurements. One powerful tool to filter out noise from ultrasonic signals is the discrete wavelet transform, which can be used in high speed applications due to its computational efficiency. In this paper, discrete wavelet transform filter bank processing is applied to two noncontact ultrasonic systems for rail testing, namely a hybrid laser/air coupled setup for the detection of transverse cracks and a purely air coupled setup for the detection of longitudinal cracks. The discrete wavelet transform proves effective in filtering noise from the signals by eliminating the need for averaging in the hybrid setup and reducing to three the number of averages needed in the air coupled setup. Data compression is an added product of the discrete wavelet transform processing. The study shows the importance of selecting the proper mother wavelet function for best processing performance. In particular, the shape of the wavelet should be similar to that of the signal being analyzed. The discrete wavelet transform emerges as a powerful tool to ease the transition to the field of noncontact ultrasonic rail testing.