A novel defect recognition method of ultrasonic guided wave monitoring signals based on optical flow estimation

Ultrasonic-guided waves (UGWs) can be used to monitor the long-term health status of large-sized engineering structures. However, the recognition of minor defects in UGW monitoring signals is still challenging, especially in the case of temperature drift. Here, we presented a novel method of ultrasonic wave monitoring signals based on the principle and algorithm of optical flow estimation. We introduced a set of operation steps, quantitative indices, and characteristic analysis. With two sets of actual measured signals from a pipeline and a switch rail, we validated that the proposed method’s indices had the high sensitivity to minor defects than time domain indices. Based on the processing results of signals with diverse structures, detection frequencies, and defect positions, we also demonstrated the adaptability of the proposed method. Furthermore, we classified the characteristics of the optical flow field distribution with respect to temperature drift and defect occurrence. The results indicated that the local perturbation characteristics in the optical flow field were still valid in the temperature drift status, whereas the classical baseline signal stretching method requires a precise temperature. The proposed optical flow-based method demonstrates great engineering potential.