Physics-informed deep filtering of ultrasonic guided waves for incipient defect inspection of large-scale square tube structures

Abstract

For large-scale structure inspection, the traditional ultrasonic inspection methods are time-consuming and inefficient. Recently, many approaches based on ultrasonic guided waves have been developed. However, during the propagation of guided waves, as the propagation distance increases, the signal amplitude attenuates gradually, as a result, the reflected wave packets from the incipient defects are always seriously contaminated or even buried in noise, and the defect signatures cannot be effectively identified in this circumstance. To detect incipient defect as early as possible to ensure structure safety, it is very essential to suppress the background noise and enhance the reflected wave packet from defect. To address this issue, a physics-informed deep filtering method is proposed for ultrasonic guided waves enhancement. In the proposed method, a deep learning model is constructed and trained based on template signals of ultrasonic guided waves simulated by digital model with different parameters, and then generalized to enhancement of weak ultrasonic guided waves from practical cases. The simulation test and experimental results on large-scale square tube structures indicate that the proposed ultrasonic guided waves enhancement method is effective for early-stage defect inspection and outperforms the traditional guided waves enhancement approaches.