[3B3] A novel time-of-flight auto-diagnosis model for ultrasonic signal using a sliding window technique based on empirical mode decomposition
F Yang¹, Q Mao¹ and K Lam¹,²
¹The Hong Kong Polytechnic University, China
²University of Glasgow, UK
Ultrasonic signal analysis is an essential technique used in various fields, including non-destructive testing (NDT), structural health monitoring and medical diagnosis. Accurate auto-diagnosis of time-of-flight (TOF) for ultrasonic signal is crucial in defect locating, ultrasound imaging, etc. However, the requisite signal extraction for TOF estimation can suffer from noise interference in the time and frequency domain and there is a tendency for interruption by spikes in the ultrasound signal envelope for wave crest auto-location. In this study, a novel auto-diagnosis model for ultrasonic signals is proposed based on the empirical mode decomposition (EMD) technique and sliding window approach. The model utilises the EMD method to extract intrinsic mode functions (IMFs) and obtain time-frequency representations of the signal. Then, the ultrasound signal envelope is extracted by Hilbert transform. The sliding window technique is employed to monitor the time-varying characteristics of the ultrasonic signal and capture the changes in TOF. A spline interpolation algorithm was embedded in the sliding window technique for reducing TOF error and computer burden. The proposed model can automatically estimate TOF for the ultrasonic signal without professional assistance. The performance of the proposed model is evaluated using steel plates with a thickness of 10 mm and pipelines with a wall thickness of 7 mm. The results demonstrate high accuracy in detecting signal abnormalities, with maximum errors of 0.25% and 1.25% for steel plates and pipelines under constant pressure, respectively. The proposed approach holds promising potential for fully automated applications in various fields in the future, such as structural health monitoring and medical diagnosis.
