Convolutional neural networks for ultrasound corrosion profile time series regression

• Published in: NDT & E international : independent nondestructive testing and evaluation Vol. 133; p. 102756
• Main Authors: Cantero-Chinchilla, Sergio, Simpson, Christopher A., Ballisat, Alexander, Croxford, Anthony J., Wilcox, Paul D.
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01.01.2023
• Subjects: Convolutional neural networks; Corrosion; Deep learning; Non-destructive evaluation; Ultrasound; Wall thickness; Deep learning; Corrosion; Convolutional neural networks; Wall thickness; Ultrasound; Non-destructive evaluation
• ISSN: 0963-8695; 1879-1174
• DOI: 10.1016/j.ndteint.2022.102756

A customised convolutional neural network (CNN) architecture is proposed in this paper to make estimations about the thickness values (minimum and mean) of corroded profiles from an ultrasonic time-series measurement (A-scan). The CNN architecture is determined after a hyper-parameter optimisation which leads to the best performing network. The model is trained using synthetic data and tested on both synthetic and experimental datasets. A comparison is made with (1) a state-of-the-art network for time series analysis and (2) conventional thickness estimation techniques such as peak envelope and threshold crossing. The proposed network provides an accurate estimation of the thickness values and outperforms the conventional techniques in both synthetic and experimental datasets. When compared to a conventional threshold-crossing technique for minimum thickness prediction, the proposed network is more consistent and less sensitive to changes in the threshold. •A CNN architecture for thickness estimation of corroded profiles is proposed.•The deep learning model uses raw ultrasonic time series as input.•The method is trained using FE simulated data from a wide parameter space.•More accurate and reliable thickness prediction than conventional techniques.