Neural network data analysis for ultrasonic non-destructive testing of main gas pipelines

Download article in PDF format

Authors: Borovskoy I. G., Matolygin A. A., Ilin E. P.

Annotation: The corrosion process stems pipeline wall thinning to the criti-cal values. Considering this the article aimed to look into a problem of detection a corrosion zone for a main gas pipeline metal wall using an ultrasonic non-destructive testing data. The proposed solution is to use one of the machine learning algo-rithms – convolutional neural network – to handle data obtained by an in-tube ultrasonic analyzer.

Keywords: ultra-sonic non-destructive testing, convolutional neural networks, neural network analysis