Designing a Deep Neural Network for an Acousto-Ultrasonic Investigation on the Corrosion behaviour of CORTEN Steel

The integrity of the CORTEN steel exposed to corrosion agents is studied using Acousto-Ultrasonic approach. The Acousto-Ultrasonic approach was tested before and after exposing the CORTEN steel to the corrosion agent. The waveforms recorded from the Acousto-Ultrasonic tests are analysed using Mel Spectrogram. The attenuation in the wave propagation due to the extent of corrosion and due to the geometrical configuration of the CORTEN steel test specimens is studied in time-frequency domain. The Mel scale is used for analysing the time-frequency characteristics of the recorded waveforms. A Deep Learning Neural Network is constructed for analysing the waveforms recorded from the Acousto-Ultrasonic test. Deep learning is used to classify the characteristic acoustic emission waveforms propagated through the CORTEN steel and attenuated due to the geometrical configuration and corrosion formation. The Convolutional Neural Network (CNN) is built in MATLAB (R) and is trained to classify the acoustic emission waveforms recorded from the Acousto-Ultrasonic test. (C) 2022 The Authors. Published by Elsevier B.V.

Automated summary

The integrity of CORTEN steel exposed to corrosion agents was studied using the Acousto-Ultrasonic approach. The attenuation in wave propagation due to corrosion and geometrical configuration of CORTEN steel was analyzed in the time-frequency domain using Mel spectrogram. A Deep Learning Neural Network was employed to classify acoustic emission waveforms propagated through the CORTEN steel and attenuated due to corrosion formation and geometrical configuration.