Consistency of in-situ brass corrosion in HCl solution image fluctuations and electrochemical potential noise revealed through NARX neural network

Abstract

Micro-scale corrosion imaging in conjunction with electrochemical potential noise (EPN) provides a promising, comprehensive corrosion monitoring technique that can precisely visualise specific aspects of corrosion activity. In this work, rapid, economical, and real-time monitoring through a self-designed device, captured simultaneously with the electrochemical potential and micrographs, was employed to analyse the corrosion activity of H57 brass in an HCl solution using the gray-level co-occurrence matrix (GLCM). Conducted simultaneously in the time-frequency domain, the continuous wavelet transform (CWT) analysis of the electrochemical potentials and image features shows EPN fluctuations corresponding to feature noises. Similarly, the consistency of them in-situ Brass corrosion in the HCl solution, as revealed through a nonlinear autoregressive exogenous (NARX) neural network. The results indicate that the image features extracted by GLCM can quantify the changes in corrosion images and correspond to the corrosion behaviour of brass. Notably, the anomalous peaks of feature noises accompany or lag behind EPN. Leveraging these image features, the optimized NARX model demonstrates a high predictive capability for EPN. These findings are expected to guide future corrosion monitoring techniques and even enable corrosion prediction via micron-scale corrosion images.

Keywords

• Electrochemical potential noise
• In situ observation
• Artificial neural network
• Gray-level co-occurrence matrix
• Wavelet analysis
• corrosion monitor