To read this content please select one of the options below:

Prognostication of half-cell potential for slabs cathodically protected with AZ91D using explainable and interpretable machine learning

Shikha Pandey (Department of Civil Engineering, Jaypee University of Engineering and Technology, Guna, India)
Yogesh Iyer Murthy (Department of Civil Engineering, Jaypee University of Engineering and Technology, Guna, India)
Sumit Gandhi (Department of Civil Engineering, Jaypee University of Engineering and Technology, Guna, India)

Anti-Corrosion Methods and Materials

ISSN: 0003-5599

Article publication date: 26 November 2024

6

Abstract

Purpose

This study aims to investigate the use of 20 commonly applied regression methods to predict concrete corrosion. These models are assessed for accuracy and interpretability using SHapley Additive Explanations (SHAP) and Local Interpretable Model-agnostic Explanations (LIME) analysis to provide structural health monitoring prognostic tools.

Design/methodology/approach

This study evaluated model performance using standard measures including root mean square error (RMSE), mean square error (MSE), R-squared (R2) and mean absolute error (MAE). Interpretability was evaluated using SHAP and LIME. The X and Y distances, concrete age, relative humidity and temperature were input parameters, whereas half-cell potential (HCP) values were considered output. The experimental data set consisted of observations taken for 270 days.

Findings

Gaussian process regression (GPR) models with rational quadratic, square exponential and matern 5/2 kernels outperformed others, with RMSE values around 16.35, MSE of roughly 267.50 and R2 values near 0.964. Bagged and boosted ensemble models performed well, with RMSE around 17.20 and R2 values over 0.95. Linear approaches, such as efficient linear least squares and linear SVM, resulted in much higher RMSE values (approximately 40.17 and 40.02) and lower R2 values (approximately 0.79), indicating decreased prediction accuracy.

Practical implications

The findings highlight the effectiveness of GPR models in forecasting corrosion in concrete buildings. The use of both SHAP and LIME for model interpretability improves the transparency of predictive maintenance models, making them more reliable for practical applications.

Social implications

Safe infrastructure is crucial to public health. Predicting corrosion and other structural problems improves the safety of buildings, bridges and other community-dependent structures. Public safety, infrastructure durability and transportation and utility interruptions are improved by reducing catastrophic breakdowns.

Originality/value

This study reduces the gap between model accuracy and interpretability in predicting concrete corrosion by proposing a data-driven method for structural health monitoring. The combination of GPR models and ensemble approaches provides a solid foundation for future research and practical applications in predictive maintenance. This comprehensive approach underscores the potential of data-driven methods for predictive maintenance in concrete structures, with implications for broader applications in various industries.

Keywords

Acknowledgements

The author thanks Jaypee University of Engineering and Technology, Guna Department of Civil Engineering faculty and staff for technical support.

Funding: This research did not receive any specific grant from funding agencies in the public, commercial or not-for-profit sectors.

Conflict of interest: The author declares that they have no conflict of interest.

Data availability statement: The data that support the findings of this study are available from the corresponding author upon reasonable request.

Citation

Pandey, S., Murthy, Y.I. and Gandhi, S. (2024), "Prognostication of half-cell potential for slabs cathodically protected with AZ91D using explainable and interpretable machine learning", Anti-Corrosion Methods and Materials, Vol. ahead-of-print No. ahead-of-print. https://doi.org/10.1108/ACMM-09-2024-3094

Publisher

:

Emerald Publishing Limited

Copyright © 2024, Emerald Publishing Limited

Related articles