AI-driven detection of alkali-silica reaction in concrete structures using feature-enhanced deep learning models (February 2025)

Wu, Yujie, Wu, Mengze, Cui, Tianyi, Lin, Jiani, Liao, Qingke and Shu, Jinqiu 2025. AI-driven detection of alkali-silica reaction in concrete structures using feature-enhanced deep learning models (February 2025). IEEE Access 10.1109/access.2025.3591865

PDF - Accepted Post-Print Version Available under License Creative Commons Attribution. Download (1MB)

Abstract

Concrete’s affordability, adaptability, and resilience make it a cornerstone of construction, yet its vulnerability to degradation, particularly Alkali-Silica Reaction (ASR), poses significant challenges. ASR induces cracking and structural instability, necessitating efficient detection methods to mitigate its impacts. This study explores the use of Artificial Intelligence (AI) techniques for ASR crack identification, employing image enhancement and advanced models such as ResNet-18, InceptionV3, and AlexNet. A dataset of ASR-affected images was developed and augmented through feature enhancement processes, improving crack visibility and classification accuracy. Among the tested models, InceptionV3 demonstrated superior performance with high accuracy and robustness. The findings reveal that AI-based approaches, combined with image enhancement, effectively identify ASR cracks without requiring structural surface treatment. This research offers a scalable, automated solution to ASR detection, advancing structural health monitoring technologies and contributing to the preservation of critical infrastructure.

Item Type:	Article
Date Type:	Published Online
Status:	In Press
Schools:	Schools > Engineering
Publisher:	Institute of Electrical and Electronics Engineers
ISSN:	2169-3536
Date of First Compliant Deposit:	28 July 2025
Last Modified:	28 Jul 2025 14:00
URI:	https://orca.cardiff.ac.uk/id/eprint/180075

Actions (repository staff only)

Edit Item