Deep learning-integrated electromagnetic imaging for evaluating reinforced concrete structures in water-contact scenarios

Assessing the resilience of overall reinforced concrete (RC) structures against micro- and macrocracks, especially in water-contact scenarios, poses a significant challenge due to the minute nature of these defects. To address this, a data-driven framework is introduced. It utilizes electromagnetic-wave (EM-wave) spectrum-encoded images alongside hybrid machine learning-deep learning (ML-DL) for predicting overall structural conditions. This methodology is reinforced by rigorous experimental validation and advanced image-processing techniques. Applied to a dataset of 7078 images, categorized into undamaged and damaged RC structures, the framework demonstrates its effectiveness by achieving an 83% prediction accuracy and an 81% F1-score. The promising results highlight the effectiveness of the presented approach in evaluating dam structures, offering a viable alternative to traditional assessment methods.