Electromagnetic wave-driven deep learning for structural evaluation of reinforced concrete strength

Abstract

Monitoring the performance of reinforced concrete structures, particularly in terms of strength reduction, presents significant challenges due to the practical limitations of traditional detection methods. This study introduces an innovative framework that incorporates a non-destructive technique using electromagnetic waves (EM-waves) transmitted via Radio Frequency Identification (RFID) technology, combined with two-dimensional (2-D) Fourier transform, fractal dimension analysis, and deep learning techniques to predict reductions in structural strength. Experiments were conducted on three reinforced concrete beam (RCB) specimens exhibiting various levels of reinforcement corrosion. From these, a dataset of 1,800 EMwave images was generated and classified into “normal” and “reduced strength” categories. These categories were used to train and validate a Convolutional Neural Network (CNN), which demonstrated robust performance, achieving a high accuracy of 0.91 and an F1-score of 0.93 in classifying instances of reduced structural strength. This approach offers a promising solution for detecting strength reduction in reinforced concrete infrastructures, enhancing both safety and maintenance efficiency.