Damage Evaluation of RC Structures Using Simplified Wave Propagation Analysis and Machine Learning

Abstract

Local vibration testing can be used to identify cracks within reinforced concrete (RC) structures; however, achieving high accuracy in the damage evaluation is a challenge. The amount of wave propagation across a crack is affected by several factors such as aggregate contact at the crack face. This study proposes a crack model that simplifies those factors in the wave propagation analysis based on the finitedifference time-domain (FDTD) method. The simplification is achieved by blocking the wave propagation across macro-crack that have a width larger than approximately 0.1 mm. The proposed crack model is validated by comparing numerical analysis and experimental results. Furthermore, a machine learning classifier is applied to the experimental and analytical data to estimate the degree of damage in the RC beams. The analytical resonant frequencies show good agreement with the experimental results of local through-thickness vibration tests on the damaged RC beam specimens. In addition to the analysis, the cracks in the RC beams are well detected by machine learning. This study shows that the proposed crack model is effective for crack identification in local vibration testing. Furthermore, machine learning contributed to improving the accuracy of damage detection.