Data Driven Framework for Structural Health Monitoring of Concrete using Elastic Wave Method

In this research, we explored the characteristics of elastic waves in relation to damage within heterogeneous and dispersive solid structures, specifically focusing on concrete. Current elastic wave-based Structural Health Monitoring (SHM) and Non-Destructive Evaluation (NDE) techniques, such as Acoustic Emission (AE) and Ultrasonic Testing (UT), are widely utilized for assessing damage in concrete structures. However, the linear nature of ultrasonic waves and the heterogeneous characteristics of concrete often limit the effectiveness of these methods, especially for detecting defects smaller than half wavelength. To address these limitations, we developed a data-driven method that was both numerically and experimentally validated through tests on concrete samples ranging from small cubes to large-scale slabs. Our approach to enhancing the detection of defects in concrete structures includes: (i) Integrating linear and nonlinear ultrasonics into a single measurement system to detect multi-scale subsurface defects.(ii) Combining supervised and unsupervised learning techniques to classify entire signal patterns as physical mechanisms with increased accuracy. (iii) Implementing SHAP (SHapley Additive exPlanations) analysis to identify the most relevant single AE parameter that represents damage, thereby reducing data size. (iv) Employing cluster-based analysis to select signals that show the highest similarity, improving AE source localization accuracy in heterogeneous media. This proposed data-driven framework demonstrated significant improvements in defect detection within concrete structures.