A baseline-free approach for damage detection and severity assessment in concrete structures using modal curvature and outlier detection techniques

This study presents a novel baseline-free method for structural damage detection that integrates unsupervised outlier analysis, clustering techniques, and the Modal Curvature Method (MCM) to enhance the accuracy of identifying damage in concrete structures. Unlike conventional methods that require baseline data from undamaged states, the proposed approach operates exclusively on measurements from potentially damaged structures, making it suitable for aging infrastructure with unknown or unavailable baselines. The method combines density-based outlier detection, specifically the Local Outlier Probability (LoOP) algorithm with LOESS regression to identify and remove curvature anomalies. By analyzing deviations in mode shape curvature and systematically isolating outliers, the technique provides robust localization and characterization of both discrete and distributed damage while preserving critical boundary information. Numerical simulations and experimental validation on Ultra-High-Performance Fiber-Reinforced Concrete (UHPFRC) beams demonstrate the method’s ability to detect progressive distributed damage with high sensitivity and noise resilience. The approach is further tested experimentally on an UHPFRC beam for increasing levels of damage. Results demonstrate the effectiveness in detecting and characterizing damage progression in UHPFRC beams subjected to bending, even in the presence of significant noise.