Structural Damage Identification: A Review of Deterministic and Uncertainty-based Methods

Abstract

Structural damage identification is vital for ensuring the safety and predictive maintenance of large-scale civil infrastructure. This review systematically compares two methodological paradigms: deterministic methods—such as dynamic fingerprinting, model updating, and data-driven techniques—and uncertainty-based methods, including Bayesian inference and interval analysis. A comparative framework is established by analyzing their respective capabilities in handling measurement noise, model error, and incomplete data. Persistent challenges are examined, including environmental variability, nonlinear structural behavior, and sensor deployment limitations. The paper further identifies emerging directions, such as physics-informed intelligent algorithms and digital twin systems, which bridge theoretical advances with engineering practice. Aimed at researchers and practitioners, this work provides a structured reference to support the transition of damage identification from theory to real-world implementation.