Safety Assessment Method for Engineering Structures Based on Modal Curvature

To address the limitations of traditional structural damage identification methods in terms of reliance on high-fidelity baseline models and sensitivity to minor damage, this paper proposes a novel physics-informed and data-driven approach based on the modal curvature variation coefficient. A damage-sensitive feature derived from the rate of change in the radius of curvature is established, providing a clear mathematical and physical interpretation to reduce model error interference and enhance local damage localization. The effectiveness of the proposed method is validated through a 1:20 scale model experiment of a main truss from a large stadium steel roof. A total of 33 experimental cases were designed, simulating single and multiple damage scenarios with varying severity levels (large, medium, and small). Multi-source monitoring techniques, including millimeter-wave radar interferometry, laser displacement sensors, high-resolution vision-based measurement, and accelerometers, were integrated. Modal parameters were extracted using the Stochastic Subspace Identification (SSI) method, and the finite element model was updated via a high-order response surface methodology. Numerical simulations and experimental results demonstrate that the proposed modal curvature variation coefficient is highly sensitive to local stiffness degradation and accurately locates both single and multiple large/medium damage regions. In cases involving multiple minor damages, the method effectively identifies the damaged areas but exhibits a risk of false positives in undamaged sections. The millimeter-wave radar measurements exhibit strong agreement with laser displacement data, confirming its viability for non-contact structural health monitoring. This research provides a robust technical framework and experimental foundation for condition assessment and early damage warning in large-scale engineering structures.