Abstract
Weld connections are critical elements in steel structures that are vulnerable to damage from fatigue, corrosion, or crack propagation. Therefore, monitoring these connections for damage is crucial to ensure long-term structural integrity. Conventionally, frequency-based modal analysis approaches have been used to detect the weld damage in steel connections. However, the localization and severity estimation of such damages need more comprehensive exploration. Therefore, this study proposed a novel dynamic strain-based sensitivity analysis methodology with a sparse regularization technique to identify and localize weld crack damage in asymmetrical steel portal frames. A full-scale simulation of the frame structure, representing the actual instrumented structure, is performed using commercial finite element software to produce strain data with and without weld cracks. Additionally, a simplified finite element model is employed to express the stiffness reduction as FE variables. The change in strain values and FE variables is utilized to create an objective function, and solved using a regularization technique. Single and multiple damage cases, including cracks in welded connections, are also simulated to obtain damage state sensor data. The proposed methodology demonstrates that it can identify dual and multiple damage cases of weld cracks. An experimental validation employing a laboratory-scale steel portal frame is also conducted. The proposed damage detection methodology identified 2 out of 4 damaged members for a 2 mm damage case and 3 out of 4 damaged members for a 3 mm damage case, thereby correctly identifying the majority of damaged members for accurate localization of the weld crack. This vibration-based methodology has effectively identified damaged weld joints when damage is localized to a single connection, offering potential as an early warning system for critical steel structures with welded connections.
Get full access to this article
View all access options for this article.