A finite element model-based two-level method incorporating recommendation algorithm for structural damage detection

• A FEM-based two-level approach combining pattern matching and model updating for hierarchical damage detection. • Element grouping and automatic clustering techniques for identifying potentially damaged regions in the first level and reducing the dimensionality of subsequent problem. • Element damage assessment in the second level by model updating with pattern matching-determined iteration and one-at-a-time updating strategy to address challenges in model updating. • Pattern matching enhancement with recommendation algorithm to guide damage localization and parameter adjustment. • Validation through numerical simulations with the ASCE Benchmark model and experimental studies on the truss bridge model. Accurate and efficient assessment of structural damage is crucial for the safety and longevity of civil engineering structures. This paper presents a finite element model-based, two-level approach that combines pattern matching and model updating for hierarchical damage detection. The recommendation algorithm is utilized to enhance the pattern matching process. In the first level, elements are classified into several groups, and potentially damaged groups are identified using pattern matching principles and automatic clustering methods, reducing the dimensionality of subsequent problem. In the second level, the extent of the damage of the elements within these groups is assessed based on model updating principles, with iteration direction and step length determined by pattern matching, while a one-at-a-time updating strategy addresses the ill-posed challenges in model updating. Throughout both levels, the recommendation algorithm, functioning as a relaxed pattern matching method, guides the localization of damaged regions and the adjustment of damage parameters of elements based on maximum similarity, effectively reducing the computational costs and inefficiencies of traditional pattern matching operations. Validation through numerical simulations on the ASCE Benchmark model and laboratory experiments on a 6-meter span truss bridge model demonstrates that the method accurately identifies damaged regions and quantifies the extent of the damage with high precision using limited natural frequencies and mode shapes, enhancing efficiency and accuracy compared to conventional approaches.