Multi-Level Substructure-Based Model Updating for Structural Damage Detection

Finite element model updating (FEMU) is critical for structural damage identification. However, applying it to large intricate structures with abundant structural parameters is often computationally prohibitive. The substructuring method enhances computational efficiency by combining model reduction with a divide-and-conquer strategy. Its integration with FEMU can significantly accelerate the updating procedure. Nevertheless, even substructure strategies remain computationally inefficient when abundant unknown parameters are updated. This study develops a multi-level substructure-based FEMU framework for efficient and accurate structural damage identification. The proposed FEMU method first partitions the entire structure into a series of local substructures. It then condenses substructures using modal reduction and assembles them into a reduced-order global model. Subsequently, substructure-level updating is applied to localize damaged substructures. This is followed by element-level updating within the identified substructure to pinpoint specific damaged elements. Numerical validation on a frame and a high-rise building demonstrates that the proposed multi-level framework drastically improves computational efficiency compared to the conventional one-level substructure-based method. It decreases the parametric quantity in each iterative updating step and simultaneously preserves a high damage detection accuracy.