Ground Settlement-Induced Building Damage Assessment With Modified Lanczos Algorithm and Extreme Learning Machine

Abstract

Construction, tunneling, and other urban anthropogenic activities strain neighboring buildings through distortion and rotation on both the surface and underground, resulting in instability of the local geological structure. This may cause devastating structural damage to buildings. Therefore, quantitative assessment of building structural damage is essential for the safety of local communities. In this study, a novel data-driven approach was applied to assess the building damage risks in urban areas. Data collected from over 50 buildings adjacent to the construction site were analyzed. The extreme learning machine (ELM) algorithm was applied to predict building structural risks. A modified Lanczos algorithm was used to regularize the ELM and improve the overall prediction performance. The computational results demonstrate the robustness and efficiency of the proposed Lanczos algorithm-regularized ELM.