Fine-Grained Seismic Damage Assessment of Buildings Based on Machine Learning and Evidential Reasoning

Existing seismic damage assessment methods for buildings often suffer from coarse results (based on groups of buildings), complex processes, and poor adaptability. To address these limitations, this paper leverages the advantages of machine learning in complex data processing and algorithmic adaptability, integrates expert knowledge, and combines qualitative and quantitative information to propose a fine-grained seismic damage assessment method for individual buildings based on machine learning and evidential reasoning. Validation using historical seismic damage cases demonstrates that this method enables rapid, accurate, and intelligent assessment of building seismic damage. It outputs all possible damage levels and corresponding probabilities for individual buildings under the combined effects of multiple seismic factors. These quantitative, fine-grained results provide critical support for pre-earthquake risk perception, post-earthquake rapid damage mapping, and precision disaster management.