Structural Damage Identification Using Machine Learning: A Case Study

Alcantara E.A.M.; Bong M.D.; Saito T. · AIP Conference Proceedings (2023) · DOI: 10.1063/5.0154368

Resumen

Making decisions by the government, owners, stakeholders, etc., immediately after an earthquake occurs is essential for either evacuating a building or resuming its activities to reduce the consequences produced by future earthquakes and minimize social disruption. The authors previously reported a methodology to predict the structural responses for damage-condition identification by using wavelet spectra as images in a convolutional neural network model. It was applied to a lumped mass model of a regular structural configuration building. In this study, the proposed methodology is improved and applied to a local government building of the Aichi prefecture in Japan called Tahara City Hall, which has an irregular structural configuration, hence a 3D frame model was used. Finally, the results show that the accuracy of the evaluation of the usability of the building was 90%, the accuracy of the evaluation of the total damage was 76.1%, the accuracy of the evaluation of the damage on each floor was more than 75%, and the R-square coefficient in the total comparison of the story drift was 0.825.

Instituciones: Universidad Nacional de Ingeniería