Reinforced concrete (RC) wall buildings are used worldwide for their demonstrated seismic performance in high seismicity areas. Presently, the nonlinear models used to calculate maximum roof drifts and acceleration during ground motions for these buildings are labor‐intensive and complex. This study proposes a machine learning‐based methodology to estimate these demand parameters. Random forests and neural networks were trained and tested on a dataset of over 160,000 ground motion analyses, assembled from 57 RC wall building archetypes modeled in OpenSeesPy, and each one subjected to 3000 ground motions of different intensities. Notably, the inputs use design information available from the building design process and ground motion parameters that are simple to estimate. The models obtained an R 2 coefficient of 0.72 and 0.76 predicting roof drifts and acceleration on the test buildings, respectively, underscoring its ability to provide rapid estimations of the response of RC wall buildings. The trained models can be accessed via the Python Package Index and a GitHub repository.
Arroyo et al. (Sun,) studied this question.