Selection of suitable earthquake ground motions for the development of best performing AI model for seismic structural health monitoring of shear wall buildings

This study highlights the importance of selecting suitable earthquake ground motions to enhance Artificial Intelligence (AI) models used in seismic structural health monitoring (SSHM) for shear wall buildings. The primary focus is identifying and selecting ground motion records that directly optimise AI model performance. A diverse dataset of earthquake ground motions, representing realistic seismic conditions, was compiled and applied to simulate the dynamic response of shear wall buildings with varying structural properties. These simulations were utilised to train, validate and test AI models designed to detect, predict and assess structural damage. Specifically, the models evaluate key structural parameters, including maximum displacement, maximum interstorey drift (MISD), base shear and maximum overturning moment under seismic loading. The research evaluates specific ground motion selection criteria—such as Peak Ground Acceleration (PGA), Peak Ground Velocity (PGV), Peak Ground Displacement (PGD), time duration and frequency content—to ensure optimised input data. The findings demonstrate that carefully selected ground motions significantly improve the accuracy, generalisation and robustness of AI models in identifying structural damage. Ultimately, this work advances SSHM systems for shear wall buildings and contributes to more effective, reliable post-earthquake damage evaluations in active seismic zones.