Improvement of the seismic resilience of regional buildings: A multi-objective prediction model for earthquake early warning

Earthquake early warning is one of the methods by which to improve structural resilience. However, uniform earthquake warning information is not highly targeted for regional buildings. Therefore, this work proposes a multi-objective earthquake warning process to provide more targeted earthquake avoidance measures for regional buildings. A multi-objective prediction model based on long short-term memory (LSTM) is established. The "Di Ting" dataset provided by China's National Earthquake Data Center is currently the largest artificial intelligence seismology training dataset. The proposed model uses 3 s of P-wave data from the dataset to predict the single-station magnitude, epicenter distance, and peak ground acceleration (PGA) of earthquakes. To achieve better predictive performance, the mean squared error (MSE) is redesigned. The average prediction errors of the magnitude and epicenter distance are found to be reduced, and the multi-objective model can achieve good predictive performance and meet the requirements of all the objectives. The predicted magnitude is used to determine whether an earthquake warning should be issued, the epicenter distance is used to determine the response time, and the PGA is used to determine the engineering status of the building. More targeted risk avoidance measures can be chosen for different building based on their disaster situation. The proposed multi-objective warning process can provide more targeted advice on how to avoid risks to buildings and reduce unnecessary panic and casualties. This is a useful attempt to improve the seismic resilience of regional buildings.