Early Peak Ground Acceleration Prediction for On-Site Earthquake Early Warning Using LSTM Neural Network

On-site earthquake early warning techniques, which issue alerts based on seismic waves measured at a single station, are promising, and have performed quite successfully during some damaging earthquakes. Conventionally, most existing techniques extract several P-wave features from the first few seconds of after trigger to predict intensity or destructiveness an incoming earthquake. This type technique neglects behavior temporal varying within P waves. In other words, characteristics data sequences not considered. this study, long short-term memory (LSTM) neural network, is capable learning order dependence in waves, employed peak ground acceleration (PGA) coming A dense LSTM architecture proposed large set earthquakes used train model. The general performance model indicated that predicted PGA values promising but generally overestimated. However, Chi-Chi set, whose fault rupture complex long, using more accurate than previous study support vector regression approach. addition, alternative alert criterion, issues when exceeds threshold successive time windows, presented, different thresholds considered also discussed.