VGG Model for Peak Ground Acceleration Prediction

Many studies on the topic of earthquake early warning systems (EEWS) have been proposed over the past ten years. Deep learning techniques can be used to determine peak ground acceleration (PGA). Earthquake catalogs are very important in that they contain information for the history of fault systems, seismic modeling, and the detriment of seismic hazards, predicting them, and eventually reducing seismic risk. To improve earthquake-resistant structures, it is important to predict earthquakes to select peak ground acceleration (PGA), which is provided in this work as a study of seismic hazard analysis. Propose to use the waveforms of weak motion velocity recording in Japan to inform the VGG-19 convolutional neural networks for PGA prediction. 1720 earthquakes (velocities) from 4 stations, whose magnitudes are in the 1-to-9 range, are used in this study. As a result, the mean absolute error (MAE) for the test model is 4.03.