Datensatz

Zusammenfassung

In order to identify the phase of seismic waves, deep learning-based phase identification technologies have recently been developed.It is known that deep learning-based phase identification technology can accurately determine the arrival time of the phase and also identify S waves, which are relatively more difficult to detect than P waves. Approximately 2.4 million seismic data from STEAD and INSTANCE were used to develop deep learning-based phase identification techniques (Sheen, 2021).In this study, a deep learning-based phase identification technology was applied to 871 earthquakes with a magnitude of 2.0 or higher on the Korean Peninsula and compared to the manual seismic wave identification results provided by the Korea Meteorological Administration. As a result, it was confirmed that most of the P-waves of the seismic wave could be identified successfully, and the S-waves were also identified. In addition, the optimal criteria for determining the phase status of P and S-waves were established using deep learning-based seismic wave identification technology. To this end, the learning model predicted the probabilities of P&S-waves, and noise from the three-component seismic waveform, and judged the section where the probabilities of P&S-waves exceeded a specific threshold to be a phase. It was confirmed that the performance of the artificial intelligence model changed according to this specific threshold, and the optimal threshold yielding the highest performance of the AI model was selected. As a result, it was confirmed that as the threshold value increased, the model accuracy increased while the reproduction rate decreased.