Research on Earthquake Detection Based on Machine Learning

Accurate and efficient detection of each earthquake is an important foundation for earthquake work. However, at present, due to the lack of observation data information, the accuracy of parameter estimation is low, especially for seismic detection of major projects, the current situation of sparse station layout makes it impossible to use multiple averages like dense network seismic detection to reduce the discreteness of parameter estimation, and the accuracy of seismic detection parameters based on a single station needs to be improved. In recent years, there has been an increasing focus on how machine learning can be used to improve seismic detection performance. In this paper, the principle of AI seismic detection and waveform mask matching is analyzed, and then a new deep learning method, TransQuake, is proposed based on the frontier sequence model Transformer for seismic wave detection. TransQuake combines STA/LTA algorithms for feature enhancement of seismic waveform data and interpretable model learning using a multi-head attention mechanism. To validate the performance of the model, this paper conducts an extensive evaluation on the 2008 Wenchuan MW7.9 earthquake aftershock dataset. The results show that TransQuake can achieve the best detection performance beyond the baseline of leading-edge algorithms.