Development of Software for Automated Marking of Seismic P- and S-waves

The paper presents a novel software solution for the automatic classification and detection of seismic P- and S-waves using neural networks. The research focuses on addressing the limitations of manual methods for seismic wave identification and the need for more reliable and efficient automation. The developed software leverages a modified ResNet34 neural network, trained on real seismic event data from Southern California. The approach involves transforming seismic waveforms into spectrograms using Short-Time Fourier Transform, which are then processed by the neural network for classification. Additionally, a post-processing algorithm filters the output to reduce noise and improve accuracy. Testing on real seismic data reveals the effectiveness of the software in identifying P-waves, achieving a Matthews correlation coefficient of 0.82 when using a band-pass filter. However, S-wave detection proved more challenging, with a maximum coefficient of 0.38 using a high-pass filter. The paper identifies this difficulty as a key area for improvement, suggesting that further refinement of neural network architecture and post-processing techniques could enhance performance. Recommendations for the practical application of the software in seismic data analysis and future research directions are also provided.