Prediction of On-Site Peak Ground Motion Based on Machine Learning and Transfer Learning
Zhu Jingbao
Liu Heyi
Luan Shicheng
Liang Kunzheng
Song Jindong
Li Shanyou
Abstract：To improve the accuracy of peak ground motion(peak ground acceleration(PGA)and peak ground velocity(PGV)prediction in Chinese instrument seismic intensity calculation for on-site earthquake early warning(EEW),a prediction method of on-site peak ground motion based on machine learning and transfer learning is proposed.A pretrained on-site peak ground motion prediction model was established using neural networks based on strong motion data recorded by the K-NET network in Japan.Based on strong motion data from China and the pretrained on-site peak ground motion prediction model,an on-site peak ground motion prediction model for China was established through transfer learning.For the Japanese and Chinese test dataset and Luding M6.8 eathquake,at 3 s after the arrival of P-wave,compared to traditional on-site peak ground motion prediction method,the method proposed in this study has smaller mean absolute error and standard deviation for PGA prediction and PGV prediction.The results indicate that the method proposed in this study can improve the reliability of predicting peak ground motion in on-site EEW to a certain extent,which is of great significance for the development of on-site EEW systems.
Keywords：earthquakesmachine learningtransfer learningP wavepeak ground motionartificial intelligence
Publication Date：2025-05-30
Online Publishing Date：2025-08-15(First online date of this platform, not the publication date of the document)
Pages：19( 1842-1860 )