Article Details

Abstract

This research proposes an anomaly detection method in earthquake precursor gravity data based on the CNN-LSTM-CBAM model. The anomaly detection in earthquake precursor gravity data is crucial for improving the timeliness of earthquake predictions. It extracts spatial features of the gravity data using CNN, and uses the LSTM to capture long-term dependency relationships in the time series. The CBAM is introduced to enhance the model's ability to focus on important features, thereby improving anomaly detection performance. Experimental comparisons with the anomaly detection methods such as AutoEncoder, CNN, LSTM, and CNN-LSTM methods show that the proposed model in this paper outperforms others in metrics such as MAE, MSE, RMSE, and R2. This model effectively identifies potential and abnormal data and provides a reliable foundation for earthquake risk management and early warning. This research offers new insights into the analysis of earthquake precursor data.

Keywords

earthquake precursor anomaly; gravity data; time series; LSTM; Attention Mechanism

References

1. [1] Li Yan'e Research on Strong Earthquake Prediction Method Based on Seismic Process Tracking [D]. Beijing: Institute of Geophysics, China Earthquake Administration, 2023
2. [2] Xu Xitong Analysis of Multi Circle Anomalies of the Earth under Earthquake Disturbance and Study on Their Evolutionary Models [D]. Changchun: Jilin University, 2024
3. [3] Chi Chengquan Analysis and Anomaly Extraction of Precursor Observation Data of Drilling Strain [D]. Changchun: Jilin University, 2020
4. [4] Xin Jiancun, Fang Wei, Yang Yihai, et al. [J]. Earthquake, 2023, 43 (4): 153-168. Analysis of changes in the geoelectric field before the 2021 MS (S subscript) 5.5 earthquake in Aksai County, Jiuquan
5. [5] Liu Jie Research on the Identification Method of Underground Fluid Precursors for Wenchuan Earthquake Based on Recurrent Neural Networks [D]. Chengdu: Chengdu University of Technology, 2021
6. [6] Sun Yishu Research on denoising and inversion methods for urban drag transient electromagnetic data [D]. Changchun: Jilin University, 2023
7. [7] Wang Hongzhou Research on deep learning based denoising and velocity model inversion techniques for seismic exploration data [D]. Changchun: Jilin University, 2024
8. [8] KAVIANPOUR P，KAVIANPOUR M，JAHANI E，et al. A CNN-BiLSTM Model with Attention Mechanism for Earthquake Prediction [J].Journal of Supercomputing，2023，79（17）：19194-19226.
9. [9] ZHANG H，ZHAI H，LIU T T，et al. Short-term Comprehensive Prediction Method for Regional Earthquakes Based on Multi-source Information Fusion [J].Proceedings of the Indian National Science Academy，2024（prepublish）：1-11.
10. [10] BUDAK C，GIDER V. LSTM Based Forecasting of the Next Day's Values of Ionospheric Total Electron Content （TEC） as an Earthquake Precursor Signal [J].Earth Science Informatics，2023，16（3）：2323-2337.
11. [11] LI H Y，LI J H，LI X G，et al. Seismic Picking Attention Module [J].IEEE Transactions on Geoscience and Remote Sensing，2024，62：1-6.
12. [12] MOUDDEN T EL，AMNAI M，CHOUKRI A，et al. Artificial Intelligence for Assessing the Planets' Positions as a Precursor to Earthquake Events [J].Journal of Geodynamics，2024，162：102057.
13. [13] Luo Xuchen Research on earthquake prediction model based on multi-source deep learning [D]. Chengdu: University of Electronic Science and Technology of China, 2023
14. [14] ZHANG J Y，SUN K，ZHU J Q，et al. Application of Model-based Time Series Prediction of Infrared Long-wave Radiation Data for Exploring the Precursory Patterns Associated with the 2021 Madoi Earthquake [J].Remote Sensing，2023，15（19）：1-22.
15. [15] Shi Xiangyu Research on Earthquake Prediction Based on Machine Learning Regression Algorithm and Its Application in China Earthquake Science Experimental Field [D]. Beijing: Institute of Earthquake Prediction, China Earthquake Administration, 2021

About the authors

Xing Qianlong (1999-), male, Han ethnicity, from Shangqiu, Henan Province, is currently pursuing a master's degree. His research interests include earthquake precursor anomalies, big data processing, and deep learning; Corresponding author: Liu Qingjie (1978-), male, Han ethnicity, from Luoyang, Henan, professor, master's degree, research direction: big data processing.