Microseismic data denoising using CNN–BiLSTM model
Journal of Geophysics and Engineering · 26.06.2026 · 23 (5) · 1614-1627
DOI: 10.1093/jge/gxag072 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Geophysics and Engineering, online erschienen am 21.05.2026. Verfasst von Yifei Bao und 1 weiteren Autor:innen. Die Arbeit nennt 26 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: National Natural Science Foundation of China.
Abstract (Originalsprache)
Microseismic data exhibit low energy, weak signals, and extreme vulnerability to noise, making their denoising process especially important. By combining a convolutional neural network (CNN) with a bidirectional long short-term memory network (BiLSTM), we proposed a deep learning based denoising method CNN–BiLSTM for microseismic data. We adopted Bayesian optimization to select the optimal hyperparameter combination and enhance computational efficiency through automatic mixed precision (AMP) training. Based on the subsurface structural models and microseismic data from the Zigong and Neijiang regions in Sichuan, China, we generated a dataset to train our model, achieving excellent denoising results. The proposed CNN-BiLSTM framework effectively integrates local feature extraction and global temporal modeling, achieving robust denoising performance under low signal-to-noise conditions. In addition, Bayesian optimization and AMP training significantly improve training efficiency and reduce computational cost, making the method suitable for large-scale microseismic applications. We applied the method to microseismic observed data from the Zigong and Neijiang areas. The results demonstrate that the CNN-BiLSTM model can effectively suppress complex noise components in field records while retaining the phase characteristics and useful high-frequency information of microseismic signals. Compared with conventional denoising methods, the proposed method achieves a better balance between noise attenuation and signal preservation, especially under low-SNR conditions. These results suggest that the proposed method has strong applicability to real microseismic monitoring scenarios and can provide more reliable waveform data for subsequent seismic interpretation.
Förderung
- National Natural Science Foundation of China
Zitieren
APA
Bao, Y., & Wang, T. (2026). Microseismic data denoising using CNN–BiLSTM model. Journal of Geophysics and Engineering, 23(5), 1614-1627. https://doi.org/10.1093/jge/gxag072
Vancouver
Bao Y, Wang T. Microseismic data denoising using CNN–BiLSTM model. Journal of Geophysics and Engineering. 2026;23(5):1614-1627. doi:10.1093/jge/gxag072
BibTeX
@article{Bao2026,
author = {Bao, Yifei and Wang, Tairan},
title = {Microseismic data denoising using CNN–BiLSTM model},
journal = {Journal of Geophysics and Engineering},
year = {2026},
volume = {23},
number = {5},
pages = {1614--1627},
doi = {10.1093/jge/gxag072}
}
EUVIMED