Research on automatic segmentation method of microseismic monitoring signal under low signal-to-noise ratio conditions
- 1School of Smart Construction and Civil Engineering, Zhongyuan University of Technology, Zhengzhou, China
- 2School of Earth Science and Engineering, North China University of Water Resources and Electric Power, Zhengzhou, China
- 3School of Civil and Transportation Engineering, South China University of Technology, Guangzhou, China
Abstract
To address the issues of reduced microseismic signal detection accuracy and difficulty in identifying event boundaries under low signal-to-noise ratio conditions, we propose SEU-Net, an automatic microseismic signal segmentation model that integrates a Squeeze-and-Excitation (SE) attention mechanism. This model embeds an SE module within the U-Net encoder, which enhances feature responses related to microseismic event boundaries through adaptive re-scaling of channel features. Simultaneously, it employs a temporal signal dimension expansion method to convert one-dimensional waveforms into an input format suitable for two-dimensional convolutional networks, thereby enabling end-to-end detection of microseismic event start and end times. Validation results using rock uniaxial compression acoustic emission test data show that SEU-Net achieves an overall segmentation accuracy of 98.97% on the test set, representing an improvement of approximately 1 percentage point over U-Net. Under a signal-to-noise ratio (SNR) of 5 dB, the proportions of samples with start and end point detection errors falling within the 0–5 range were 79.98% and 89.93%, respectively. Compared with EQTP, SEU-Net achieved an initial F1 score of 88.88% and an intersection-to-union ratio of 79.98%, which are higher than EQTP’s 86.81% and 76.69%, respectively. The error distribution and various evaluation metrics indicate that SEU-Net can maintain high accuracy in event boundary detection even under noise interference. Three-dimensional source localization based on the detection results shows that the spatial distribution of sources corresponding to SEU-Net is more concentrated and consistent with the actual fracture evolution process of the test specimen. This method provides technical support for the automatic detection of microseismic signals and source localization in complex rock mass engineering environments with low signal-to-noise ratios.
Related
- A novel automatic real-time motion tracking method in MRI-guided radiotherapy using enhanced Tracking-Learning-Detection framework with automatic segmentation.
- Conditions for implementing projective measurements through continuous monitoring
- Overexposure-free automatic exposure algorithm for division-of-aperture polarization camera
- Penalized PET/CT Reconstruction Algorithms With Automatic Realignment for Anatomical Priors
- Derivative Computation in PINNs: Automatic Differentiation, Finite Differences and Beyond
BibTeX
@article{liu2026research,
title = {{Research on automatic segmentation method of microseismic monitoring signal under low signal-to-noise ratio conditions}},
author = {Weijian Liu and Jianbo Li and Zhizeng Zhang and Zhenxia Yuan and Shuai Teng and Danqing Song and Mingrui Zhu and Wenlong Lv},
journal = {PLOS One},
year = {2026},
month = sep,
doi = {10.1371/journal.pone.0358562},
url = {https://doi.org/10.1371/journal.pone.0358562},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Research on automatic segmentation method of microseismic monitoring signal under low signal-to-noise ratio conditions
AU - Weijian Liu
AU - Jianbo Li
AU - Zhizeng Zhang
AU - Zhenxia Yuan
AU - Shuai Teng
AU - Danqing Song
AU - Mingrui Zhu
AU - Wenlong Lv
PY - 2026
DA - 2026/09/18
JO - PLOS One
DO - 10.1371/journal.pone.0358562
UR - https://doi.org/10.1371/journal.pone.0358562
AB - To address the issues of reduced microseismic signal detection accuracy and difficulty in identifying event boundaries under low signal-to-noise ratio conditions, we propose SEU-Net, an automatic microseismic signal segmentation model that integrates a Squeeze-and-Excitation (SE) attention mechanism. This model embeds an SE module within the U-Net encoder, which enhances feature responses related to microseismic event boundaries through adaptive re-scaling of channel features. Simultaneously, it employs a temporal signal dimension expansion method to convert one-dimensional waveforms into an input format suitable for two-dimensional convolutional networks, thereby enabling end-to-end detection of microseismic event start and end times. Validation results using rock uniaxial compression acoustic emission test data show that SEU-Net achieves an overall segmentation accuracy of 98.97% on the test set, representing an improvement of approximately 1 percentage point over U-Net. Under a signal-to-noise ratio (SNR) of 5 dB, the proportions of samples with start and end point detection errors falling within the 0–5 range were 79.98% and 89.93%, respectively. Compared with EQTP, SEU-Net achieved an initial F1 score of 88.88% and an intersection-to-union ratio of 79.98%, which are higher than EQTP’s 86.81% and 76.69%, respectively. The error distribution and various evaluation metrics indicate that SEU-Net can maintain high accuracy in event boundary detection even under noise interference. Three-dimensional source localization based on the detection results shows that the spatial distribution of sources corresponding to SEU-Net is more concentrated and consistent with the actual fracture evolution process of the test specimen. This method provides technical support for the automatic detection of microseismic signals and source localization in complex rock mass engineering environments with low signal-to-noise ratios.
SN - 1932-6203
ER -