Robust microseismic arrival picking for hydraulic fracturing monitoring using a deep orthogonal collaborative ensemble network

Accurate microseismic arrival picking is critical for hydraulic fracturing monitoring and subsequent fracture characterization. In recent years, deep learning-based pickers have achieved promising performance by automatically learning complex waveform features. However, these pickers still face challenges in low signal-to-noise ratio (SNR) environments and in changing monitoring scenarios. These challenges are closely related to the restricted feature representation capability of individual network architectures, which may hinder their ability to capture diverse waveform features. To overcome this issue, we propose a deep orthogonal collaborative ensemble network (DOCE-Net) for robust microseismic arrival picking in hydraulic fracturing monitoring. DOCE-Net employs three heterogeneous base learners, including U-Net, Res-LSTM, and TransNet, to capture complementary waveform representations from multi-trace microseismic data. An orthogonality constraint is introduced to reduce feature redundancy and enhance diversity among different learners. The outputs of these learners are further combined through a semblance-based decision-level fusion strategy, which exploits prediction consistency to improve picking reliability. The proposed method is evaluated using two field hydraulic fracturing microseismic datasets. In the in-domain tests, DOCE-Net consistently outperforms the three individual base learners, particularly under low SNR conditions, and achieves higher accuracy than its single-trace counterpart, confirming the contribution of multi-trace spatial coherence. In the cross-domain tests, DOCE-Net also achieves superior performance after fine-tuning with limited labeled samples from an unseen monitoring domain. These results demonstrate the effectiveness of DOCE-Net for robust microseismic arrival picking and cross-domain generalization. This study indicates that DOCE-Net can provide a reliable strategy for developing robust and transferable microseismic phase arrival pickers.

Authors

Institutions

Publication Details

Journal
Scientific Reports
Published
2026-11-21
DOI
https://doi.org/10.1038/s41598-026-76189-2
Primary Topic
Hydraulic Fracturing and Reservoir Analysis
Type
article
Field-Weighted Citation Impact
0.00

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Robust microseismic arrival picking for hydraulic fracturing monitoring using a deep orthogonal collaborative ensemble network

Fanghuang Chen, Yilun Zhang, Xifan Chen
1 links
Scientific Reports
Hydraulic Fracturing and Reservoir Analysis
article

Robust microseismic arrival picking for hydraulic fracturing monitoring using a deep orthogonal collaborative ensemble network

Fanghuang Chen, Yilun Zhang, Xifan Chen
article en

Abstract

Accurate microseismic arrival picking is critical for hydraulic fracturing monitoring and subsequent fracture characterization. In recent years, deep learning-based pickers have achieved promising performance by automatically learning complex waveform features. However, these pickers still face challenges in low signal-to-noise ratio (SNR) environments and in changing monitoring scenarios. These challenges are closely related to the restricted feature representation capability of individual network architectures, which may hinder their ability to capture diverse waveform features. To overcome this issue, we propose a deep orthogonal collaborative ensemble network (DOCE-Net) for robust microseismic arrival picking in hydraulic fracturing monitoring. DOCE-Net employs three heterogeneous base learners, including U-Net, Res-LSTM, and TransNet, to capture complementary waveform representations from multi-trace microseismic data. An orthogonality constraint is introduced to reduce feature redundancy and enhance diversity among different learners. The outputs of these learners are further combined through a semblance-based decision-level fusion strategy, which exploits prediction consistency to improve picking reliability. The proposed method is evaluated using two field hydraulic fracturing microseismic datasets. In the in-domain tests, DOCE-Net consistently outperforms the three individual base learners, particularly under low SNR conditions, and achieves higher accuracy than its single-trace counterpart, confirming the contribution of multi-trace spatial coherence. In the cross-domain tests, DOCE-Net also achieves superior performance after fine-tuning with limited labeled samples from an unseen monitoring domain. These results demonstrate the effectiveness of DOCE-Net for robust microseismic arrival picking and cross-domain generalization. This study indicates that DOCE-Net can provide a reliable strategy for developing robust and transferable microseismic phase arrival pickers.

Scientific Reports
Minzu University of China (CN), Yunnan University (CN), Peking University (CN)
National Science and Technology Major Project
Hydraulic Fracturing and Reservoir Analysis
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon