Scholarlink

<p>Accurate arrival-time picking is essential for downhole microseismic event localization and hydraulic-fracturing evaluation. Traditional methods (e.g., STA/LTA) and deep-learning networks are typically designed for single-event picking, making it difficult to reliably identify all phases when multiple events occur within the same time window. To address this challenge, this study proposes a fully convolutional network (FCN)-based dual-event detection and arrival-time picking model for downhole microseismic monitoring. The model processes fixed-length three-component waveform data, and P-waves and S-waves are encoded with opposite Gaussian probability distributions to enhance phase discrimination. Trained on 80 downhole microseismic events with extensive data augmentation, the model is validated on field data and compared with the STA/LTA method, the PhaseNet model, and a single-event FCN model. Results demonstrate superior dual-event recognition and arrival picking performance, achieving F1 scores of approximately 95% for both P-waves and S-waves and maintaining strong robustness under low-SNR conditions. This study provides an efficient and reliable solution for multi-event microseismic monitoring, offering valuable support for underground engineering safety and shale gas well early-warning applications.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Dual-event arrival-time picking for downhole microseismic monitoring using a fully convolutional network

  • 期刊论文

  • 发表日期：

    2026年5月21日

查看全文
  • Jiali Pan,
  • Xiao Tian,
  • Yichong Chen,
  • Yuxing Pan,
  • Xiong Zhang

摘要

Accurate arrival-time picking is essential for downhole microseismic event localization and hydraulic-fracturing evaluation. Traditional methods (e.g., STA/LTA) and deep-learning networks are typically designed for single-event picking, making it difficult to reliably identify all phases when multiple events occur within the same time window. To address this challenge, this study proposes a fully convolutional network (FCN)-based dual-event detection and arrival-time picking model for downhole microseismic monitoring. The model processes fixed-length three-component waveform data, and P-waves and S-waves are encoded with opposite Gaussian probability distributions to enhance phase discrimination. Trained on 80 downhole microseismic events with extensive data augmentation, the model is validated on field data and compared with the STA/LTA method, the PhaseNet model, and a single-event FCN model. Results demonstrate superior dual-event recognition and arrival picking performance, achieving F1 scores of approximately 95% for both P-waves and S-waves and maintaining strong robustness under low-SNR conditions. This study provides an efficient and reliable solution for multi-event microseismic monitoring, offering valuable support for underground engineering safety and shale gas well early-warning applications.