Scholarlink

<p>The construction environment of deep underground projects is complex and variable, and the resulting interference signals with different characteristics reduce the long-term recognition accuracy of microseismic signals. In this study, a microseismic signal recognition method that adapts to environmental changes is proposed. The recognition accuracy of microseismic signals in complex changing environments is improved by constructing an adaptive vision transformer (Ada-VIT) model that integrates feature extraction, and multi-label classifier with domain adaptive capability, and optimizes the features using adaptive mechanisms. The clean microseismic signals and pure noise signals collected from the Hanjiang-to-Weihe River Diversion project are source-domain samples for model training. More complex and noisier microseismic signals are used as target-domain samples for identification and classification. The results show that the Ada-VIT model significantly outperforms ResNet50 and the improved convolutional neural network-recurrent neural network (CNN-RNN) model in microseismic signal recognition and classification with a high mean accuracy (mAP) of 0.98. Furthermore, this Ada-VIT model is instance validated on the microseismic monitoring dataset of the Jinping II Hydropower Station and exhibits excellent adaptive classification performance. This research provides significant theoretical support and a robust technical foundation for the long-term identification and precise early warning of microseismic signals in deep engineering 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

Adaptive Classification Model Based on Dynamic Variations in Microseismic Signal Characteristics

  • 期刊论文

  • 发表日期：

    2025年10月22日

查看全文
  • Jinglan Zhang,
  • Jiaming Li,
  • Beichang Tang,
  • Shuguang Zhang,
  • Shibin Tang

摘要

The construction environment of deep underground projects is complex and variable, and the resulting interference signals with different characteristics reduce the long-term recognition accuracy of microseismic signals. In this study, a microseismic signal recognition method that adapts to environmental changes is proposed. The recognition accuracy of microseismic signals in complex changing environments is improved by constructing an adaptive vision transformer (Ada-VIT) model that integrates feature extraction, and multi-label classifier with domain adaptive capability, and optimizes the features using adaptive mechanisms. The clean microseismic signals and pure noise signals collected from the Hanjiang-to-Weihe River Diversion project are source-domain samples for model training. More complex and noisier microseismic signals are used as target-domain samples for identification and classification. The results show that the Ada-VIT model significantly outperforms ResNet50 and the improved convolutional neural network-recurrent neural network (CNN-RNN) model in microseismic signal recognition and classification with a high mean accuracy (mAP) of 0.98. Furthermore, this Ada-VIT model is instance validated on the microseismic monitoring dataset of the Jinping II Hydropower Station and exhibits excellent adaptive classification performance. This research provides significant theoretical support and a robust technical foundation for the long-term identification and precise early warning of microseismic signals in deep engineering applications.