Scholarlink

<p>‌ Conventional rock coring faces limitations in efficiency, cost, and labor demand. This study develops an intelligent real-time rock classification system based on multivariate drilling parameters. Firstly, five different strength rock layers and specimens were poured, and the mechanical parameters of each rock layer were obtained through laboratory mechanical testing. Next, based on the self-designed drilling device, the borehole drilling was simulated and multiple drilling parameter indicators were obtained. Then, the Out of Bag principle was used to evaluate the importance of various drilling parameter indicators in rock classification, and key parameter indicators for rock identification were selected. Finally, based on the particle swarm optimization algorithm, the back propagation neural network and support vector machine algorithm were optimized for rock layer recognition. The research results show that revolution per minute has the greatest impact on the classification results of rock layers. Compared to the optimized back propagation neural network model, the optimized support vector machine algorithm significantly improves the prediction effect of rock type, with an accuracy increase of 21.68%.</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

Recognition of rock formation while drilling based on optimized machine learning algorithms

  • 期刊论文

  • 发表日期：

    2025年8月13日

查看全文
  • Cancan Liu,
  • Xigui Zheng,
  • Haifeng Wang,
  • Niaz Muhammad Shahani,
  • Yankun Ma,
  • Zike Meng

摘要

‌ Conventional rock coring faces limitations in efficiency, cost, and labor demand. This study develops an intelligent real-time rock classification system based on multivariate drilling parameters. Firstly, five different strength rock layers and specimens were poured, and the mechanical parameters of each rock layer were obtained through laboratory mechanical testing. Next, based on the self-designed drilling device, the borehole drilling was simulated and multiple drilling parameter indicators were obtained. Then, the Out of Bag principle was used to evaluate the importance of various drilling parameter indicators in rock classification, and key parameter indicators for rock identification were selected. Finally, based on the particle swarm optimization algorithm, the back propagation neural network and support vector machine algorithm were optimized for rock layer recognition. The research results show that revolution per minute has the greatest impact on the classification results of rock layers. Compared to the optimized back propagation neural network model, the optimized support vector machine algorithm significantly improves the prediction effect of rock type, with an accuracy increase of 21.68%.