Scholarlink

<p>An orebody model is essential for effective planning of both short- and long-term mining. Traditionally, the demarcation between the orebodies and the surrounding host rocks is determined through field observations by mining geologists and chemical analyses of rock samples. However, this method is notably time-intensive, particularly for underground mines. Portable X-ray fluorescence (XRF) has emerged as a potent method for geological mapping. Therefore, we explored the feasibility of utilizing supervised learning for lithology classification based on XRF data, aiming to expedite geological mapping in underground mines. Furthermore, a dataset was compiled using laboratory measurements of rock samples obtained from a vanadiferous titanomagnetite mine in South Korea. To refine the supervised learning, we evaluated the performance of six distinct supervised learning models. The one-dimensional convolution neural network model demonstrated superior performance, achieving the highest accuracy (0.95) and prediction accuracy (0.94). The implementation of this approach in the mining industry could significantly improve exploration and extraction efficiency, reduce mining expenditure, and increase safety through faster and more accurate geological assessments.</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

Vanadiferous titanomagnetite ore rock classifier using machine learning from portable X-ray fluorescence spectra

  • 期刊论文

  • 发表日期：

    2025年6月18日

查看全文
  • Jieun Baek,
  • Seongjun Cho,
  • Seungwook Shin

摘要

An orebody model is essential for effective planning of both short- and long-term mining. Traditionally, the demarcation between the orebodies and the surrounding host rocks is determined through field observations by mining geologists and chemical analyses of rock samples. However, this method is notably time-intensive, particularly for underground mines. Portable X-ray fluorescence (XRF) has emerged as a potent method for geological mapping. Therefore, we explored the feasibility of utilizing supervised learning for lithology classification based on XRF data, aiming to expedite geological mapping in underground mines. Furthermore, a dataset was compiled using laboratory measurements of rock samples obtained from a vanadiferous titanomagnetite mine in South Korea. To refine the supervised learning, we evaluated the performance of six distinct supervised learning models. The one-dimensional convolution neural network model demonstrated superior performance, achieving the highest accuracy (0.95) and prediction accuracy (0.94). The implementation of this approach in the mining industry could significantly improve exploration and extraction efficiency, reduce mining expenditure, and increase safety through faster and more accurate geological assessments.