Scholarlink

<p>The rock mass rating (RMR) system is commonly used in mining and excavation activities to gather information on rock mass quality and support measures. Researchers have offered empirical methods to classify rock masses based on intact rock and discontinuity parameters. This study addresses the application of machine learning in predicting the RMR<sub>b</sub> value. Three different ML algorithms were trained and tested using the input and output data gathered from the existing RMR table and were later verified on two case studies in the northern Himalayas and Eastern Ghats, India. The results were later compared using the Pearson coefficient and statistical <i>t</i>-test and were found plausible. All the parameters for finding RMR are field-based except uniaxial compressive strength (UCS), which requires laboratory instrumentation. The work also implemented ML algorithms to predict the RMR<sub>b</sub> without the input of UCS value. The algorithms were trained with all input features except UCS, but the output RMR<sub>b</sub> value used in training and testing includes the value of the UCS parameter for each observation. The ML model scales the rating of the UCS parameter within other input parameters to predict the final RMR<sub>b</sub>. The results reveal that the ML algorithms can predict the RMR<sub>b</sub> with considerable accuracy. More careful consideration should be taken near class boundaries while using ML for the prediction.</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

Application of machine learning tools in predicting basic rock mass rating of mountainous road cut slopes

  • 期刊论文

  • 发表日期：

    2025年2月21日

查看全文
  • Jaspreet Singh,
  • Sarada Prasad Pradhan,
  • Madhvendra Pratap Singh

摘要

The rock mass rating (RMR) system is commonly used in mining and excavation activities to gather information on rock mass quality and support measures. Researchers have offered empirical methods to classify rock masses based on intact rock and discontinuity parameters. This study addresses the application of machine learning in predicting the RMRb value. Three different ML algorithms were trained and tested using the input and output data gathered from the existing RMR table and were later verified on two case studies in the northern Himalayas and Eastern Ghats, India. The results were later compared using the Pearson coefficient and statistical t-test and were found plausible. All the parameters for finding RMR are field-based except uniaxial compressive strength (UCS), which requires laboratory instrumentation. The work also implemented ML algorithms to predict the RMRb without the input of UCS value. The algorithms were trained with all input features except UCS, but the output RMRb value used in training and testing includes the value of the UCS parameter for each observation. The ML model scales the rating of the UCS parameter within other input parameters to predict the final RMRb. The results reveal that the ML algorithms can predict the RMRb with considerable accuracy. More careful consideration should be taken near class boundaries while using ML for the prediction.