Scholarlink

<p>The accurate estimation of uniaxial compressive strength (UCS) is critical for the design stability analysis and safety assessment of civil and mining engineering projects worldwide. The conventional laboratory testing methods to measure UCS are often expensive, time-consuming, and require specialized equipment and expertise, which limits their applicability for rapid or large-scale assessments. Therefore, this study proposes a novel hybrid model that integrates the Group Method of Data Handling with Levenberg–Marquardt optimization (GMDH-LM) and with two other models, including improved whale optimization algorithm–artificial neural network (IWO-ANN) and imperialist competitive algorithm–artificial neural network (ICA-ANN), to predict UCS based solely on petrographic mineralogical data, significantly reducing the reliance on extensive laboratory testing. The GMDH-LM and other models were trained on an independent dataset encompassing 150 dolerite rock samples acquired from the Kirana Hills, Panjab, Pakistan. Furthermore, the model’s accuracy and validity were evaluated utilizing a distinct set of 55 rock samples for validation. The results show that the developed novel hybrid GMDH-LM model revealed superior achievement with high prediction accuracy, achieved values of <i>R</i><sup>2</sup> above 0.99, and low root mean square errors (MSE), signifying strong agreement between predicted and measured values of UCS. Furthermore, sensitivity analysis indicated that key mineralogical components, specifically K-feldspar, plagioclase, and chlorite, are the prime dominating parameters influencing rock strength (UCS). These findings substantiate existing geotechnical knowledge and deliver insights into the underlying mechanisms. The GMDH-LM model constantly demonstrated superior precision and robustness against IWO-ANN and ICA-ANN models. Thus, the proposed technique serves as a reliable, cost-effective, and competent substitute for UCS estimation, easing faster judgment and resource evaluation in geotechnical and mining 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

Prediction of Uniaxial Compressive Strength from Petrographic Features of Kirana Dolerites Using a Hybrid Group Method of Data Handling-Levenberg–Marquardt (GMDH-LM) Model

  • 期刊论文

  • 发表日期：

    2025年9月15日

查看全文
  • Javid Hussain,
  • Nafees Ali,
  • Xiaodong Fu,
  • Jian Chen,
  • Naveed Ahmad Khan,
  • Sartaj Hussain,
  • Sayed Muhammad Iqbal

摘要

The accurate estimation of uniaxial compressive strength (UCS) is critical for the design stability analysis and safety assessment of civil and mining engineering projects worldwide. The conventional laboratory testing methods to measure UCS are often expensive, time-consuming, and require specialized equipment and expertise, which limits their applicability for rapid or large-scale assessments. Therefore, this study proposes a novel hybrid model that integrates the Group Method of Data Handling with Levenberg–Marquardt optimization (GMDH-LM) and with two other models, including improved whale optimization algorithm–artificial neural network (IWO-ANN) and imperialist competitive algorithm–artificial neural network (ICA-ANN), to predict UCS based solely on petrographic mineralogical data, significantly reducing the reliance on extensive laboratory testing. The GMDH-LM and other models were trained on an independent dataset encompassing 150 dolerite rock samples acquired from the Kirana Hills, Panjab, Pakistan. Furthermore, the model’s accuracy and validity were evaluated utilizing a distinct set of 55 rock samples for validation. The results show that the developed novel hybrid GMDH-LM model revealed superior achievement with high prediction accuracy, achieved values of R2 above 0.99, and low root mean square errors (MSE), signifying strong agreement between predicted and measured values of UCS. Furthermore, sensitivity analysis indicated that key mineralogical components, specifically K-feldspar, plagioclase, and chlorite, are the prime dominating parameters influencing rock strength (UCS). These findings substantiate existing geotechnical knowledge and deliver insights into the underlying mechanisms. The GMDH-LM model constantly demonstrated superior precision and robustness against IWO-ANN and ICA-ANN models. Thus, the proposed technique serves as a reliable, cost-effective, and competent substitute for UCS estimation, easing faster judgment and resource evaluation in geotechnical and mining engineering applications.