Scholarlink

Estimating the unrestricted crushing resistance (UCS) of treated soils using additives such as calcium oxide and cement is essential for developing precise geo-mechanical models. Conventional approaches for determining UCS are often resource-intensive, requiring significant time, labor, extensive laboratory testing, and intricate well-log analysis. This paper introduces a robust hybrid machine learning approach employing a probabilistic classifier integrated with meta-heuristic optimization techniques, specifically the Simulated Annealing Optimization (SAO) and Dynamically Algebraic Optimization Approach (DAOA), to enhance the efficiency and accuracy of UCS predictions. The proposed methodology was evaluated using various evaluation metrics, including correctness, specificity, sensitivity, and the F1-measure, demonstrating a high prediction accuracy of 92.5%, with a precision of 91.2%, recall of 90.8%, and an F1-score of 91.0%. The approach was validated using a diverse set of UCS samples from stabilized soil experiments, highlighting the superiority of the hybrid model compared to traditional methods. Future enhancements include integrating additional environmental and geotechnical data, exploring more sophisticated machine learning techniques such as ensemble and deep learning, and developing intuitive user interfaces for practical field applications. This work presents a significant step forward in real-time UCS prediction, providing a rapid, scalable, and accurate solution for geotechnical engineering practices.

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

Optimizing Unrestricted Compressive Strength of Stabilized Soil Using Naive Bayes and a Multimodal Machine Learning Approach for Calcium Oxide and Cement Prediction

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Manorama Chouhan,
  • Partha Sarathy Banerjee,
  • Kunal Mishra

摘要

Estimating the unrestricted crushing resistance (UCS) of treated soils using additives such as calcium oxide and cement is essential for developing precise geo-mechanical models. Conventional approaches for determining UCS are often resource-intensive, requiring significant time, labor, extensive laboratory testing, and intricate well-log analysis. This paper introduces a robust hybrid machine learning approach employing a probabilistic classifier integrated with meta-heuristic optimization techniques, specifically the Simulated Annealing Optimization (SAO) and Dynamically Algebraic Optimization Approach (DAOA), to enhance the efficiency and accuracy of UCS predictions. The proposed methodology was evaluated using various evaluation metrics, including correctness, specificity, sensitivity, and the F1-measure, demonstrating a high prediction accuracy of 92.5%, with a precision of 91.2%, recall of 90.8%, and an F1-score of 91.0%. The approach was validated using a diverse set of UCS samples from stabilized soil experiments, highlighting the superiority of the hybrid model compared to traditional methods. Future enhancements include integrating additional environmental and geotechnical data, exploring more sophisticated machine learning techniques such as ensemble and deep learning, and developing intuitive user interfaces for practical field applications. This work presents a significant step forward in real-time UCS prediction, providing a rapid, scalable, and accurate solution for geotechnical engineering practices.