Scholarlink

<p>The article is based on the presentation at the international conference “Soil Mechanics and Geotechnics in High-Rise and Underground Construction,” which was held in honor of Prof. Z. G. Ter-Martirosyan (Moscow State University of Civil Engineering, September, 2024). The problem of determining the deformation modulus of cohesive soil based on the data of the cone penetration test (CPT) with the use of machine-learning methods has been considered. The database including the results of CPT, as well as data on soil characteristics in the vicinity of boreholes were compiled to train the predictive models. Using free Python libraries, a series of machine-learning models was trained to assess the deformation modulus based on various sets of input data. The results showed that adding physical properties of soils to the input data can significantly improve the accuracy of prediction. However, only more advanced models, such as neural networks and gradient-boosting methods, are capable of fully leveraging the benefits of incorporating additional data.</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

Estimation of Soil Deformation Modulus from Cone Penetration Test Data Using Machine-Learning Methods

  • 期刊论文

  • 发表日期：

    2025年3月1日

查看全文
  • I. V. Ofrikhter,
  • A. B. Ponomarev

摘要

The article is based on the presentation at the international conference “Soil Mechanics and Geotechnics in High-Rise and Underground Construction,” which was held in honor of Prof. Z. G. Ter-Martirosyan (Moscow State University of Civil Engineering, September, 2024). The problem of determining the deformation modulus of cohesive soil based on the data of the cone penetration test (CPT) with the use of machine-learning methods has been considered. The database including the results of CPT, as well as data on soil characteristics in the vicinity of boreholes were compiled to train the predictive models. Using free Python libraries, a series of machine-learning models was trained to assess the deformation modulus based on various sets of input data. The results showed that adding physical properties of soils to the input data can significantly improve the accuracy of prediction. However, only more advanced models, such as neural networks and gradient-boosting methods, are capable of fully leveraging the benefits of incorporating additional data.