Scholarlink

<p>The utilization of secant pile to support soil excavation is common in earthwork because requiring less construction space and efficient installation time. However, a fast method for preliminary design has yet to be revealed in previous studies, especially under extreme climate conditions. This study provides an alternative method to analyse secant pile stability due to extreme rainfall conditions using machine learning technique. Prior to this analysis, the determination of pore water pressure (PWP) using the same technique was also proposed in this study. To obtain the research aim, 672&#xa0;models have been analysed using finite element method to investigate PWP due to extreme rainfall condition and to analyse factor of safety (FOS), considering the variation of soil cohesion (c), excavation and depth ratio (D/H), and soil unit weight (γ). Afterwards, two machine learning methods, i.e. support vector machine (SVM) and artificial neural network (ANN), have been deployed to generate the model. Regarding the result, three variations of return period (RP) of rainfall trigger different water infiltration depth, attributed to the discrepancy of the depth of persistent zone. It can also be reported that ANN produced more accurate result, compared to SVM method, indicated by higher value of coefficient of determination (R<sup>2</sup>) for both PWP and FOS analysis. Moreover, the alteration in c provides a more notable impact on FOS value than that the change in D/H and γ. In term of extreme weather condition, 25-year and 50-year RP of rainfall promotes the decline in FOS by around 6.1 and 9.9%, respectively, compared to 5-year RP of rainfall. This research is expected to provide a fast and accurate design method for secant piles to prevent stability problems, especially under extreme weather conditions.</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

The Use of Machine Learning for Analysing Hydrological Behaviour and Stabilization of Soil Excavation, Supported by Secant Pile Due to Extreme Rainfall

  • 期刊论文

  • 发表日期：

    2025年7月8日

查看全文
  • Arwan Apriyono,
  • Sumiyanto,
  • Thitinan Indhanu,
  • Sony Pramusandi,
  • Priswanto,
  • Paulus Setyo Nugroho

摘要

The utilization of secant pile to support soil excavation is common in earthwork because requiring less construction space and efficient installation time. However, a fast method for preliminary design has yet to be revealed in previous studies, especially under extreme climate conditions. This study provides an alternative method to analyse secant pile stability due to extreme rainfall conditions using machine learning technique. Prior to this analysis, the determination of pore water pressure (PWP) using the same technique was also proposed in this study. To obtain the research aim, 672 models have been analysed using finite element method to investigate PWP due to extreme rainfall condition and to analyse factor of safety (FOS), considering the variation of soil cohesion (c), excavation and depth ratio (D/H), and soil unit weight (γ). Afterwards, two machine learning methods, i.e. support vector machine (SVM) and artificial neural network (ANN), have been deployed to generate the model. Regarding the result, three variations of return period (RP) of rainfall trigger different water infiltration depth, attributed to the discrepancy of the depth of persistent zone. It can also be reported that ANN produced more accurate result, compared to SVM method, indicated by higher value of coefficient of determination (R2) for both PWP and FOS analysis. Moreover, the alteration in c provides a more notable impact on FOS value than that the change in D/H and γ. In term of extreme weather condition, 25-year and 50-year RP of rainfall promotes the decline in FOS by around 6.1 and 9.9%, respectively, compared to 5-year RP of rainfall. This research is expected to provide a fast and accurate design method for secant piles to prevent stability problems, especially under extreme weather conditions.