Scholarlink

<p>Predicting consolidation settlement under staged loading is challenging due to the nonlinear and time-dependent response of soft soils. Traditional methods oversimplify soil parameters during back-analysis and rely on trial-and-error approaches, which are time consuming and prone to inaccuracies. To address these limitations, this study proposes a physics-informed neural network (PINN) framework for predicting consolidation settlement under staged loading conditions. The framework integrates field-measured settlement data from magnetic extensometers with governing equations derived from fundamental consolidation theories. In contrast to conventional data-driven models, the PINN incorporates physical laws directly into the loss function, enabling inverse modeling to back-analyze essential soil parameters, including the coefficient of consolidation and final settlement. The PINN’s capability to approximate governing partial differential equations was first validated by comparison with finite element method results. Subsequently, the accuracy of predictions was assessed using 51 test cases obtained from a study site at Busan Newport, South Korea. Results showed that the PINN achieved an average root mean square error of 6.85&#xa0;cm, outperforming the long short-term memory and artificial neural network models. Additionally, final settlement predictions by the PINN showed a lower error compared with conventional methods. A comparative analysis with the finite element method (FEM) showed that, although FEM successfully reproduced staged loading behavior, its predictive accuracy was lower than that of the PINN. These findings confirm that the PINN framework offers improved accuracy and generalizability, especially in staged loading scenarios. This physics-integrated, data-driven model provides a robust solution for practical applications in construction management.</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

Physics-informed neural networks for back-analysis and consolidation settlement prediction using field measurements

  • 期刊论文

  • 发表日期：

    2025年12月24日

查看全文
  • Seongho Hong,
  • Sung-Ryul Kim

摘要

Predicting consolidation settlement under staged loading is challenging due to the nonlinear and time-dependent response of soft soils. Traditional methods oversimplify soil parameters during back-analysis and rely on trial-and-error approaches, which are time consuming and prone to inaccuracies. To address these limitations, this study proposes a physics-informed neural network (PINN) framework for predicting consolidation settlement under staged loading conditions. The framework integrates field-measured settlement data from magnetic extensometers with governing equations derived from fundamental consolidation theories. In contrast to conventional data-driven models, the PINN incorporates physical laws directly into the loss function, enabling inverse modeling to back-analyze essential soil parameters, including the coefficient of consolidation and final settlement. The PINN’s capability to approximate governing partial differential equations was first validated by comparison with finite element method results. Subsequently, the accuracy of predictions was assessed using 51 test cases obtained from a study site at Busan Newport, South Korea. Results showed that the PINN achieved an average root mean square error of 6.85 cm, outperforming the long short-term memory and artificial neural network models. Additionally, final settlement predictions by the PINN showed a lower error compared with conventional methods. A comparative analysis with the finite element method (FEM) showed that, although FEM successfully reproduced staged loading behavior, its predictive accuracy was lower than that of the PINN. These findings confirm that the PINN framework offers improved accuracy and generalizability, especially in staged loading scenarios. This physics-integrated, data-driven model provides a robust solution for practical applications in construction management.