Scholarlink

<p>Uncertainty quantification (UQ) is essential for the reliable design of offshore plate anchors, where prediction accuracy is influenced by uncertainties in model parameters, soil properties, loading conditions, and numerical modelling. This paper presents a unified UQ framework that integrates Bayesian inference with stochastic Polynomial Chaos Expansion (PCE) surrogate modelling to efficiently quantify multiple sources of uncertainty in plate anchor capacity under combined vertical–horizontal–moment (V–H–M) loading. While surrogate modelling techniques have been applied to selected offshore foundation problems, their application to comprehensive uncertainty quantification of plate anchors under combined loading remains limited. In particular, addressing model parameter, soil strength and loading uncertainties, as well as numerical approximation error effects within a single framework has not previously been investigated. The proposed UQ framework is demonstrated through a simplified plate anchor case study and comprises two complementary workflows: (1) Workflow A, this method integrates Bayesian calibration of an analytical yield-surface model with a PCE surrogate, providing a controlled validation of the PCE methodology prior to its deployment on finite element simulations. This method enables evaluation of the impacts of yield surface model parameter uncertainty, undrained shear strength uncertainty, and numerical approximation error (mesh discretisation) on the estimated anchor capacities; (2) Workflow B, this method applies PCE directly to replace the computationally expensive finite element simulation of plate anchor for efficient uncertainty quantification, particularly focusing on the effects of undrained shear strength and loading inclination uncertainty on the estimated anchor capacities. The PCE surrogates in both workflows achieved excellent predictive performance with RMSE between 0.5 to 2.1&#xa0;kN/m for workflow A and less than 0.5&#xa0;kN/m for workflow B, over both vertical and horizontal capacities, while significantly reducing computational cost. The results demonstrate that uncertainties in both soil strength and loading inclination led to significant uncertainty in the predicted anchor capacities. Also, mesh discretisation (numerical approximation error) influenced the inferred phenomenological yield-surface parameters and introduced considerable uncertainty into the estimated anchor capacities. Although demonstrated for offshore plate anchors, the proposed framework is readily transferable to uncertainty quantification problems across a broad range of geotechnical 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

Stochastic Polynomial Surrogate Models for Uncertainty Quantification of Offshore Plate Anchor Capacity

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月21日

查看全文
  • Negin Yousefpour,
  • Bo Wang,
  • Changjian Zhou,
  • Alessio Mentani

摘要

Uncertainty quantification (UQ) is essential for the reliable design of offshore plate anchors, where prediction accuracy is influenced by uncertainties in model parameters, soil properties, loading conditions, and numerical modelling. This paper presents a unified UQ framework that integrates Bayesian inference with stochastic Polynomial Chaos Expansion (PCE) surrogate modelling to efficiently quantify multiple sources of uncertainty in plate anchor capacity under combined vertical–horizontal–moment (V–H–M) loading. While surrogate modelling techniques have been applied to selected offshore foundation problems, their application to comprehensive uncertainty quantification of plate anchors under combined loading remains limited. In particular, addressing model parameter, soil strength and loading uncertainties, as well as numerical approximation error effects within a single framework has not previously been investigated. The proposed UQ framework is demonstrated through a simplified plate anchor case study and comprises two complementary workflows: (1) Workflow A, this method integrates Bayesian calibration of an analytical yield-surface model with a PCE surrogate, providing a controlled validation of the PCE methodology prior to its deployment on finite element simulations. This method enables evaluation of the impacts of yield surface model parameter uncertainty, undrained shear strength uncertainty, and numerical approximation error (mesh discretisation) on the estimated anchor capacities; (2) Workflow B, this method applies PCE directly to replace the computationally expensive finite element simulation of plate anchor for efficient uncertainty quantification, particularly focusing on the effects of undrained shear strength and loading inclination uncertainty on the estimated anchor capacities. The PCE surrogates in both workflows achieved excellent predictive performance with RMSE between 0.5 to 2.1 kN/m for workflow A and less than 0.5 kN/m for workflow B, over both vertical and horizontal capacities, while significantly reducing computational cost. The results demonstrate that uncertainties in both soil strength and loading inclination led to significant uncertainty in the predicted anchor capacities. Also, mesh discretisation (numerical approximation error) influenced the inferred phenomenological yield-surface parameters and introduced considerable uncertainty into the estimated anchor capacities. Although demonstrated for offshore plate anchors, the proposed framework is readily transferable to uncertainty quantification problems across a broad range of geotechnical engineering applications.