Scholarlink

<p>Corrosion can significantly impact the structural integrity of marine and industrial vessels. While numerous studies have investigated and quantified corrosion, accurately estimating its effects on structural performance remains difficult. This research considered the influence of the pitting corrosion parameters on the ultimate strength of unstiffened metal plates subjected to uniaxial in-plane compression. Therefore, the dependable published results from finite element method (FEM) simulations were utilized to create data-driven models to predict the reduction in ultimate strength (<i>R</i>) based on different pit geometries and plate slenderness. To achieve this, three white box data-driven approaches, including classification and regression trees (CART), gene expression programming (GEP), and group method of data handling (GMDH), were employed to derive explicit formulas for estimating <i>R</i>. The developed models exhibited strong predictive capabilities. However, the GMDH approach has the best predictive accuracy, followed by GEP and CART methods. In addition, SHapley Additive exPlanations approach and sensitivity analysis were performed to identify the contribution of each variable to the prediction of <i>R</i>. The uncertainty analysis evaluated the reliability of the developed models. The proposed formulas derived from white box models offer practical expressions for estimating the residual strength of corroded plates, providing essential support for assessments of structural safety.</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

Application of CART, GMDH, and GEP Methods for Estimation of Ultimate Strength of Unstiffened Plates with Pitting Corrosion

  • 期刊论文

  • 发表日期：

    2026年2月24日

查看全文
  • Ning Li

摘要

Corrosion can significantly impact the structural integrity of marine and industrial vessels. While numerous studies have investigated and quantified corrosion, accurately estimating its effects on structural performance remains difficult. This research considered the influence of the pitting corrosion parameters on the ultimate strength of unstiffened metal plates subjected to uniaxial in-plane compression. Therefore, the dependable published results from finite element method (FEM) simulations were utilized to create data-driven models to predict the reduction in ultimate strength (R) based on different pit geometries and plate slenderness. To achieve this, three white box data-driven approaches, including classification and regression trees (CART), gene expression programming (GEP), and group method of data handling (GMDH), were employed to derive explicit formulas for estimating R. The developed models exhibited strong predictive capabilities. However, the GMDH approach has the best predictive accuracy, followed by GEP and CART methods. In addition, SHapley Additive exPlanations approach and sensitivity analysis were performed to identify the contribution of each variable to the prediction of R. The uncertainty analysis evaluated the reliability of the developed models. The proposed formulas derived from white box models offer practical expressions for estimating the residual strength of corroded plates, providing essential support for assessments of structural safety.