Scholarlink

<p>Accurate prediction of the burst pressure of corroded pipes is crucial for preventing their failure. This study integrates finite element method (FEM) simulations with machine learning (ML) models to predict the burst pressure with high accuracy. FEM analyses, employing both solid and shell elements, were performed on offshore API 5L X65 steel pipes with real-world multiple corrosion defects. The burst pressure was estimated by applying a factor of 1.05 to the instability criterion. Shell elements yielded results comparable to those of solid elements, while also significantly reducing computational time from 55.82 h to 25.75 h. Three ML models—multilayer perceptron, Gaussian process, and support vector machine—were developed based on three different input groups. Among them, the support vector machine demonstrated the best performance, achieving the highest coefficient of determination (<i>R</i><sup>2</sup> = 0.95). SHapley Additive exPlanations (SHAP) analysis identifies average defect depth as the most influential parameter, contributing 56.21% to predictions and exhibiting an inverse correlation with burst pressure, aligning with real-world behavior.</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

Burst Pressure Prediction of Corroded Pipes Using Finite Element Analysis–Machine Learning Models Considering Pigging Data: A Case Study of Offshore Pipelines

  • 期刊论文

  • 发表日期：

    2025年12月3日

查看全文
  • Sina Kooshamanesh,
  • Mohammad Mahdi HajiAbadi,
  • Vahid Salari

摘要

Accurate prediction of the burst pressure of corroded pipes is crucial for preventing their failure. This study integrates finite element method (FEM) simulations with machine learning (ML) models to predict the burst pressure with high accuracy. FEM analyses, employing both solid and shell elements, were performed on offshore API 5L X65 steel pipes with real-world multiple corrosion defects. The burst pressure was estimated by applying a factor of 1.05 to the instability criterion. Shell elements yielded results comparable to those of solid elements, while also significantly reducing computational time from 55.82 h to 25.75 h. Three ML models—multilayer perceptron, Gaussian process, and support vector machine—were developed based on three different input groups. Among them, the support vector machine demonstrated the best performance, achieving the highest coefficient of determination (R2 = 0.95). SHapley Additive exPlanations (SHAP) analysis identifies average defect depth as the most influential parameter, contributing 56.21% to predictions and exhibiting an inverse correlation with burst pressure, aligning with real-world behavior.