Scholarlink

<p>Concrete structures exposed to elevated temperatures undergo progressive thermal and mechanical degradation, compromising serviceability and safety. This study examines plain cement concrete (PCC) and reinforced cement concrete (RCC) beams across multiple grades subjected to hydrocarbon fire, integrating finite element modelling (FEM) with calibrated surrogate models. A curated dataset spanning temperature, grade, and beam type supports the training of Gaussian process regression (GPR), support vector regression (SVR), and gradient boosted machines (GBM). Cross validation with standardized residuals provides diagnostic calibration, while external benchmarking against independent FEM simulations ensures physical consistency. Results indicate that the effective thermal conductivity indicator <InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(k_{\textrm{eff}}\)</EquationSource> </InlineEquation>, derived from transient thermal field outputs under Eurocode-consistent temperature-dependent conductivity inputs, decreases with temperature in a manner consistent with moisture loss and microstructural evolution in heated concrete, while thermo-mechanical response reveals sharp escalation of displacements beyond <InlineEquation ID="IEq2"> <EquationSource Format="TEX">\(500^{\circ }\textrm{C}\)</EquationSource> </InlineEquation>, rapid crack width expansion after <InlineEquation ID="IEq3"> <EquationSource Format="TEX">\(350^{\circ }\textrm{C}\)</EquationSource> </InlineEquation>, and concentration of von Mises stresses at beam edges. Across most targets and materials, GPR achieves the strongest overall fidelity and provides calibrated predictive dispersion, while GBM yields the lowest PCC crack-width error and SVR exhibits the largest deviations in strongly nonlinear regimes. Design charts and three-dimensional response surfaces, overlaid with FEM samples, provide practitioner tools for risk-aware screening. The integrated FEM-surrogate framework offers computational efficiency while preserving accuracy, enabling rapid assessment of fire-induced behaviour and aligning with performance-based design strategies for resilient infrastructure.</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

A Hybrid FEM-Surrogate Enhanced Framework for Thermo-Mechanical Degradation in Concrete Beams Under Hydrocarbon Fire

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月30日

查看全文
  • Anshu Sharma,
  • Basuraj Bhowmik

摘要

Concrete structures exposed to elevated temperatures undergo progressive thermal and mechanical degradation, compromising serviceability and safety. This study examines plain cement concrete (PCC) and reinforced cement concrete (RCC) beams across multiple grades subjected to hydrocarbon fire, integrating finite element modelling (FEM) with calibrated surrogate models. A curated dataset spanning temperature, grade, and beam type supports the training of Gaussian process regression (GPR), support vector regression (SVR), and gradient boosted machines (GBM). Cross validation with standardized residuals provides diagnostic calibration, while external benchmarking against independent FEM simulations ensures physical consistency. Results indicate that the effective thermal conductivity indicator \(k_{\textrm{eff}}\) , derived from transient thermal field outputs under Eurocode-consistent temperature-dependent conductivity inputs, decreases with temperature in a manner consistent with moisture loss and microstructural evolution in heated concrete, while thermo-mechanical response reveals sharp escalation of displacements beyond \(500^{\circ }\textrm{C}\) , rapid crack width expansion after \(350^{\circ }\textrm{C}\) , and concentration of von Mises stresses at beam edges. Across most targets and materials, GPR achieves the strongest overall fidelity and provides calibrated predictive dispersion, while GBM yields the lowest PCC crack-width error and SVR exhibits the largest deviations in strongly nonlinear regimes. Design charts and three-dimensional response surfaces, overlaid with FEM samples, provide practitioner tools for risk-aware screening. The integrated FEM-surrogate framework offers computational efficiency while preserving accuracy, enabling rapid assessment of fire-induced behaviour and aligning with performance-based design strategies for resilient infrastructure.