Detail publikačního výsledku

Isogeometric sensitivity-enhanced polynomial chaos expansion under geometric uncertainty

MANQUE, N.; VALDEBENITO, M.; LIEDMANN, J.; FAES, M.; NOVÁK, L.

Originální název

Isogeometric sensitivity-enhanced polynomial chaos expansion under geometric uncertainty

Anglický název

Isogeometric sensitivity-enhanced polynomial chaos expansion under geometric uncertainty

Druh

Článek WoS

Originální abstrakt

Geometric uncertainty remains a significant challenge in engineering applications, as shape variations can substantially affect mechanical response and make response-only polynomial chaos expansion (PCE) surrogates sample-demanding. Accordingly, this paper proposes an isogeometric sensitivity-enhanced PCE for systems characterized by a small set of uncertain geometric parameters. The key idea is to enrich the surrogate with geometry-consistent derivatives obtained from the same isogeometric analysis (IGA) model used to solve the mechanical problem. Within this framework, geometry, mechanical analysis, and variational sensitivity evaluation share a common NURBS-based representation. The approach is assessed through two- and three-dimensional linear elasticity benchmarks, a nonlinear Neo-Hookean problem, and a case combining geometric and material uncertainty. Across these examples, sensitivity enrichment generally improves global response and gradient prediction relative to response-only sparse PCE, with the clearest gains observed in low- and intermediate-data regimes. Cost-aware comparisons with gradient-enhanced Kriging show competitive accuracy while avoiding the additional high-fidelity response evaluations required for finite-difference gradients. Overall, the results indicate that IGA-derived variational sensitivities can improve the data efficiency of PCE-based uncertainty quantification under low-dimensional geometric variability, with gains varying across benchmarks and metrics.

Anglický abstrakt

Geometric uncertainty remains a significant challenge in engineering applications, as shape variations can substantially affect mechanical response and make response-only polynomial chaos expansion (PCE) surrogates sample-demanding. Accordingly, this paper proposes an isogeometric sensitivity-enhanced PCE for systems characterized by a small set of uncertain geometric parameters. The key idea is to enrich the surrogate with geometry-consistent derivatives obtained from the same isogeometric analysis (IGA) model used to solve the mechanical problem. Within this framework, geometry, mechanical analysis, and variational sensitivity evaluation share a common NURBS-based representation. The approach is assessed through two- and three-dimensional linear elasticity benchmarks, a nonlinear Neo-Hookean problem, and a case combining geometric and material uncertainty. Across these examples, sensitivity enrichment generally improves global response and gradient prediction relative to response-only sparse PCE, with the clearest gains observed in low- and intermediate-data regimes. Cost-aware comparisons with gradient-enhanced Kriging show competitive accuracy while avoiding the additional high-fidelity response evaluations required for finite-difference gradients. Overall, the results indicate that IGA-derived variational sensitivities can improve the data efficiency of PCE-based uncertainty quantification under low-dimensional geometric variability, with gains varying across benchmarks and metrics.

Klíčová slova

Polynomial chaos expansion (PCE); Isogeometric analysis (IGA); Geometric uncertainty; Variational sensitivity analysis; Gradient-enhanced surrogate modeling;Constrained regression

Klíčová slova v angličtině

Polynomial chaos expansion (PCE); Isogeometric analysis (IGA); Geometric uncertainty; Variational sensitivity analysis; Gradient-enhanced surrogate modeling;Constrained regression

Autoři

MANQUE, N.; VALDEBENITO, M.; LIEDMANN, J.; FAES, M.; NOVÁK, L.

Vydáno

01.10.2026

Nakladatel

Elsevier BV

Periodikum

Probabilistic engineering mechanics

Svazek

86

Číslo

October

Stát

Spojené království Velké Británie a Severního Irska

Strany od

104068

Strany počet

27

URL

BibTex

@article{BUT212948,
  author="{} and  {} and  {} and  {} and Lukáš {Novák}",
  title="Isogeometric sensitivity-enhanced polynomial chaos expansion under geometric uncertainty",
  journal="Probabilistic engineering mechanics",
  year="2026",
  volume="86",
  number="October",
  pages="27",
  doi="10.1016/j.probengmech.2026.104068",
  issn="0266-8920",
  url="https://www.sciencedirect.com/science/article/pii/S0266892026001797?dgcid=SSRN_redirect_SD&ssrnid=7370861"
}