CSM9: CRE’s contributions to 9th Computational Stochastic Mechanics Conference in Corfu

From June 21–25, 2026, Nataly Manque Roa from the Chair for Reliability Engineering participated in the 9th International Conference on Computational Stochastic Mechanics (CSM9), held in Corfu, Greece. Representing CRE, she presented recent research on uncertainty quantification under geometric variability, with a focus on combining isogeometric analysis and sensitivity-informed surrogate modelling. 

Isogeometric-Sensitivity-Informed Polynomial Chaos Expansion 

Abstract:

Surrogate models, such as polynomial chaos expansion (PCE), offer an efficient alternative to direct numerical simulations for uncertainty quantification. Nevertheless, when uncertainty arises from geometric variations, the physical domain itself changes, making traditional finite element analysis (FEA) workflows computationally expensive due to mesh updates or remeshing. This study explores the combination of isogeometric analysis (IGA) with sensitivity-informed PCE to address geometric uncertainty. IGA provides a geometry-consistent framework in which the same NURBS-based representation connects the geometry, analysis model, integration maps, response evaluation, and variational sensitivities. The resulting IGA-derived sensitivities are incorporated into the PCE construction, where response values fit the surrogate, while sensitivity information constrains its gradients. The methodology is demonstrated on two- and three-dimensional linear elasticity problems with uncertain geometric parameters, using stress triaxiality as a stress-based quantity of interest. The results show that sensitivity-enhanced PCE improves response accuracy and gradient fidelity, especially in low-sample regimes, while requiring only moderate additional computational cost. The study also motivates future extensions toward active learning and adaptive polynomial enrichment guided by IGA sensitivities.

Conference details sourced from: CSM9 Official Website