Computational update of a statistical surrogate model for nonlinear stochastic dynamics using partial target dataset in the context of aerospace nozzle analysis - Université Gustave Eiffel
Communication Dans Un Congrès Année : 2024

Computational update of a statistical surrogate model for nonlinear stochastic dynamics using partial target dataset in the context of aerospace nozzle analysis

Résumé

The present research deals with a computational application of a methodology that consists in identifying a stochastic surrogate model with respect to a small incomplete target dataset [1]. The investigated structure is a three-dimensional engine nozzle, made of an elastic homogeneized material and subjected to an internal stochastic pressure jet and that is assumed to undergoes large displacements. A small target dataset consisting in a subset of normal accelerations located at the exit of the nozzle and expressed in the frequency domain is assumed to be available. A parameterized stochastic nonlinear computational model (SNLCM) of the nozzle for which controlled parameters describe the isotropic part of the elastic material and the spectrum dispersion of the stochastic load and for which uncontrolled parameters describe the anisotropic part of the material is constructed. Given the complexity of such highly nonlinear SNLCM, high computational costs are needed to get one response for a given set of parameters. First, a rough grid of controlled parameters is generated and the SNLCM is used without uncontrolled parameters to get the quantities of interests (QoI) corresponding to the small incomplete target dataset. The parameterized SNLCM is also used with random values of controlled and uncontrolled parameters for constructing a small training dataset. This latter one describes the realizations of the controlled parameters, of the corresponding random responses located at the exit of the nozzle and of the corresponding QoI related to the target. Then, the PLoM algorithm, which is based on a purely probabilistic approach [2,3] is used and adapted to the constraint of an existing incomplete target data et [1] in order to construct a surrogate computational model whose learning set is constituted of realizations of controlled parameters and of the corresponding QoI. The updating of such surrogate computational model allows then to match for the best with the available incomplete target data set. REFERENCES: [1] C. Soize, R. Ghanem, Probabilistic-learning-based stochastic surrogate model from small incomplete datasets for nonlinear dynamical systems, Computer Methods in Applied Mechanics and Engineering, doi:10.1016/j.cma.2023.116498, 418, 116498, pp.1-25 (2024). [2] C. Soize, R. Ghanem, Data-driven probability concentration and sampling on manifold, Journal of Computational Physics, doi:10.1016/j.jcp.2016.05.044, 321, 242-258 (2016). [3] C. Soize, R. Ghanem, Probabilistic learning on manifolds (PLoM) with partition, International Journal for Numerical Methods in Engineering, doi:10.1002/nme.6856, 123(1), 268-290 (2022).
Fichier non déposé

Dates et versions

hal-04683398 , version 1 (02-09-2024)

Identifiants

  • HAL Id : hal-04683398 , version 1

Citer

Evangéline Capiez-Lernout, Olivier Ezvan, Christian Soize. Computational update of a statistical surrogate model for nonlinear stochastic dynamics using partial target dataset in the context of aerospace nozzle analysis. The 9th European Congress on Computational Methods in Applied Sciences and Engineering (ECCOMAS 2024), Jun 2024, Lisboa, Portugal. ⟨hal-04683398⟩
10 Consultations
0 Téléchargements

Partager

More