ESTECO sent two researchers o Inria in July 2023 to investigate how to combine Uncertainties Quantification (UQ) techniques with Reduced Order Model (ROM) and study the use of ROM within an MDO framework for industrial applications.
Knowledge about the theoretical background of the MDO tools (modeFRONTIER) and SPDM platforms (VOLTA) has been transferred to the other secondees and the hosts. Similarly, knowledge about the industrial ROM tools has been gained exploiting the collaborations with the other partners (Optimad).
An UQ method based on nonintrusive Snapshot Proper Orthogonal Decomposition (POD) ROM and Polynomial Chaos Expansion (PCE) has been developed to efficiently compute the uncertainty propagation on a vectorial field of interest due to uncertain input parameters. The method is applied to the CFD problem of the flow over an airfoil with parameterized uncertain angle of attack and inflow velocity to estimate the uncertainties of the pressure field in each mesh cell. The CFD models have been integrated in the MDO platform modeFRONTIER and VOLTA to automate the sampling of the snapshots used to train the ROM and to share them among the users who need to use the results for their own studies.
The method has been tested also on a thermal model of a rotating disk/heat source system for which an analytical solution is available. The model exposes several (possible) stochastic variables allowing to study the effect of the number of variables on the UQ strategy and the computational time required.
ROM tools from other partners (OPTIMAD) have been integrated into VOLTA and modeFRONTIER framework. This enables to share, projects, document and knowledge and so to democratize the design process among the research teams. This enables also to use the sampling strategies (Lipschitz sampling and Gaussian Mixture Distribution) developed during the previous secondees by ESTECO and integrated in the same platform VOLTA.