The 8th European Congress on Computational Methods in Applied Sciences and Engineering (ECCOMAS Congress 2022) is held in Oslo, Norway, from 5 to 9 June, 2022.
Our expert Alberto Clarich will be at the event to present Multi-fidelity simulations for multidisciplinary design optimization. He will be in the room Hedmark (GF) of the congress venue NOVA Spektrumfrom on 6 June from 11AM to 1PM.
The presentation illustrates efficient methodologies for the training of RSM (Response Surfaces), by taking advantage of multi-fidelity simulation models.
From one side, the Dataset Reducer DOE available in modeFRONTIER can be used to select from a dataset of LF (low-fidelity) simulation samples, the optimal subset where the HF (high-fidelity) simulation samples can be defined, by an internal optimization process that minimizes the regression error in the selection of the LF subset.
In a second stage, the two series of LF and HF samples can be used to feed the Cokriging RSM, which guarantees to obtain an higher accuracy than the one given by each one of the single series.
Finally, the multi-fidelity approach is applied to ROM (Reduced Order Models), to obtain a CFD model equivalent to the original one, but defined by a reduced number of parameters (POD or Proper Orthogonal Decomposition method).
The three methodologies, that can be fully integrated in a modeFRONTIER and VOLTA optimization environment, are introduced and validated by the application to aeronautical and automotive CFD test cases.
Authors of this presentation are: Luca Battaglia from ESTECO, Lucia Parussini and Carlo Poloni from the University of Trieste.
ECCOMAS is a scientific organization founded in 1992, grouping together European associations with interests in the development and applications of computational methods in applied sciences and technology.