Exploring the behavior of complex industrial problems might become burdensome, especially in high-dimensional design spaces. Reduced Order Models (ROMs) aim to minimize the computational effort needed to study different design choices by exploiting already available data. In this work, we propose a methodology where the full-order solution is replaced with a Proper Orthogonal Decomposition based ROM, enhanced by a multi-fidelity surrogate model. Multi-fidelity approaches allow to exploit heterogeneous information sources, and consequently reduce the cost of creating the training data needed to build the ROM. To explore the multi-fidelity ROM capabilities, we present and discuss results and challenges for an automotive aerodynamic application, based on a geometric morphing of the DrivAer test case with multi-fidelity fluid-dynamics simulations.
Integration of multi-fidelity methods in parametrized non-intrusive reduced order models for industrial applications
Dicech, Fausto
Primo
;Parussini, Lucia;Spagnolo, Anna;
2025-01-01
Abstract
Exploring the behavior of complex industrial problems might become burdensome, especially in high-dimensional design spaces. Reduced Order Models (ROMs) aim to minimize the computational effort needed to study different design choices by exploiting already available data. In this work, we propose a methodology where the full-order solution is replaced with a Proper Orthogonal Decomposition based ROM, enhanced by a multi-fidelity surrogate model. Multi-fidelity approaches allow to exploit heterogeneous information sources, and consequently reduce the cost of creating the training data needed to build the ROM. To explore the multi-fidelity ROM capabilities, we present and discuss results and challenges for an automotive aerodynamic application, based on a geometric morphing of the DrivAer test case with multi-fidelity fluid-dynamics simulations.File | Dimensione | Formato | |
---|---|---|---|
1-s2.0-S1877750324003041-main.pdf
accesso aperto
Tipologia:
Documento in Versione Editoriale
Licenza:
Creative commons
Dimensione
3.75 MB
Formato
Adobe PDF
|
3.75 MB | Adobe PDF | Visualizza/Apri |
Pubblicazioni consigliate
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.