Non-intrusive reduced order modeling of fluid flows via finite element inspired graph neural network

Key Points

• Enhanced fluid dynamics modeling using a graph neural network approach improves performance, allowing for faster simulations.
• The study demonstrates reduced order modeling techniques with non-intrusive methods that maintain accuracy in computational fluid dynamics.
• Employing a graph neural network framework, the analysis exhibits effective reduction of complexity in simulations for fluid flows.
• Implications highlight the potential for broader applications in engineering, benefiting from more efficient and scalable modeling methods.