PRIN 2022 PNRR - ROMEU

ROMEU (Reduced-Order Modeling for Environmental and Urban flows) is a research project funded by the Italian Ministry of Universities and Research and jointly led by the BioRobotics Institute – Scuola Superiore Sant’Anna (SSSA), the IMT School for Advanced Studies Lucca, and the Università Cattolica del Sacro Cuore (UCSC). The goal is to develop next-generation reduced-order models (ROMs) that enable accurate, real-time simulations of complex urban and environmental flows, characterised by slow Kolmogorov n-width decay and high Reynolds numbers.

The consortium has already produced several journal papers on non-intrusive ROMs based on artificial neural networks, segregated hybrid ROMs, and data-augmented intrusive ROMs, in addition to validated urban/environmental benchmarks associated with a hierarchical validation strategy.

Overall, the project’s approach is highly interdisciplinary, merging mathematical model reduction, machine learning, and computational fluid dynamics. By automatically generating detailed urban domains and applying advanced ROM techniques, ROMEU strives to drastically reduce computation time for large-scale problems (like city-wide pollutant dispersion or environmental flows in porous media) while maintaining accuracy. This enables faster-than-real-time simulations, which are crucial for applications such as urban air quality forecasting, micro-climate assessment, and emergency response planning.

Here follows some community efforts:

• SSSA has introduced non-linear manifold transformations (RADON-CDT and neural-network mappings) and an automated pipeline that converts public geo-referenced data into detailed 3-D urban geometries using City4CFD.
• UCSC has designed domain-decomposition ROMs and regularised / filtered models for Navier–Stokes and thermo-poroelasticity and has explored Forward Sensitivity approaches for optimal control and hierarchical ROM strategies.
• IMT has developed a Galerkin-projection reduced Navier–Stokes solver for high-Reynolds-number flows past flexible structures, integrated with an Empirical Lagrangian Interpolation Method (ELIM) for hyper-reduction, and is building a hybrid solver that couples Navier–Stokes and potential-flow domains.

Project goals

Develop Next-Generation Reduced Order Models: Formulate new ROM methodologies tailored to urban and environmental flows, which often involve slow-decaying energy spectra. This includes non-intrusive ROMs using neural networks and other non-linear transforms to capture phenomena that linear subspace methods struggle with. The goal is to improve the efficiency of simulations without sacrificing accuracy in complex scenarios.

Automate Urban Flow Simulations: Create an automated pipeline to go from geospatial city data to simulation-ready models. Using tools like City4CFD, the project reconstructs 3D city geometries (including buildings, terrain, and surface features) directly from public data. This objective ensures that detailed computational domains can be generated with minimal manual effort, facilitating widespread use of CFD in urban planning.

Handle High Reynolds and Multi-Physics Regimes: Extend ROM techniques to turbulent flows and coupled physics. For high-Reynolds-number flows in large computational domains (e.g. turbulent flow field of a city environment), ROMEU aims to efficiently incorporate turbulence modeling into a domain-decomposed ROM and to apply regularization or filtering to stabilize POD-Galerkin ROMs.The project explores also segregated multi-physics simulations connected by reduced-order coupling strategies.

Validate and Demonstrate in Real Cases: Test the developed ROMs on realistic benchmarks – such as urban pollutant dispersion and vortex-induced vibrations of structures. One objective is to achieve accurate real-time predictions of an entire day’s urban air quality by reducing a full 3D unsteady simulation to a low-dimensional model. Another is to show that a ROM can capture the key dynamics of an FSI problem (like a flexible cylinder in cross-flow) including the forces on structures. These demonstrations will validate that the ROM technology developed in ROMEU is ready for practical, large-scale scenarios.

ENTE PROMOTORE: Ministero dell’Università e della Ricerca

NOME PROGETTO: ROMEU

CUP: H53D23008900001

PERIODO E DURATA: 29/09/2023 - 29/09/2025 (24 mesi)

FINANZIAMENTO: 99.923 €

COORDINATORI: SSSA Prof. Giovanni STABILE