Samir Yahiaoui

@conference{rabia_4343,

title = {Physics-Regularized Residual Learning for Nonlinear Urban Turbulence Reconstruction},

author = {Aminallah Rabia and Samir Yahiaoui and Hani Hamdan and Jinan Charafeddine},

url = {https://rnl2026.sciencesconf.org/data/pages/programme
2026.pdf},

year  = {2026},

date = {2026-03-01},

booktitle = {29ème rencontre du non-linéaire},

address = {Paris, France},

abstract = {Airflow in urban environments is governed by the nonlinear dynamics of the turbulent flow, including vortex shedding, shear-layer instabilities, and wake interactions across multiple spatial scales. While Reynolds-averaged Navier-Stokes (RANS) models provide computational efficiency, they systematically damp nonlinear turbulent structures resolved by Large Eddy Simulation (LES). Bridging this fidelity gap remains a central challenge in the efficient approximation of complex flow dynamics. We formulate this problem as the learning of a correction operator that maps low-fidelity RANS solutions to LES-consistent fields. Instead of directly approximating the high-fidelity solution, we adopt a residual formulation that models the discrepancy between RANS and LES predictions. This approach preserves large-scale flow structure while focusing the learning process on unresolved interactions. To maintain physical consistency, the surrogate is regularized by divergence-free constraints and gradient consistency of turbulent kinetic energy, embedding structural properties of incompressible flow into the learning objective. Across multiple urban geometries and Reynolds regimes, the proposed framework significantly reduces velocity and turbulence errors   while recovering coherent structures such as recirculation zones and shear layers. Leave-one-geometry-out experiments demonstrate robust transfer to unseen layouts. These results suggest that physics-regularized residual operator learning provides an efficient pathway toward approximating LES-level turbulence dynamics while preserving key nonlinear flow mechanisms.},

note = {24 au 26 mars 2026

Université Paris Cité (campus Grands Moulins, Bâtiment Buffon, 15 rue Hélène Brion, 75013 Paris},

keywords = {},

pubstate = {published},

tppubtype = {conference}

}

@conference{rabia_4344,

title = {Computational Fluid Dynamics of the wind in the city},

author = {Aminallah Rabia and Amine Ammar and Alessandro Biancalani and Francisco Chinesta and Samir Yahiaoui},

url = {https://www.esilv.fr/en/icto-2024-on-augmented-intelligence-for-smarter-societies-hosted-at-dvrc/},

year  = {2024},

date = {2024-06-01},

booktitle = {ICTO 2024:"on Augmented Intelligence for Smarter Societies"},

address = {Nanterre, France},

abstract = {The nonlinear dynamics of turbulence formed by the wind flowing near solid objects can be

studied with a variety of different physical models, more or less numerically demanding. An application

is the study of the formation of turbulence past buildings, with the goal of determining

the no-fly zones for drones in smart cities.

Our purpose is to find a compromise between the hardware requirements and the result's accuracy

while exploiting the RANS (Reynolds-averaged Navier-Stokes) and hybrid based turbulence

models.

Specifically the k-Omega Shear Stress Transport model, which provides accurate prediction of

flow separation, is compared to the results with those of the Delayed Detached Eddy Simulation

(DDES, specifically Spalart-Allmaras DDES) model, more physically comprehensive, but more

numerically demanding. Both successfully predicted the average velocity distribution and the average

Reynolds stress distribution behind the obstacle [[2], [3]]. However, the predicted loads for

simplified geometry for the drone modeled in this study as a cube at certain locations differed by

a factor of 4 depending on the turbulence model employed. The analysis of the effect of different

turbulence models on a group of simplified drone models with fixed positions close to buildings

allows us to check and develop the no-fly zones. In this work, the difference between the DDES

and URANS models is investigated for the estimation of the turbulence intensity and the total

forces acting on the bluff bodies.

Despite certain limitations in the URANS model, it is still a practical alternative considering

the cost advantages and the great demand for meshes in urban simulation. Subsequent studies

could develop upon this to correct the results, such as adding a safety factor in predicting the

level of turbulence, to ensure the accurate use of drones in urban environments. Additionally,

the complexity of urban models could be simplified by using Reduced Order Modelling to allow

for the possibility of urban simulation by DDES.},

keywords = {},

pubstate = {published},

tppubtype = {conference}

}

@conference{li_4342,

title = {Numerical computation of a turbulent wind flow over buildings and estimation of its effect on drone's model},

author = {Chaozhen Li and Amine Ammar and Alessandro Biancalani and Francisco Chinesta and Aminallah Rabia and Samir Yahiaoui},

url = {http://nonlineaire.univ-lille1.fr/SNL/media/2024/programme/ProgrammeRNL2024.pdf},

year  = {2024},

date = {2024-03-01},

booktitle = {27e Rencontre du Non-Linéaire},

address = {Paris, France},

abstract = {The nonlinear dynamics of turbulence formed by the wind flowing near solid objects can be studied

with a variety of different physical models, more or less numerically demanding. An application is the

study of the formation of turbulence past buildings, with the goal of determining the no-fly zones for

drones in smart cities.

In this paper, we examine the Unsteady Reynolds-averaged Navier-Stokes (URANS, specifically kOmega Shear Stress Transport) model, which provides accurate prediction of flow separation than other

RANS models. We compare the results with those of the Delayed Detached Eddy Simulation (DDES,

specifically Spalart-Allmaras DDES) model, more physically comprehensive, but more numerically demanding. Both successfully predicted the average velocity distribution and the average Reynolds stress

distribution behind the obstacle [2,3]. However, the predicted loads for simplified geometry for the drone

modeled in this study as a cube at certain locations differed by a factor of 4 depending on the turbulence

model employed. The analysis of the effect of different turbulence models on a group of simplified drone

models with fixed positions close to buildings allows us to check and develop the no-fly zones. In this

work, the difference between the DDES and URANS models is investigated for the estimation of the

turbulence intensity and the total forces acting on the bluff bodies.

Despite certain limitations in the URANS model, it is still a practical alternative considering the cost

advantages and the great demand for meshes in urban simulation. Subsequent studies could develop

upon this to correct the results, such as adding a safety factor in predicting the level of turbulence, to

ensure the accurate use of drones in urban environments. Additionally, the complexity of urban models

could be simplified by using Reduced Order Modelling to allow for the possibility of urban simulation by

DDES},

note = {18 - 20 mars 2024

LIEU : Université Paris Cité, amphithéâtre

Buffon, 15 rue Hélène Brion, Paris 13e.},

keywords = {},

pubstate = {published},

tppubtype = {conference}

}