Session: 05-02-02 CFD Methods

Paper Number: 86921

86921 - Flow Simulation and Investigation Around a Estate Vehicle Using Hybrid Methods 

Flow simulation and investigation around a Estate vehicle using hybrid methods Francois Delassaux (1,2), Iraj Mortazavi (1), Vincent Herbert (2) & Charles Ribes (2)

(1) M2N Lab, CNAM Paris, France. (2) Stellantis, Velizy-Villacoublay, France.

In this work, numerical simulations are performed and validated with homemade experiments in order to accurately predict the flow around a real estate car. Numerical parameters and grids have been fitted and compared to experiments in order to reproduce the flow accurately.

The automotive industry is facing drastic restrictions regarding greenhouse gas emissions, with the objective of 95 g/km of CO2 emission from 2020. To reach this goal, reducing fuel consumption from aerodynamics drag is one of the main issues for engineers. Bluff body flows are characterized by regions of separated flows containing wide spectra of turbulent scales. These regions are mainly responsible for drag and lift forces applied on the body. Turbulence modeling must be capable of giving a fair prediction of separation to accurately reproduce the global flow features. Hence, the aim of this work is to study the flow around a realistic car geometry (estate vehicle), using massively parallel numerical simulations and validating them with homemade experiments. It is widely known that Reynolds-Averaged Navier-Stokes (RANS) methods failed to predict the reattachment area on the rear slanted surface of the body. The hybrid RANS/LES (Large Eddy Simulations) methods are then necessary to accurately predict such a flow topology while controlling the numerical cost of these methods. The Delayed Detached Eddy Simulation (DDES) model is used in this work. This method allows a quite affordable computational cost in attached boundary layers using RANS modeling, and accurate prediction when flow separation occurs with LES resolution. In order to get the best numerical procedure, the grid design is as critical as the model influence. In order to achieve an accurate understanding of the flow behavior around the car geometry, a wide range of experiments are performed measuring forces, pressure coefficients, using 2D-PIV and tomography for 3D description of the flow in the wake of the body. Such a numerical/experimental comparative study is crucial, as the flow features around a real car are very complex. Thus, a robust numerical method is required to predict different complex flow topologies (attached/detached areas, A-pillar and C-pillar vortices, etc.), tuned to ensure a good compromise between computational cost and physical accuracy.

A finite volume method is used in this work. Mesh is designed as follows: the surface grid is made of triangles, the boundary layer is discretized using layers of prisms cells (around the body, on the ground and the roof) and the domain is filled with tetrahedron cells. Refinement boxes for tetrahedron cells are used in the wake of the body, especially to capture flow separation. Accordingly, the tetrahedrons cells are refined around the back of the body to ensure a smooth transition between prism and tetrahedron cells.

The inlet boundary condition is defined as velocity inlet with Vinf = 33 m/s for the vehicle due to measurements. A pressure outlet condition is applied to the exit surface, with gauge pressure equal to 0 Pa. No-slip wall boundary conditions have been applied on the body based on the integration of the governing equations down to the wall itself. The numerical time step dt = 5.10−5s ensures a Courant number around 1 in LES regions of the flow. As mentioned in the previous section, DDES model is used for prediction and analysis of the flow. RANS underlying model is the SST k-ω model. Concerning numerical schemes, 2nd order schemes are used both in space and time. The SIMPLE algorithm resolves the pressure/velocity coupling equations.

Numerical results will be shown at the conference.

Presenting Author: Iraj Mortazavi M2N-CNAM