A combined approach to optimize by simulation the aerodynamic function of the fan system used for engine cooling in automotive application

SC/TETRA  is an all-in-one general purpose  Computational Fluid Dynamic (CFD) software using unstructured mesh. It was developed with the specific purpose of "Easily enabling calculation of complex geometries".

Automotive is among the area which mainly use CFD calculation, like at Valeo, an automotive Tier 1 supplier whose thermal system department is responsible for the design, the development and the sale of fan systems for engine cooling (among others components for cooling systems).

Valeo is committed to the company Fluorem for many years for the development of parameterized techniques for simulations. Many advances have been obtained on both the aerodynamic profiles and the blade design. Requirements for energy efficiency, compactness and noise reduction led the company to seek for and use advanced numerical simulation techniques to shorten design time and integrate optimization process.

SC-tetra tools are used in this context for theirs abilities to explore innovative concepts, among which is the new semi-radial fan design aimed to provide high efficiency under low torque. The code has produced predictions of aerodynamic overall performance that were on-line with experimental results obtained later, probably thanks to the features of automatic mesh refinement: complex phenomena are detected and predicted even with uncertain locations, and it has been the case for instance for flow separations and tip recirculations between the rotating fan and the fixed shroud.

The possibilities of the code in terms of high power computing were also explored through a program supported by a national research fund aimed to develop collaborative works and remote computing. Many highly parallelized simulations were conducted for fan studies, and for which Design of Experiment (DOE) techniques were used.

The use of the code is furthermore extended to perform LES simulations for aeroacoustic prediction thanks to the parallelization which allows obtaining source for extruded profiles in a short timeframe. The extracted statistics of turbulence are in addition used to calibrate some analogic models based on the amiet theory, whereas the wake characteristics are integrated in a Sear’s model for stator noise.


The Author

Dr Macoumba N’DIAYE