F2008-12-087
3D Fan Modelling Strategies for Heavy Duty Vehicle Cooling Installations - CFD with Experimental Validation
The need for low fuel consumption and lower emissions is for every automotive OEM a reality. With a component in today´s trucks that typically consumes of the order of magnitude 50bHp fully engaged - reducing this loss is very important. In addition a lot of the new after treatment systems requires a higher airflow through the engine bay, and new legislations requires quieter fans, i.e. today there is a need for fans with higher efficiency.
To reach these needs and goals CFD offers great opportunities and potential. However since computer power is still limited for resolving transient Navier Stokes equations without any turbulence models and in a mesh-independent domain - one relies heavily on the performance of these models. This is what is addressed in this paper - an investigation of different fan modelling strategies and how these interfere with turbulence models and mesh-size.
Main focus is kept on modelling the fan with the use of MRF (Multiple reference frames, frozen rotor, stationary simulations) and its behaviour compared to resolving the fan rotation in a rigid body rotation (transient). The turbulence models investigated is a number of the standard two equation k-epsilon and k-omega models. This work is done by comparing numerical simulations to experimental testing in an as near as possible identical set-up.
For this paper we have also focused on a complete vehicle installation, not a component study.
In this paper it was found that it is very hard to get the fan MRF model to perform well in a complete vehicle installation, this due to a limited space of choosing a rotational domain for the fan - independent of choice of turbulence model and mesh size. Fully transient (URANS) sliding mesh approach performs well, even for small case sizes. It was found in this paper that a small case of 3 million cells sliding mesh performed superior to a case of 16 million polyhedral cells with MRF, noteworthy is that the latter consumed twice as many CPU-hours for converging.
This paper does not claim to deliver the final answer to fan modelling techniques, for the sliding mesh runs we managed to get within 2% of measured pressure for a fixed air-flow, except for the fan stalling and fan transition region. However these areas are, to the authors´ knowledge and experience hard to measure with good repetition, so a numerical issue in these regions is to be expected. The transient runs were for RNG k-epsilon model. More effort needs still to be made on better experimental data, not just measured for the system, but also component data such as radiator coefficients, larger meshes and more turbulence models should be tried transient in order to predict these regions better. However this paper gives a strong point on what type of fan modelling strategy can be worth considering for UTM simulations in the upcoming future