CFD modelling of axial fans for thermal management applications

Abstract: Reducing CO2 emissions is one of the key challenges in todays automotive industry. Different strategies to achieve this goal are, for example, to electrify the powertrain and to reduce the aerodynamic drag by reducing the amount of cooling air through the underhood. In order not to sacrifice the cooling performance, a better understanding of the air flow through the underhood compartment is necessary. Computational Fluid Dynamics (CFD) is an important tool for the investigation of the underhood flow, since it gives the possibility to look at the flow field even in areas where measurement equipment cannot reach. This work focusses on the simulation of the main driving factor of underhood flow: the axial cooling fan. In CFD multiple methods of varying complexity are available to simulate the fan rotation. The more computational expensive approach, called the Rigid Body Motion (RBM), physically resolves the rotation, while the less expensive approach, called the Multiple Reference Frame (MRF), rotates the air in proximity to a stationary rotor around the blades. In the first study presented in this thesis, the flow field downstream of an axial fan is investigated for these two methods and in addition, a third hybrid approach is also evaluated. The results are compared to experimental data, which were obtained in the Volvo Cars Model-scale Wind Tunnel by using Laser Doppler Anemometry. In the second study, the interaction between the MRF approach and different upstream and downstream conditions were investigated. For this study, a temperature variation upstream of the rotor is introduced, as well as different geometric obstacles up- and downstream. As a result of the presented investigations it could be confirmed that the RBM approach gives the best representation of the flow field downstream of the fan. The MRF on the other hand showed severe limitations that the users need to be aware of. Especially the rotation of an upstream temperature source, that was found in Paper II, can lead to misleading results of component temperatures. However, when the temperature field is uniformly distributed or of minor importance, the MRF is still the most feasible approach for large applications.