A Design of Experiment for Evaluating Installation Effects and the Influence of Blade Loading on the Aeroacoustics of an Automotive Engine Cooling Fan

D2 - Aerodynamic Design

Aeroacoustic performances of cooling fans in automotive industry have become a quality factor. However, tighter packaging and the quest for higher performance have made it more difficult to create quieter fans.
A specific design of experiment (DOE) has been carried out to assess the influence of various, commonly encountered geometrical parameters on the aeroacoustics of such cooling fans. It provided the necessary data to help find the best compromise between aerodynamics and aeroacoustics for a given packaging.

The work was conducted in two phases, the first one being a numerical study on blade loading, and the second one being an experimental study on the fan's interaction with its support:

• 3 fans delivering the same pressure rise at the target operating point were designed by the means of CFD. The first blade is loaded mostly near the bottom, whereas the second has greater load near the tip, and the third is designed to operate at lower rotational speed.

• 17 supports were designed according to a Nearly Orthogonal Latin Hypercube (NOLH) design of experiment based on one physical parameter (flow rate) and 6 geometrical ones. They include the rotor-strut distance, the strut's sweep angle, aspect ratio and cross-sectional area, the tip clearance and the rotating ring's immersion in the shroud.

The fans and the supports were manufactured and tested in an anechoic chamber equipped with a flow-rate-control rig. Each of the fans was tested with all the supports at the target operating point and several other flow rates. In total, 3 designs of experiment were carried out, each corresponding to a specific type of blade loading. The collected data was then used to create response surface models for efficiency and sound power.
Maximum efficiency, minimum noise, and a compromise between both configurations were then sought via the response surface models integrated into optimization loops. The results show that loading the blade near the tip makes it less sensitive to installation effects. Furthermore, maximum efficiency and minimum noise do not occur for the same combination of parameters. Future work should include a validation of the trends obtained by optimization.