Aeroacoustics at BEHR

Noise generated by Heating, Ventilation and Air Conditioning (HVAC) systems has long been a design concern throughout the automotive, rail and aerospace industries. Exposure to extraneous noise generated by acoustically non-optimal systems not only detracts from ride enjoyment, but also contributes significantly to driver fatigue.

BEHR (in collaboration with CD-adapco through the DESTINY:AAC and DESTINY:2 projects), have undertaken a comprehensive analysis of computational methods for predicting this phenomenon on industrial test cases.

The test case discussed here is that of a production HVAC duct (Figure 01), with some minor modifications to aid direct comparison between CFD and experiment. The experimental monitor locations ca be seen in Figure 01b

STAR-CD offers the unique capability of predicting the maximum resolvable acoustic source frequency from a steady-state calculation: the Mesh Frequency Cut-off Analysis (Figure 02). In conjunction with the Lilley source term (also Figure 02), the user is able to optimize the case set-up, purely by running steadystate calculations to ensure that the frequency range of interest will be captured. This offers a significant improvement over successive transient CFD calculations being used to adjust the mesh resolution.

Figure 02 shows the steady-state Mesh Frequency Cut-off analysis and Lilley source term for a slice through the center of the final mesh.

The mesh was produced using pro-STAR’s automatic meshing methodology, known as ‘trimmed cell technology’, and contains 3.5 million cells. A section through the center plane of the mesh can be seen in Figure 03. The mesh was locally refined in the regions in which the acoustic sources (as predicted by the Lilley source term) are strongest, such that frequencies up to approximately 1000 Hz would be captured (as indicated by the Mesh Frequency Cut-off analysis). Firstly, the advantage of this approach is that it enables the user to focus mesh refinement (which will yield higher frequency resolution) in only the most important areas. This is pertinent toAeroacoustics since mesh size contributes heavily to transient run times. Secondly, the user is able to establish the frequency range within which the model is valid. The Mesh Frequency Cutoff values for the four monitor locations were between 800 and 1000 Hz.

In order to predict accurately both narrow- and broad-band acoustic sources, transient calculations utilizing advanced turbulence models are required. Here, users of STAR-CD benefit from a transient solver that has repeatedly been shown to be up to an order of magnitude faster than conventional solvers. In this case, the PISO solver was deployed along with the DES/k-? turbulence model. The run time was 6.5 days on eight 2.8 GHz Linux processors to complete 0.5 seconds of simulated time, using a time-step of 0.4x10-4 seconds. With current CPU speeds, this remains a considerable computational cost. This limits the use of such techniques to in-depth analysis of specific problem test cases. A snapshot of the Velocity Magnitude from the DES calculation can be seen in Figure 04.

Figure 04 illustrates the very fine scales that are being resolved in the wake of the flap. Throughout the time-varying calculation, pressure values were monitored at the pressure tapping locations from the experiment. Figure 05 shows the Sound Pressure Level (SPL) against Frequency at two of the monitor locations.



Firstly, the agreement between the computed result and experiment is excellent in the low frequency range. Secondly, the deviation of the CFD result from experiment occurs at approximately 1000 Hz, beyond which point an increase in mesh resolution would be required in order to capture the turbulence scales. This agrees very well with the value predicted by STAR-CD’s steady-state Mesh Frequency Cut-off analysis.

The resolution of complex problems, such as aeroacoustic noise, will be greatly assisted by the detailed understanding achieved through studies such as these. This work is therefore part of BEHR’s ongoing quest to produce market leading HVAC systems: the aeroacoustic characteristics of which are an integral part.

Figures
01: Test case geometry and experimental monitor points
02: Mesh Frequency Cut-off analysis and Lilley source term
03: Section slice through centre of mesh
04: Snapshot of velocity magnitude from the DES calculation
05: SPL against Frequency for monitor points 2 and 4
(Experiment STAR-CD)