Pressure-velocity measurements of a small automotive fan at different working conditions. A noise generation perspective. Luana Georgiana Stoica 1 University of Roma Tre, Engineering Department Via Vito Volterra, 62, 00146 Rome, Italy Alessandro Di Marco 2 University of Roma Tre, Engineering Department Via Vito Volterra, 62, 00146 Rome, Italy Elisa de Paola 3 University of Roma Tre, Engineering Department Via Vito Volterra, 62, 00146 Rome, Italy Enrico Mollica 4 SPAL Automotive Srl Via per Carpi n. 26/B, 42015 Correggio, Italy

ABSTRACT Although noise from large fans has been extensively studied by the scientific community, few experi- mental studies have been performed on small-scale fans, where aerodynamic effects due to low Reyn- olds numbers need to be taken into account. For this reason, some preliminary experimental investi- gations have been carried out on cooling fans for automotive applications. The experimental cam- paign has been focused on the characterization of the flow outlet, and its correlation with the noise generation on a 5.5-inch five-bladed fan. Simultaneous measurements of velocity with hot-wire ane- mometry (HWA) and pressure fluctuations with microphones have been carried out, the HWA probe being located downstream, two microphones positioned in the far-field of the fan, and another mi- crophone measuring the wall pressure fluctuations on the shroud. The fan has been also studied at different working conditions, varying the rotational speed, and applying specific fine grids at the intake, in order to well reproduce the presence of the cooler. Analyses in the time and frequency domain have been performed on the measured data, including cross-correlation between pressure and velocity signals, in order to find the main driving mechanisms that are related to the noise gen- eration.

1. INTRODUCTION

Automotive noise has become a relevant focal point in automotive industry, and a major contrib- utor to the overall noise is the cooling fan. Consequently, the design process of automotive cooling fan systems cannot exclude their acoustic performances. Noise radiated by fans can be divided in two

1 luana.stoica@uniroma3.it 2 alessandro.dimarco@uniroma3.it 3 elisa.depaola@uniroma3.it 4 e.mollica@spal.it

worm 2022

main components: the tonal component occurring at the blade passage frequency harmonics, and the broadband noise component. Several mechanisms concur to this latter component. The blade or fan self-noise component is the one generated by blades operating in a clean undisturbed flow [1], thus it represents the minimum noise a fan would emit, even when no installation effects are involved. An- other contribution is represented by the noise due to the upstream turbulence ingested by the rotor, shed for instance by other parts of the automotive cooling module. An important mechanism involved is the trailing-edge noise, caused by the interaction of the blade turbulent boundary layer with the geometrical discontinuity represented by the trailing edge. In addition, blade tip vortices and leakage flows may also contribute significantly to fan noise [2]. Fan design often involves numerical simulations, such as RANS ([3]), Lattice-Boltzmann [4], LES [5], DES [6], or BEM [7], usually coupled with the Ffowcks Williams and Hawkings analogy to solve the acoustic part. Due to the complexity of the three-dimensional flow generating inside the fan, numerical simulations are very computationally expensive, thus some approximations are required to be able to analyse different design solutions, especially when noise investigation in involved. It is therefore very important to have experimental data for the validation of these codes. Few studies can be found in literature with acoustic experimental data from automotive shrouded fans. Among these, Rynell et al. [8] have carried out acoustic measurements on a cooling module composed of an automotive radiator, a shroud, a fan, a frame and a hydraulic motor, at different rotational speeds. Tip-leakage flow noise was investigated in Canepa 2016 [9] and Canepa 2019 [10], using acoustic measurements and particle image velocimetry. Amoiridis et al. [11] investigated the acoustic emissions from an automotive cooling module by means of sound source localization using beamforming and directivity measurements. In this paper, fan noise is investigated varying the radiator grid type and rpm and correlating the acoustic data with the velocity measured in the propeller wake and with the wall pressure fluctuations on the propeller shroud. Section 2 gives an overview of the experimental set-up, while in section 3, results are presented and discussed in terms of aerodynamics and aeroacoustics. 2. EXPERIMENTAL SET-UP

worm 2022

The fan under investigation is a shrouded automotive cooling fan, with a diameter of 0.14 m (5.5 in.) and five forward-skewed evenly spaced blades. Its hub to tip ratio is 0.43, and the tip clearance ratio 3%.

Figure 1: Experimental set-up. Red points indicate HWA measurement points.

The design point corresponds to a flow coefficient 𝜑= 0.162 , a head coefficient 𝜓= 0.225 , and a rotational speed of 5500 rpm. The fan rotates clockwise seen from upstream and is driven by a brushed DC motor. Velocity fluctuations were acquired using a hot wire anemometer (HWA), while acoustic pressure and wall pressure fluctuations on the shroud were acquired by three Microtech Ge- fell M360 electret microphones. All the instruments were acquired simultaneously using a NI PXI- 6143 DAQ, with a sampling frequency of 65536 Hz and an acquisition time of 20 seconds. A sketch of the experimental set-up can be seen in Figure 1.

A B

worm 2022

a) b) Figure 2: Grids used to simulate the radiator: a) type A; b) type B.

Three rotational speeds were considered and two different grid types simulating the radiator (Figure 2). Grid A is a PLA 3D printed radiator-type grid, with 4x1.3 mm cells and 5 mm thickness, while grid B is an aluminium grid with 0.5x0.6 mm cells and 0.5 mm thickness. A summary of the config- urations tested can be found in Table 1. Radial positions are expressed in terms of the fan radius R.

Table 1: Parameters varied during the tests.

rpm Grid type Radial position HWA (r/R)

5500, 5700, 5900 No grid, A, B 1.10, 1.06, 1.02, 0.97, 0.83, 0.69, 0.54, 0.49

To. ie ee Wee | MAAAN NANAAN WN) ,pAAAAG

The fan performance curves at the three rotational speeds considered are shown in Figure 3, together with the resistance determined by the two grid types. The fan curve at 5700 rpm was measured on a test bench built in agreement with the AMCA 210 [12] and ISO 5810 [13] standards. The fan curves at different rpm were built according to the scaling laws of the flow rate and of the static pressure. The grid curves were analytically approximated with parabolic laws using the velocity data measured by the HWA. The operating points in the configuration tested are thus identified. It is worth noting that when grid B is installed, the fan finds itself operating with a more centrifugal flow. This can explain the velocity trends measured.

aN dodatateteteterolnty

Figure 3: Fan performance curve at rotational speeds considered and grid resistance curves. 3. RESULTS AND DISCUSSION

In this section, the acoustic and aerodynamic data are processed, and results analysed. Power spectral densities of pressure and velocity fluctuations are computed using the Welch method, using a Ham- ming window, 50% overlap and with a Δ𝑓 of 1 Hz.

3.1. Aerodynamic characterisation The mean velocity profiles at a distance of 5 mm downstream the fan are shown in Figure 4. Meas- urements are carried out in 8 radial positions from hub to tip as shown in Figure 1. The probe was positioned in order to measure mainly the axial speed component. The rotational speed raise causes an increase in the outlet velocity, without affecting the profile shape, in all configurations.

Figure 4: Mean velocity profiles, 5 mm downstream the fan. When no grid is mounted, the outlet velocity assumes an almost parabolic profile, with the maximum values at radial positions between 0.7R and 0.9R. When grid A is installed, the profile is skewed outwards, thus the speed is lower in the area of the blade comprised between the hub and 0.8R and higher from 0.8R to 1.1R. The finer grid (B) leads to speed values up to 80% lower with respect to the other configurations. This can be explained looking at Figure 3, noting that the flow becomes

worm 2022

more radial for this configuration. The speed profile for grid B is almost uniform, except for a rise of 50% with respect to the mean speed in radial positions outside the blade tip.

The velocity fluctuation spectra measured at the fan outlet (Figure 5) have a decreasing trend with frequency, and some tonal contributions are visible, at the shaft frequency and the first two blade passing frequency (BPF) harmonics. The B type grid velocity spectrum has a smaller amplitude than the other two configurations, at all frequencies, while for the A grid type, the amplitude is comparable with the non-obstructed fan.

worm 2022

Figure 5: Velocity fluctuations spectra at the blade tip (r/R=1), grid comparison: 5500 rpm. The velocity fluctuations at the fan outlet are represented against frequency and radial position in Figure 6 for frequencies up to 50 times the shaft frequency. When no grid is installed and for the coarser grid (A), the speed trend along the blade can be divided into three areas: a velocity relative maximum is registered adjacent to the fan hub, a minimum occurs around 0.7R and 0.9 R respectively in the no grid and in the grid A configurations, and an absolute maximum occurs in the area corre- sponding to the blade tip, tip clearance and shroud. The finer grid (B) leads to an almost uniform velocity profile along the blade, and to an increasing value outside the blade.

Figure 6: Velocity fluctuations frequency spectra in different radial positions, at 5500 rpm.

4

These trends are confirmed at all frequencies, except the tonal components at the shaft frequency and the first two BPF harmonics which are visible at all radial positions. The difference between the mean velocity profiles in Figure 4 and the trends in Figure 6 are ascribable to these tonal components.

3.2. Aeroacoustic characterisation The wall pressure fluctuations and acoustic data spectra are examined in this section. Regarding the wall pressure fluctuations (Figure 7 a)), contributions at the BPF and at shaft frequencies of order nB- 2, nB-1 and nB+1 (n=harmonic number; B=number of blades) are noticeable. Looking at the noise spectra (Figure 7 b) and c)), there are evident tonal contributions at the first 6 shaft frequency har- monics and at BPFs up to the 20 th . It is worth noting here that the fan has evenly space blades, thus several BPF harmonics are clearly visible. For frequencies up to 3000 Hz, the B type grid proves to be noisier, while the A type grid emits a slightly lower noise than the clean configuration. At higher frequencies, the A type grid has the highest noise emissions. The low frequency noise in the fan wake (b) is probably due to the fan flow being partially directed towards the microphone.

a)

b) c) Figure 7: Spectra at 5500 rpm, grid effect comparison: a) wall pressure fluctuations; b) acoustic data downstream the fan; c) acoustic data upstream the fan.

worm 2022

3.2. Pressure velocity correlations In this section, the microphones and the hot-wire anemometer signals correlation is computed, in order to understand which radial positions are more involved in the generation of the noise perceived by the farfield microphones. In particular, the magnitude-squared coherence estimate ( 𝛾 2 ) is used, which is a function of frequency with values between 0 and 1. These values indicate how well one input signal ( 𝑥 ) corresponds to the other ( 𝑦 ) at each frequency ( 𝑓 ). This indicator is a function of the power spectral densities, P xx (f) and P yy (f) , and the cross power spectral density, P xy (f) , of 𝑥 and 𝑦 :

2

𝛾 2 (𝑓) = ห𝑃 𝑥𝑦 (𝑓)ห

𝑃 𝑥𝑥 (𝑓)𝑃 𝑦𝑦 (𝑓)

a)

b) Figure 8: Magnitude squared coherence estimate of the velocity at each radial station with respect to the farfield microphones: a) downstream mic, b) upstream mic. In Figure 8, the magnitude square coherence is computed for the velocity and farfield pressure signals downstream and upstream the fan respectively, for each configuration and radial position, at the shaft frequency and the first two BPF harmonics. In the configuration without grid, radial positions between 0.7 and 0.97 are well correlated both at the shaft frequency and the BPF. The second harmonic has a lower 𝛾 2 estimate, but the same trend with the radial position.

worm 2022

When the coarser grid (A) is mounted, positions closer to the hub (up to 0.7R) are less correlated with noise generation than in the previous configuration. Velocities measured at r/R above 0.8 have a very high coherence with the farfield pressure at the shaft frequency, while at the BPF harmonics, there is a coherence drop outside the blade (1.1R). Moreover, at the BPF, 𝛾 2 is slightly higher than at the shaft frequency for r/R between 0.8 and 1.06. When mounting the finer grid (B), acoustic pressure and velocity are no longer correlated at the shaft frequency, while at the BPF the coherence assumes very high values when r/R is between 0.7 and 0.97. Regarding the second BPF harmonic, 𝛾 2 has a maximum value of about 0.4 at r/R between 0.7 and 0.85 with respect to the wake microphone and around r/R=0.85 when computed with respect to the upstream microphone. 4. CONCLUSIONS

An extensive test campaign is carried out on an automotive cooling fan, to gain insight on the noise generated in different operating conditions. At this aim, different grid types are mounted on the fan, simulating the radiator and introducing different pressure drop values in order to evaluate its influence on the generated noise. Wall pressure fluctuations on the fan shroud, acoustic pressure and velocity fluctuations are measured simultaneously and their coherence computed in order to understand which mechanisms dominates the noise generation. The main outcomes of the tests carried out are that the features of the installed grids highly influence the shape and values of the velocity profiles at the fan outlet and changes the relative importance of the different noise generation mechanisms involved. Both grids have an effect of shifting the flow outwards. Regarding the velocity profile at the flow outlet, the coarser grid (A) gives rise to a skewed profile, with similar velocity values with respect to the configuration without the grid, while the finer grid (B) significantly lowers the speed and leads to a uniform profile along the blade, with a speed rise moving outside the blade. In this latter configuration, the flow direction becomes more radial. Relevant tonal contributions are registered at the shaft frequency and at the first two BPF harmonics, either in the velocity and microphones spectra. The magnitude square coherence is thus computed at these frequencies. The radial positions with the highest coherence values between velocity fluctua- tions and acoustic pressure are:  r/R between 0.7 and 0.97 for all frequencies considered, when no grid is installed;  for the coarse grid, the highest coherence is registered for r/R between 0.8 and 1.1, at all frequen-

cies considered;  for the finer grid, signals are not coherent at the shaft frequency; at the BPF, there is a good cor-

relation in the range 0.7 to 0.97, and a lower coherence at 2 ⋅ BPF, with a value of 0.4 around r/R=0.83. Therefore, the outward half of the blade proves to be the main responsible for the noise generation, although the proportions and the frequencies involved change when different grid types are involved. 5. ACKNOWLEDGEMENTS

The authors kindly acknowledge SPAL Automotive srl for having provided the tested fan. 6. REFERENCES

1. Wright, S.E., The acoustic spectrum of axial flow machines. Journal of Sound and Vibration ,

45(2) , 165-223 (1976).

worm 2022

2. Moreau S., Roger, M., Competing Broadband Noise Mechanisms in Low-Speed Axial Fans. AIAA

Journal , 45(1) , (2007). 3. Buisson, M., Ferrand, P., Soulat, L., Aubert, S., Moreau, S., Rambeau, C., Henner, M., Optimal

design of an automotive fan using the Turb’Opty meta-model. Computers & Fluids , 80 , 207–213 (2013). 4. Lallier-Daniels, D., Piellard, M., Coutty, B., Moreau, S., Aeroacoustic study of an axial engine

cooling module using Lattice-Boltzmann simulations and the Ffowcs Williams and Hawkings’ analogy. European Journal of Mechanics B/Fluids , 61 , 244–254 (2017). 5. Hashim, H. M., Dogruoz, M. B., Arik, M., Acoustic analysis of an axial fan . 16th IEEE Interso-

ciety Conference on Thermal and Thermomechanical Phenomena in Electronic Systems, 2017, 645-651. 6. Rynell, A., Chevalier, M., Abom, M., Efraimsson, G., A numerical study of noise characteristics

originating from a shrouded subsonic automotive fan. Applied Acoustics , 140 , 110-121 (2018). 7. Zhong, Y., Li, Y., Li, P., Chen, J., Xia, T., Kuang, X., Blade structure design based on multi-

objective optimization of automotive fan. Journal of Physics: Conference Series , 1952 , 032022 (2021). 8. Rynell, A., Efraimsson, G., Chevalier, M., Abom, M., Acoustic characteristics of a heavy-duty

vehicle cooling module. Applied Acoustics , 111 , 67–76 (2016). 9. Canepa, E., Cattanei, A, Mazzocut Zecchin, F., Milanese, G., Parodi, D., An experimental inves-

tigation on the tip leakage noise in axial-flow fans with rotating shroud. Journal of Sound and Vibration , 375 , 115–131 (2016). 10. Canepa, E., Cattanei, A, Mazzocut Zecchin, F., Leakage noise and related flow pattern in a low-

speed axial fan with rotating shroud. International Journal of Turbomachinery Propulsion and Power , 4(3) , 17 (2019). 11. Amoiridis, O., Zarri, A., Zamponi, R., Pasco, Y., Yakhina, G., Christophe, J., Moreau, S., Schram,

C., Sound localization and quantification analysis of an automotive engine cooling module. Jour- nal of Sound and Vibration , 517 , 116534 (2022). 12. Amca, B., 210-laboratory methods of testing fans for aerodynamic performance rating. Air Move-

ment and Control Association International, Inc. (2007). 13. Organization, I. S., Industrial fans - performance testing using standardized airways. ISO

5801:2017 (2017).

worm 2022