See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/292208625

CFD Prediction with LES for Psycho Acoustic Relevance in Ventilation

Article  in  Energy Procedia · January 2016

DOI: 10.1016/j.egypro.2015.12.238

CITATIONS READS

5 168

4 authors, including:

Victor Hodor Lucian Nascutiu

Universitatea Tehnica Cluj-Napoca Universitatea Tehnica Cluj-Napoca
15 PUBLICATIONS   29 CITATIONS    37 PUBLICATIONS   71 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Advanced strategies for high performance indoor Environmental QUAliTy in Operating Rooms View project

Monitoring hydraulic devices View project

All content following this page was uploaded by Lucian Nascutiu on 02 February 2016.

The user has requested enhancement of the downloaded file.

 Available online at www.sciencedirect.com

ScienceDirect
Energy Procedia 85 (2016) 252 – 259

Sustainable Solutions for Energy and Environment, EENVIRO - YRC 2015, 18-20 November
2015, Bucharest, Romania

CFD prediction with LES for psycho acoustic relevance in

ventilation
Victor Hodora*, Dan Birlea, Lucian Nascutiua and Mircea Diudeaa
a
Technical University of Cluj Napoca, Muncii Bulevard, no.103-105, Postcod: 400641, Cluj-Napoca, România

Abstract

The classical adage: “Less Noise __ More Sound” challenged our interest -whether if it is or not possible to model by numerical
simulation a suitable prediction in terms of sound quality to discriminate between Consonant _ Dissonant _Neutral sound
patterns. Ansys_Fluent CFD_CAA code was used (with a 1e -6 time-step for 12milon of hexahedral cells grid) in order to
perform related gas-dynamic numerical prediction – in ventilation. We have started our investigations with a simplest designed
flat plated blades (for a revers spinning centrifugal fan) –prescribing three different incremented blade disposal. We have set
some monitoring points (at certain transversal sections). The LES model with FW-H mode predicted results, were first Fast
Fourier Transformation converted, than we have processed them with Matlab –in order find the most relevant group
(octaves)/musical notes, and finally to decide by the use of an dedicated Wolfram code –whether the predicted sound should
experience psycho acoustic supportability (i.e.-perception).
©2015 The Authors.
© 2016 Published Published
by Elsevier by Elsevier
Ltd. This Ltd.
is an open access article under the CC BY-NC-ND license
Peer-review under responsibility of the organizing committee EENVIRO 2015.
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer-review under responsibility of the organizing committee EENVIRO 2015
Keywords: Computational aeroacoustics; LES; psycho acoustic; perception

1. Introduction

From the very beginning we have to notice that a pure tone should not distress our perception, even it might act at
a relative high Sound pressure level. The human ear, react different to certain annoying patterns i.e. scream or
chainsaw, etc. An inherent source with uniform inflow conditions is the tonal noise associated with the blade passing
frequency (BPF), which is the rotational frequency times the number of impeller blades. An important additional

* Corresponding author.
E-mail address: victor.hodor@termo.utcluj.ro.

1876-6102 © 2016 Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer-review under responsibility of the organizing committee EENVIRO 2015
doi:10.1016/j.egypro.2015.12.238
 Victor Hodor et al. / Energy Procedia 85 (2016) 252 – 259 253

acoustic noise is the narrow-banded tip clearance noise, which can become dominant over the tonal blade passing
frequency noise [1].
If direct noise computation (DNC) is not reachable for industrial applications, hybrid methods based on
acoustical analogy or stochastic modelling have already demonstrated good aptitudes [2]. The elevated sound
levels cause trauma to cochlear structure in the inner ear, which gives rise to irreversible hearing loss [3]. A very
loud sound in a particular frequency range can damage the cochlea's hair cells that respond to that range, thereby
reducing the ear's ability to hear those frequencies in the future, however, loud noise in any frequency range has
deleterious effects across the entire range of human hearing. The outer ear (visible portion of the human ear)
combined with the middle ear amplifies sound levels by a factor of 20 when sound reaches the inner ear [4]. Noise
has been associated with important cardiovascular health problems [4]. In 1999, the World Health Organization
concluded that the available evidence suggested a weak correlation between long-term noise exposure above 67-70
dB (A) and hypertension. More recent studies have suggested that noise levels of 50 dB (A) at night may also
increase the risk of myocardial infarction by chronically elevating cortisol production [5].
There are several methods for implementing the Maximum Permissible Sound Pressure Levels (i.e. standard ISO
1910.95): It may be a fixed level limit, such as 55 dB (A), or it may be a level relative to the ambient sound, such as
5 dB (A) above the ambient. Also, it may require measurement of the frequency spectrum, such as one octave bands,
or A-weighting, such as dB(A). Also, it may define different maximum levels based on zoning criteria, such as
residential, commercial, or industrial. For every additional 3 dB (A), the maximum exposure time is reduced by a
factor 2, e.g. 20 hours per week at 88 dB (A).
Recently researches underline consistent health care impacts coming from psychoacoustic relevance’s. Estimates
of sound annoyance typically rely on weighting filters, which consider some sound frequencies to be more important
than others based on their presumed audibility to humans. The older dB (A) weighting filter described above is used
widely in the U.S., but underestimates the impact of frequencies around 6000 Hz and at very low frequencies. The
newer ITU-R 468 noise weighting filter is used more widely in Europe. The propagation of sound varies between
environments; for example, low frequencies typically carry over longer distances. Therefore different filters, such as
dB (B) and dB(C), may be recommended for specific situations.
The subsequent correlation analysis between the preference values and psychoacoustic parameters showed that
the parameters sharpness, roughness and the acoustic parameter articulation index determine the sound quality for
HVAC noise of equal loudness [6]. Sound F, which stands out due to its very high preference within the test group,
also shows much higher values in all identified parameters [7]. The conclusive spectral analysis supports the
importance of sharpness for determining sound quality on HVAC noise. High frequencies above 3 kHz and low
frequencies below 100-200 Hz seem to have a major impact on sound quality of HVAC noise. Nevertheless, the
eliminated parameter loudness still constitutes the dominant parameter and needs to be considered in acoustic
optimization of HVAC noise [8].

2. Fundamentals in CAA prediction

LES results for instantaneous pressure were used as source of aero-acoustic noise (surface dipole). LES solutions
can be used to perform SPL calculation to identify high SPL region and compare alternative duct design. The
Ffowcs Williams-Hawkings method is limited since no reflection, deflection or transmission is considered. It does
provide upper and lower bounds which could be employed for engineering design purposes [9]. An aeroacoustics
model based on the Ffowcs Williams and Hawkings equation is used to predict dipole and monopole tonal noises in
the frequency domain. Showing the importance of the monopole source in this kind of fans constitutes the main
contribution in these research tasks [10]. The most accurate model of turbulence used by Ansys Fluent is LES
(Large Eddy Simulation). This model calculate all the large swirls only the little swirls are modelled [11]. LES
coupled with Ffowcs Williams and Hawking acoustic similarity criterion create a new method named CAA or
Computational Aeroacoustics which can predict the level of sound produce in flow.
Sound propagation can be calculated in different ways [12]: CAA (Computational Aeroacoustics); Direct sound
computation; SSPM (Segregated Source-Propagation Methods) - Propagation is decoupled from source: Source and
propagation are treated as mutually independent; Models may be used for computing propagation: Lighthill-Curle
Method, Ffowcs-Williams-Hawkings Method, FEM/BEM (Solution of Lighthill’s equation/Wave equation).
254 Victor Hodor et al. / Energy Procedia 85 (2016) 252 – 259

1. Direct calculation - Computational Aeroacoustics (CAA) – Resolve acoustic pressure fluctuations as a part of
the CFD solution. 2. Coupled CFD with specialized acoustics codes, Boundary Element Methods (BEM), Hybrid
zonal methods: Acoustic waves are not tracked with CFD solution; Use special acoustics codes to calculate wave
propagation. 3. Acoustic Analogy modelling: Use CFD to calculate source field; Use analytical solution to
propagate sound from source to receiver location. 4. Steady RANS based noise source modelling: Use empirical
correlations to estimate acoustic radiation based on mean flow solutions
FW-H Model - Lighthill’s Theory of Aerodynamic Noise
For uniform medium, the equation use is

డమ ఘ
െ ܽ଴ଶ ߘ ଶ ‫ ݌‬ൌ Ͳ (1)
డ௧ మ

For turbulent sound source the equation use is

డమ ఘ డమ ்೔ೕ
െ ܽ଴ଶ ߘ ଶ ‫ ݌‬ൌ (2)
డ௧ మ డ௫೔ డ௫ೕ

Complex fluid process can be represented by equivalent acoustic sources (∂2Tij / (∂xi ∂xj) thought to be in
uniform medium at rest. Ffowcs Williams-Hawkings (FW-H) derived a generalized Lighthill equation valid for
moving surfaces [12].

డమ ுሺ௙ሻఘ డమ డி೔ ఋሺ௙ሻ డொ೔ ఋሺ௙ሻ

െ ܽ଴ଶ ߘ ଶ ሾߩ‫ܪ‬ሺ݂ሻሿ ൌ ൣܶ௜௝ ‫ܪ‬ሺ݂ሻ൧ ൅ ൅ (3)
డ௧ మ డ௫೔ డ௫ೕ డ௫೔ డ௧

డ௙
‫ܨ‬௜ ൌ െሾߩ‫ݑ‬௜ ሺ‫ݑ‬௝ െ ‫ݒ‬௝ ሻ ൅ ‫ߜ݌‬௜௝ െ ߬௜௝ ሿ (4)
డ௫೔

డ௙
ܳ௜ ൌ ሾߩ൫‫ݑ‬௝ െ ‫ݒ‬௝ ൯ ൅ ߩ௢ ‫ݒ‬௝ ሿ (5)
డ௫೔

ui -fluid velocity, un -fluid velocity component normal to the surface, v i -surface velocity component, vn -surface
velocity component normal to the surface, δ - Dirac delta function, H -Heaviside function.

3. Interest domain: Mesh and RMS

Interest domain is represented by the three different incremental blades disposal of a generic centrifugal fun, and
the surrounding space. The analysed field has three domains: two stationary domain and one rotational domain.

Fig.1a Overall Mesh Fig.1b Detail in the computational grid

 Victor Hodor et al. / Energy Procedia 85 (2016) 252 – 259 255

The two stationary domains are represented by the air between the case of the fan and the air that enclose the
electric motor. The rotational domain is represented by the air in the region near the fan.
All the faces that are in contact between the domains must be declared as interfaces. This must be done in order
to have a bundle of crossing between those domains. The mesh that we used in our analysis is one of 12 million
tetrahedral elements with high refinement inside the fan case. In the boundary layer we have imposed 12 elements
with a cumulative high of 2 mm, with a grow rate of 1.12. The computed case is characterized by the two boundary
conditions: The angular velocity of the fan at 3000 revolutions per minute, and, the Opening condition -for outer
surfaces of the stationary surrounding domain [13]. After generating the mesh and the setup of the boundary
conditions, we initial start the computing procedure with RANS model (Fig. 2a) [14, 15], in order to achieve
stability and convergence for a proper begin of the unsteady simulation. The unsteady simulation was carried out
with LES model (Fig. 2b) [16-18] combined with Ffowcs-Williams Hawking acoustic analogy in order to extract
sound level produced by the each fan.

a) b)

Fig. 2 a) RMS for quasi-steady regime; b) RMS for transient regime

In the acoustic model options we have imposed 18 monitoring points (virtual microphones) for the computation
of the sound level characteristics. Acoustic analyses requires an adequate time discretization, depending on the time
step size the sound is computed on higher or lower frequencies ranges. In this case we have set the time step at 1e-6
second, in order to reveal the sound up to 10 kHz. Running time of the numerical simulation relate to 3 complete
spinning’s of the fan. We have achieved the convergence in terms of RMS after 61 days of calculation on a dual
processor machine (E5 2620) with 128GB RAM. To be noticed that at 16200 iterations we have changed the time
step from 1e-4s to 1e-6s.The related results were exported to the postprocessor were we have extracted the data base
for sound relevance of the each centrifugal fan.

4. Predicted Field Contours by LES model

In figure 3 we have the pressure fields on the fan surface, at high resolution between 2600 Pa (i.e. red colour) and
-6800 Pa (i.e. blue colour), for the three different incremented blades disposal (Fig.3)

a b c

Fig.3 Pressure Contour on Fan surface: a-Var.1, b-Var.2, c-Var.3 blades disposal
256 Victor Hodor et al. / Energy Procedia 85 (2016) 252 – 259

The collision between the fan blade (solid interface) and the air surrounding the fan creates dipole sources, which
propagate in the air flow pass volume as is shown in the figure 4, for each of the three incremental blades disposal.

a b c

Fig.4 Dipole Sources in volume: a-Var.1, b-Var.2, c-Var.3 blades disposal

The three different blade disposal, give us the following pressure fields surrounding the fan, at high resolution
between 2600 Pa (i.e. red colour) and -6800 Pa (i.e. blue colour). (Fig.5).

a b c

Fig.5 Overall Pressure Contours: a-Var.1, b-Var.2, c-Var.3 blades disposal

In figure 6 we have the acoustic power level fields in the air surrounding the fan, at high resolution between 0 dB
and 110 dB, for each of the three different incremented blades disposal (Fig.6)

a b c

Fig.6 Entire Ambient Contours of Acoustic Power Level (dB): a-Var.1, b-Var.2, c-Var.3 blades disposal
The related Sound pressure levels (in between 40_110 dB) are presented in the following FFT format (from 10
Hz to 10000 Hz), (Fig. 7)

a b c

Fig.7 SPL(dB vs.Frequencies): a-Var.1, b-Var.2, c-Var.3

 Victor Hodor et al. / Energy Procedia 85 (2016) 252 – 259 257

The first discrimination procedure might be express in terms of octave (Fig. 8)

a b c

Fig.8 a. Octava A, b. Octava B, c. Octava C

5. Experimental results

We have first manufactured three prototypes in order to reach a starting reasonable gas-dynamic configuration –
for the fan magnitude and deflector position. Tests performed on these fan geometries prototypes were done at 3000
revolutions per minute. All these configurations of different blades disposal fan were manufactured with a
3DLeapFrog HS Create print machine.

Fig. 9 Generic centrifugal fan Fig. 10 Experimental ensemble

a b c

Fig.11 Blades disposal a. Symmetric dissonant, b. Asymmetric neutral, c. Asymmetric consonant

The predicted aerodynamic fields, may be experimentally validated by using (at least) two different techniques:

1). By image processing analogy - like classic Schlieren imaging method as we focus in a previous work
2). By aeroacoustics using sound (20Hz to 20 kHz) logistic/devices and dedicated codes.

The peaks were discriminated and selected with MatLab (Fig.12)

258 Victor Hodor et al. / Energy Procedia 85 (2016) 252 – 259

Fig.12 Peaks selection in Matlab

In order to be used in the Wolfram_OpenSource [19] code, to evaluate the sound quality (Consonant, Dissonant
or Neutral) (Fig.13).

a b c

Fig 13. a. Neutral sound; b. Dissonant sound; c. Consonant sound [19]

Experimental data acquisitions for the three incremented blades disposal ware done by using Minilyzer ML1 NTI
audio AG and E-MU 0404 audio/MIDI interface.

a.
symmetrical blades disposal b. asymmetrical blades disposal c. asymmetrical blades disposal

Fig.14 Experimental SPL

6. Conclusions

The present project concern with the issue (whether or not) LES model might give enough satisfaction in CFD
prediction for aero acoustic noise. Signal processing -pattern investigations attributes might be subjected to the
classical adage: “Less Noise __ More Sound”
 Victor Hodor et al. / Energy Procedia 85 (2016) 252 – 259 259

LES prediction offer us the possibility to pre-investigate several incremented blades disposal, and in this way to
predict the consonant or neutral gas-dynamic geometry in order to reduce psychoacoustic inconveniences.
The three different incremented blades disposal -tests (that are presents above), give us the capability to predict
and confirm that the third one (c. var. 3 disposal) -generating neutral perception, will ensure less psychoacoustic
discomfort.
Further work goes-on with the idea to add (to design) a Slightly Swinging Deflector –in the gas-dynamic flow
path. By simple acting it (sometimes -with a simple programmable chip), in some very restrictively controlled limits
to prevent significant flow path depreciation, it can provoke the desired sound effect. The swinging deflector should
be very easy time-position balanced.

Acknowledgement

Present researches were done using some financial support from Romanian National Authority for Scientific
Research, PN-II-PT-PCCA-2011-3.2-1212, Grant PARTENERIATE of the, CNCS – UEFISCDI, 2012-2016
“Advanced strategies for high performance indoor Environmental Quality in Operating Rooms” (EQUATOR).

References

[1] E. Sundstrom, B. Semlitsch, M. Mihaescu, Centrifugal Compressor: The Sound of Surge, American Institute of Aeronautics and Astronautics,
Royal Institute of Technology (KTH), Stockholm, (2004).
[2] H. Dechipre, H. Michael, Aeroacoustics Simulation of an Automotive A-Pillar Rain Gutte, Automotive Techniques, CAE Volkswagen AG,
Munich, EASC European Automotive Simulation Conf, (2009).
[3] S. Rosen, P. Olin, Hearing Loss and Coronary Heart Disease, Archives of Otolaryngology, 82 (1965).
[4] Office of Noise Abatement and Control, Noise: A Health Problem United States Environmental Protection Agency, Washington, DC 20460,
(1978).
[5] H. Ising, W. Babisch, Noise-Induced Endocrine Effects and Cardiovascular Risk, (1999) 37–48
[6] M. Mihaescu, L. Fuchs, R.Z. Szasz, E.J. Gutmark, Flow and Noise Investigations of a Separate Flow Exhaust System, (2005) 1221-1229.
[7] P. Lercher, J. Hörtnagl, WW. Kofler, Work noise annoyance and blood pressure: combined effects with stressful working conditions, Int.
Arch Occup Environ Health, 23-8 (1993).
[8] H. Silke, B. Thomas, B. Ralf, Psychoacoustic analysis of HVAC noise with equal loudness, University of Erlangen-Nuremberg, Inter-noise
Conference (2014).
[9] M. Mihaescu, S. Murugappan, M. Kalra, S. M. Khosla, E. J. Gutmark, Large Eddy Simulation and Reynolds Averaged Navier-Stokes
Modeling of Flow in a Realistic Pharyngeal Airway Model; an Investigation of Obstructive Sleep Apnea, Journal of Biomechanic, 41
(2008).
[10] S. Khelladi, S. Kouidri, F. Bakir, Predicting tonal noise from a high rotational speed Centrifugal fan, (2006) 113-133.
[11] F. Bode, R. Benea, V. Hodor, Investigation on thermo-acoustical instabilities related to a confined swirling burner, Mechanika, (2010) 37-
42.
[12] UCSD, Modeling Airborne Noise using ANSYS CFD, (2014).
[13] V. Hodor, R. Benea, F. Bode, E. Lupu, R. Szasz, Acoustic analogy using LES on a partially premixed turbulent flame, OpenFOAM
international Conf, Berlin, (2008).
[14] A. Meslem, F. Bode, C. Croitoru, I. Nastase, Comparison of turbulence models in simulating jet flow from a cross-shaped orifice, Eur. J.
Mech. B-Fluids, 44 (2014) 100-120.
[15] A. Meslem, I. Nastase, F. Bode, C. Beghein, Optimization of a Lobed Perforated Panel Diffuser - A Numerical Study of Orifice
Arrangement, Int. J. Vent., 11 (2012) 255-270.
[16] F. Bode, I. Nastase, A. Meslem, C. Croitoru, RANS and LES models comparison for a cross shaped jet flow with application in personalized
ventilation, Joint Conference 33rd AIVC Conference and 2nd TightVent Conf., Copenhagen, Denmark, (2012) 113-117.
[17] F. Bode, I. Nastase, Meslem A., Croitoru C., RANS and LES Models of a Lobed Jet Flow Integrated in a Personalized Ventilation System,
Proceedings of CLIMA 2013, The 11th REHVA World Congress and the 8th International Conference on Indoor, Air Quality, Ventilation
and Energy Conservation in Buildings, Prague, Czech Republic,, (2013) p.3047-3056(3010p).
[18] F. Bode, A. Meslem, C. Croitoru, Numerical simulation of a very low Reynolds cross-shaped jet, Mechanika, (2013) 512-517.
[19] J. Kiehl, ‘Jazz Voicings’ with Wolfram Alpha (Computational knowledge engine) http://demonstrations.wolfram.com/JazzVoicings

View publication stats