2006 01 1192

Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

SAE TECHNICAL

PAPER SERIES

2006-01-1192

Broadband Noise Source Models as


Aeroacoustic Tools in Designing
Low NVH HVAC Ducts
Omar M. Mohamud and Perry Johnson
Visteon Corporation

2006 SAE World Congress


Detroit, Michigan
April 3-6, 2006
400 Commonwealth Drive, Warrendale, PA 15096-0001 U.S.A. Tel: (724) 776-4841 Fax: (724) 776-5760 Web: www.sae.org

The Engineering Meetings Board has approved this paper for publication. It has successfully completed
SAE's peer review process under the supervision of the session organizer. This process requires a
minimum of three (3) reviews by industry experts.
All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or
transmitted, in any form or by any means, electronic, mechanical, photocopying, recording, or otherwise,
without the prior written permission of SAE.
For permission and licensing requests contact:
SAE Permissions
400 Commonwealth Drive
Warrendale, PA 15096-0001-USA
Email: permissions@sae.org
Tel:
724-772-4028
Fax:
724-776-3036

For multiple print copies contact:


SAE Customer Service
Tel:
877-606-7323 (inside USA and Canada)
Tel:
724-776-4970 (outside USA)
Fax:
724-776-0790
Email: CustomerService@sae.org
ISSN 0148-7191
Copyright 2006 SAE International
Positions and opinions advanced in this paper are those of the author(s) and not necessarily those of SAE.
The author is solely responsible for the content of the paper. A process is available by which discussions
will be printed with the paper if it is published in SAE Transactions.
Persons wishing to submit papers to be considered for presentation or publication by SAE should send the
manuscript or a 300 word abstract to Secretary, Engineering Meetings Board, SAE.
Printed in USA

2006-01-1192

Broadband Noise Source Models as Aeroacoustic Tools in


Designing Low NVH HVAC Ducts
Omar M. Mohamud and Perry Johnson
Visteon Corporation
Copyright 2006 SAE International

ABSTRACT
Computational Fluid Dynamics (CFD) is an integral part
of product development at Visteon Climate Systems with
a validated set of CFD tools for airflow and thermal
management processes. As we increasingly build CAE
capabilities to design not only thermal comfort, but quiet
systems, developing noise prediction capabilities
becomes a high priority.
Two Broadband Noise Source (BNS) models will be
presented, namely Proudmans model for quadrupole
source and Curles boundary layer model for dipole
source. Both models are derived from Lighthills acoustic
analogy which is based on the Navier-Stokes equations.
BNS models provide aeroacoustic tools that are effective
in screening air handling systems with higher noise
levels and identifying components or surfaces that
generate most of the noise, hence providing
opportunities for early design changes.
In this paper, BNS models were used as aeroacoustic
design tools to redesign an automotive HVAC center
duct with high levels of NVH. The design direction
suggested by BNS tools were later supported by physical
test data. These models were found to be valuable and
cost effective tools in providing reliable design direction
early in the development cycle.

INTRODUCTION
The ability of an analysis tool to characterize a systems
level of noise and its spectral distribution at a receiver
location is a powerful design tool. It could be used as a
virtual test lab to measure whether a given system meets
its design requirements. Currently, such capability is not
available within a development cycle for complex
industrial applications. This is due to the fact that noise
information at receiver locations requires the solution of
both the flow field and the acoustic field. However,
during early stages of the product development, noise
information at receiver locations is not critical. Rather,
reliable system noise information at the source may be
sufficient to provide design direction during development.

Noise information at the source requires only a CFD


solution that captures unsteady flow structure. At early
stages of the development cycle, a CFD tool that
provides a measure of noise levels at the source for a
given system in order to provide a design direction would
be very useful. In addition, if that tool is able to identify
components or surfaces that generate most of the noise
it would provide, to the development team, a valuable
means to derive design optimization.
Broadband Noise Source (BNS) models that require
steady state solution can be used as effective and
inexpensive design tools [1, 2]. Here, Proudmans
quadrupole model and Curles boundary layer model are
presented [3, 4].
Both models assume isotropic
turbulence, low Mach number flows, and broadband
noise and can only provide noise information at the
source.
The intent of this paper is to show the effectiveness of
BNS models during the development process. However,
before presenting these models, a brief description of
aeroacoustic modeling and options available to a product
design team in industrial environments is discussed.

AEROACOUSTIC MODELING
Aeroacoustic behavior can be completely characterized
by solving the compressible Navier-Stokes equations. In
other words, aeroacoustic phenomena can be explained
through the use of the principles of mass, momentum,
and energy conservation. For simple problems, Direct
Numerical Simulation (DNS) that solves the
compressible Navier-Stokes equations is able to solve
for both the aerodynamic flow field and the acoustic field.
However, for problems of industrial applications, this
approach becomes impossible due to the fact that the
acoustic energy is several levels of magnitude smaller
than the hydrodynamic energy. In addition, the acoustic
pressure perturbations are several levels of magnitude
smaller than the hydrodynamic pressure. Further, length
and time scales of the two fields are not compatible.
Therefore, the need to separate the two flow fields
becomes apparent.

THEORY OF AERODYNAMIC NOISE

dipole, and quadrupole sources. These applied external


forces can be obtained from a CFD solver.

In 1952, Lighthill [5] rearranged the Navier-Stokes


equations and derived the well known wave equation and
developed the theory of aeroacoustic sound.

Tij

a02 2 =
2
xi x j
t
2

(1)

With

Tij = vi v j + pij a02ij

(1.2)

Where

Tij is the Lighthills tensor, pij is the stress

tensor,

ij

is Kronecker delta, and

a0 is the speed of

sound in a medium at rest.


Equation 1 basically decouples the hydrodynamic field
and the sound field and is the basis of Lighthills acoustic
analogy. As a result, the process of noise generation
and propagation can be separately addressed. The
practical implications of this theory are significant as
propagation of sound waves amounts to solving a
hyperbolic wave equation in a medium at rest with
externally applied forces.
Lighthill derived the sound propagation equation for
unbounded flows. In 1955, Curle [4] added the effects of
non-moving boundaries. In 1969, Ffowcs-Williams and
Hawkings (FWH) [6] derived a generalized Lighthill's
equation with moving boundaries.
2
Q Fi Tij
2
2 2
a0 =

+
t xi xi x j
t 2

Where

(2)

Q
is due to unsteady volume displacement of
t

the fluid volume by moving walls and represents the


contribution of the monopole source.

Fi
is due to interaction of the flow with the rigid body
xi
and represents the contribution of the dipole noise.

2Tij
xi x j

is Lighthills tensor due to the structure of the

flow and represents the quadrupole source contribution.


Equation 2 is the most general sound wave equation
based on Lighthills analogy theory. In predicting the
aerodynamic noise of industrial problems, Equation 2
can be used to solve a wave equation for the noise
propagation in a medium at rest with applied monopole,

COUPLED APPROACHES
A coupled approach utilizes the Lighthills analogy and
FWH equation implemented in its most general form.
Here, a CFD solver is used to characterize the
fluctuations in the unsteady flow field, and an acoustic
code is utilized to solve for the propagation equation.
Two separate mesh domains are used in this approach
with great potential in simulating industrial aeroacoustic
problems. Commercial acoustic solvers that can be
coupled with commercial CFD solvers are currently
available. Kim et al [7] and Ayar et al [8] used this
approach for automotive applications and found
reasonable correlation with experimental data.
This coupled approach can be simplified further if there
is a line of sight between the source and the receiver.
Under this assumption, the FWH equation is solved
analytically and no acoustic solver is required. This
approach is suited for external flows in industrial
applications with low to medium complexity.
Implementation of a coupled approach based on
Lighthills analogy has increased the role of aeroacoustic
modeling for industrial applications. However, it still
requires unsteady turbulence models that can capture a
detailed fluid flow structure and an acoustic solver for the
wave equation. These two requirements still present a
challenge in incorporating this approach into the
development process.
During early stages of product development, a measure
of a systems noise level at the source is sufficient to
drive the design optimization process. However, a
coupled aeroacoustic approach with detailed transient
CFD analysis and a noise propagation solver becomes a
burden to the early stages of a fast paced product
development cycle. Instead, approximate noise source
strength extracted from a steady state CFD solution
could provide a useful and practical aeroacoustic design
tool.

BROADBAND NOISE SOURCE MODELS


All previous approaches required a transient CFD
solution that captures the flow structure and presents
some challenge in terms of resource requirements and
solution time. Although the coupled approach is within
reach for automotive applications, it cannot be
incorporated, at the present, into a fast paced
development process. Ultimately, the coupled approach
is expected to be deployed, within the next few years, as
a design tool at some point during the development
cycle, particularly before tooling is committed.
When the generated sound does not have any distinct
tone and the radiated sound energy is distributed over
the entire range of frequencies, statistical turbulence
quantities extracted from steady state RANS solution can

be used together with semi-empirical correlations to


provide a measure of the broadband source noise.
Aeroacoustic models that quantify the broadband source
noise generated by the flow, per unit surface or volume,
are termed Broadband Noise Source (BNS) models.
Though BNS models are attractive aeroacoustic design
tools they present major limitations. These models do
not provide any tonal noise information or noise spectra
at receiver locations. Instead they provide only an
approximate measure of the radiated noise at the
source.
There are a number of BNS models [1]. In this paper,
Proudmans model for quadrupole source [3] and Curles
boundary layer model for dipole source [4] will be
presented. Both models are derived from Lighthills
acoustic theory which is based on Navier-Stokes
equations.
PROUDMANS NOISE SOURCE MODEL
In 1952, immediately after Lighthills acoustics theory,
Proudman [3] using Lighthills theory considered noise
generated by a homogenous isotropic turbulence. Using
statistical models he derived the following analytical
expression for the Acoustic Power (AP) per unit volume:

AP = 0 (

is a constant,

velocity defined as:

u2 =

2
k
3

Equation 3 can be written in terms of k and


5
t

as:

(4)

Where

Mt =

p ( x, t ) =

1
4a 0

( xi y i )ni p
( y, )dS ( y )
t
r2

(5)

p is the surface
pressure, n is the wall normal direction, p is the
acoustic pressure, and a0 the speed of sound.
Where,

is the emission time,

Including the correlation area and manipulating Equation


5, the Surface Acoustic Power (SAP) radiated from the
body surface can be computed by:
2

p
Ac ( y) t dS ( y)

(6)

Where Ac is the correlation area and

is the density.

p
t is the mean-square time derivative of the source
surface pressure, x is the receiver coordinate, y is the
coordinate of the source surface, and S is the source
surface.

Here, k is the turbulent kinetic energy.

AP = 0 M

The second model is based on Curles [4] integral where


the radiated acoustic pressure was derived as a function
of the fluctuating surface pressure of the rigid body
surface. The following formulation assumes low Mach
number and only accounts for dipole contribution of the
source noise.

(3)

0 is the density, a0 is the


speed of sound, l is length scale, and u is turbulent

Where

BOUNDARY LAYER NOISE SOURCE MODEL

1
SAP =
12 0 a 03

u u
)
l a 05

quadrupole noise generated at the source from an


isotropic turbulence.

2k
a0

Later Lilley [9] reexamined Proudmans expression and


added a correction for time correlation effects previously
neglected by Proudman. Sarkar and Husseini [10]
computed numerically radiated acoustic noise using DNS
and found results that are in agreement with Proudmans
analytical expression. However, they recalibrated the
constant for better agreement with DNS data. Here,
Fluent 6.2 Solver [1], used for this work, implements the
calibrated constant and provides a measure of the

The mean-square derivative is computed from turbulent


kinetic energy, turbulent dissipation rate, and wall shear
stress.
Leclercq and Symes [11] compared experimental work
and numerical implementation of Curles integral of bluff
bodies in fluctuating flow field. They found that Curles
theoretical development, Equation 6, predicted the
radiated noise. Khondge et al [2] applied BNS models to
predict aeroacoustic performance of a duct and found
that predicted noise was directionally supported by
experimental data.
Proudmans quadrupole model and Curles dipole model
were selected in this paper. These two models, based
on Lighthills analogy, require little computational effort in
extracting, from a steady state CFD solution, the
Acoustic Power (AP) and the Surface Acoustic Power
(SAP). Both aeroacoustic variables are accessed within
the Fluent Solver as an aeroacoustic post-processing
tool.
AP provides Proudmans quadrupole noise

generation, whereas SAP provides Curles surface dipole


noise generation.
These two aeroacoustic variables are well suited to Air
Handling Subsystem (AHS) applications which represent
a complex internal flow that includes rotating parts
(blowers), heat exchangers, and distribution ducts. In
these low Mach number flows, surface dipole noise
radiation generally contributes most of the system noise.
However, turbulent generated quadrupole generation
becomes important around the blower blades.

APPLICATION - HVAC PANEL DUCTS


Automotive HVAC systems are used to provide
conditioned air into the cabin for comfort and safety.
Outside fresh air is drawn in at the base of the vehicle
windshield by the AHS that moves conditioned air into
the vehicle cabin. The AHS, including a blower, scroll,
two heat exchangers, housing, and blend and mode
doors, is packaged below the instrument panel. In panel
mode, outlet panel registers mounted on the instrument
panel and the panel distribution ducts supply conditioned
air into the cabin. Noise requirements are prescribed at
the vehicle level and microphones are placed near the
driver and passenger ear locations to measure noise
levels. However, during development, component noise
contribution to the overall system noise becomes
important.

In automotive HVAC systems, panel ducts supply air into


the cabin through the instrument panel registers. Panel
duct air supply design in terms of noise and comfort is
critical due to its proximity to the front seat occupants. In
this paper, BNS models were used as design tools to
optimize the panel duct design for noise and airflow.
CFD MODEL
A CFD model, based on steady state RANS with k
turbulence model, of the entire AHS and distribution
ducts was simulated. First, baseline AHS performance
was established in terms of noise and airflow using the
initial design. Then panel ducts and registers were
modeled separately with a velocity profile applied at the
inlet. This sub-model, shown in Figure 1 and referred
here to as Panel Duct (PD) model, behaves as if it were
part of the entire system. This modeling method was
necessary during the design optimization process and
saved computer resources and execution time.
A cross-section of the PD model mesh is shown in
Figure 2. Figures 3 and 4 show mesh resolution of the
center duct neck and the vent registers, respectively.

Center Duct Neck

Here, a case will be studied where the initial level design


of an automotive HVAC panel duct did not meet its
design requirements due to high levels of NVH, poor
overall sound quality, and low airflow to the outboard
registers. Confidential customer requirements will not be
shared here. However, CFD predicted airflow and NVH
levels will be presented and compared with test airflow
and NVH data. The geometry of the panel duct with flow
inlet and outlets is presented in Figure 1. As shown in
Figure 1, the panel duct has three major parts, namely
the center duct, the outboard duct, and the registers.

Vent Registers

Cross section @ y = 60 mm

Figure 2: Panel duct model mesh

Airflow Outlet

Registers

Center Duct

Cross section @ y = 60 mm
Center Duct Mesh Close-up
Outboard

Airflow Inlet

Figure 3: Panel duct model mesh-center duct neck close-up


Figure 1: Panel duct geometry - Initial

The PD model had 900,000 tetra cells and required an


execution time of 104 minutes on a 4 node HP Itanium 2
processor with 1 Gb RAM per node.

Vent Registers
Mesh Close-up

Utilizing BNS tools, noise characteristics of the initial duct


were extracted from the CFD solution of the panel duct
sub-model in terms of total SAP and AP generated
source noise. Further, components and surfaces with
high noise generation were identified. Guided by the
CFD results, high noise source regions of the center duct
were redesigned. To save design time and cost, design
modifications were performed within the CFD postprocessor tool. After only few iterations, a center duct
design with a baffle at the inlet was proposed.

RESULTS
Figure 5 shows that most of the dipole noise was
generated from the center duct with very little
contribution from the outboard ducts and registers. The
contour plot of the quadrupole noise generation in Figure
6 shows that the neck area of the center duct radiates
most of the noise. In addition, Figure 7 shows surface
acoustic power level iso-surfaces of radiated dipole noise
with significant generation at the neck.

Cross section @ y =60 mm

Figure 4: Panel duct model mesh-vent registers zone close-up

The accuracy of BNS aeroacoustic models will depend


strongly on the accurate characterization of the main
flow.
To validate the steady state CFD solution,
predicted and measured system airflow in panel recirc
mode (MAX AC) is presented in Table 1. The data
shows good agreement between measured and
predicted total airflow and airflow distribution and
suggests that the CFD solution characterizes the
physical system.

3.0E-05
2.5E-05

SAP (Watts)

Fluent 6.2, used for this work, offers both Proudmans


quadrupole model, Equation 4, and the boundary layer
dipole model, Equation 6, as post-processing tools.
Physical quantities extracted from steady state solution
such as turbulence velocity, turbulence length scale, and
mean-square surface pressures were used to compute
3
AP and SAP. AP (w/m ) is the turbulent generated
2
quadrupole noise, whereas SAP (w/m ) is the surface
generated dipole noise. Both of these quantities are
evaluated at the source and provide an approximate
measure of the local contribution to the total system
generated noise.

3.5E-05

2.0E-05
1.5E-05
1.0E-05
5.0E-06
0.0E+00
Center

Outboards

Registers

Component
Figure 5: Initial design component source noise

Table 1: AHS airflow in panel recirc mode, m3/hr


High Noise Generation

LHO

LHC

RHC

RHO

Total

CFD

36.85

34.79

36.67

34.61

142.91

Test

35.32

34.61

40.61

34.61

145.15

Change

-1.53

-0.18

+3.94

0.00

+2.24

Cross Section @ y=60 mm

Figure 6: Acoustic Power level contours - Initial

Low Noise
High Noise

Figure 9: Surface Acoustic Power level iso-surfaces - Redesign


Figure 7: Surface Acoustic Power level iso-surfaces - Initial

All noise data of the initial design indicates that the


center duct should be redesigned. Guided by the CFD
results, high noise source regions of the center duct
were redesigned. After several iterations, a center duct
design with a baffle at the inlet was proposed, see Figure
8. The function of the baffle is to split incoming airflow
between outboard and center vent registers. This design
is referred to as Redesign.

Figure 11 shows a comparison of dipole noise radiation


(SAP) between the Initial design and the Redesign panel
ducts. The figure shows a significant reduction in center
duct noise generation with slight noise reduction of the
outboard ducts and registers for the Redesign.

Noise Reduced

Center Duct

Cross-section @ y = 60 mm

Figure 10: Acoustic Power level contours - Redesign


Split Baffle

Figure 8: Panel duct geometry - redesign with baffle

Results of the Redesign duct are shown in Figures 9 and


10 and show significant reduction in noise generation
relative to the initial design.

Table 2 presents BNS predicted source noise and


compares a sub-model of the initial center duct with a
sub-model of the redesigned center duct. The data
shows that computed Surface Acoustic Power at the
source indicates a 60% noise reduction for the
redesigned duct. In addition, Surface Acoustic Power
level at the source is reduced by 4.0 dB.
Table 3 presents Lab NVH bench data measured at the
AHS level and compares a system with the initial center
duct and a system with the redesigned center duct. The
data shows a 4.9 dB reduction in recirc mode (MAX AC)
and 4.4 dB reduction in panel fresh mode for the
redesigned duct.

components or surfaces with high noise generation,


hence providing opportunities for early design changes.

3.5E-05

Aeroacoustic variables at the source were used to drive


the design optimization process which resulted in
reduced noise levels, better sound quality, and a more
efficient airflow supply into the vehicle cabin.

3.0E-05

Current

SAP (Watts)

2.5E-05
2.0E-05

BNS models of the panel duct sub-model predicted a 4.0


dB noise reduction at the source, whereas a lab NVH
test of the entire AHS measured a 4.9 dB reduction at
the receiver. This indicates that BNS models provided
reliable design direction early in the development
process and proved to be valuable and cost effective
aeroacoustic design tools.

1.5E-05
1.0E-05

Redesign
5.0E-06
0.0E+00

Center

Outboards

Registers

REFERENCES

Component

1.
Figure 11: Initial and Redesign component source noise

It is important to emphasize that Table 2 and Table 3


data should not be directly compared. Table 2 presents
CFD predicted aeroacoustic noise at the source with a
sub-model that included only outboard ducts, center
ducts, and registers. Whereas, Table 3 data refers to
the entire AHS noise measured at the driver ear location.
However, it is worthy to note that measured noise in
Table 3 shows 4.9 dB reduction at the receiver and
confirmed the design direction originally suggested by
the BNS models early in the development process.
Table 2: BNS predicted source noise- PD sub-model

Initial

SAP
Source
(Watts)
4.34e-5

SAP Level
Source
(dB)
66.4

Redesign

17.32e-6

62.4

60 %

-4.0

Model

Change

Table 3: AHS NVH bench data Driver side

Recirc.
(dB)

Fresh
(dB)

Initial

71.3

66.9

Redesign

66.4

62.5

Change

-4.9

-4.4

CONCLUSION
BNS models are found to be effective tools in screening
systems with higher noise levels and identifying

Fluent 6.2 Users Guide, Fluent Inc., Lebanon NH,


2005.
2. Khondge, A.D, Sovani, S.D., Kim, S., Guzy, S.C.,
Farag, A.A. On Predicting Aeroacoustic
Performance of Ducts with Broadband Noise
Source Models, SAE Paper, 2005-01-2495, 2005.
3. Proudman, I., The Generation of Noise by Isotropic
Turbulence, Pro. Roy. Soc. A214, pp. 219, 1952.
4. Curle, N., The Influence of Solid Boundaries upon
Aerodynamic Sound, Proc. Roy. Soc. A2314, pp.
505-514, 1955.
5. Lighthill, M.J. On the Sound Generated
Aerodynamically: Part I General Theory , Proc.
Roy. Soc. A211, pp.564-587, 1952.
6. Ffowcs Williams, J.E., Hawkings, D.L., Sound
Generation by Turbulence and Surfaces in Arbitrary
Motion, Phil. Trans. Roy. Soc., A264 (1151), 321243, 1969.
7. Kim, S-E., Dai, Y., Koutsavdis, K., Sovani, S.,
Kadam, N., Ravuri, M.R. A Versatile
Implementation of Acoustic Analogy Based Noise
Prediction Method in a General-Purpose CFD
Code AIAA Paper N0. 2003-3202, 2003.
8. Ayar, A, Ambs, R., Capellmann, C., Schillemeit, B.,
Matthes, M. Prediction of Flow-Induced Noise in
Automotive HVAC Systems Using a Combined
CFD/CA Approach, SAE Paper, 2005-01-0509,
2005.
9. Lilley, G.M., The Radiated Noise From Isotropic
Turbulence Revisited, NASA Contract Report No.
93-75, NASA Langley Research Cnter, Hampton,
VA 24681, 1993.
10. Sarkar, S, Hussaini, M.Y. Computation of the
Sound Generated by Isotropic Turbulence, NASA
Contract Report No. 93-74, NASA Langley
Research Center, Hampton VA 24681, 1993.
11. Leclercq, D. J. J., Symes, M. K., Predicting the
Noise Generated by a Bluff Body in a Fluctuating
Flow, AIAA Paper 2004-20917, 2004.

You might also like