Professional Documents
Culture Documents
Door - Honeycomb Construction
Door - Honeycomb Construction
double-walled structures
SATHISH KUMAR
Doctoral Thesis
Stockholm June 2012
TRITA-AVE-2012:20
ISSN-1651-7660
ISBN 978-91-7501-334-3
c Sathish Kumar, June 2012
Tryck:US-AB
Abstract
Sandwich panels with aluminium face sheets and honeycomb core material have
certain advantages over panels made of wood. Some of the advantages of these con-
structions are low weight, good moisture properties, fire resistance and high stiffness-
to-weight ratio etc. As product development is carried out in a fast pace today, there is
a strong need for validated prediction tools to assist during early design stages. In this
thesis, tools are developed for predicting the sound transmission through honeycomb
panels, typical for inner floors in trains and later through double-walled structures
typical for rail-vehicles, aircrafts and ships.
The sandwich theory for wave propagation and standard orthotropic plate theory
is used to predict the sound transmission loss of honeycomb panels. Honeycomb is an
anisotropic material which when used as a core in a sandwich panel, results in a panel
with anisotropic properties. In this thesis, honeycomb panels are treated as being or-
thotropic and the wavenumbers are calculated for the two principal directions. The
wavenumbers are then used to calculate the sound transmission using standard or-
thotropic theory. These predictions are validated with results from sound transmission
measurements. The influence of constrained layer damping treatments on the sound
transmission loss of these panels is investigated. Results show that, after the damping
treatment, the sound transmission loss of an acoustically bad panel and a normal panel
are very similar.
Further, sound transmission through a double-leaf partition based on a honeycomb
panel with periodic stiffeners is investigated. The structural response of the periodic
structure due to a harmonic excitation is expressed in terms of a series of space harmon-
ics and virtual work theory is applied to calculate the sound transmission. The original
model is refined to include sound absorption in the cavity and to account for the or-
thotropic property of the honeycomb panels. Since the solution of the space harmonic
analysis is obtained in a series form, a sufficient number of terms has to be included
in the calculation to ensure small errors. Computational accuracy needs to be balanced
with computational cost as calculation times increases with the number of terms. A
new criterion is introduced which reduces the computational time by up to a factor
ten for the panels studied. For all the double-leaf systems analysed, the sound trans-
mission loss predictions from the periodic model with the space harmonic expansion
method are shown to compare well with laboratory measurements.
vii
Dissertation
The work presented in this Doctoral thesis was carried out within The Centre for ECO2
Vehicle Design at the Department of Aeronautical and Vehicle Engineering, The Royal
Institute of Technology (KTH) in Stockholm, Sweden.
This thesis consists of two parts. The first part gives an overview of the research with
a summary of the performed work. The second part collects the following scientific
articles:
Paper A. Leping Feng and Sathish Kumar, On Application of radiation loss factor in
the prediction of sound transmission loss of a honeycomb panel. International Journal of
Acoustics and Vibration, 17 (2012) 47-51.
Paper B. Sathish Kumar, Leping Feng and Ulf Orrenius, Predicting the sound trans-
mission loss of honeycomb panels using the wave propagation approach. Acta Acustica
united with Acustica, 97 (2011) 869-876.
Paper C. Sathish Kumar, Leping Feng and Ulf Orrenius, Modelling the sound trans-
mission through rib-stiffened double-leaf partitions with cavity absorption. Submitted
to Journal of Sound and Vibration, January 2012.
Paper D. Sathish Kumar, Leping Feng and Ulf Orrenius, Modelling the sound trans-
mission through rib-stiffened sandwich double-leaf partitions using space harmonic
analysis. Submitted to Journal of Sound and Vibration, May 2012.
ix
Contents of this thesis has been presented in the following
conferences
• The 17th International Congress on Sound and Vibration, Cairo, Egypt, July 2010.
• Noise and Vibration: Emerging Methods, NOVEM- 2012, Sorrento, Italy, April
2012.
Paper A. Sathish performed the measurements and provided data for computations,
Leping performed the computations and wrote the paper.
Paper B. Sathish performed the measurements, computations and wrote the paper
with inputs from Leping and Ulf.
Paper C. Sathish developed the model, performed the computations and wrote the
paper with inputs from Leping and Ulf.
x
Acknowledgments
The work presented in this thesis has been carried out within the Centre for ECO2 Vehi-
cle Design at the Department of Aeronautical and Vehicle Engineering, KTH, Sweden.
This thesis is a part of the research project “Coupling shape-vehicle body structure”
within the centre. The financial support provided by Vinnova, KTH and the industrial
partners is gratefully acknowledged.
To begin with, I would like to thank my main supervisor, Leping Feng for accepting
me as a PhD student and for his guidance, support and comments since the start of this
thesis. In addition many thanks to my co-supervisor Ulf Orrenius for his support, com-
ments on this work and for proofreading the manuscript of the papers and this thesis.
I would like to thank Per Wennhage for the useful discussions during the earlier stages
of this thesis. I would also like to thank Gunnar Ziwes from EURO-COMPOSITES
R
Sathish Kumar
xi
Contents
1 Introduction 3
5 Measurements 29
5.1 Loss factor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
5.2 Sound reduction index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
5.3 Vibration velocity level . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
5.4 Bending stiffness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
xiii
7 Conclusions and future work 51
II Appended papers 65
Part I
1
Chapter 1
Introduction
3
Sathish Kumar
Figure 1.1: Noise and vibration transmission paths in a railway compartment [3].
1 Air-borne noise through floors, walls, windows, roofs and auxiliary equipments
like fans, motors, gears, HVAC units.
2 Structure-borne noise from bogie, diesel engines.
Air-borne sound often originates from external sources and propagates into the
vehicle interior through the floor, wall panels and also through the holes in the body-
work, door seals, etc. Whereas, structure-borne sound is the result of mechanical vi-
brations propagating through the vehicle structure and eventually causing localised
displacements in air. The noise from wheel-rail region is often the major source of both
air-borne and structure-borne sound. On vehicles with underfloor diesel engines, noise
from the engine can be significant and so is the noise from air conditioning systems.
Vibrational energy from these sources can be transmitted to the passenger compart-
ment. Thus, the floor structure of a passenger compartment is expected to possess
good sound insulation properties. For most passenger trains, floating floors are ap-
plied to obtain sufficient noise reduction from the sources under the floor. In addition,
lightweight and thin floor designs are desired for increased weight reduction.
The inner flooring of a passenger compartment made of sandwich panels with alu-
minium face sheets and honeycomb core material have certain advantages over floor
panels made of wood. For this reason, traditional floor panels made of wood is being
replaced by light weight sandwich structures. Some of the advantages of these con-
4
Overview and Summary
structions are low weight, good moisture properties, fire resistance and high stiffness-
to-weight ratio etc. As mentioned earlier, one main limitation of such light weight struc-
ture is the poor sound insulation property. In the transportation industry, automotive,
aircraft and rail vehicles are treated with damping materials to reduce structure-borne
sound and its effectiveness mainly depends upon parameters such as materials, loca-
tion and size of the damping treatment. The basic principle of damping is to convert vi-
bratory energy into heat. Traditional damping treatments using viscous damping lay-
ers are typically of two types, unconstrained and constrained layer damping [8]. Due to
the shear deformation occurring in the visco-elastic layer, constrained layer damping
treatments are known to yield significantly larger system damping compared to uncon-
strained layer damping, for the same mass of damping material used [19]. The damp-
ing treatments are effective in reducing structural vibration and it is often thought that
reduced vibration results in noise reduction. However, when the radiation efficiency
increases more than the vibration reduction, the total noise radiated from the struc-
ture increases [8, 17, 18]. This increased radiation due to the damping treatment does
not provide the desired effects on sound insulation. The lack of effective predictive
methods for assessing the sound insulation effects due to added damping on complex
industrial structures leads to excessive use of damping materials. Examples are found
in the railway industry where sometimes the damping material applied per carriage is
more than one ton.
As product development is carried out in a fast pace today, there is a strong need
for validated prediction tools to assist during early design stages. There is a need for
computationally efficient models for calculating the sound insulation properties of the
floors accounting all practical design aspects of typical train structures e.g. orthotropic-
ity of the corrugated plates, short circuiting due to boundary conditions and the effect
of stiffeners. In this thesis, prediction models for sound transmission through inner
floor panels used in floor structures are developed and validated. The effect of damp-
ing treatment on the sound insulation of inner floors is investigated. Also, a sound
transmission model for rib-stiffened double-leaf partitions is developed and extended
to account for complex design aspects such as orthotropicity, sound absorption in the
cavity, stud stiffness and placement. The aim of this thesis is to provide simple analyt-
ical models of sufficient accuracy and speed to predict the sound insulation of parti-
tions. The models should be useful for quick parameter studies at early design stages
and to be used together with optimisation techniques for better sound proofing in en-
gineering structures like airplane fuselage and railway floors. A brief introduction to
these structures is presented in the following chapter.
5
Chapter 2
Floating floors are used in vehicles to reduce noise and vibration levels. There are sev-
eral studies on the successful use of floating floors for structure-borne noise in the ship
and aircraft industry [33, 32, 26, 4]. For passenger trains, floating floors are applied to
obtain a sufficient noise reduction from sources under the floor. A typical floating floor
construction consists of an inner floor isolated from the car body with discrete elastic
inner layers (studs) resting on supporting beams (stiffeners) as shown in Figure 2.1.
For thermal and acoustic insulation, the cavity between the inner floor and the car
body framework’s corrugated plate is filled with glass or mineral wool. Traditionally
the inner floor is made of plywood, but in recent years metal and composite sandwich
panels are increasingly used. Among other properties, sandwich structures provide
solutions for lightweight and thin floor designs. A properly designed floating floor
should: (i) Provide a vibration isolation for the inner floor for high frequencies. (ii)
Reduce the acoustical power radiated by the inner floor. (iii) Reduce transmission of
air-borne sound into the car.
7
Sathish Kumar
Sandwich structures due to their low density in combination with high flexural stiff-
ness are extremely popular in aerospace and marine applications where weight is a
major issue e.g. in commercial planes, pleasure boats, space shuttles and satellites. In
ground transportation they are increasingly found in cars, buses and trains. A classical
sandwich structure consists of two stiff, strong, thin face sheets bonded to either side of
a relatively thick, weaker, light weight core material. The faces are usually made from
a high performance material such as steel, aluminium or fibre composite, whereas the
core is usually a structural solid foam, balsa wood or honeycomb (this can again be
made of aluminium, kraft paper, etc). The structural properties of the face sheets and
the core are less significant as individual panels, but when glued together to form a
sandwich, they produce a structure of high stiffness and high strength-to-weight ratio,
a property which is of great interest to the industry.
The good stiffness properties of a sandwich construction can be illustrated by the
following example. A structure made up of a homogeneous material with a given
Young’s modulus and strength having unit width and thickness 0 t0 will have a cer-
tain bending stiffness which is normalised as 1. Then the beam is cut into two halves
of thickness 0 t/20 and a core material of thickness 0 2t0 is bonded between these two
halves. The corresponding stiffness and strength is 0 120 and 0 60 times more than the
homogeneous beam respectively. This is called the sandwich effect. The core material is
assumed to have a surface density much lower than the face sheets and therefore any
addition in weight to the structure is considered negligible.
t 1 1 1
t/2
2t ~1 12 6
t/2
4t ~1 48 12
8
Overview and Summary
9
Sathish Kumar
10
Chapter 3
11
Sathish Kumar
θ
pr
pi
E1, t 1
y
E2, G2 , t 2
x
E1, t 1
pt
Figure 3.1: Geometry and material parameters of a sandwich panel with an isotropic
core.
For a sandwich panel with symmetric face sheets and an isotropic core as shown in
Figure 3.1, the wavenumbers can be derived as a solution to the following expression
[31],
The functions U and V are given in Appendix- Part A. A detailed account about
deriving these functions can be found in the original Reference [31]. The solutions for
k are obtained when F = 0.
Due to the manufacturing of honeycomb cores, the cell wall thickness doubles in the
direction of cell orientation and the actual shape of the honeycomb cell can be irregular
making the core anisotropic as shown in Figure 3.2.
For simplicity, the honeycomb core can be treated as orthotropic and the whole
12
Overview and Summary
Panels G2x (Pa) G2z (Pa) Loss factor Loss factor Poisson’s
HCP HCP + CLD ratio of core
Panel 2 430 × 106 220 × 106 0.005 0.04 0.33
Panel 3 201 × 106 54 × 106 0.005 0.04 0.33
∂4 W ∂4 W ∂4 W
Dx + 2D xz + D z − ω 2 mW = jωΦ, (3.3)
∂x4 ∂x2 ∂z2 ∂z4
where Dx and Dz are the bending stiffness in x − and z− directions respectively and
√
Dxz can be approximated by Dxz ≈ Dx Dz .
13
Sathish Kumar
!2 −2
2
cos2 φ
ωm cos θ sin φ 4 4
τ (θ, φ) = 1 + j 1− + 2 k0 sin θ , (3.4)
2ρ0 c0 k2Bx k Bz
p p
where k Bx = 4 ω 2 m/Dx and k Bz = 4 ω 2 m/Dz are the wavenumbers in x − and
z−directions respectively. For small loss factors, the damping can be included by re-
placing Dx and Dz in Equation 3.4 with Dx (1 + jη ) and Dz (1 + jη ) respectively [17].
Assuming the same loss factor for waves in all the directions in the panel, Equation 3.4
can be re-arranged as
" #2 " !#2 −1
ωm cos θ k40 sin4 θ ωm cos θ k40 sin4 θ
τ (θ, φ) = 1+η + 1− , (3.5)
2ρ0 c0 k4φ,eqv. 2ρ0 c0 k4φ,eqv.
1 sin2 φ cos2 φ
= + 2 . (3.6)
k2φ,eqv. k2Bx k Bz
The diffuse field sound transmission coefficient over all angles of incidence θ and trans-
mission φ is calculated by [36]
14
Overview and Summary
´ 2π ´ θlim
0 0 τ (θ, φ) sin (θ ) cos (θ ) dθdφ
τ= ´ 2π ´ θlim , (3.7)
0 0 sin ( θ ) cos ( θ ) dθdφ
where θlim , is the limiting angle above which it is assumed that no sound is incident
on the partition and according to Mulholland et al. [30] θlim can be limited to 78◦ . The
sound transmission loss is then calculated by
1
STL = 10 log10 . (3.8)
τ
ρ0 c0 σ
ηrad = , (3.9)
ωm
where ηrad is the radiation loss factor and σ is the radiation efficiency which has
been investigated by Beranek [2] and Maidanik [27]. According to Leppington et al.
[24], a better estimation of the averaged radiation efficiency of a rectangular plate can
be given as
a+b µ+1 2µ
σ̄ ≈ ln + 2 µ > 1,
πµkab (µ2 − 1) µ−1 µ −1
√
σ̄ ≈ kaH ( x ) µ = 1, (3.10)
q
σ̄ ≈ 1/ 1 − µ2 µ < 1,
p
where a and b are the width and length of the plate with a < b, µ = k B /k = f c / f
with f c being the critical frequency of the plate, and H ( x ) is a function of the panel
aspect ratio ( a/b) and in our case it can be approximated to 0.41 [24].
For an isotropic panel, there is one critical frequency and the radiation loss factor
can be applied as in Reference 31. But for orthotropic panels, there exists different crit-
ical frequencies depending on the transmission angle φ. For anisotropic panels, there
exists no expressions for calculating the wavenumbers and critical frequency as a func-
tion of φ. Since the honeycomb panel is treated as being orthotropic , the equivalent
critical frequency in the ribbon (x) and transverse (z) directions can be calculated using
15
Sathish Kumar
the equivalent wavenumber from Equation 3.6 and the related radiation loss factor can
be obtained.
The loss factor is normally measured in 1/3rd octave bands because of the large
fluctuations in the measurements. However, Feng and Kumar [13] showed that by us-
ing a measured loss factor in 1/3rd octave band, the sound transmission loss will be
underestimated around the critical frequency because of the extremely large radiation
loss factor concentrated in a very small frequency band as indicated in Equation 3.10.
As a consequence, the radiation loss factor calculated in narrow band has to be added
to the measured loss factor for better sound transmission loss predictions around the
critical frequency [13]. The total loss factor can be expressed as sum of the measured
loss factor and the radiation loss factor.
16
Chapter 4
17
Sathish Kumar
Figure 4.1: Double-leaf partition with stiffeners and absorbing materials in cavity.
take into account the effect of the structural response on the acoustic system. Lee and
Kim [21] adopted the method of Mead and Pujara to a 2-D single plate with stiffeners
and included the effect of structural response on the acoustic system. The stiffeners
were modelled as periodically distributed lumped masses mounted to the two pan-
els, together with a series of uniformly distributed translational and rotational springs
and a detailed approach was presented to calculate the sound transmission loss of the
structure. As an extension of Lee’s work [21], Wang et al. [39] studied double-walled
rib-stiffened panels and following Lee, Wang disregarded many complicating features
(e.g. sound absorbing material in the cavity). In this thesis, Wang’s model has been
extended to include the effects of different fluid properties in the centre cavity which
accounts for the sound propagation through an absorbing material. Also, the model is
modified so that orthotropic sandwich panels can be included as a part of the double-
walled partition. Naturally, the same model can be simplified and used to investigate
the sound transmission properties of a double-leaf partition with standard sandwich
panels.
18
Overview and Summary
given by Kr0 = EIz /l where Iz is the second moment of area about the z-axis. For “C”
shaped stiffeners studied in this thesis, Iz = t3o /12.
The plane wave pi , incident on the periodic panel with stiffeners induces a reflected
wave pr , a transmitted wave pt and the panel motion. Since the structure is periodic,
the system response is also expected to be periodic [28], the motion of each panels can
be expressed in terms of space harmonic expansions. The plane wave is assumed to
incident along the x-y plane at an angle θ as shown in Figure 4.2b and the incident
wavenumber components k x and k y are given as
where k = ω/c0 and ω is the angular frequency (2π f ) and c0 is the speed of sound in
air. Similarly, the velocity potentials in the incident, cavity and the transmitted space
can be expressed in terms of space harmonic series [28]. The wavenumber in the y-
direction in, k yn is given by the expression
s
2
2 2nπ
k yn = (k) − k x + . (4.2)
L
Since the cavity is filled with an absorbing material, the wavenumber in the y-direction
in the cavity, k ync is given by the expression
19
Sathish Kumar
s
2
2 2nπ
k ync = (k c ) − k x + . (4.3)
L
where k c is the complex wavenumber in the absorbing material and is related to the
dynamic density ρc (ω ) and the dynamic bulk modulus Kc (ω ) of the absorbing material
as
s
ρc (ω )
kc = ω . (4.4)
Kc ( ω )
The dynamic density and the dynamic bulk modulus are obtained through Allard-
Champoux model for rigid frame materials [1]
1 σ ρo f
ρ c ( ω ) = ρ0 1+ G1 , (4.5)
j2π ρ0 f σ
! −1
γ−1
Kc (ω ) = γP0 γ− . (4.6)
1 + 1/j8πNpr (ρo f /σ )−1 G2 (ρ0 f /σ )
p
The functions G1 (ρ0 f /σ) and G2 (ρ0 f /σ) are given by G1 (ρ0 f /σ) = [1 + jπ (ρ0 f /σ )]
and G2 (ρ0 f /σ ) = G1 (ρ0 f /σ) 4Npr , where ρ0 is the density of air, σ is the static air
flow resistivity, γ is the specific heat ratio, P0 is the air equilibrium pressure and Npr
is the Prandtl number. A real part of the wavenumber greater than unity indicates a
lower speed of sound through the absorbing material and a negative imaginary part
indicates the dissipation of energy. Applying the boundary conditions at the fluid-
panel interface, the modal amplitudes can be found in terms of the amplitudes of the
space harmonics. The amplitudes of the space harmonics can be found by applying the
principle of virtual work as explained in Appendix-Part B. Making use of the coupling
relations between the modal amplitudes of waves in air and the flexural motion of
panel, the following relations are obtained.
20
Overview and Summary
2ρ 2jk ync H
4 jω jω 2 ρ
c 1 + e
D1 k x + 2mπ − m p1 ω 2 + 0
− α1,m
L k yn k ync 1 − e 2jk ync H
+∞ +∞
! " #
0 0
0 2nπ 2mπ
2
+ Kt − ω M ∑ α1,n + Kr ∑ α1,n k x + L kx +
L
n=−∞ n=−∞
+∞
! "
+∞ # (4.7)
0 0 2nπ 2mπ
− Kt ∑ α2,n − Kr ∑ α2,n k x + L kx +
L
n=−∞ n=−∞
2jω 2 ρc e jkync H 2jωρ I, f or m = 0,
0
+ α2,m =
k 1 − e2jkync H
ync
0, m 6= 0
4 jk yn H jk ync H
jω 2 ρ 2jω 2 ρ
D2 k x + 2mπ 0e ce
− m p2 ω 2 + − − α2,m
L k yn
k ync 1 − e 2jk ync H
+∞ +∞
! " #
0 0
0 2nπ 2mπ
+ Kt − ω 2 M ∑ α2,n + Kr ∑ α2,n k x + L kx +
L
n=−∞ n=−∞
+∞
! "
+∞ #
(4.8)
0 0 2nπ 2mπ
− Kt ∑ α1,n − Kr ∑ α1,n k x + L kx +
L
n=−∞ n=−∞
jω 2 ρc 1 + e2jkync H
+ α1,m = 0.
2jk ync H
k ync 1 − e
These equations are solved for αi,m which can be used to find the modal amplitudes.
The power transmission coefficient is then calculated by
∞
∑nn=+ 2 Re k
=−∞ | ξ n | yn
τ (θ ) = . (4.9)
| I |2 k y
The diffuse field sound transmission coefficient over all angles of incidence θ within
the incident plane is calculated by [36]
´ θlim
0 τ (θ ) sin (θ ) cos (θ ) dθ
τ= ´ θlim . (4.10)
0 sin ( θ ) cos ( θ ) dθ
21
Sathish Kumar
The plane wave pi , incident on the periodic panel with stiffeners induces a reflected
wave pr , a transmitted wave pt and the panel motion. Since the structure is periodic,
22
Overview and Summary
the system response is also expected to be periodic [28], the motion of each panels can
be expressed in terms of space harmonic expansions. The plane wave is assumed to
incident along the x-y plane at an angle θ as shown in Figure 4.3b and the incident
wavenumber components k x , k y and k z are given as
where k = ω/c0 and ω is the angular frequency (2π f ) and c0 is the speed of sound in
air. The wavenumbers in the y-direction, k yn and k ync are given by the expressions
s
2
2 2nπ
k yn = (k) − k x + − ( k z )2 and (4.12)
L
s
2
2 2nπ
k ync = (k c ) − k x + − ( k z )2 . (4.13)
L
Applying the boundary conditions at the fluid-panel interface, the relationship be-
tween modal amplitudes and the amplitudes of the space harmonics are obtained. Ap-
plying the principle of virtual work, making use of the coupling relations between the
modal amplitudes of waves in the fluid and the flexural motion of panel, the following
relations are obtained.
" 4 2
2mπ 2mπ
Dx1 kx + + Dxz1 kx + (k z )2 + Dz1 (k z )4 − m p1 ω 2
L L
jω 2 ρc 1 + e2jkync H +∞
!
jω 2 ρ 0 0
∑ α1,n
0
+ − α1,m + Kt − ω 2 M
k yn k ync 1 − e 2jk ync H n=−∞
+∞ +∞
" # !
0 2nπ 2mπ 0
+Kr ∑ α1,n k x + kx + − Kt ∑ α2,n (4.14)
n=−∞ L L n=−∞
+∞
" #
0 2nπ 2mπ
−Kr ∑ α2,n k x + kx +
n=−∞ L L
2jω 2 ρc e jkync H 2jωρ I, f or m = 0,
0
+ α2,m =
k 1 − e2jkync H
ync
0, m 6= 0
23
Sathish Kumar
" 4 2
2mπ 2mπ
Dx2 kx + + Dxz2 kx + (k z )2 + Dz2 (k z )4 − m p2 ω 2
L L
+∞
!
jω 2 ρ0 e jkyn H 2jω 2 ρc e jkync H 0 0
+
k yn
−
2jk ync H
− α2,m + Kt − ω 2 M ∑ α2,n
k ync 1 − e n=−∞
+∞ +∞
" # !
0 2nπ 2mπ 0
+Kr ∑ α2,n k x + kx + − Kt ∑ α1,n (4.15)
n=−∞ L L n=−∞
+∞
" #
0 2nπ 2mπ
−Kr ∑ α1,n k x + kx +
n=−∞ L L
jω 2 ρc 1 + e2jkync H
+ α1,m = 0.
2jk ync H
k ync 1 − e
Where, Dxi and Dzi are the bending stiffness of the sandwich panels in two orthotropic
√
directions and Dxzi can be approximated by Dxzi ≈ Dxi Dzi . For a honeycomb panel,
Dxi and Dzi can be obtained using the method described in Chapter 3.
These equations are solved for αi,m which can be used to find the modal amplitudes.
The power transmission coefficient is then calculated by
∞
∑nn=+ 2
=−∞ | ξ n | Re k yn
τ (θ, φ) = . (4.16)
| I |2 k y
The diffuse field sound transmission coefficient over all angles of incidence θ and trans-
mission φ is calculated by [36]
´ 2π ´ θlim
0 0 τ (θ, φ) sin (θ ) cos (θ ) dθdφ
τ= ´ 2π ´ θlim . (4.17)
0 0 sin (θ ) cos (θ ) dθdφ
Where θlim , the angle above which it is assumed that no sound is received is limited to
78◦ [30]. The sound transmission coefficients from Equations 4.10 & 4.17 can be used
to calculate the sound transmission loss by
1
STL = 10 log10 . (4.18)
τ
The model in this thesis is developed for a double-leaf partition made of two or-
thotropic sandwich panels but naturally, isotropic material properties can also be ap-
plied.
24
Overview and Summary
4.3 Convergence
Since the solution to this space harmonic analysis is obtained in a series form, a conver-
gence check is performed to establish the number of terms required for accurate sound
transmission loss predictions.
70
Convergence line
60
50
STL (dB)
40
200 Hz
1000 Hz
30 5000 Hz
10000 Hz
20
10
0
0 10 20 30 40 50 60 70 80 90 100
Number of Terms
The standard procedure is to find the number of terms required at the highest fre-
quency of interest and thereafter calculate the sound transmission for all other frequen-
cies with the same number of terms. The cost of using a fixed number of terms for the
series expansion is the high computation time. To reduce computation time, a new cri-
terion is developed here for selecting the space harmonic terms for quick convergence.
25
Sathish Kumar
The basis of the new convergence criterion is that the pressure field at a given fre-
quency can be expressed with a limited number of propagating and evanescent waves.
At low frequencies this number of propagating waves is low, but it increases with fre-
quency and a larger number of terms is needed for convergence. In this new method
only the space harmonics corresponding to the propagating and a few evanescent
waves are included in the solution for each frequency. Neglecting cavity absorption,
propagating and evanescent waves can be separated by checking the wavenumbers
in the y-direction. For double-leaf partitions with isotropic panels, the wavenumber
in the y-direction is given by Equation 4.2. When ω/c0 >| k x + 2nπ/L | the wave is
propagating and when ω/c0 <| k x + 2nπ/L | the wave is evanescent (wavenumber
has only an imaginary part). This is illustrated in Figure 4.5 where the wavenumber
k yn is plotted against the space harmonics at selected frequencies. Since the evanescent
waves do not possess a real part, their amplitude is shown as 0.
200
180 1000 Hz
5000 Hz
160 10000 Hz
Space Harmonics
140 corresponding to
Propagating waves
120
real (k yn)
60
40
20
0
−40 −30 −20 −10 0 10 20 30 40
n
Figure 4.5: Real k yn plotted against the space harmonic terms, θ = 0◦ (Old criterion).
Lf
nmin = − (1 + sinθ ) , (4.19)
c0
Lf
nmax = + (1 − sinθ ) . (4.20)
c0
26
Overview and Summary
By evaluating α1,n and α2,n in Equations 4.7 & 4.8 with n ranging from nmin to
nmax , includes only the propagating waves in the solution. Since the panels are closely
spaced, some evanescent waves do not decay sufficiently and have considerable am-
plitudes when reaching the second panel. The space harmonic terms corresponding
to these evanescent waves cannot be ignored and therefore, a number of evanescent
waves (∆n) should be included in the solution (n = nmin − ∆n to nmax + ∆n). A nu-
merical investigation revealed that evanescent waves with amplitude decays of more
than 60 dB when reaching the second panel, do not significantly affect the sound trans-
mission loss and therefore need not be included in the calculation. The old and the new
criterion for convergence can be illustrated in Figures 4.6 & 4.7.
200 200
Propagating waves Propagating waves
150 150
real (kyn)
real (kyn)
50 50
0 0
10000 10000
8000 8000
6000 20 6000 20
4000 10 10
4000
0 0
2000 2000 −10
−10
100 −20 100 −20
Frequency (Hz) Frequency (Hz) Space Harmonic terms
Space Harmonic terms
Figure 4.6: Real k yn plotted against the space harmonics and frequency, θ = 0◦ .
100
100 Propagating waves
Propagating waves
80
80
60
real (kyn)
Evanescent waves
real (kyn)
60
40 Evanescent waves
40
20
20
0
90 0
80
70 80
60 20
50 60
40 20 10
30 40
10 0
20 0 20
10 −10 −10
0 −20 0 −20
Angle of Incidence (θ) Angle of Incidence (θ) Space Harmonic terms
Space Harmonic terms
Figure 4.7: Real k yn plotted against the space harmonics and angle of incidence, f =
5000 Hz.
In Figure 4.6, the wavenumbers are shown as a function of frequency with normal
incidence, wave heading (θ = 0◦ ), whereas in Figure 4.7, the wavenumbers are plotted
at 5000 Hz as a function of wave heading angle. For the new criterion, only the space
harmonics which are included in the solution are plotted against the real part of the
27
Sathish Kumar
wavenumber k yn for different frequencies. The non-zero wavenumbers in the plot cor-
respond to propagating waves and the others corresponds to evanescent waves. The
number of evanescent waves included in the solution, ∆n = 5.
And for orthotropic plates, from Equation 4.12, it can be determined that the waves
are propagating when ω/c0 >| (k x + 2nπ/L) + k z | and evanescent when ω/c0 <|
(k x + 2nπ/L) + k z |. Hence, for a given frequency f , angle of incidence θ, and angle of
propagation in the plate φ and stiffener separation distance L, the propagating waves
are limited to the region nmin ≤ n ≤ nmax where
q
Lf 2
nmin =− 1 − (sinθsinφ) + sinθcosφ , (4.21)
c0
q
Lf 2
nmax =+ 1 − (sinθsinφ) − sinθcosφ . (4.22)
c0
28
Chapter 5
Measurements
Loss factor is a good measure of the structural damping present in a system. From the
several methods available, a simple and robust decay method was used to measure
the loss factor of the test panels in 1/3rd octave bands. The test panel was suspended
in springs to achieve free-free boundary conditions minimizing boundary losses apart
from radiation. For the same reason, impulse excitation was used to avoid unwanted
external damping due to a shaker mounting. The decay method uses the fact that free
vibrations decay with time and by measuring the reverberation time of the structure,
the loss factor can be calculated by [8]
2.2
η= , (5.1)
fn T
where f n is the frequency and T is the reverberation time in seconds. Here, random
accelerometer positions and excitation points were used to get a spatial average of the
measured vibration response and to reduce the influence from individual modes.
After the damping layers were attached, the reverberation time decreased signifi-
cantly making decay measurements difficult. Therefore, power injection method was
used to measure the loss factor of the structures with damping treatment. This method
is not ideal for complex built-up structures, but works very well for freely hanging
simple structures as in our case. For a point excited system, the input power is related
to the spatial average of the vibration velocity as P = Sm0 ωηv2∆ /2 [8]. This equation
can be re-arranged to calculate the loss factor, as
2P
η= , (5.2)
Sm0 ωv2∆
29
Sathish Kumar
where P is the input power and S, m0 , v2∆ the area, surface density and the mean
square vibration velocity of the test structure respectively.
In practice, using harmonic excitation, the loss factor is determined according to
Feng [12], as
− Img( G f a )
η= , (5.3)
Sm0 Gaa
with G f a being the cross-spectrum of the excitation force and response acceleration
and Gaa being the power spectrum of acceleration.
Anechoic Room
Anechoic Room Reverberation Room
Reverberation Room
Test Panel
Test Panel
Rotating
Boom
Loud-
speaker
Intensity probe
Intensity probe Rotating Boom Loud Speaker
PC &
Signal Generator Amplifier
The panel was mounted in such a way that flanking transmission was minimised
30
Overview and Summary
thus ensuring that the sound transmission was only through the test panel. Then the
air-borne sound reduction index was calculated as given below
Where R I is the sound reduction index, L P is the sound pressure level in the rever-
beration room, L I is the sound intensity measured in the anechoic room, Sm is the mea-
sured surface area and S is the area of the test specimen. Weighted (apparent) sound
reduction index RW 0 , a single numbered quantity used to describe the sound insulation
where Lv is the vibration velocity level, < v > is the average panel normal velocity
and vre f = 10−9 m s-1 is the reference value for vibration velocity.
31
Sathish Kumar
ure 2.3) and suspended using rubber strings to achieve free-free boundary conditions.
The measurement setup is shown in Figure 5.2.
Charge Amplifier
Rubber Strings
PC & DAQ
Beam Accelerometer
Impact Hammer
An impact hammer was used to excite the beam and an accelerometer was used
to measure the response. The inertial effects due to accelerometer mass was neglected
since the mass of the accelerometer was too small (2.4 grams) compared to the test
beams (975 grams and 889 grams). Different accelerometer positions were selected and
measured to make sure to find all relevant modes. From the measured natural frequen-
cies the bending stiffness was calculated by [42]
ωn m0 L4x ωn m0 L4z
Dx = , D z = , (5.6)
α4n α4n
where ωn is the natural angular frequency, L the length of the beam, m0 the beam
surface density, n is the order of the natural frequency and α1 = 4.73, α2 = 7.85, α3 =
11.00 , α4 = 14.14 , αn = nπ + π/2 for n > 4 . From the measured bending stiffnesses,
the flexural wavenumbers of the beams were calculated from the relations
s s
4 ωn2 m0 4 ωn2 m0
kx = , kz = . (5.7)
Dx Dz
32
Chapter 6
The loss factor of the honeycomb panels were measured using the decay method as
explained in Section 5.1. The measured loss factors for all three honeycomb panels
without and with the damping treatment are shown in Figure 6.1.
0.1 0.1
Panel 1 Panel 2
0.09 Panel 1 + CLD 0.09 Panel 2 + CLD
0.08 0.08
0.07 0.07
0.06 0.06
Loss Factor
Loss Factor
0.05 0.05
0.04 0.04
0.03 0.03
0.02 0.02
0.01 0.01
0 0
100 125 160 200 250 315 400 500 630 800 1000 100 125 160 200 250 315 400 500 630 800 1000
Frequency (Hz) Frequency (Hz)
0.1
Panel 3
0.09 Panel 3 + CLD
0.08
0.07
0.06
Loss Factor
0.05
0.04
0.03
0.02
0.01
0
100 125 160 200 250 315 400 500 630 800 1000
Frequency (Hz)
(c) Panel 3
Figure 6.1: Measured loss factor of the honeycomb panels without and with damping
treatment.
33
Sathish Kumar
From the figures it can be concluded that there is a significant increase in the struc-
tural loss factor with damping treatment for all the three panels. The decay measure-
ments were limited to a frequency of 1000 Hz above which, the reverberation time was
too short to get a useful result. Due to the previously mentioned problem of obtaining
loss factor results using the decay method with the damping treatment, power injection
method was used later. The principal use of damping treatments is to reduce vibrations
in the structure. To see the effect of damping treatments on vibration attenuation, the
vibration velocity level of the panels were compared without and with the damping
treatment. Figures 6.2a, 6.3a & 6.4a show the averaged vibration velocity level of the
panels to an acoustic excitation without and with the damping treatment. A diffuse
sound field was created in the reverberation room and it was ensured that the acoustic
excitation levels were same for all panels and measurements. As expected, the damp-
ing treatments reduce the vibration levels for all three panels. Figures 6.2b, 6.3b & 6.4b
shows the vibration level difference of the panels before and after the treatment.
120 14
Panel 1
Panel 1 + CLD
12
110
Vibration Velocity Level (dB)
10
100
8
∆ Lv (dB)
90
6
80
4
70
2
60 0
125 250 500 1000 2000 4000 125 250 500 1000 2000 4000
Frequency (Hz) Frequency (Hz)
120 14
Panel 2
Panel 2 + CLD
12
110
Vibration Velocity Level (dB)
10
100
8
∆ Lv (dB)
90
6
80
4
70
2
60 0
125 250 500 1000 2000 4000 125 250 500 1000 2000 4000
Frequency (Hz) Frequency (Hz)
34
Overview and Summary
120 14
Panel 3
Panel 3 + CLD
12
110
Vibration Velocity Level (dB)
10
100
∆ Lv (dB)
90
6
80
4
70
2
60 0
125 250 500 1000 2000 4000 125 250 500 1000 2000 4000
Frequency (Hz) Frequency (Hz)
The sound reduction index of the panels was measured using the sound intensity
technique as explained in Section 5.2. The procedure was repeated after the damping
treatments were applied. The measured sound reduction index of panel 1 without and
with the damping treatment is shown in Figure 6.5. Just as observed in the loss factor
results, the damping treatment increased the sound reduction index of the panel. How-
ever, this increase is not entirely due to the damping treatment. A part of this increase
is due to the mass added to the panel due to the damping treatment (masslaw).
45
Panel 1
40 Panel 1 + CLD
35
Sound Reduction Index (dB)
30
25
20
15
10
0
125 250 500 1000 2000 4000
Frequency (Hz)
Figure 6.5: Measured sound reduction index of panel 1 without and with damping.
For this reason, mass normalisation was done to remove any effect of added mass
on the results. This was done by subtracting the mass contribution ∆ = 20 log(m0 /m00 )
from the measured sound reduction indices in each frequency bands, where m0 and m00
are the mass densities with and without damping layers attached. The measured sound
35
Sathish Kumar
reduction index of panel 1 compared to the mass normalised sound reduction index
with damping treatment is shown in Figure 6.6a. The result shows that the damping
treatment has very little influence on the sound reduction index of the panel. Similarly,
the sound reduction index of panel 2 and panel 3 are mass normalised and the results
are shown in Figure 6.6b and Figure 6.6c respectively. As observed for panel 1, the
damping treatment on panel 2 has very little influence. On the other hand, for panel 3
a significant increase in sound transmission loss can be seen above 500 Hz.
45 45
Panel 1 Panel 2
40 Panel 1 + CLD 40 Panel 2+ CLD
35 35
Sound Reduction Index (dB)
25 25
20 20
15 15
10 10
5 5
0 0
125 250 500 1000 2000 4000 125 250 500 1000 2000 4000
Frequency (Hz) Frequency (Hz)
45
Panel 3
40 Panel 3 + CLD
35
Sound Reduction Index (dB)
30
25
20
15
10
0
125 250 500 1000 2000 4000
Frequency (Hz)
(c) Panel 3
Figure 6.6: Measured sound reduction index of the honeycomb panels without and
with damping treatment (mass normalised).
To investigate and quantify this influence, weighted sound reduction index RW0 ,a
single number quantity characterising the sound insulation of a partition was calcu-
lated according to ISO 717 − 1 : 2006 and is presented in Table 6.1. For panel 1, the
damping treatment increases the sound insulation by 5 dB whereas the influence from
added mass is 4.9 dB which shows that the increased losses does not increase sound
insulation. On the other hand for panel 3, the sound insulation increases by 8 dB out
of which only 3.9 dB is due to the added mass, clearly showing that damping treat-
ment increases sound insulation. From Table 6.1, it can also be inferred that without
the damping treatment, panel 2 has a weighted sound reduction index 3 dB more than
panel 3. But, once the damping material is attached, both panels have the same sound
36
Overview and Summary
insulation rating. It should be noted that the only structural difference between panel
2 and 3 is the honeycomb cell size affecting the shear stiffness of the core.
Table 6.1: Weighted sound reduction index of the honeycomb panels without and with
damping treatment.
The lowest frequency at which the wavelength of bending waves in plate is equal to
the wavelength in air is known as the critical frequency. For a single-leaf isotropic panel,
there exists one critical frequency f c whereas, for a honeycomb panel there exists a
range of critical frequencies between f cL and f cW (where the subscripts L and W are
the two principal orthotropic directions). This critical frequencies in the honeycomb
panels can be better illustrated using dispersion curves, where the wavenumbers of
the panels are plotted together with the wavenumbers in air.
Using the sandwich theory, the wavenumbers of the honeycomb panels were cal-
culated in the two orthotropic directions (L and W) and plotted in Figures 6.7a and
6.7b. In these figures, critical frequency is observed where the plate wavenumbers
(k L and kW ) intersect the wavenumbers in air (k air ). In Figure 6.7, the parallel lines are
the asymptotes to the wavenumbers. The lower asymptote (k lower ) corresponds to the
wavenumber for pure bending of the entire construction whereas, the upper asymp-
tote (k upper ) corresponds to the wavenumber of flexural waves propagating in one of
the laminate. It can be seen in Figure 6.7b that the wavenumbers in W-direction (kW )
matches with the wavenumbers in air (k air ) over a large frequency range (Figure 6.8).
This implies that the wavelength of the bending wave in the plate is close to the wave-
length of sound in air which leads to coincidence over the entire frequency range.
37
Sathish Kumar
2 2
10 10
Wave Number (1/m)
125 250 500 1000 2000 4000 8000 125 250 500 1000 2000 4000 8000
Frequency (Hz) Frequency (Hz)
2
10
Wave Number (1/m)
1
10 kair
klower
kupper
kL
kw
Later, the panels were cut in the W-direction into beams (2050 mm x 70 mm) and
their bending wavenumbers were measured. The predicted wavenumbers are com-
pared with the measurements in Figure 6.9 and a good agreement is achieved. The
exact values of the measured and predicted critical frequencies for the two panels are
shown in Table. 6.2. The measured and predicted critical frequencies in W-direction
were found to be within a range of 2%.
38
Overview and Summary
Panels f cL Hz f cW Hz f cW Hz
prediction prediction measurement
Panel 2 792 888 904
Panel 3 851 3632 3556
2
10
Wave Number (1/m)
1
10
Air
0
Panel 2−Measured
10 Panel 3−Measured
Panel 2−Predicted
Panel 3−Predicted
Further, the vibration velocity level difference of the panel is compared with the
sound reduction index difference before and after the application of damping. A typi-
cal effect of damping treatments can be seen in Figure 6.10a. For panel 1, even though
significant vibration reduction is achieved with damping treatment, the sound radiated
from the structure is not reduced to the same extent. For panel 2, the increased sound
reduction index is approximately equal to the vibration reduction as seen in Figure
6.10b. On the other hand for panel 3 in Figure 6.10c, an interesting phenomenon is ob-
served. For panel 3, the increase in sound reduction index is more than the vibration
reduction at certain frequencies. The reason for this effect is the extended coincidence
range for panel 3 as observed before. It is well know that damping is more effective in
and around the critical frequency. Naturally, when damping treatments are applied to
a panel with a wide critical frequency range, their effect is greater than for a standard
panel.
39
Sathish Kumar
14 14
∆ Lv ∆ Lv
12 ∆ SRI 12 ∆ SRI
10 10
8 8
∆ (dB)
∆ (dB)
6 6
4 4
2 2
0 0
125 250 500 1000 2000 4000 125 250 500 1000 2000 4000
Frequency (Hz) Frequency (Hz)
14
∆ Lv
12 ∆ SRI
10
8
∆ (dB)
0
125 250 500 1000 2000 4000
Frequency (Hz)
(c) Panel 3
Figure 6.10: Comparison of the difference in vibration velocity level and sound reduc-
tion index without and with the damping treatment.
The sound reduction index of the honeycomb panel was predicted using the model
described in Section 3. Figure 6.11 shows the measured SRI of panel 2 compared with
the predictions made using the measured structural loss factor averaged over fre-
quency. It can be seen that for panel 2, the predictions with the average loss factor holds
good for frequencies below critical frequency ( f < f c ) but at ( f = f c ) and above the
critical frequency ( f > f c ), the sound reduction index is under predicted. According to
Heckl [17], the radiation loss factor can have a significant contribution for structures
such as composite structures that are very lightly damped similar to those investigated
in this thesis. As a consequence, the radiation loss factor calculated in narrow band
frequency as in Equation 3.10 is added to the measured loss factor for better sound re-
duction index predictions around the critical frequency as shown by Feng and Kumar
[13].
40
Overview and Summary
45
Panel 2 − Measured
40 Panel 2 − Prediction
35
Sound Reduction Index (dB)
30
25
20
15
10
0
125 250 500 1000 2000 4000
Frequency (Hz)
Figure 6.12a shows the predicted sound reduction index calculated with an added
radiation loss factor. A good agreement can be seen between the measured curve and
the predictions when the theoretical radiation loss factor is added. The same compari-
son for panel 3 is shown in Figure 6.12b.
45 45
Panel 2 − Measured Panel 3 − Measured
40 Panel 2 − with Radiation Lossfactor 40 Panel 3 − with Radiation Lossfactor
35 35
Sound Reduction Index (dB)
30 30
25 25
20 20
15 15
10 10
5 5
0 0
125 250 500 1000 2000 4000 125 250 500 1000 2000 4000
Frequency (Hz) Frequency (Hz)
Figure 6.12: Measured and predicted sound reduction index with added radiation
losses.
41
Sathish Kumar
The periodic model derived in Section 4.1 is used to examine the vibro-acoustic perfor-
mance of a double-leaf partition with homogenous aluminium panels and aluminium
stiffeners.The predicted transmission loss is then compared to the experimental data
from Legault and Atalla [23]. The parameters listed in Table 6.3 are used to describe the
partition and the sound absorbing material in the cavity. Assuming the air in the cavity
to be at standard temperature and pressure we have ρ0 = 1.21 kg m-3, Npr = 0.702,
and P0 = 101325 Pa and the wavenumbers in the cavity are obtained using Equation
4.4.
Table 6.3: Dimensions and material data of panels and fibrous material [23].
Using the traditional convergence criterion, the number of terms required at the
highest frequency of interest is determined and the sound transmission is calculated
for all other frequencies with the same number of terms. From the convergence curve in
Figure 6.13a, it is found that at least 65 terms are needed for convergence at 10 kHz. The
sound transmission loss of the partition predicted using the traditional convergence
criterion is compared to the measured data in Figure 6.13b. The calculations are made
in narrow bands with a frequency step of 10 Hz and later averaged over 1/3rd octave
bands. From the figure it can be concluded that an overall agreement is achieved.
42
Overview and Summary
60
70
Convergence line
60 50
40
200 Hz 30
1000 Hz
30 5000 Hz
10000 Hz 20
20
10 Measured
10
Predicted
n= −32 to +32
0 0
0 10 20 30 40 50 60 70 80 90 100 125 250 500 1000 2000 4000 8000
Number of Terms Frequency (Hz)
Figure 6.13: Convergence analysis and sound reduction index of the partition.
60
50
Sound Reduction Index (dB)
40
30
20
∆n = 2
10
∆n = 5
∆n = 10
0
125 250 500 1000 2000 4000 8000
Frequency (Hz)
Figure 6.14: Predicted sound reduction index using the new convergence scheme with
different ∆n.
It can be observed that the sound transmission results obtained with ∆n = 5 is very
similar to that obtained with ∆n = 10 but with a smaller computational expense. A
43
Sathish Kumar
numerical investigation revealed that when ∆n > 5, the evanescent waves decayed by
60 dB when reaching the second panel. Therefore, for the partition investigated here it
can be concluded that ∆n = 5 is enough to achieve numerical convergence. Figure 6.15
shows the measured sound transmission loss compared to the predictions obtained by
the traditional fixed term method and the new adaptive method. For this partition,
the new method takes one tenth of the computation time of the traditional technique
without compromising the accuracy of the result.
60
50
Sound Reduction Index (dB)
40
30
20
10 Measured
Traditional Convergence (n = −32 to +32)
New Convergence with ∆n = 5
0
125 250 500 1000 2000 4000 8000
Frequency (Hz)
Figure 6.15: Measured and predicted sound reduction index with different conver-
gence schemes.
44
Overview and Summary
(a) Stiffener glued to the honeycomb (b) Partition with mineral wool but without aluminium panel
panel
Table 6.4: Dimensions and material data of panels and mineral wool.
The sound transmission loss of the partitions were measured using the sound in-
tensity technique between 100 Hz and 5000 Hz as described in Section 5.2. Figure 6.17
shows the measured sound transmission loss of the two partitions. A difference in the
sound transmission loss can be seen in the higher frequencies. The lower transmis-
sion loss of the partition with honeycomb panel 3 is due to its softer core. The sound
transmission loss of the two partitions were calculated using the periodic model for
sandwich panels as described in Section 4.2 and the results are shown in Figures 6.18
& 6.19. The bending stiffness of the honeycomb panels used in the calculation were
obtained analytically using the wave propagation theory as described in Reference 31.
45
Sathish Kumar
60
50
Sound Reduction Index (dB)
40
30
20
10
Partition with HC panel 2 (6.4 mm core)
Partition with HC panel 3 (19.2 mm core)
0
125 250 500 1000 2000 4000
Frequency (Hz)
Using the traditional convergence criterion, it was determined that at least 37 space
harmonic terms were necessary for the solution to converge at 5000 Hz . Whereas, us-
ing the new convergence criterion with ∆n = 5 , the computation time for the partitions
was reduced by a factor of 10. Figures 6.18 & 6.19 shows that the predictions using the
periodic model agree well with the measurements.
60
50
Sound Reduction Index (dB)
40
30
20
10
Measured
Prediction
0
125 250 500 1000 2000 4000
Frequency (Hz)
Figure 6.18: Measured and predicted sound reduction index of the partition with hon-
eycomb panel 2.
46
Overview and Summary
60
40
30
20
10
Measured
Prediction
0
125 250 500 1000 2000 4000
Frequency (Hz)
Figure 6.19: Measured and predicted sound reduction index of the partition with hon-
eycomb panel 3.
47
Sathish Kumar
70
60
Sound Reduction Index (dB)
50
40
30
20
10
Measured
Periodic model
0
125 250 500 1000 2000 4000
Frequency (Hz)
Figure 6.21: Comparison of the measured and predicted sound reduction index.
Figure 6.21 shows a comparison of the predictions, using the new convergence cri-
terion, to the test data from Reference 16 and it can be concluded that an overall agree-
ment is achieved. The space harmonics corresponding to all propagating waves and
a few evanescent (∆n = 3) waves were included in the calculation. For intermediate
stud thicknesses, the stiffness is rather complex but can be obtained using experiments
or through detailed stiffener vibration models [37].
48
Overview and Summary
Table 6.5: Dimensions and material data of panels and mineral fibre [16].
49
Chapter 7
The measurements carried out during this thesis illustrates that constrained layer damp-
ing treatments increase the loss factors for all the honeycomb panels tested. As ex-
pected, the vibration velocity levels also decreased after damping treatment. The effect
on the sound transmission properties is more complex. For panels 1 and panel 2 the
damping treatment does not affect the sound transmission loss apart from the effects
of the added mass. However, for panel 3, the same damping treatment increases the
sound insulation due to the panel’s extended coincidence range as demonstrated by
the dispersion curves. It is known that damping has its most influence around the crit-
ical frequency where the transmission is resonant. Naturally, panel 3 with its extended
coincidence range had its sound insulation improved when the damping treatment
was applied. The weighted sound reduction index, R0w of panel 2 was 3 dB higher
than that of panel 3 but with the damping treatment applied both panels have the
same weighted sound reduction index (R0w ). In-short, the visco-elastic damping treat-
ment increases the sound insulation of an acoustically bad panel (in our case panel 3).
Whereas, the same treatment on a normal panel (panel 1 and 2) does not improve the
sound insulation apart from the effects of added mass.
Radiation losses can be a significant contributor to the total losses for lightly damped
composite structures. Since the measured loss factor is averaged over 1/3rd octave
bands, the abrupt increase of loss factor around the critical frequency is not captured in
the measurements. Therefore, the measured structural loss factor alone is not enough to
characterise the total losses of a light weight panel around the critical frequency. Apart
from the measured structural loss factor, the radiation loss factor calculated in narrow
bands must also be included in the model to accurately predict the sound transmis-
sion loss of lightweight panels like those discussed in this thesis. The sound reduction
index calculated using the wave propagation approach, combined with orthotropic
theory, agrees very well with the measurements.
51
Sathish Kumar
52
Chapter 8
8.1 Paper A
On Application of radiation loss factor in the prediction of sound transmission loss
of a honeycomb panel
Leping Feng and Sathish Kumar.
This article shows the importance of radiation loss factor in predicting the sound
transmission property of lightly damped panels, especially around the critical fre-
quency region. When the sound transmission properties of a panel is investigated,
the radiation loss factor is often neglected for panels unless the structural damping
is very less or when the sound transmission around the critical frequency of the panel
is of interest. This article shows that by including the theoretical radiation loss factor
in narrow bands to the measured structural loss factor, the sound reduction index of
honeycomb sandwich panels can be predicted with good accuracy around the criti-
cal frequency. The wave propagation constants are calculated as given by Nilsson [31]
and the sound reduction index of the panels are predicted using the orthotropic panel
theory. The predictions are validated through laboratory measurements.
8.2 Paper B
Predicting the sound transmission loss of honeycomb panels using the wave propa-
gation approach
Sathish Kumar, Leping Feng and Ulf Orrenius.
The wave propagation approach can be used to predict the sound transmission
properties of sandwich panels having an isotropic core. In this article, the wave propa-
gation approach is used to predict the sound transmission loss of honeycomb panels. A
honeycomb panel is a lightweight sandwich panel with a honeycomb core of hexagon
53
Sathish Kumar
cell. Due to the manufacturing process of honeycomb cores, the cell wall thickness
doubles in the direction of cell orientation and the actual shape of the honeycomb cell
can be irregular making the core anisotropic. As an approximation, the honeycomb
core can be treated as orthotropic and the whole panel can be treated as a orthotropic
sandwich panel. The wavenumbers are calculated for the two principal orthotropic di-
rections. The orthotropic panel theory is used to predict the sound transmission loss
of panels with two different core structures. Visco-elastic damping with constraining
layers (CLD) are attached to these panel and the sound transmission loss is calculated.
The predicted wavenumbers and the sound transmission loss of the panels agree well
with laboratory measurements.
8.3 Paper C
Modelling the sound transmission through rib-stiffened double-leaf partitions with
cavity absorption
Sathish Kumar, Leping Feng and Ulf Orrenius.
The sound transmission loss of infinite, periodically rib-stiffened double-leaf parti-
tions with different fluids in the cavity is investigated. The sound transmission model
for double-leaf partitions with periodic stiffeners has been extended to include sound
absorbing materials in the cavity. The vibro-acoustic response of the panels is expressed
as a series of space harmonics and virtual work theory has been successfully applied
for calculating the sound transmission loss of the partition. The sound propagation
in the cavity is defined by assuming an equivalent fluid model. The number of space
harmonic terms required for convergence of the sound transmission loss is studied. To
reduce the computation time, a new criterion is introduced to select the appropriate
space harmonics based on frequency and angle of incidence. This new criterion en-
sures quick convergence and reduced computation time. For the partitions analysed,
the computation time was reduced by a factor of ten. The model is verified through
comparison with experimental data available in the literature.
8.4 Paper D
Modelling the sound transmission through rib-stiffened sandwich double-leaf par-
titions using space harmonic analysis
Sathish Kumar, Leping Feng and Ulf Orrenius.
An analytical model is developed for the sound insulation of infinite, periodically
rib-stiffened double-leaf partitions consisting of sandwich panels. The two panels are
54
Overview and Summary
separated by “C” shaped stiffeners and the cavity between the panels is filled with
mineral wool. The sandwich panel investigated in this paper has a honeycomb core
which makes the panels orthotropic. The apparent bending stiffness of the sandwich
panels in the two orthotropic directions are derived using sandwich theory. The cavity
between the sandwich panels is filled with absorbing material and the sound propaga-
tion in the cavity is defined by assuming an equivalent fluid model. The vibro-acoustic
response of the panels is expressed as a series of space harmonics and virtual work
theory has been applied for calculating the sound transmission loss of the partition. A
new criterion is employed for quick convergence which reduces the computation time.
For the partitions analysed, the computation time was reduced by a factor of ten. The
predicted transmission loss is validated through comparison with laboratory measure-
ments.
55
Appendix
Part A
h i h i
U1 = (1 − eα ) L1 ζ 2L − k2 ν2 / (1 − ν2 ) ; U2 = (1 + eα )ζ L k4 − k41
h i
U3 = (1 − e− β )ζ T kL1 [(1 − 2ν2 ) / (1 − ν2 )] ; U4 = (1 + e− β )k k4 − k41
h i
V1 = (1 − eα )k k2L1 − k2 ; V2 = (1 + eα )2ζ L M1 k
h i h i
−β −β
V3 = (1 − e )ζ T k2L1 −k 2
; V4 = (1 + e )2M1 k 2
− k2L1 / (1 − ν2 )
α = ζ L λ L ; β = ζ L λT
s s
ω2 ρ 2 (1 + ν2 ) (1 − 2ν2 ) 2ω 2 ρ2 (1 + ν2 )
ζL = k2 − ; ζT = k2 −
E2 (1 − ν2 ) E2
s s
ω2 ρ 2 ω 2 ρ2
λL = ; λT =
E20 − k 2 G2 − k2
56
Overview and Summary
Part B
The principle of virtual work states that the sum of the works done by all the elements
in one period of the system must do no virtual work when moved through any one of
the virtual displacements [28]
The total virtual work is the sum of virtual works of the two panels elements, the
virtual work of the translational and rotational springs and from the lumped mass of
the studs. The principle of virtual work requires that
The virtual work contributed by the two panel elements in one period of a panel ele-
ment can expressed as
ˆL
∂4 W1 ∂2 W1
δΠ p1 = D1 + m p1 2 − jωρo Φ1 + jωρc Φ2 δW1∗ dx at y = 0, (8.3)
∂x4 ∂t
x =0
ˆL
∂4 W2 ∂2 W2
δΠ p2 = D2 + m p2 2 − jωρc Φ2 + jωρo Φ3 δW2∗ dx at y = H. (8.4)
∂x4 ∂t
x =0
The virtual work contributed by the two panel elements in one period of a panel ele-
ment can expressed as
57
Sathish Kumar
ˆL
∂4 W1 ∂4 W1 ∂4 W1 ∂2 W1
δΠ p1 = Dx1 + 2D xz1 + D z1 + m p1 − jωρo Φ1 + jωρc Φ2
∂x4 ∂x2 ∂z2 ∂z4 ∂t2
x =0
δW1∗ dx at y = 0,
(8.5)
ˆL
∂4 W2 ∂4 W2 ∂4 W2 ∂2 W2
δΠ p2 = Dx2 + 2D xz2 + D z2 + m p2 − jωρc Φ2 + jωρo Φ3
∂x4 ∂x2 ∂z2 ∂z4 ∂t2
x =0
δW2∗ dx at y = H.
(8.6)
+∞ +∞
" #
∗
δΠt1 = Kt (W1 (0) − W2 (0)) δα1,m = Kt ∑ α1,n − ∑ ∗
α2,n δα1,m , (8.7)
n=−∞ n=−∞
+∞ +∞
" #
δΠt2 = ∗
Kt (W2 (0) − W1 (0)) δα2,m = Kt ∑ α2,n − ∑ ∗
α1,n δα2,m . (8.8)
n=−∞ n=−∞
0
0 2mπ
∗
δΠr1 = Kr W1 (0) − W2 (0) j k x + δα1,m
L
+∞ +∞
" #
2nπ 2nπ 2mπ
= Kr ∑ α1,n k x + − ∑ α2,n k x + kx + ∗
δα1,m ,
n=−∞ L n=−∞ L L
(8.9)
58
Overview and Summary
0 0
2mπ
∗
δΠr2 = Kr W2 (0) − W1 (0) j k x + δα2,m
L
+∞ +∞
" #
2nπ 2nπ 2mπ
= Kr ∑ α2,n k x + − ∑ α1,n k x + kx + ∗
δα2,m .
n=−∞ L n=−∞ L L
(8.10)
+∞
" #
∗
δΠm1 = −ω 2 MW1 (0) δα1,m = −ω 2 M ∑ ∗
α1,n δα1,m , (8.11)
n=−∞
+∞
" #
δΠm2 = −ω 2 ∗
MW2 (0) δα2,m 2
= −ω M ∑ ∗
α2,n δα2,m . (8.12)
n=−∞
59
Bibliography
[1] Jean F. Allard and Yvan Champoux. New empirical equations for sound propa-
gation in rigid frame fibrous materials. Journal of the Acoustical Society of America,
91:3346–3353, 1992.
[2] Leo L. Beranek. Noise and Vibration Control Engineering. John Wiley & Sons, 1992.
[3] U. Carlsson. Acoustical and vibrational aspects of future railway vehicles. Tech-
nical report, KTH-FKT, 1997.
[4] S. I. Cha and H. H. Chun. Insertion loss prediction of floating floors used in ship
cabins. Applied Acoustics, 69:913–917, 2008.
[11] Frank J. Fahy. Sound and Structural Vibration: Radiation, Transmission and Response.
Academic Press, 1987.
61
Sathish Kumar
[13] L. Feng and S. Kumar. On application of radiation loss factor in the prediction of
sound transmission loss of a honeycomb panel. International Journal of Acoustics
and Vibration, 17:47–51, 2012.
[14] S. Finnveden and M. Barbagallo. A cavity-wall element for statistical energy anal-
ysis of double walls. Technical report, KTH-AVE, 2012.
[17] M. Heckl. The tenth sir richard fairey memorial lecture: Sound transmission in
buildings. Journal of Sound and Vibration, 77:165–189, 1981.
[18] L. Kari, K. Lindgren, L.Feng, and A.C. Nilsson. Constrained polymer layers to
reduce noise: reality or fiction?- an experimental inquiry into their effectiveness.
Polymer esting, 21:949–958, 2002.
[20] G. Kurtze and B. Watters. New wall design for high transmission loss or high
damping. Journal of the Acoustical Society of America, 31:739–748, 1959.
[21] J. H. Lee and J. Kim. Analysis of sound transmission through periodically stiff-
ened panels by space-harmonic expansion method. Journal of Sound and Vibration,
251:349 – 366, 2002.
[22] J. Legault and N. Atalla. Sound transmission through a double panel structure
periodically coupled with vibration insulators. Journal of Sound and Vibration, 329:
3082 – 3100, 2010.
[23] J. Legault and N. Atalla. Numerical and experimental investigation of the effect of
structural links on the sound transmission of a lightweight double panel structure.
Journal of Sound and Vibration, 324:712 – 732, 2009.
62
Overview and Summary
[25] R. H. Lyon and R. G. DeJong. Theory and Application of Statistical Energy Analysis.
Butterwirth-Heinemann, 1995.
[34] E. Nilsson and A. C. Nilsson. Prediction and measurement of some dynamic prop-
erties of sandwich structures with honeycomb and foam cores. Journal of Sound
and Vibration, 251:409–430, 2001.
[35] A. Pellicier and N. Trompette. A review of analytical methods, based on the wave
approach, to compute partitions transmission loss. Applied Acoustics, 68:1192 –
1212, 2007.
[36] Allan D. Pierce. Acoustics : an introduction to its physical principles and applications.
McGraw-Hill Book Co., New York, 1981.
63
Sathish Kumar
[38] A. J. Price and M. J. Crocker. Sound transmission through double panels using
statistical energy analysis. Journal of the Acoustical Society of America, 47:683–693,
1970.
64