STUDY OF SIMULATED RAINFALL NOISE ON ROOFS AND GLAZINGS

PACS Reference : 43.40-r

C. Guigou-Carter1 ; M. Villot 1 and C. Horlaville2

CSTB
1
CSTB, 24 Rue Joseph Fourier, 38400 St Martin d’Hères, France
Tel : (33) 04 76 76 25 25
Fax : (33) 04 76 44 20 46
E-mail : [email protected] ; [email protected]
2
CSTB, LABE, Champs sur Marne, 77421 Marne-La-Vallée – Cedex 2, France
Tel : (33) 01 64 68 84 87
Fax : (33) 01 64 68 83 14
E-mail : [email protected]

ABSTRACT
In this paper, a theoretical model and a laboratory set-up for the evaluation of rainfall noise on complex
systems such as roofs and glazings are discussed. First, the impact force associated to a raindrop
size and a statistical model for the distribution of natural raindrop size allowing the determination of
the structural rainfall power spectrum injected to the test structure are presented. Several 3-
dimensional theoretical models describing the vibrational response and transmitted acoustic field for a
planar complex (multi-layered) system are then briefly developed. The laboratory set-up for measuring
rainfall noise is under construction; preliminary experimental and theoretical results will soon be
compared.

INTRODUCTION
In modern buildings, the use of light weight elements (generally made of steel sheets) or the use of
large skylight type glazings on roofs has become more extensive. The noise produced inside
buildings by rainfall impacting on these roof surfaces is raising significant concern. Indeed, in spaces
with such roof surfaces, rainfall increases the interior background noise levels, inducing a decrease of
speech intelligibility.
A new international standard prescribing a laboratory method for the measurement of sound generated
by rainfall on buildings elements using artificial raindrops under controlled conditions is being
discussed. Other mechanical simulation methods (impact hammer) have been investigated but have
not been able to adequately simulate real rain both in terms of sound levels and spectra generated.
The rain can be described in terms of rainfall rate, drop diameters and fall velocities. The characteristic
parameters for the artificial raindrops have been chosen to be 15 mm/h rainfall rate, 2 mm drop
diameter and 4 m/s fall velocity for an intense rain and up to 40 mm/h rainfall rate, 5 mm drop
diameter and 7 m/s fall velocity for a heavy rain. Artificial raindrops can be generated by hydraulic
spray nozzle generating water drops of uniform diameter.
In this context, the CSTB has decided to implement a laboratory set-up for artificial rainfall noise
measurement as well as to develop the associated analytical tools to help achieving design
improvement with respect to rainfall noise of roof elements.
In this paper, the impact force associated to a raindrop size and a statistical model for the distribution
of natural raindrop size allowing the determination of the structural rainfall power spectrum injected to
the test structure are presented. Three 3-dimensional theoretical models describing the vibrational
response and transmitted acoustic field for a planar complex (multi-layered) system are then
introduced and their results are compared. At the time this paper is being written, the laboratory set-
up for measuring rainfall noise is still under construction. Experimental results should be soon
available.

ANALYTICAL MODEL

Rainfall Excitation Spectrum

The impact force F(t) caused by a drop can be expressed by [1]
 3 Vt 
F( t ) = ρ π r 2 Vt2  1 − t  for 0 ≤ t ≤ 8r/3V t (1)
 8r 
where ρ is the water density, r the water drop radius, and Vt the drop terminal velocity.
The excitation force of natural rainfall can be evaluated by adopting a statistical model for the
distribution of natural rainfall proposed by [2]. This distribution N(D) in mm-1m-3 is
N( D) = 8000 exp{− 4.1R −0.21 D} (2)
where R is the rainfall rate in mm/h and D the water drop diameter in mm.
The terminal velocity of a raindrop (in m/s) is given by an empirical formula [3]
[ ]
Vt (D ) = 9 .58 {1 − exp − (D / 1.77 ) 1.147 } (3)
Consequently, the number n of raindrops having a diameter between D and D+δD impacting a unit area
(1 m2) of roof per unit time (1 s) is
n(D ) = N(D) δ D Vt (D ) (4)
The power spectrum of the excitation force associated with a raindrop of diameter D is then expressed
~ 2
Pexc (D, ω) = n(D ) F(D, ω) T0 (5)
~
where T0 is the reference time (T0=1 s), and F(D,ω) is the impact force spectrum, i.e., the Fourier
transform of the impact force given in Equation (1).
For artificial rainfall, all drops are assumed to have the same diameter Da and the same terminal
velocity Vta (usually obtained by measurement); the number na of drops impacting a unit area of roof
per unit time is then for a fall rate Ra
10 −3 R a 6
na = (6)
3600 π D a
For a flexible structure excited by rainfall, it is necessary to take into account the point impedance of
the structure Zstruc and the flow impedance of the water drop Zdrop. The impact force spectrum is then
changed to
~ Z struc ~
Fflex( D, ω) = F(D, ω) with Z flow = ρ πr 2 V(D ) (7)
Z struc + Z flow

Figure 1 shows the impact force spectrum for drop diameters from 0.5 mm to 5.0 mm. It can be
observed that with increasing raindrop diameter, the impact force spectrum level increases and the
frequency at which the impact force spectrum is maximum decreases. This is due to the fact that
larger sized raindrops have more mass, a higher terminal velocity and a longer impact duration.
Table 1 shows some typical characteristics of different natural and artificial (Intense 1 and Heavy 1)
rainfall types as given in [4]. Figure 2 presents in third octave band the power spectrum associated
with the natural as well as the artificial rainfalls described in Table 1 (for the natural rainfall, δD=0.1 mm
was considered). For natural rainfall, even if the number of drops per m3 decreases sharply with drop
diameter (see Equation (2)), the larger sized raindrops contributes the most to the power spectrum
because they are associated with higher impact force spectrum as explained previously. It can be
seen that the power spectra corresponding the artificial rainfall Intense 1 and Heavy 1 recommended
by [4] do not well represent those of their associated natural intense and heavy rainfall. The power
spectra for the artificial rainfall Intense 2 and Heavy 2 provide a better agreement with their associated
natural rainfall over the entire frequency range.

Table 1 : Characteristics of different rainfall types.

Rainfall type Rainfall rate (mm/h) Drop diameter (mm) Fall velocity (m/s)
Natural Moderate Up to 4 0.5 – 1.0 1–2
Natural Intense Up to 15 1.0 – 2.0 2–4
Natural Heavy Up to 40 2.0 – 5.0 5–7
Artificial Intense 1 15 2.0 4.0
Artificial Heavy 1 40 5.0 7.0
Artificial Intense 2 15 1.5 5.0
Artificial Heavy 2 25 3.5 7.5

-40

-50

-60

-70
Impact Force (dB)

-80

-90

-100

-110

-120

Natural D=0.5mm Natural D=1mm

-130 Natural D=2mm Natural D=3mm
Natural D=4mm Natural D=5mm
Artificial D=2mm Artificial D=5mm
-140
10

16

25

40

63

100

160

250

400

630

1000

1600

2500

4000

6300

10000

Frequency (Hz)
Figure 1: Impact force spectrum for different drop diameters.

Infinite Multi-layered Structure Model

The model for infinite multi-layered structures is based on a transfer matrix approach (see [5] for
example). The different layers of constant thickness, constituting the structure can be either solid,
fluid, or porous (following Biot’s theory) elements. Furthermore, the layers can be bounded or
unbounded with each others. The computer program CASC based on this approach is already used at
CSTB to predict sound transmission, sound absorption and impact noise, to obtain propagation
constants. It was then modified to include rainfall type excitation as described in the previous section
and to calculated the associated sound power radiation. It should also be noted that a spatial
windowing technique can be used to take into account the effect of the finite size structure as
presented in [6].

-20

-30

-40
Power Spectrum (dB)

-50

-60

-70

-80 Natural R=4mm/h Natural R=15mm/h

Natural R=40mm/h Artificial Intense 1
Artificial Heavy 1 Artificial Intense 2
Artificial Heavy 2
-90
10

16

25

40

63

100

160

250

400

630

1000

1600

2500

4000

6300

10000
Frequency (Hz)
Figure 2: Power spectrum for different rainfall types.

Modal Model
The modal model allows the consideration of a single plate, or two plates coupled by an air cavity. The
model is based on a superposition of a simply supported thin plate solution for the structural
component and duct type solutions with rigid boundary conditions for the coupling cavity if present.
Such a model is described in [7]. An average of 100 forces randomly distributed on the structure
surface was used for the prediction of sound power radiation under rainfall type excitation.

SEA Model
A SEA model for a single component was also implemented. The derivations follow those presented in
[8], allowing to write the radiated sound power as
ρ c
Π SEA = 0 0 σ Π in (8)
ms ωη
where ρ 0c 0 is the characteristic acoustic impedance of air, ms the structure density per unit area, η
the structure loss factor, σ the radiation efficiency (given by Leppington, see in [8]) and Πin the injected
power. For rainfall type excitation, the injected power can be expressed as

Π in = Pexc ( ω) Real {Ys } with Ys = 1 8 D s m s

1
2
( (9) )
where Ds is the structure bending stiffness and Pexc(ω) the power spectrum associated with the rainfall
(deduced from Equation (5)).
RESULTS
Results are presented for a single glazing (10 mm thick glass) and a double glazing (composed of 10
mm thick glass / 6 mm thick air / 4 mm thick glass). The characteristics of the glass component are
given in Table 2. The considered glazings have a surface area of 1.48x1.23 m2. The calculations
presented below are conducted for the artificial heavy rainfall noted Heavy 2 in Table1 (which is very
close in terms of the excitation power spectrum to the natural rainfall with a 40 mm/h rate as seen in
the previous section).

Table 2 : Structural characteristics of the glass component.

Density (kg/m3) Young Modulus (GPa) Poisson Coefficient Loss Factor (%)
Glass 2500 62 0.22 5

Single Glazing
The radiated power for the single glazing system is presented in Figure 3 for the different models
considered. First it should be noticed that for this system and for the rainfall excitation considered, the
maximum acoustic radiation occurs close to the critical frequency of the panel (1250 Hz) since, in that
frequency range, its radiation efficiency is important and the excitation spectrum close to its
maximum. As expected, above the critical frequency of the system the different models agree very
well since the radiation efficiency is close to 1. Below the critical frequency, the infinite size CASC
results show low radiated power levels as the radiation efficiency for infinite structure is quite low but
not null since supersonic waves exist in the structure due to the damping in the structure (see [6]).
The results from the SEA and modal approaches agree fairly well. Note that the high level of radiated
power at the critical frequency for the SEA modal comes from the evaluation of the radiation efficiency
with Leppington formula (this formula does not take into account the loss factor in the structure).
Finally, the radiated power obtained with a spatial filtering technique to take into account the finite
size of the structure (noted CASC – Finite Size in Figure 3) is over-evaluated below the critical
frequency compared to that obtained with modal and SEA models. The difference is about 3 dB
around 100 Hz.

50

45

40
Radiated Power (dB)

35

30

25
CASC - Infinite Size
20
Modal Model
SEA Model
15
CASC - Finite Size
10
1000

1250

1600

2000

2500

3150

4000

5000
50

63

80

100

125

160

200

250

315

400

500

630

800

Frequency (Hz)
 Figure 3: Sound radiated power for the single glazing.

Double Glazing
The radiated power for the double glazing system is presented in Figure 4. As expected, above the
critical frequency of the panels constituting the double glazing (1250 Hz) the different models
presented agree very well. In this case, the maximum sound radiation occurs in the frequency range
from the system resonance frequency (associated with the air cavity around 315 Hz) to the first panel
critical frequency. The spatial filtering technique applied to the infinite multi-layered structure model to
take into account the finite size of the structure is associated to an increase of radiated power
compared to the infinite structure case only below the cavity resonance frequency. However, the
predicted radiated power between the resonance frequency and the critical frequency is unchanged by
the use of this technique. Therefore, it is under-evaluated compared to the results obtained with the
modal model in that frequency range. More work is needed regarding the spatial filtering technique in
the case of structural excitation with the presence of a cavity coupling structural elements.

45

40

35
Radiated Power (dB)

30

25

20

15

10 CASC - Infinite Size

5 Modal Model
0 CASC - Finite Size

-5

-10
1000

1250

1600

2000

2500

3150

4000

5000
50

63

80

100

125

160

200

250

315

400

500

630

800

Frequency (Hz)
Figure 4: Sound radiated power for the double glazing.

REFERENCES
[1] H. Suga and H. Tachibana, Sound radiation characteristics of lightweight roof constructions
excited by rain, Journal of Building Acoustics, 1(4), pp. 249-270, 1994.
[2] J.S. Marshall and W.M. Palmer, The distribution of raindrops with size, Journal of Meteorology,
Vol. 5, pp. 165-166, 1948.
[3] A.C. Best, Empirical formulae for the terminal velocity of water drops falling through the
atmosphere, Quarterly Journal of the Royal Meteorological Society, Vol. 76, pp. 302-311, 1950.
[4] Draft version 4 of ISO/TC 43/SC 2/WG 18 N 198, Laboratory measurement of sound generated by
rainfall in buildings and of buildings elements, April 2002.
[5] M.L. Munjal,, Response of a multi-layered infinite plate to an oblique plane wave by means of
transfer matrices, Journal of Sound and Vibration, 162, 333-343,1993.
[6] M. Villot, C. Guigou and L. Gagliardini, Predicting the acoustical radiation of finite size multi-
layered structures by applying spatial windowing on infinite structures, Journal of Sound and Vibration,
245(3), 433-455, 2001.
[7] L. Gagliardini, Simulation numérique de la transmission acoustique par les parois simples et
multiples, Ph.D. Dissertation, 1991.
[8] R. J. M. Craik, Sound Transmission through Buildings using Statistical Energy Analysis, Gower,
1996.