Composites Science and Technology 59 (1999) 1735±1743

Acoustic emission wave propagation in a viscoelastic plate

M. Giordano a, L. Condelli b, L. Nicolais b,*
a
Institute for the Composite Materials Technology, National Research Council, P.le Tecchio 80125 Naples, Italy
b
Department of Materials and Production Engineering, University of Naples ``Federico II'', P.le Tecchio 80125 Naples, Italy

Received 13 October 1998; received in revised form 20 January 1999; accepted 5 February 1999

Abstract
In loaded materials the strain-energy release due to microstructural changes results in stress-wave propagation. Acoustic emission
deals with the recognition of such waves over the material surfaces and for this reason the application of acoustic emission techni-
ques to the monitoring of the mechanical integrity of materials and structures is appropriate. Nevertheless, the use of such tech-
niques in the case of polymer-based composite materials needs a clearer understanding of the relationships between the recorded
signal, the damage process and the structure geometry. In this work a model is proposed of the propagation in a viscoelastic plate of
waves due to a damage event. The microfailure event has been represented as a point source and ray theory has been applied to
develop the physical model of wave propagation. The spectral analysis technique has been applied to the solution of the wave
equations. Model calculations in the frequency domain have been compared with experimental spectra resulting from ®ber break-
age in di€erent ®ber/matrix systems from single-®ber fragmentation tests. # 1999 Elsevier Science Ltd. All rights reserved.
Keywords: A. Polymer-matrix composites; C. Damage mechanics; D. Acoustic emission; Wave propagation

1. Introduction mechanism, i.e. the failure mechanism, or the wave

propagation through a ®nite medium. A theoretical
Several types of microfailure event occur when poly- model needs to elucidate the in¯uence of the failure
mer composite structures are loaded. These include ®ber dynamics and the propagation medium upon the
pull-out, ®ber breakage, matrix cracking and interfacial received acoustic signal features.
debonding. The damage process leads to a degradation Several attempts have been made in AE process
of the mechanical properties and to catastrophic failure modeling. Starting from the models conceived for the
of the whole structure. Acoustic emission is the only earthquakes, Otsu and Ono [1] developed a generalized
`nondestructive' technique able to monitor the damage theory of AE by using the integral formulation of the
growth and accumulation continuously and, con- elastodynamics. A key role in their formulation is
currently, to provide information on the nature of the played by a Green function of the second kind in a half
damage. space. They have shown that some Green functions in
The acoustic emission (AE) process deals with: a the half space are applicable to the problem of a plate:
microfailure event (source) generating transient waves, that is the most realistic case. The same authors [2] have
the propagation of such waves in a ®nite medium and developed a source-representation model investigating
their detection at the transducer position. In particular, tensile and shear cracks. Suzuki et al. [3] extended the
a microfailure event within a loaded material induces a analysis to a dissipative medium including a relaxation
sudden redistribution of the strain energy around the function. Simulations were compared with experimental
damaged area, generating transient dynamical waves. results from few-®bers model composites.
Waves propagate through the medium to the surfaces, By contrast, the generalized ray theory [4] has been
where AE is detected by a transducer. The recorded used to model AE in a elastic plate [5±7]. The wave
signals bring information about either the source motion is decomposed into disturbances travelling
along a multitude of ray-paths. Laplace transforms were
* Corresponding author. Tel.:+39-81-768-2400; fax:+39-81-768-
used in developing solutions for several types of sources
2404. placed in any location of the plate. No viscoelastic
E-mail address: [email protected] (L. Nicolais) e€ects have been accounted for. Moreover, transient

0266-3538/99/$ - see front matter # 1999 Elsevier Science Ltd. All rights reserved.
PII: S0266 -3 538(99)00035 -4
1736 M. Giordano et al. / Composites Science and Technology 59 (1999) 1735±1743

Nomenclature general model of the AE process in a viscoelastic plate is

stated in Section 2; Section 3 deals with the application
a attenuation (mÿ1) of the physical model drawn in Section 2; the model
c phase velocity (m/s) testing through comparison with experimental results
C source spectrumÐdisplacement (m s) and the main remarks are in Section 4.
C" source spectrumÐstrain (s)
ce elastic velocity (m/s)
E tensile modulus (Pa) 2. Model statement
f frequency (sÿ1)
h plate half-thickness (m) 2.1. Wave equation and spectral solution
Hg geometry transfer function
Hm material transfer function Wave motion in a homogeneous medium is described
Ht transducer transfer function by the momentum balance equation, in the case of neg-
i imaginary unit ligible Lagrangian inertia that is:
k wave number (mÿ1)
kI imaginary part of wave number (mÿ1) @2 s
 ˆ r
 1†
kR real part of wave number (mÿ1) @t2
L source-transducer horizontal distance (m)
P power per unit volume (W/m3) where s is the displacement and  is the stress tensor.
Rn distance covered by the nth ray (m) The closure equation needed to solve the problem
s displacement (m) must be provided by the constitutive equation of the
t time (s) medium. It relates the stress  with the strain history
T experimental signal duration (s) experienced by the material.
Te emission time (s) In the plane wave case, the most general wave motion
U strain energy per unit volume (J/m3) equation for a linearly viscoelastic material is:
x space variable (m)
n impact angle of the nth ray @s @s @2 s @2 s @2 s
s‡a ‡b ‡c 2‡d ‡ e 2 ‡


ˆ 0 2†
" strain @x @t @x @x@t @t
"0 strain before cracking
l attenuation length (m) The spectral analysis technique [10] can be used to
 viscosity (Pa s) evaluate the general solution of Eq. (2), by taking the
 density (kg/m3) Fourier Transform:
 stress (Pa)
 relaxation time (s) d^s d2 s^
A1 !†^s ‡ A2 !† ‡ A3 !† 2 ‡


ˆ 0 3†
! angular frequency (rad/s) dx dx
à Fourier transform symbol (s)
Ä dimensionless variable symbol Equations of this kind have solutions of the form eÿikx
and this leads to the characteristic algebraic equation

A1 ‡ ÿik†A2 ‡ ÿik†2 A3 ‡


ˆ 0 4†
ultrasonic waves in a viscoelastic plate have been stu-
died by Weaver et al. [8,9]. Double transform technique Eq. (4) de®nes the wavenumber k and its relationship
has been used to solve wave equation. A closed form in with frequency ! is known as the spectrum relation.
the frequency domain for the high-frequency response Each value of k that satis®es Eq. (4) represents a mode
to a step load has been obtained. of propagation of the wave and the general solution, in
In this work a model of the AE process in a ®nite the frequency domain, is given by the superposition of
viscoelastic medium has been developed. The strain- the di€erent modes.
energy release at the failure event has been analyzed to X
recover the source model. The ray theory has been s^ x; !† ˆ Cm !†eÿikm !†x 5†
applied to model the propagation problem in a linearly m
viscoelastic plate with reference to a Maxwellian med-
ium. As spectral analysis technique has been used in The time, space solution s x; t† can be obtained by tak-
solving the wave equation. Model results have been ing the inverse transform Eq. (5)
compared with experimental data from single-®ber frag-
mentation tests, taking into account the transducer trans- s x; t† ˆ s^ x; !†ei!t d! 6†
fer function. The paper has been organized as follows: a
 M. Giordano et al. / Composites Science and Technology 59 (1999) 1735±1743 1737

The same result is achieved by using the Fourier series !  !0 , and by `low frequency' !  !00 , with a wide
representation of s x; t† range of frequencies which are comparable with some of
X the characteristic ones.
s x; t† ˆ s^ x; !n †ei!n t 7† We assume that the viscoelastic behavior of the med-
ium is represented by the Maxwell model accordingly
where s^ x; !n † are the Fourier coecients, that can be with experimental evidence of an elastic behavior at
evaluated from Eq. (5) at ! ˆ !n . high frequencies:
Finally
XX d d" 
ˆE ÿ 13†
s x; t† ˆ Cm !n †ei!n t eÿikm !n †x 8† dt dt 

@s
The coecients Cm !n † must be evaluated by imposing where " ˆ @x .
the appropriate boundary conditions. The wave equation [Eq. (1)] becomes the well known
telegraph equation:
2.2. Linear viscoelastic media
@2 s 1 @s E @2 s
‡ ÿ ˆ0 14†
Analysis of the wave propagation has been restricted @t2  @t  @x2
to linear viscoelastic media: in this case the spectrum
relationship k ˆ k !† is derived from Eq. (4) once the and the wave number k results in
constitutive equation has been assigned. s
 
The wave number, k, de®nes the phase velocity, c,  i 1 p
and the attenuation, a, of a propagating wave. k !† ˆ ! !ÿ ˆ ! ! ÿ i† 15†
E  ce 
!
k !† ˆ kR !† ‡ ikI !† ˆ ‡ ia !† 9† The phase speed depends on the frequency leading to a
c !† dispersive medium. In fact di€erent frequency compo-
In the case of an elastic solid theqwave number is real nents of the wave travel at di€erent speeds changing

wave shape as it propagates. The Maxwell constitutive
and the wave phase velocity c ˆ E is independent on
equation shows two limiting behaviors:
the frequency. No dissipation occurs in an elastic solid;
as a consequence the wave frequency content and . ! ! 0 viscous behavior
amplitude do not change during propagation. Both phase velocity c and attenuation a increase with
In the case of a Newtonian ¯uid both attenuation and the square root of frequency.
wave speed change with frequency [11];
s . ! ! 1 elastic behavior
r
! 2! Phase velocity c is not dependent on frequency while
jaj ˆ cˆ 10†
2  attenuation attains a constant value becoming negligible
as relaxation time  increases. Fig. 1 shows the depen-
Energy is dissipated and the wave is smoothed out dence of the wave number on frequency.
changing its frequency content during propagation.
As we move to viscoelastic materials, for which there 2.3. Source model
are at least three parameters, E;  and , one can de®ne
an elastic velocity ce and an intrinsic frequency ! (or The boundary conditions must be speci®ed at the
relaxation time ): damage localization within the material. During failure,
p the strain energy stored in the material is dissipated in
ce ˆ E= 11†
the creation of new surfaces, in heat generation and in
the generation of stress waves propagating through the
! ˆ E= ˆ 1= 12†
medium. However, we assumed a point defect, neglecting
The existence of an intrinsic frequency characterizes also the heat generation.
viscoelastic materials; for such materials (but not for Let us consider the strain-energy release caused by the
elastic solids or Newtonian ¯uids) there is an objective failure, that is:
criterion for distinguishing between high frequencies,
!  ! , and low frequencies, !  ! . Of course, a real 1

U ˆ ÿ E"2o 16†
material is endowed with a spectrum of relaxation 2
times, spanning characteristic frequencies from some
minimum frequency !0 to some maximum one !00 ; hence where "0 is the strain before cracking that is fully
perhaps by `high frequency' one should understand released at the end of failure process.
1738 M. Giordano et al. / Composites Science and Technology 59 (1999) 1735±1743

We assume that the failure process is characterized by of longitudinal waves on the boundary surfaces has
an emission time Te and all the stored strain energy been assumed. Fig. 3 shows the plate geometry
1 2
2 E"o is released during the failure. The model aims to evaluate the control point P L; h†
The accurate dynamics of damage process is descri- displacement in response of a strain longitudinal wave
bed by the power P as a function of time. In fact, the generated in the source point P(0,0). It is based on the
failure process brings the material from an equilibrium consideration that the signal detected at the control
state (before the damage) to a new equilibrium state point P L; h† is the superposition of the di€erent rays
(damaged), when the released power is exactly zero. All reaching this position after their re¯ection at the
the information regarding the source mechanism are boundary surfaces.
collected in the expression of the power P during the The free surface boundary condition
emission time Te (Fig. 2). Considering no dissipative
mechanism acting during failure: yy ˆ 0 wy ˆ 0 19†

"
1 ÿ
t
leads to the generation of two re¯ected waves at each
E "d" ˆ E "2 ÿ "20 ˆ P t†dt 17†
"o 2 0 ray impact point. A wave is re¯ected in a wave of the
same type and with the same angle plus a wave of dif-
The strain history at failure location can be evaluated ferent type with an angle given by the Snell law [12]. For
from the released power. Eq. (17) results: our analysis the secondary waves have been neglected
s considering their slowness. Fig. 4 shows the picture of
t this mechanism for the primary waves.
" t† P t†
ˆ 1 ‡ 1 2 dt 18† It can be easily shown that the total distance covered
"0 0 2 E"0
by the nth ray is:
q
2.4. Wave propagation in a plate Rn ˆ h2 2n ÿ 1†2 ‡ L2 20†

In this section a model of the wave propagation based In addition it must be considered that the nth ray
on the ray theory has been developed. Acoustical rays impacts the control point with an angle n
have been described as plane waves and total re¯ection
L
n ˆ arctan 21†
2n ÿ 1†h

where n is the angle between the nth ray and the y axis.
Total re¯ection hypothesis allows to evaluate the con-
trol point response as the superposition of rays travelling
distances Rn in a pseudo-in®nite medium. The y compo-
nent of the displacement in the frequency domain is:
X
s^y RL;h ; !† ˆ s^n Rn ; !† cos n 22†
n

Fig. 1. Dimensionless phase speed co ˆ c=ce (solid line) and attenua- 3. Model calculations
tion koI ˆ kI ce  (dotted line) as functions of frequency.
In general terms, the relation between the frequency
spectrum of the signal detected by the transducer and
the source spectrum can be expressed [13] as:

Fig. 2. Strain energy release rate due to a failure event. Fig. 3. Plate geometry.
 M. Giordano et al. / Composites Science and Technology 59 (1999) 1735±1743 1739

Fig. 4. Re¯ection of the rays on the boundary surfaces.

s !† ˆ C !†Hm !†Hg !†Ht !† 23†

where Hm !† is the transfer function of the material,
Hg !† is the transfer function of the specimen geo-
metry, Ht !† is the frequency response of the transducer
and C !† is the spectrum of the source event. It is worth
noting that each term in Eq. (23) can be, in principle,
separately evaluated.

3.1. Plane wave propagation in an in®nite Maxwell

medium

The plane wave propagation in an in®nite medium is

described by Eq. (5), where the wave number is speci®ed
by the Maxwell constitutive equation [Eq. (13)]. In this
case an unique mode of propagation applies.
The source spectrum must be speci®ed. According to Fig. 5. Dimensionless power and normalized strain versus dimension-
less time.
the consideration in Section 2.3, the power release P t†
during the emission time has been assumed to be:
  function of the dimensionless frequency !~ ˆ !Te . It is
E"2  t worth noting that the phase spectrum of the source has a
jP t†j ˆ 0 sin  24†
2Te 2 Te smooth behavior at low frequency !Te  1. On the con-
trary, a severe fragmentation of the curve, identi®ed by
From Eq. (18) we evaluate the strain at the source point: sudden changes of the phase values versus frequency,
  characterizes the phase spectrum for frequencies !Te  1.
" t†  t Strain at x position within the in®nite medium is:
ˆ cos 25†
"0 2 Te
"^ !; x; Te ; ; ce † ˆ C" !; Te †eÿik !;;ce †x 27†
Fig. 5 shows dimensionless power and normalized
strain. In terms of dimensionless variables:
The source spectrum in terms of strain is:
~ ˆ Te !~ ˆ !Te
ÿi!Te 28†
C" !; Te † ˆ ÿ2"0 Te
e ‡ 2i!Te
26† x~ ˆ x
l k~ ˆ kl
4! Te ÿ 2
2 2

where the dimensionless number ^ is the Deborah num-

Fig. 6 shows the dimensionless amplitude C~ " !† ~
~ ˆ C"0 T!†e
ber, i.e. the ratio between the characteristic relaxation
~ †
Im C !†
and phase  ˆ arctan Re C !†
~ † spectra of the source as a time of the viscoelastic medium  and the characteristic
1740 M. Giordano et al. / Composites Science and Technology 59 (1999) 1735±1743

Fig. 6. Dimensionless amplitude and phase spectra of the source.

time of the damage phenomenon, the emission time Te

and l ˆ ce  is the attenuation length.
Eq. (27) becomes:

k~ !~ †
"^~ !; ~ ˆ C~ e !†
~ †
~ x; ~ ÿi
e
~ x~
ˆ C~ e !†H ~ x†
~ m !~ ; ~ 29†

Fig. 7 shows the dimensionless amplitude spectra for

di€erent Deborah number, di€erent dimensionless dis-
tances as a function of the dimensionless frequency !. ~
The dimensionless amplitude spectrum of the source is
reported for comparison.
Fig. 7a shows the case of ^  1. The viscous behavior is
characterized by the lowering of the high frequency com-
ponents, the energy content of the wave is continuously Fig. 7. Dimensionless strain amplitude spectra for di€erent dimen-
sionless distances within an in®nite medium at di€erent Deborah
dissipated as the wave propagates through the medium.
numbers.
In the case of ^  1 (Fig. 7b) the dimensionless
amplitude spectrum does not change with propagation,
no dissipation mechanism acts. Fig. 7c shows the inter- h~ ˆ Lh plate shape factor. Dimensionless spectrum equa-
mediate case ^  1 where viscoelasticity allows a reduc- tion at control point position is:
tion of the amplitude of the spectrum but the spectrum   X  
C~ !~ †eÿik !~ ~†Rn L;h† cos n h~
~ ~ ~ ~
shape is almost retained. s~^y !; ~ ~ ˆ
~ h;
~ L;

n
 30†
3.2. Wave propagation in a Maxwellian plate ˆ C~ !~ †Hmg !~ ; ~ h~
~ L;
The plane wave propagation in a plate is described by
Eq. (22). A new dimensionless variable must be introduced: or equivalently in terms of strain:
 M. Giordano et al. / Composites Science and Technology 59 (1999) 1735±1743 1741

"^~ !;
~ L; ~ †
~ h; ~ ˆ C~ " !†H ~ L;
~ mg !~ ; ~
~ h† 31† Deborah numbers , ^ as a function of dimensionless fre-
quency !~ for di€erent plate sizes L. ~
where L~ ˆ Ll . In the viscous case, ^  1, (Fig. 8a) the dimensionless
A ®rst consideration is that it is not possible to sepa- strain amplitude spectrum is smooth at low frequency.
rate the e€ect of geometry from the e€ect of the mate- It fragments for frequencies !Te 51 showing sudden
rial on the wave propagation. The transfer functions changes of the amplitude values versus frequency. The
Hm !† and Hg !† from Eq. (23) are grouped into the fragmentation of the amplitude spectrum spreads
transfer function Hmg !†. toward lower frequencies as the plate size increases.
Fig. 8 shows the dimensionless strain amplitude spec- In the elastic case, ^  1, (Fig. 8b) the main e€ect of
tra for a ®xed value of the shape factor h, ~ for di€erent the boundary re¯ection is the fragmentation of
the dimensionless strain amplitude spectrum, both in
low and high frequency zone. The viscoelastic case,
^  1; (Fig. 8c) shows characteristics of both the limit-
ing cases.
A clear explanation of the sudden changes of the
amplitude values can be provided. The di€erent arrival
times of each ray induces phase modi®cation in fre-
~ ~ R~ n L; ~
~ h†
quency domain according with the term eÿik !~ † of
the transfer function. So the summation over rays cor-
responds to adding components with di€erent phase.
This interference phenomenon produces the fragmenta-
tion of the amplitude spectra. The more severe frag-
mentation of the high frequency components can be
explained with reference to the phase spectrum of the
source. In this region the phase spectrum is very irre-
gular, as a consequence a time shift produces larger
e€ect on the amplitude evaluation. On the other hand in
the viscous region of the spectrum (!~  1) the inter-
ference is reduced due to the slowness of the waves.

4. Results and discussion

4.1. Model testing

The model has been tested by comparison with

experimental data. Acoustic emission signals due to
®ber failure propagating in a plate have been obtained
from single-®ber fragmentation tests for three di€erent
®ber-matrix systems: polyester-matrix/carbon-®ber (6
mm diameter) (PCA), epoxy-matrix/carbon-®ber (6 mm
diameter) (ECA) and epoxy-matrix/carbon-®ber (12 mm
diameter) (ECB). Fig. 9 shows a scheme of the test. A
two transducers (resonant-wide band) based system has
been used to identify and record the AE event. The
resonant probe has been used as a trigger for ®ber-fail-
ure identi®cation, while the AE signal has been recorded
by the wide-band probe. The apparatus allows the
source localization along the ®ber through the di€erence
of the signal arrival times at the two transducers. More
detailed information on the experimental set-up can be
found in a previous paper [14].
Fig. 10 shows representative signals for (PCA),
Fig. 8. Dimensionless strain amplitude spectra for ®xed shape factor (ECA) and (ECB) systems. Each spectrum spans from
~
of the plate h=0.033 for di€erent Deborah numbers as a function of 10 to 1000 kHz because of the acquisition window of the
the plate size. wide band transducer. The spectra were obtained by
1742 M. Giordano et al. / Composites Science and Technology 59 (1999) 1735±1743

Fig. 9. Scheme of the ®ber fragmentation test.

Table 1
Experimental data from Single Fiber Fragmentation Tests for di€erent
matrix/®ber systemsa

f (kHz) ce (m/s) L (mm) h (mm)

PCA 3 2300 6.5 1

ECA 12 1960 35 1
ECB 6 1960 18 1
a
PCA, polyester/carbon, 6 mm; ECA, epoxy/carbon, 6 mm; ECB,
epoxy/carbon, 12 mm.

sampling
f ˆ 1=T, the phase velocity ce as measured,

the source position L and the half-thickness of the plate
h for each test.
On the other hand, the strain amplitude spectrum
j"^ !†j of the signal recorded at the transducer position is
the result of the source event C" !†, the propagation of
the signal through the specimen Hmg !† and the transfer
function of the transducer Ht !†:

j"^ !†j ˆ jC" !†Hmg !†Ht !†j 32†

The synthesis of the signal as recorded at the transducer

position needs the knowledge of the transfer function of
the transducer Ht !† itself. Fig. 11 shows the transducer
transfer function as provided by the manufacturer.
Theoretical predictions have been performed by using
the spectrum equation [Eq. (32)] in the dimensionless
form. In principle, the ®tting parameters are the source
emission time Te and the relaxation time  for each
®ber/matrix system, but we assumed the same source
emission time for PCA and ECA systems having the 6
mm diameter ®ber and the same matrix relaxation time
for ECA and ECB systems having the epoxy matrix.
Fig. 10a shows the comparison between PCA experi-
mental amplitude spectrum and the calculated one. In
this case both the relaxation time  and the emission
time Te have been used as the ®tting parameters. Fig.
Fig. 10. Experimental (arbitrary units) versus calculated dimension- 10b shows the comparison between ECA experimental
less strain amplitude spectra for the di€erent matrix/®ber systems. amplitude spectrum and the calculated one. In this case
the emission time Te has been assumed the same of the
discrete Fourier Transform (via Fast Fourier Trans- PCA case, due to the same ®ber used, the relaxation
form) of the time-domain recorded signals. The dis- time  was the only ®tting parameter. Fig. 10c shows the
cretization of the frequencies
! ˆ 2 T is set by the comparison between ECB experimental amplitude
duration T of the recorded signal [10]. Di€erent signals spectrum and the calculated one. In the last case the
have di€erent durations. Table 1 reports the frequency relaxation time  was the same of the ECA case, due to
 M. Giordano et al. / Composites Science and Technology 59 (1999) 1735±1743 1743

. The Maxwell model accounts for the propagation

medium viscoelasticity, but the whole formulation
holds for a generic linearly viscoelastic medium.

In spite of those assumptions, the model seems to

predict amplitude spectra whose shapes are very close to
experimental signals. The ®tting parameters in wave
synthesis are the emission time Te describing the failure
event and the relaxation time of the polymer  account-
Fig. 11. Transfer function of the WD transducer as provided by the
ing for the matrix viscoelasticity. As well, the estimated
manufacturer, PAC. values of these parameters have physical coherence. The
relaxation time  for the polyester resin is of the same
order (=100 ms) of that in Felix et al. [15]. Moreover,
Table 2 the ®ber having the biggest size (12 mm) shows the
Model parameters used in simulations greatest emission time as expected. The calculated
Deborah numbers suggest that viscoelasticity plays a
Te (ms)  (ms) ~ L~ h~
key role in the acoustic emission signals due to the ®ber
PCA 1.0 108 108 2.8 0.15 breakage in the structure we analyzed. In conclusion, a
ECA 1.0 80 80 17.9 0.03 simple tool to investigate both the failure dynamic and
ECB 1.3 80 61.5 7.1 0.06
the role of viscoelasticity in waves propagation through
a plate has been developed.

References
the same propagation medium, while the emission time
[1] Ohtsu M, Onu K. A generalized theory of acoustic emission and
Te is the ®tting parameter. Table 2 summarizes the
Green's functions in a half space. J Acoustic Emission 1984;3:27±40.
estimated parameters together with the values of the [2] Ohtsu M, Onu K. A generalized theory and source representa-
~ L;
dimensionless variables ; ~
~ h. tions of acoustic emission. J Acoustic Emission 1986;5:124±33.
[3] Suzuki H, Takemoto M, Onu K. The fracture dynamics in a dis-
4.2. Conclusions sipative glass-®ber/epoxy model composite with AE source simu-
lation analysis. J Acoustic Emission 1996;14:35±50.
[4] Cagniard L. Re¯ection and refraction of progressive seismic
The proposed model is able to reproduce the main waves. New York: McGraw±Hill, 1962.
features of the acoustic emission signals from single ®ber [5] Ceranoglu AN, Pao YH. Propagation of elastic pulses and
fragmentation tests varying the ®ber size and the propa- acoustic emission in a plate Part 1: theory. J Appl Mech
gation medium. In the following the main assumptions 1981;48:125±32.
[6] Ceranoglu AN, Pao YH. Propagation of elastic pulses and
of the model have been recalled and criticized. acoustic emission in a plate Part 2: epicentral responses. J Appl
. The acoustic emission point source has been mod- Mech 1981;48:133±8.
[7] Ceranoglu AN, Pao YH. Propagation of elastic pulses and
eled by means of simple considerations on the
acoustic emission in a plate. Part 3: general responses. J Appl
strain energy release due to the failure event. At Mech 1981;48:139±47.
this elementary stage, the source is characterized [8] Weaver RL, Sachse W, Niu L. Transient ultrasonic waves in a
by a single parameter, the source emission time Te . viscoelastic plate: theory. J Acoust Soc Am 1989;85:2255±61.
In the real cases failure involves an extended [9] Weaver RL, Sachse W, Niu L. Transient ultrasonic waves in a
viscoelastic plate: applications to materials characterization. J
damaged area. So, a more precise source model
Acoust Soc Am 1989;85:2262±7.
should consider a deeper analysis of the failure [10] Doyle JF. Wave propagation in structures. New York: Springer,
micromechanics. 1997.
. Wave propagation in a plate has been analyzed [11] Landau LD, Lifshitz EM. Fluid mechanics. Oxford: Pergamon
with reference to the ray theory. Response at the Press, 1987.
[12] Kolsky H. Stress waves in solids. New York: Dover, 1963.
transducer position is due to the superposition of [13] Berthelot JM, Souda MB, Robert JL. Frequency analysis of
rays re¯ected on the plate surfaces. Rays have acoustic emission signals in concrete. J Acoustic Emission
been described as plane waves but, in principle, the 1993;11:11±18.
same formulation can be extended to spherical [14] Giordano M, Calabro A, Esposito C, D'Amore A, Nicolais L.
waves. In addition the secondary waves has been An acoustic emission characterization of the failure modes in
polymer-composite materials. Comp Sci Tech 1999;58:1923±8.
neglected according to their slowness: the last [15] Felix MP. Attenuation and dispersion characteristics of various
hypothesis released leads only to a cumbersome plastics in the frequency range 1±10 MHz. J Comp Mater
problem. 1974;8:273±87.