Leaky Lamb waves in fibrous composite

laminates
Cite as: Journal of Applied Physics 58, 4531 (1985); https://doi.org/10.1063/1.336268
Submitted: 09 May 1985 . Accepted: 03 September 1985 . Published Online: 04 June 1998

D. E. Chimenti, and Adnan H. Nayfeh

ARTICLES YOU MAY BE INTERESTED IN

Leaky Lamb waves in an anisotropic plate. I: An exact solution and experiments

The Journal of the Acoustical Society of America 85, 2268 (1989); https://
doi.org/10.1121/1.397772

The interaction of Lamb waves with delaminations in composite laminates

The Journal of the Acoustical Society of America 94, 2240 (1993); https://
doi.org/10.1121/1.407495

Leaky Lamb waves in an anisotropic plate. II: Nondestructive evaluation of matrix cracks in
fiber-reinforced composites
The Journal of the Acoustical Society of America 89, 1590 (1991); https://
doi.org/10.1121/1.401017

Journal of Applied Physics 58, 4531 (1985); https://doi.org/10.1063/1.336268 58, 4531

© 1985 American Institute of Physics.

leaky lamb waves in fibrous composite iaminates
D. E. Chimenti
Materials Laboratory, Air Force Wright Aeronautical Laboratories, Wright-Patterson AFB, Ohio 45433
Adnan H. Nayfeh
Depanment 0/Aerospace Engineering and Engineering Mechanics, University o/Cincinnati, Cincinnati, Ohio
45221
(Received 9 May 1985; accepted for publication 3 September 1985)
Results of experimental measurements and theoretical calculations on ultrasonic leaky Lamb-
wave propagation in fiber-reinforced, unidirectional composite laminates are presented. With the
Lamb wave vector oriented parallel to the fiber direction, dispersion curves of phase velocity
versus frequency and plate thickness have been constructed from measurements of ultrasonic
reflection from fluid-loaded composite plates. The experimental results are supported by a
theoretical model of Lamb-wave propagation in the composite plate. The model begins with an
approximate calculation of the effective, homogeneous, transversely isotropic elastic behavior of a
unidirectional composite laminate in the long-wavelength limit, using a two-step procedure based
on alternating layered media. This intermediate continuum result is then incorporated into a
calculation of the ultrasonic reflection coefficient of a fluid-loaded anisotropic plate, which is
assumed to approximate the fibrous composite laminate. Good quantitative agreement with the
model is found if the fiber volume fraction is taken to be an adjustable parameter. However, not all
portions of the dispersion curves predicted by the model can be observed in the data. It is
conjectured that relative differences in mode coupling account for this discrepancy.

INTRODUCT~OIN momentum and constitutive-relation term is given. This the-

Elastic-wave propagation in fibrous composite materi- ory leads to simple wave-propagation equations which con-
als has been the subject of numerous investigations in recent tain the full influence of the microstructure.
years. However, the morphology of fiber-reinforced compo- Interest in the acoustic properties of these inhomogen-
sites can seriously complicate the calculation of their wave eous and generally anisotropic materials is heightened by
propagation properties. Typically, these materials are com- their increasing use in applications requiring high stiffness-
posed of brittle, high stiffness fibers embedded in a matrix of to-weight ratios. Ultrasonic nondestructive evaluation is one
more ductile material which bonds the fibers together and useful mean to insure the structural integrity of fibrous com-
acts as a load-transfer medium. Since it is clearly not practi- posites. In order to take fun advantage of the information
cal to attempt a solution of the completely general elastic- contained in an ultrasonic test, it is helpful to have a clear
wave problem, most prior \vork has employed various ap- understanding of wave-propagation characteristics in the
proximations to render the calculations tractable. The material under examination. Accordingly, the current study
simplest analytical descriptions of composites are the so- focuses on the elucidation of plate (Lamb) waves in fluid-
called "effective Inodu]us" theories, in which the approxi- loaded, unidirectional, fibrous composite materials, where
mate composite material constants are a geometrically the wave-propagation direction coincides with the fiber axis.
weighted average of the constituent properties. Such a the- Experimental results of numerous measurements on several
ory has been formulated by Postma 1 to define the gross me- samples of glass-epoxy composite plates are presented. In
chanical behavior of composites. However, the applicability addition, the ultrasonic reflection properties of a composite
of the effective modulus theories to dynamic loading is some- plate are derived, from whose behavior the existence of
what restricted. A model designed to account for dynamic Lamb waves can be inferred. Section I contains the theory,
effects must reflect the influences of the microstructure and while Sec. II gives the experimental details, Sec. HI reports
the anisotropy. the results of experiments and analysis, and conclusions are
One model satisfying these criteria is the "effective stiff- drawn in Sec. IV.
ness" theory proposed by Achenbach, Sun, and Herrmann. 2
Here, the real composite is transformed into a homogeneous i. THEORY
higher-order continuum with microstructure. Bedford and We consider the case of an acoustic wave incident on a
Stern3 and Lempriere4 have employed mixture theories as fiber-reinforced composite plate which is immersed in a flu-
models of composite e1astodynamics. Composite constitu- id. In practice, the plate consists of small, nearly parallel
ents are superimposed in space and allowed to undergo indi- fibers of circular cross section on an approximately hexagon-
vidual deformations. Other models based on an examination al array, as illustrated in Fig. l(a). A coordinate system is
of Floquet modes have been advanced by Nemat-Nasser. 5 chosen such that the x axis coincides with the average fiber
Our own approach6-9 to interacting continua offers an alter- direction and the z axis is normal to the plate-fluid boundar-
native procedure for modeling the response of composites, ies. We further assume that the plate has infinite extent in the
where in particular, a rational construction of the mixture y direction. The plan is to approximate the highly structured

4531 J. Appl. Phys. 58 (12), 15 December 1985 0021-8979/85/244531-08$02.40 © 1985 American Institute of PhysiCS 4531
 0000
(0)
0000
00000

(b) Matenal2
----~------~------

------~------~------ FIG. 2. Bilayered structure with relevant boundary conditions indicates an

intermediate step in the calculation.

the two-dimensional result we seek. First, the layered struc-

ture of Fig. 2 is analyzed, deriving from this parallel-stress
FIG. \. Schematic representation of a fibrous composite. Idealized actual model the properties of "compound" layer I in Fig. lib). In
composite is shown in (a); model composite upon which calculations are the second step, the final model for the fibrous composite is
based is in (bl. Layer 1 is a "compound" of fiber and matrix properties,
established by treating the composite of Fig. lIb) as consist-
whereas layer 2 includes only the matrix properties.
ing of the two-dimensional "compound" layer 1 stacked in
series with the (matrix) layer 2. This procedure extends our
composite by a continuum, retaining the appropriate elastic previous calculations6-9 dealing with the construction of
anisotropy, and to analyze the ultrasonic reflection from one-dimensional continuum models for composite materi-
such a plate to study the behavior of Lamb waves in these als.
materials.
A. Layered mode~
To achieve this end we must solve the field equations in
the plate and in both upper and lower portions of the fluid, Consider the bilayered composite of Fig. 2, composed of
satisfying the mechanical continuity conditions at the inter- two types oflaminates in welded contact and stacked normal
faces. Because of the complicated microstructure of compo- to the y direction. The translational invariance of the prob-
sites, there exist continuity conditions on both macroscopic lem permits us to isolate a repeating unit, as indicted in Fig.
and microscopic scales. In the first instance, boundary con- 2. The relevant field equations are given by
ditions at the fluid-plate interfaces must be satisfied, while a(Iij ..
on the microscale there are additional, conditions at the fiber- --=pv., (1)
ar j J
matrix interfaces. Due to the number and nature of these
conditions it would be impractical to obtain exact solutions (Iik = ADi , Jv, + fL (aVj + aVk ),
of the field equations. Our alternative is to formulate an ap- ar, ark a0
proximate analysis in which the composite plate is replaced j, k, 1= 1,2,3, i = 1,3, (2)
by a homogeneous, transversely isotropic medium. In mak-
ing this approximation the microcontinuity conditions will which apply to both materials of the bilayer cell. In Eqs. (1)
be utilized. and (2) summation over repeated variable indices in the same
Since aU experimental measurements have been carried term is assumed. Here, Vi are the components of the dis-
out at frequencies such that the acoustic wavelength is much placement u, v, w; ri are the spatial coordinates x, y, z; and
larger than the fiber diameter, we can immediately avoid two (Ijk are the elements of the elastic stress tensor; the material.
theoretical complications. The first is the dispersion with density is p, and A and fL are the Lame constants. Applying
frequency of the ultrasonic wave speed, and the second con- the boundary conditions of Fig. 2 and averaging Eq. (1 ),j = 1
cerns the shape of the fibers. Square-cross-section fibers on a and Eq. (2), k = j = 1 across the cell thickness according to
regular square array, as shown in Fig. l(b), are postulated to
model the circular fibers of the real composite, preserving
(-)a '= h a- I fa ( )a dy
the correct fiber volume fraction. Because the corners of the yields
square fibers do not playa significant role in the acoustic
properties at long wavelength, this is a mild approximation. (3)
We now employ a building-block approach to arrive at

4532 J. Appl. Phys., Vol. 58, No. 12,15 December 1985 D. E. Chimenti and A. H. Nayfeh 4532
 Together with Eqs. (5)-(7) we now have a two-dimensional
description of the elastic behavior of the "compound" bi-
layered structure.
a= 1,2, (4)
where no = ha /(h I + h z ) are volume fractions of materials 1
and 2. The momentum interaction term is P and S is a consti-
tutive-equation interaction term, to be derived from the re- B. Fibrous model
maining boundary conditions. The analysis so far parallels Building on the above calculation we now construct the
that of our earlier work.6-9 effective elastic properties for a fibrous composite, consider-
Since we consider only long-wavelength excitations, the ing the "compound" layer 1 stacked in series with the matrix
instantaneous displacements change very little on the scale material of layer 2 normal to the z axis. Following the devel-
of the cell dimension. Therefore, we invoke the approxima- opment that led to Eqs. (6) and (7), we define new elastic
tion ur---+U2 = u and wI---+W Z = W as h----..G. Summing Eq. (3) constants appropriate to mixtures in parallel and series of
written for the two media, we have the properties of compound layer 1 and matrix layer 2. These
ao-x ao-xz __
relations take the form
---;;;- + a;- = Pc u, (5) (10)

where o-x, o-xz' and Pc are weighted sums of the respective for a series combination of E and P, or for E and # in parallel
quantities for materials I and 2, e.g., Pc = n l PI + n l Pz' ( )*=[nj( )c+nt/( h]-I, (11)
Note that the term P has been eliminated, a consequence of whereas
the assumption of a frequency-independent wave speed. By
the procedure outlined in our previous work6-9 we derive the ,1.* = (Ac njEc + ,1.2 nt/E2HnjEc + nt/E2)-I. (12)
following expression for the constitutive-equation interac- Here nc is the volume fraction of the compound layer in Fig.
tion term lIb), and nt is that of the matrix layer. Now a new set offield
S =~
G
[A I (auaxI + Owl
az
) _ Az (au 2 + Owz )].
ax az
(6)
equations can be constructed which are formally identical to
Eq. (1), but where instead ofEq. (2) we have
Substituting from Eq. (6) into Eq. (4) with a = 1 and a = 2 au 1 * aw
leads to a -E
- -+/1.-
x ax az'
(7a) az
-E*
-
aw ,1.* au
a;+ ax' (13)

where
a xx au+aW)
=/1* (- - .
Ec = n 1(A I + 2,utl + n 2{A z + 2,u2) - (AI - A1f/G,
,.. az ax
Ac = n 1 Al + n 2 A. z - (AI - A.1f/G, Consider the fibrous composite modeled above to be in
the form of a plate of thickness d and immersed in an infinite
with
fluid medium. The upper and lower surfaces of the plate are
G = (AI + 2,u I)ln I + (,1.2 + 2,u2)1n 2. situated at z = 0 and z = d, and an incident harmonic sound
:For shear waves we have wave propagating in the x-z plane insonifies the plate. The
incident wave vector kf makes an angle 8 with thez axis, and
(7b) all particle motion is assumed to be confined to the x-z plane.
The calculation of the plane-wave reflection coefficient from
where ,uc = n 1 ,u I + n 2#1' Propagation along the z axis is a fluid-loaded, transversely isotropic plate proceeds in a
found by permuting x and z. This procedure gives straightforward manner. Substitution of the constitutive re-
lations of Eq. (13) into the field equations yields coupled,
(8) second-order partial differential equations for the displace-
ments. Making the usual ansatz for the formal solutions
(9) leads to a system of linear simultaneous equations for the
displacement amplitudes. These equations are

01 O2 - 15 1 - 152 0 iaf I AI iaf

PI Pz PI P1 0 Pf A2 · -PI
QI Q2 -QI -Q2 0 0 A3 0
(14)
t5 l R I fJ 2 R2 - fJl/R I - fJ 1/R 2 -iaf 0 A4 0
PIR) P2 R z PJR I P2 /R 2 Pf 0 A5 0
Q)R) Q2R2 -Q)/R) - Q2/R2 0 0 A6 0

4533 J. Appl. Phys .. Vol. 58, No. 12, 15 December 1985 D. E. Chimenti and A. H. Nay1eh 4533
 5.7 4.0
TRANSMITTER RECEIVER

5.0 -~­ .~

4.3

36

2.9
FLUID
2.20~----:0:-'.2::-5----:0-:-5-:-0---""'0.-=-75:----~I.rro N LW
VOL FRACTiON
FIG. 5. Experimental geOmetry of Lamb-wave measurements in fibrous
FIG. 3. Principal wave speeds in unidirectional glass-epoxy composite as a composite plates. Transmitter is fixed with respect to sample, while receiver
function of fiber volume fraction, where Co and c~ are longitudinal speeds scans parallel to interface. Dashed lines show specular reflection. When
parallel and perpendicular to the fiber axis, and c2 is the common transverse Lamb wave is present, shaded region contains most of reflected energy. Null
speed. Particle motion is assumed to be confined to the :xz plane. zone is denoted by N, and leaky wave reflected field by L W. A similar radi-
ation pattern, not shown, exists below the plate.
where
are normalized wave speeds. The principal wave speeds cal-
~ = U *- E * aj Dj' ~i = iDj - aj'
culated for a sound wave whose particle motion is confined
Rj = exp( - a j td), j= 1,2, to the xz plane are collected in Fig. 3, as a function of the
(15) fiber volume fraction. Here, Co and c~ are the longitudinal
Pf = iiLf(a} - 1), af = (1 - k Jlt2)112, wave speeds parallel and perpendicular to the fiber axis, re-
spectively, while Cz is the common transverse wave speed.
kf = OJ/cf' c} = iLf/Pf'
This completes our analysis of ultrasonic reflection from a
Here, i = r-T, and the subscript/refers to the fluid. The fibrous composite plate.
circular frequency is OJ, t is the wavenumber of a propagat- As an illustration of this calculation Fig. 4 shows the
ing Lamb wave in the plate, and the Aj in Eq. (14) are dis- magnitude and phase of the plane-wave reflection coefficient
placement amplitUdes. Further required expressions are from a composite plate immersed in a fluid, where the mate-
aj = - [rl + (- l)i(y. - 4r2)1/2] t, rial parameters are appropriate to glass epoxy and water
with an incident angle of 20°. The sharp minima in the mag-
aj = [(k ~ - t 2) + c~o t a j ]I [it 2 a j (cfo - c~o) 1, (16)
nitude and rapid reversals in the phase of R indicate the
j= 1,2, presence of propagating modes, in this case Lamb waves, in
rl= [(k~ _tl)c~2 +(k~ _t2)C~0 the plate. 10,l1 From the angle of incidence of the ultrasonic
beam, the phase velocity of the Lamb wave can be inferred.
+ t 2(cio - c~0)(cf2 - 1)J/2c~0 202' We assume here that the Lamb wave vector equals the wave
(17) vector projection of the longitudinal wave from the fluid
r2=(k~ -t2)(k~ _t2)!c~0 202' onto the surface of the plate, cf!cL = sin 8. As discussed by
ko = OJ/Co, kl = OJ/c l , Schoch 10 this approximation is justified. for the conditions of
where the current study. It has been shown 11 that small errors are
possible in a narrow range offrequency times plate thickness
2o=Elp, ~=j.L*/p,
for the two lowest-order modes. Velocity dbpersion curves
cfo = (iL * + z,.t*)/E, ~o =/1-*/E, (1.8) can then be constructed by examining the reflection coeffi-
cfl =(iL*+z,.t*)I/1-*, C~2 =E*Ij.L*,
cient over a fi3,t;.ge of angles of incidence. Similar computa-
tions for isotropic or orthotropic plates either with or with-
1.0 - 100
out fluid loading have been carried out by others. 10-17
~ ILIEXPERiMENTAL TECHNIQUE
1\ I,
1\ " 60 An experimental study has been perfonned in conjunc-
I,
1\ I,
I' I tion with our theoretical investigations to test the range of
I I
I validity of the model calculation and its dependence on sig-
I I
I I nificant parameters. We have varied the plate thickness,
I I
\ I \ I composite fiber content, ultrasound frequency, and inci-
\ I \ I dence angle in the course of many measurements on five
\ I \ I
0.2 , I , I -60 different samples of unidirectional glass-epoxy composite
,I ,.
\'
plates. To speed acquisition and reduction of the data, the
0.~--~~~~--:~~~:--J~~~~:--~~--~·400
experiment and subsequent d.ata analysis are performed un-
0.0 08 1.6 2.4 3.2 4.0 4.8 5.6 64
FREQ (MHz)
der computer control.
Samples for this study are flat pl.ates approximately 10
FIG. 4. Ultrasonic plane-wave reflection coefficient from a composite plate
as calculated according to Eq. (14). Magnitude is solid curve; phase is plot- by 15 em, which have been fabricated from nonwoven
ted as dashed curve. Scotchply glass-reinforced plastic, type 1003, supplied by

4534 J. Appl. Phys., Vol. 58, No. 12,15 December 1985 D. E. Chimenti and A. H. Nayfeh 4534
 Sine = 18 0
10
/---. - - - 2.19 MHz
I \ - - - - - 1.99MHz

~
8 - - - 2.39MHz

6 ~
I

//-'
\
I I
I
I 'i
4 I \
I \
I , \

2 I
I
I
\\
'"'
Y
I
\~~
0
-180 -12.0 -6.0 0.0 12.0 18.0 240 30.0
X(mm)
5.6
FIG. 6. Profile of reflected beam at and near the Lamb angle for a 1. 14-mm
FREQ (MHz)
glass-epoxy plate in water. Solid curve at Lamb condition corresponds to
2.19 MHz. Chained and dashed curves are 200 kHz above and below this FIG. 7. Deconvolved reflection spectrum from a 1.35-mm composite plate.
frequency, respectively. Strong null occurs only when Lamb waves are gen- Incident angle is 24°. Deep minima indicate excitation ofLarnb waves in the
erated. plate. Transducer center frequency is 2.2 MHz.

the 3M Company. Plate thickness is determined by selecting

a suitable number of plies during the layup process. The re- Data are acquired in these experiments by exciting a
sulting structure is a unidirectional composite consisting of broadband ultrasonic transducer with low-level rf tone
long parallel glass fibers about 15 J1.m in diameter evenly bursts from a function generator. The 20-ps bursts propa-
distributed in a matrix of epoxy plastic. Because of uncon- gate through a water path long enough to place the sample in
trolled variations in the values of cure parameters during the transducer's far field at all frequencies. The reflected
production, the volume fraction of fiber can vary from sam- acoustic beam is detected by a second, matched transducer
ple to sample. In burnotf density measurements on coupons positioned at the same angle and distance to the sample. A
cut from our samples, fiber volume fractions as low as 49% piece ofO.2-mm tungsten foil is first placed over the sample
and as high as 59% were recorded. Elastic constants for the to establish the x-axis origin, by noting the reflection of a
two material constituents are collected in Ref. 18. swept-frequency tone burst. We have determined that this
A schematic of the experimenta] geometry is given in configuration effectively suppresses transducer side-lobe ef-
Fig. 5. An ultrasonic beam from a damped piston transducer fects. Removing the tungsten foil, we observe the ultrasonic
insonifies the plate at the selected incident angle. At frequen- reflection as a function of frequency in the composite plate,
cies corresponding to generation of Lamb waves in the plate, translating the receiver so as to optimize the ratio of on/off-
the reflected ultrasonic field will show the displacement and mode reflection. Then, data are accumulated by discretely
distortion characteristic of the presence of propagating stepping frequency and recording the reflected signal with a
waves in the plate, as illustrated schematically in Fig. 5. This broadband receiver and video detector, sampled by a gated
phenomenon, which is expected on the basis of the rapid- integrating amplifier. Directly thereafter, an amplitUde cali-
phase variation in the reflection coefficient, is analogous to bration is automatically performed to correct any nonlinear-
our earlier observations on leaky Rayleigh waves l9- 21 and ities in the receiver system. To insure accurate transducer
has been confirmed in isotropic plates by others. 13 Not illus- alignment and incident angle, precision optical rotators and
trated in Fig. 5 is the transmitt(~d field below the plate, which tilt tables have been integrated into the ultrasonic system.
according to Fig. 4- will in fact contain most of the energy. We estimate angular uncertainty to be 1 arc min and x-posi-
Examples of the reflected beam profiles at and near the tion error at 0.05 mm.
Lamb condition are shown in Fig. 6. Here we plot the ampli-
7.4
tude of the reflected field as a function of receiver position
scanned parallel to the surface of the plate, as indicated in
Fig. 5. At 2.19 MHz, the Lamb condition is satisfied, and the 6.3

reflected field in Fig. 6 displays the strong distortion and ]

spatial displacement mentioned earlier. Varying the fre- ~5.2
quency by 10% in either direction results in beam profiles ...J
w
>
more closely approximating the incident beam. A way from ~4.1
the resonance condition for efficient mode conversion from a «
:x:
0..
compressional wave in the fluid to a Lamb wave in the plate,
ultrasonic reflection should be nearly specuiar. We have tak- 3.0

en advantage of the nonspecular reflection at angles and fre-

quencies of strong mode conversion to enhance the detect-
2.5 5.0 7.5 10.0 12.5 15.0 17.5 20.0
ability of Lamb waves in the plate. By positioning our Fd (MHz-mm)
receiver at the strong minimum in the beam profile of Fig. 6, FIG. 8. Velocity dispersion curves for an aluminum alloy as a function of
the finite-beam distortion effect compounds the decrease in frequency times plate thickness. Solid curves are theory. while data are plot-
R (0) at the Lamb condition, detailed in Fig. 4. ted discretely.

4535 J. Appl, Phys., Vol, 58, No. 12,15 December 1985 D. E. Chimenti and A. H. Nayfeh 4535
 7.6 d = 0.56, 1.14 mm VF = .67 7.6
d = 1.35 mm

6.5 6.5
u
~
"-
~ 5.4
-' -'
w w
> >
w 4 .3 w 4.3
<f) <f)
<l <l
:I: :I:
0. 0.
3.2 3.2

2.1 L_.....L._~:.....:...._..L-_--l.:::::...._:-_~--"'=~:-----: 2.1

o 2 3 7 8 0 2
Fd (MHz-mm) Fd (MHz-mm)

FIG. 9. Velocity dispersion for Lamb waves in glass-epoxy composite plates FIG. 10. Velocity dispersion in a glass-epoxy plate 1.35 mm thick. Solid
0.56 and 1.14 mm in thickness. Model calculation assumes fiber volume curve is theory; data are discrete points.
fraction of 0.67. Theory is solid curve; experimental data are discrete points.
reflected field, corresponding to the existence of Lamb
waves. Some degree of scatter is seen in these and subsequent
In a separate stepped-frequency experiment, ultrasonic
data. We attribute this feature to a lack of uniformity in the
reflection from an isotropic bulk reflector is measured at a
glass-epoxy specimens. Rather good detailed a~reement is
similar incident angle, water-path distance, and x position.
seen between the prediction of our model calculatIOn and the
This transfer function is used to deconvolve the transducer
experimental data for the two thinnest samples in the series.
response from the raw data on the composite samples. An
The model accurately follows the trend of the data for all
example of the result of that operation is shown in Fig. 7. We
modes we observe. At the lower left in Fig. 9 the measured
plot the deconvolved signal amplitude as a function of fre-
and calculated phase velocities of mode extinction for the
quency for a composite plate 1.35 mm thick. Excellent sig-
lowest-order symmetric mode are very close. The pro-
nal-to-noise ratio and clear definition of the minima are ap-
nounced lateral shift in the data near 5.0 km/s is well. repro-
parent in these data. Figure 7 is similar to the magnitude
duced by the model and corresponds to the effective longitu-
curve of the calculated reflection coefficient seen in Fig. 4.
dinal critical angle of the composite plate. A similar, but less
Since the data are an integration over angle of the beam pro-
dramatic, effect can be detected in the calculation of Fig. 8
file with the reflection coefficient, it is expected that the ex-
on an isotropic aluminum plate.
act functional shapes of the curves in Figs. 4 and 7 will differ.
In Fig. 9 and subsequent results we have treated the
However, it can be demonstrated that the minima in R (Fd,e)
volume fraction of the fibers as a slightly adjustable param-
coincide with the onset of the null zone shown in Fig. 6,
eter, the only one in the model. The difference between the
indicating the presence of leaky Lamb waves in the plate.
measured fiber fraction and the one assumed in the theory
To verify our experimental procedure and data analysis,
curve of Fig. 9 is about 8 vol %. Experimental determina-
we have undertaken some preliminary measurements on alu-
tions of fiber fraction are performed at the edge of the speci-
minum plates. Lamb waves in aluminum plates have been
men and carry an uncertainty of 1 vol. %, but spatial varia-
studied extensively (see Ref. 22 and references contained
tions in the fiber fraction can be larger than that for glass
therein) and so offer a good opportunity for comparison. In
epoxy and can reduce the reproducibility of the data, ~s r~­
Fig. 8 we show the results of dispersion measurements on an
ferred to earlier. The experimental fiber volume fractIOn IS
aluminum plate (Al707 5-T6 alloy) compared with our model
also sensitive to the assumed densities of the composite con-
calculation, where both materials in the model are alumi-
stituents. Another factor contributing to this disparity may
num. Very good detailed agreement at incident angles from
0 0
14 to 26 gives us confidence in the validity of our approach. 7.6
d = 1.63mm VF=.58
The two curves at the extreme left are the lowest-order sym-

~
metric and antisymmetric modes, which approach the Ray- 6.5
leigh wave velocity at large Fd. The higher-order modes con- u
~
verge to the transverse wave velocity. "-
~ 5.4
III. RESULTS AND DISCUSSION .J
w
>
Experimental. measurements of leaky Lamb wave dis- '"
<f)
4.3
persion in gl.ass-epoxy composite plates, compared to the <l
l:
0.
model calculations outlined in Sec. II, are presented in the 3.2
next series of plots. Figure 9 shows dispersion data and the-
ory for unidirectional composite plates 0.56 and 1.14 mm
2.1
thick. The phase velocities appropriate to Lamb eigenmodes 0 2
of the plate are plotted as a function of frequency times plate
thickness. The continuous curves are the model, and the dis- FIG. 11. Velocity dispersion in a glass-epoxy plate 1.63 mm thick. Solid
crete points are the experimental positions of minima in the curve is theory; data are discrete points.

4536 J. Appl. Phys., Vol. 58, NO.12,15 December 1985 D. E. Chimenti and A. H. Nayfeh 4536
 7.6
d = 2.03 mm VF =.55 extrapolating the higher-order mode curves to large phase
velocity, we have identified the symmetry of each of the first
6.5 six modes. According to a well-known analysis,23 each of the
u.
'"
"-
higher-order modes approaches a standing wave at normal
~ 5.4 incidence. Integral transverse and half-integral longitudinal
-'
wavelengths in the plate correspond to symmetric modes,
'">
",4.3
while the reverse combination yields the antisymmetric
Vl
<t
modes. For the three mode overlaps we have examined, the
:x: pairs consist of one of each mode type. It is also possible that
Q.

3.2 the modes in this region are distinct but not resolvable within
the seven-digit accuracy of the numerical computations.
Since we evaluate the theoretical reflection coefficient
2 3 4 5 6 7 8 along the real Fd axis to identify the Lamb eigenmodes, cri-
Fd (MHz - mm)
teria for mode identification is a significant consideration.
FIG. 12. Velocity dispersion in a glass-epoxy plate 2.03 mm thick. Solid Initially, we searched for minima in the magnitUde of
curve is theory; data are discrete points.
R (Fd,e ) accompanied by rapid phase reversals, as illustrated
in Fig. 4. However, above a phase velocity of about 5.0 km/s,
be the small degree of fiber nonparallelism or variations in this method yielded equivocal results. A more reliable mea-
fiber-matrix wetting introduced when samples are pressed to sure of propagating modes in the plate is the large positive
a desired thickness. In spite of these material limitations and change in the phase of R (Fd,e ) by itself. In particular, this
considering the complexity of the composite, we have found phase reversal is a clear indication of passage beneath a
excellent agreement between these and subsequent measure- Lamb wave pole or between a pole-zero pair. Yet, detailed
ments and our straightforward model calculation. examination of the reflection coefficient in the range from
Data and analysis for a thicker plate (d = 1.35 mm) are 5.2 to 5.8 km/s reveals, by this criterion, no behavior sugges-
given in Fig. 10. Here, we have taken the fiber volume frac- tive of the existence of more than one Lamb-wave mode.
tion to be 0.63, or about 8 vol % higher than the density This observation leads us to conclude that the abrupt con-
determination on a coupon cut from the corner of this speci- vergence of the symmetric and antisymmetric modes
men. After this adjustment, many of the features of the pre- between 5.2 and 5.8 km/s is a fundamental property of the
vious comparison can be identified here as well. Over the model. Our data seem to support this conclusion.
entire spectrum the agreement is rather convincing, even The last aspect of the data upon which we comment is
better than the previous results. Only for some of the higher- the lack of reinstatement of both higher-order modes above
order modes is the comparison less favorable. But it should 5.8 km/s. Although the model predicts that, in every case,
be noted that these modes are highly sensitive to small varia- mode pairs reestablish distinct branches above this velocity,
tions in the material properties. Data and theory on two only in the data of Fig. 9 for the A 1 and S 1 modes is there
additional samples at plate thicknesses of 1.63 and 2.03 mm experimental evidence of this behavior. Examination of the
are presented in Figs. 11 and 12. These results display a simi- calculated reflection coefficient in this region shows marked
lar close agreement between measurement and model, al- differences in the depth of minima in the magnitude of Rand
though because of transducer bandwidth limitations, only degree of variation in the phase from one mode to the next.
Our previous work 20•21 suggests that these differences are
limited data could be collected on the lowest order symmet-
related to the strength of mode excitation. We conjecture,
ric mode. In each case the fiber volume fraction assumed in
the calculation has been about 5 vol % larger than the ex- therefore, that above 5.8 km/s the higher-order modes
which are predicted, but not observed, are much less strong-
perimental density measurements, which is consistent with
ly excited than those we could measure.
the results on the thinner specimens. Since our theoretical
model is a fairly simplified approximation to the real com-
posite, the slight adjustment (-15%) applied to the fiber iV. CONCLUSIONS
volume fraction is not unreasonable. Moreover, the values of In summary, we have calculated the effective homogen-
the constituent elastic constants we have used, which were eous anisotropic elastic behavior of a unidirectional compos-
obtained from the manufacturer. are experimental averages ite laminate in the long-wavelength limit using a two-step
over many measurements. Deviations from these average procedure based on alternating layered media. This interme-
values are certainly possible. We observe that the calculated diate result has then been incorporated into a calculation of
mode structure is relatively sensitive to the moduli of the the ultrasonic reflection coefficient of a fluid-loaded aniso-
constituents; therefore, uncertainty in the elastic constants tropic plate, which has been taken as an approximation to
must be considered a source of systematic error. the inhomogeneous composite laminate. Extensive experi-
An unusual feature of the calculations which is discern- mental results have been presented on Lamb-wave disper-
ible in all the data is the brief apparent coincidence of the sion in five composite specimens of different thickness. It is
symmetric and anti symmetric modes at the longitudinal important to state that the real composite can, and probably
wave speed in the dispersion results. These curves, extending will, deviate in several significant ways from the ideal mor-
downward from 5.2 km/s and upward from 5.8 km/s, ap- phology we have assumed in our model calculation. The fi-
parently merge into a single feature between those values. By bers will not always be perfectly aligned or evenly distribut-

4537 J. Appl. Phys., Vol. 58, No. 12, 15 December 1985 D. E. Chimenti and A. H. Nayfeh 4537
ed, leading to resin-rich and resin-starved regions in the 3A. Bedford and M. Stern, J. Appl. Mech. 38, 8 (1971).
plate. Furthermore, depending on its fabrication, some of 4B. Lempriere, Lockheed Palo Alto Research Laboratory, Report No.
LMSC-6-78-69-2I, (1969).
the composite's fibers will touch one another, or they may 5S. Minagawa, S. Nemat-Nasser, and M. Yamada, Int. J. Numer. Methods
not be in perfect welded contact with the matrix in all areas Eng. 17,1335 (1981).
of the laminate. Considering the complexity and potential 6G. A. Hegemier, G. A. Gurtman, and A. H. Nayfeh, Int. J. Solids Struct.
9,395 (1973).
nonideality of the real composite and the relative simplicity
7A. H. Nayfeh and G. A. Gurtman, J. Appl. Mech. 41,106 (1974).
of our theoretical model, the comparison between our calcu- SA. H. Nayfeh and E. A. Nasser, J. App\. Mech. 45,822 (1977).
lations and experimental measurements has yielded remar- 9A. H. Nayfeh, Fibre Sci. Technol. 10, 195 (1977).
kable agreement. lOA. Schoch, Acustica 2,1 (1952).
ilL. E. Pitts, T. J. Plona, and W. G. Mayer, J. Acoust. Soc. Am. 60, 374
(1976).
12R. Fiorito, W. Madigosky, and H. Uberall, J. Acoust. Soc. Am. 66,1857
ACKNOWLEDGMENTS (1979).
13T. J. Plona, L. E. Pitts, and W. G. Mayer, J. Acoust. Soc. Am. 59, 1324
It is our pleasure to acknowledge the expert computer (1976).
programming assistance ofK. D. Shimmin and Jeffrey Fox. 14G. Maidanik, J. Sound Vih. 3, 288 (1966).
We also wish to thank R. J. Andrews and his staff at the ISD. C. Woriton, J. Appl. Phys. 32, 967 (1961).
University of Dayton Research Institute for providing the 16M. Heckl, Acustica 10, 109 (1960).
17M. F. M. Osborne and S. D. Hart, 1. Acoust. Soc. Am. 17, 1 (1945).
characterized samples of glass-epoxy composite. Funding 18Fiber: C II = 73.1, Coo = 29.9 GPa, p = 2.55; matrix: CII = 3.45,
for this work was provided by the Air Force Office of Scienti- Coo = 1.28 GPa,p = 1.25.
fie Research and the National Science Foundation. 19 A. H. Nayfeh, D. E. Chimenti, L. Adler, and R. L. Crane, J. Appl. Phys.
52,4985 (1981).
2°D. E. Chimenti, A. H. Nayfeh, and D. L. Butler, 1. App!. Phys. 53, 170
(1982).
'G. W. Postma, Geophysics 20,780 (1955). 21A. H. Nayfeh and D. E. Chimenti. J. Acoust. Soc. Am. 75,1360 (1984).
2c. T. Sun, J. D. Achenbach, and G. Herrmann, J. App). Mech. 35, 476 Z2L. J. Derner and L. H. Fentnor, J. Nondestr. Test. 27.121 (1969).
(1968). 231. A. Viktorov. Rayleigh and Lamb Waves (Plenum, New York, 1967).

4538 J. Appl. Phys., Vol. 58, No. 12, 15 December 1985 D. E. Chimenti and A. H. Nayfeh 4538