Professional Documents
Culture Documents
0.48
0.01
__
11
22
12
_
E
3
_
d
31
d
32
0
__
+ e
33
E
3
(1)
where D
3
, e
33
and E
3
are the charge density, the dielectric
permittivity and the electric eld, respectively, along the poling
direction. The terms d
31
and d
32
are the piezoelectric constants,
and
11
,
22
and
12
are the normal and shear strain components
in the sensor. The reduced stiffness matrix of an MFC
transducer is:
[Q] =
E1
12112
12 E2
12112
0
12 E2
12112
E2
12112
0
0 0 G
12
(2)
where E
1
and E
2
are the Youngs moduli of the transducer
along 1 and 2,
12
and
21
are the respective Poissons
ratios, and G
12
is the in-plane shear modulus. Expanding
equation (1), the charge density becomes:
D
3
= (d
31
Q
11
+ d
32
Q
12
)
11
+ (d
31
Q
12
+ d
32
Q
22
)
22
_
d
2
31
Q
11
+ 2d
31
d
32
Q
12
+ d
2
32
Q
22
e
33
_
E
3
. (3)
Considering the electrical boundary conditions of the MFC
sensor to be an open circuit, the total charge over the electrode
area is zero, thus
_ _
D
3
dx dy = 0 (Tzou 1993). In this case,
the voltage can be calculated as:
V =
_ _ _
E
3
dx dy dz
lb
. (4)
Solving equation (3) for E
3
and integrating according to
equation (4) the following voltage expression is obtained:
V =
_
t
_ _
[(d
31
Q
11
+ d
32
Q
12
)
11
+ (d
31
Q
12
+ d
32
Q
22
)
22
]dx dy
_
{lb[e
33
(d
2
31
Q
11
+ 2d
31
d
32
Q
12
+ d
2
32
Q
22
)]}
1
(5)
where it is assumed that all variables are uniform through the
thickness, t (thin sensor).
Substitution of the reduced stiffness matrix of equation (2)
into equation (5) results in the following expression of the
voltage response of the MFC sensor as a function of the two
in-plane strains,
11
and
22
:
V =
_
t
_ _
[(d
31
E
1
+ d
32
12
E
2
)
11
+ (d
31
12
E
2
+ d
32
E
2
)
22
]dx dy
_
{lb[(1
21
12
)e
33
(d
2
31
E
1
+ 2d
31
d
32
12
E
2
+ d
2
32
E
2
)]}
1
. (6)
Figure 2. Rectangular MFC sensor subjected to Lamb waves in
general oblique incidence.
The sensors response to Lamb ultrasonic waves will
be examined in the following section by considering the
appropriate strain elds in equation (6). It is assumed that the
wave elds are not signicantly affected by the presence of the
sensor.
2.3. Response of MFC sensors to exural Lamb waves
Consider a rectangular MFC sensor bonded to the upper
surface of an isotropic plate of thickness 2d and subjected to
a harmonic strain eld associated to antisymmetric (exural)
Lamb waves propagating in the plane (x
, z) along direction x
(gure 2). This wave mode is commonly used for the detection
and location of damage or impacts, as discussed in section 1.
The wave propagation direction x
2
)
(7)
where A is an arbitrary constant which depends on the
magnitude of the wave and the parameters r and s are dened
as:
r =
_
k
2
2
c
2
L
, s =
_
k
2
2
c
2
T
(8)
where is the frequency, k is the wavenumber, and c
L
and
c
T
are the bulk longitudinal and shear velocities in the plate,
respectively.
For plane waves (
y
y
= 0), the only strain relevant to the
surface-mounted sensor is the in-plane component in the wave
propagation direction given by:
x
|
z=d
=
u
x
z=d
= i k
2
A
_
tanhrd
2rs
k
2
+ s
2
tanh sd
_
e
i(kx
2
)
. (9)
The in-plane strain
x
x
can be transformed into the lengthwise
and widthwise axes of the sensor using the following strain
transformation:
11
=
x
x
cos
2
22
=
x
x
sin
2
.
(10)
1491
H M Matt and F L di Scalea
Ignoring shear lag effects, the voltage expression in
equation (6) can be written as:
V =
H
lb
_ _
x
x
dx dy (11)
where H is the following frequency-independent constant:
H = {t [(d
31
E
1
+ d
32
12
E
2
) cos
2
+ (d
31
12
E
2
+ d
32
E
2
) sin
2
]}{[(1
21
12
)e
33
(d
2
31
E
1
+ 2d
31
d
32
12
E
2
+ d
2
32
E
2
)]}
1
. (12)
Thus, H depends on the electromechanical properties and
thickness of the sensor, as well as on the wave propagation
direction relative to the sensor geometrical axes.
Substituting equation (9) into (11), performing the
variable transformation x
e
i
= 2i sin ) and prosthaphaeresis formulae, the following
expression is obtained for the sensors voltage response to
antisymmetric Lamb waves:
V = iVe
i(t +
2
)
(13)
where the amplitude is:
V =
4H A
lb sin cos
_
tanhrd
2rs
k
2
+ s
2
tanh sd
_
sin
_
kb sin
2
_
sin
_
kl cos
2
_
. (14)
It should be noted that the above expression is formally
equivalent to that previously obtained by the authors for
monolithic PZT sensors (Lanza di Scalea et al 2007b), with
the constant H accounting for the differences in properties
between the isotropic PZT device and the anisotropic MFC
device.
Similarly to what was found by Lanza di Scalea et al
(2007b), wavelength tuning effects emerge through the terms
sin(kb sin /2) and sin(kl cos /2). For parallel incidence
( = 0
=0
=
2H Ak
l
_
tanhrd
2rs
k
2
+ s
2
tanh sd
_
sin
_
kl
2
_
(parallel incidence). (15)
Because sin(kl/2) = sin(l/), the response will be large
at = 2l/(2n 1) for n = 1, 2, 3, . . . corresponding to
sensor lengths equal to an odd multiple of half the wavelength.
Contrary to what was established for piezoelectric actuators
(Giutgiutiu 2005), these conditions do not necessarily
correspond to local maxima of the sensor response due to the
factor multiplying the sinusoidal term. The response will be
zero, instead, at = l/n for n = 1, 2, 3, . . . corresponding to
sensor lengths equal to an integer multiple of the wavelength.
Furthermore, from the term l in the denominator, the response
will generally increase with decreasing sensor length.
In order to eliminate the arbitrary term A in equations (14)
and (15), the response can be normalized by the Lamb wave
power ow expressed as:
N =
_
_
+d
d
P
x
dz (16)
Table 1. Electromechanical constants of the MFC sensors (direction
1 is along the piezoelectric bers).
E
1
E
2
d
31
d
32
e
33
(GPa) (GPa) (pC N
1
) (pC N
1
)
12
21
(nF m
1
)
30
a
16
a
370
a
370
b
0.31
a
0.16
c
16.37
b
a
From Schoenecker et al (2005).
b
From Smart Material Corporation, Sarasota, FL.
c
Calculated as
21
=
12
(E
2
/E
1
).
where P
x
is the x
_
u
x
t
_
+
x
z
_
u
z
t
_
_
(17)
where Re indicates the real part and * indicates complex
conjugate.
The MFC sensor response to antisymmetric Lamb waves
was validated experimentally on a 1.2 m 1.2 m 1.5 mm
(4
0.06
12
0.06
) [0/45]
2S
CFRP plate.
A rectangular MFC sensor with active area of 25.5 mm
12.2 mm (1
0.48
).
The predicted sensor responses were computed as V/N
from equations (14) and (16). For the calculation of the k =
k() dispersion solutions in these equations, a semi-analytical
nite element (SAFE) method discussed at length in a previous
work (Bartoli et al 2006) was employed for both the aluminum
plate and the CFRP plate. The SAFE analysis of the composite
plate modeled each lamina individually and then accounted for
the lay-up when assembling the global nite element matrix.
The electromechanical properties of the sensor employed to
calculate the constant H are given in table 1.
The theoretical and experimental frequency responses of
the sensors are shown in gure 3 for both plates and for
wave incident angles = 0
, 45
and 90
. The general
behavior seen in gure 3 is, again, similar to what was
found for monolithic PZT sensors (Lanza di Scalea et al
2007b). The largest response in all cases is seen at low
frequencies. This conrms that the sensor behaves as an ideal
strain detector when it is small compared to the wavelength.
At higher frequencies, where the wavelength is comparable
to the sensors dimensions, tuning effects, discussed earlier,
govern the response. The comparison between predictions and
measurements is rather satisfactory in terms of both relative
amplitude and spectral shape. The overestimation of the
predicted responses seen at higher frequencies can be attributed
to the assumption of at frequency content of the excitation
strain used in the harmonic model. Wave attenuation losses,
not considered in the study, will contribute to altering the
frequency content of the wave generated by a broadband
source.
1492
Piezoelectric rosettes for wave source location
Figure 3. Theoretical and experimental response amplitude spectra of rectangular MFC sensor subjected to antisymmetric Lamb waves in an
aluminum plate ((a)(c)) and in a [0/45]
2S
CFRP plate ((d)(f)). Waves propagating at 0
x
. (18)
From equations (9) and (14), the sensitivity factor for
antisymmetric waves is:
S =
4H
lbk
2
sin cos
sin
_
kb sin
2
_
sin
_
kl cos
2
_
. (19)
It can be demonstrated that, for l (sensor smaller
than wavelength), the sensitivity can be decomposed into a
longitudinal sensitivity factor, S
1
, and a transverse sensitivity
factor, S
2
, as:
S S
1
cos
2
+ S
2
sin
2
(20)
where is the usual wave propagation direction relative to the
sensors lengthwise direction. The longitudinal and transverse
sensitivity factors are dened for waves propagating along
the sensors lengthwise and widthwise directions, respectively.
Thus:
S
1
= S|
=0
= {2t (d
31
E
1
+ d
32
12
E
2
) sin(kl/2)}
{lk[(1
21
12
)e
33
(d
2
31
E
1
+ 2d
31
d
32
12
E
2
+ d
2
32
E
2
)]}
1
S
2
= S|
=90
= {2t (d
31
12
E
2
+ d
32
E
2
) sin(kb/2)}
{bk[(1
21
12
)e
33
(d
2
31
E
1
+ 2d
31
d
32
12
E
2
+ d
2
32
E
2
)]}
1
.
(21)
It can be further demonstrated numerically that the error
associated with the approximation in equation (20) tends to
zero for width to length (b/l) sensor ratios equal to 0.57.
Figure 4 demonstrates the validity of the approximation in
equation (20). The exact and approximate sensitivity factors
are plotted according to equations (19) and (20), respectively.
The results were calculated for the 25.5 mm 12.2 mm MFC
sensor bonded to the 1.5 mm thick aluminum plate subjected
to an a
0
wave eld at incident angles of = 30
, 45
and
60
, 45
and
60
in an aluminum plate.
where the wave strain components along the longitudinal and
the transverse sensor directions are explicitly indicated.
Equation (22) makes the MFC response formally
equivalent to that of conventional electrical resistance strain
gages. This equation also emphasizes that the response is
highly dependent on the wave propagation direction, and
thus signicant directivity behavior is expected at certain
frequencies. Figure 5 shows the longitudinal ( = 0
) and
the transverse ( = 90
. The
expected wavelength tuning effects are seen, with the zeros
and maxima of the longitudinal response occurring at lower
frequencies compared to the transverse response, owing to the
larger lengthwise dimension. More importantly, signicant
directivity behavior is apparent. At = l, for example,
the transverse sensitivity S
2
is large, whereas the longitudinal
sensitivity S
1
is zero. At frequencies lower than the =
l point, the sensitivity at = 45
12
E
2
+ d
32
E
2
) sin (kb/2)
b (d
31
E
1
+ d
32
12
E
2
) sin (kl/2)
. (25)
Let us assume that the lengthwise direction of the i th
sensor is oriented at an angle
i
from the global coordinate
axis x (gure 6(b)). In analogy with electrical resistance strain
gages, strain transformation relations can be used to determine
the Cartesian strain components from the response of the three
sensors:
_
xx
yy
x y
_
= [T]
1
V
A
S1
V
B
S1
V
C
S1
(26)
where
[T] =
_
cos
2
A
+ K
T
sin
2
A
sin
2
A
+ K
T
cos
2
A
cos
2
B
+ K
T
sin
2
B
sin
2
B
+ K
T
cos
2
B
cos
2
C
+ K
T
sin
2
C
sin
2
C
+ K
T
cos
2
C
(1 K
T
) sin
A
cos
A
(1 K
T
) sin
B
cos
B
(1 K
T
) sin
C
cos
C
_
. (27)
The principal strain angle of the wave, , measured from
the global axis x (gure 6(b)), can be nally computed as:
tan 2 =
x y
xx
yy
=
Num
Den
. (28)
1494
Piezoelectric rosettes for wave source location
Figure 6. (a) Concept of piezoelectric rosettes for the wave source location in a plane. (b) Reference system for the calculation of the
Cartesian strains of the wave from the sensors response.
The 90
< < 90
;
if Num < 0, 90
< < 0
;
if Num = 0 and
Den > 0, = 0
;
Den < 0, = 90
Den = 0, is indeterminate.
4.2. Determination of the wave source location
Evaluation of the source location in a plane is readily achieved
by the intersection of the principal directions determined by
two rosettes (gure 6(a)). The principal angles calculated
from the two rosettes,
1
and
2
, dene the equations of the
straight lines connecting the wave source to each of the rosette
centroids according to:
y
SOURCE
= (x
SOURCE
x
1
) tan
1
+ y
1
y
SOURCE
= (x
SOURCE
x
2
) tan
2
+ y
2
(29)
where (x
SOURCE
, y
SOURCE
) represent the coordinates of the
wave source in the (x, y) Cartesian system, and (x
1
, y
1
),
(x
2
, y
2
) represent the coordinates of the centroids of the two
rosettes, respectively.
The linear system of equations can nally be solved for
the coordinates of the wave source:
x
SOURCE
=
y
2
y
1
+ x
1
tan
1
x
2
tan
2
tan
1
tan
2
y
SOURCE
= (x
SOURCE
x
1
) tan
1
+ y
1
.
(30)
5. Validation of the source location
5.1. Test specimens and experimental procedure
Three test specimens were employed for the validation tests.
Shown in gure 7, these consisted of the 1.2 m 1.2 m
1.5 mm aluminum plate, the 0.3 m 0.3 m 1.5 mm [0/
45]
2S
CFRP plate, and a 0.8 m 0.3 m 5.2 mm (31.5
11.8
0.2
) and a 50.8 mm (2