612 ieee transactions on ultrasonics, ferroelectrics, and frequency control, vol. 47, no.

3, may 2000

Fracture Source Location in Thin Plates Using

the Wavelet Transform of Dispersive Waves
Hyunjo Jeong and Young-Su Jang

Abstract—A new signal processing approach was pre- velocities of the lowest symmetric (S0 ) and antisymmet-
sented for acoustic emission source location using the dis- ric (A0 ) Lamb waves reduce to the plate wave solutions.
persive waves in a thin plate. For wave propagation in The two modes of propagation are called the extensional
dispersive media, the accuracy of source location can be
improved by using the arrival times of a single frequency and the flexural modes, respectively. These modes travel
component in the output signals at an array of sensors. at different velocities and exhibit dispersion characteris-
The wavelet transform (WT) was used to resolve this prob- tics. If dispersive wave is employed for source location, the
lem. By utilizing the time-frequency data of the WT, the arrival time of the wave of a single frequency component
frequency-dependent arrival time traveling with the group should be used for accurate source location. Ziola and Gor-
velocity was shown to be easily determined. Experiments
were performed using a lead break as the simulated frac- man [1] used a cross correlation technique to determine the
ture source on the surface of an aluminum plate. Two plate time of propagation of the flexural wave in thin plates. The
S A
modes corresponding to the 0 and 0 Lamb waves were idea behind this is that, if a single frequency component in
identified, and their group velocities were accurately mea- the output waveform from each transducer could be deter-
sured. The source location results based on the WT method mined, then the time differences caused by the propagation
agreed well with the true locations. The WT method was
also compared with the cross correlation technique, and of that frequency component could be used for the location
both methods provide similar results. analysis.
The WT can also be used to detect the arrival times
of the dispersive waves propagating in plates. There have
I. Introduction been intense research activity in the application of wavelets
in various fields of science and engineering [2], [3]. Re-
he analysis of wave propagation in beams, plates, cently, the WT has been introduced to the time-frequency
T and shells is of importance in a number of nonde-
structive evaluation techniques that use acoustic waves.
analysis of transient waves propagating in dispersive me-
dia. Kishimoto et al. [4], [5] applied the WT to the time-
Much information about the generation and propagation frequency analysis of dispersive waves in beams. Suzuki
of elastic signals in these structures is needed before they et al. [6] correlated the WT of AE signals to the frac-
are fully utilized. The ability to locate fracture sources in ture modes of fiber reinforced composites. In this paper,
structures is one of the major attractions of the acoustic we apply the WT to the source location analysis of the
emission (AE) technique. In the conventional source loca- plate waves. It will be shown that, by using the peak of
tion analysis, the AE events are located based on the time the magnitude of the WT, a frequency component in the
of arrival at a number of transducers, using the known output signal can be extracted. From this, the time dif-
distances between the transducers and the bulk longitudi- ferences needed for the source location algorithm can be
nal wave propagation velocity. In commercially available determined. First, we introduce the theory of planar source
AE analyzers, the arrival time is determined by threshold location by a triangulation method. Next, the WT theory
crossing techniques. For instance, the first threshold cross- using the Gabor wavelet is briefly outlined, and its applica-
ing can be defined as the time at which the envelope of the tion to dispersive waves is explained. A method is shown to
AE signal crosses a preset threshold [1]. When the propa- calculate the group velocities of two plate modes by using
gating wave is dispersive, error could occur in the source the time-frequency data of the simulated AE waves. The
location because the AE pulse changes shape because of three source location problems are considered, and their
dispersion. results are compared with the cross correlation technique.
When the test specimen is platelike, away from the
acoustic source, the propagating waves will be governed
by Lamb’s homogeneous equation, the solutions to which II. Theory
are called Lamb waves. In the thin plate case, where the
wavelength is much greater than the plate thickness, the A. Planar Source Location

Manuscript received June 23, 1999; accepted December 2, 1999. For planar source location, three sensors must be used.
Financial support was provided by the Korea Research Foundation Analysis of arrival times at different sensors enables the
in the program year of 1998. emission source to be located by means of a triangulation
The authors are with the Division of Mechanical Engineer-
ing, Wonkwang University, Iksan, Jeonbuk 570-749, Korea (e-mail: method. A pair of differences in arrival time from an ar-
[email protected]). ray of three sensors defines two hyperbolas at which point

0885–3010/$10.00

c 2000 IEEE
Authorized licensed use limited to: UNIV ESTADUAL PAULISTA JULIO DE MESQUITA FILHO. Downloaded on February 17,2024 at 11:52:41 UTC from IEEE Xplore. Restrictions apply.
 jeong and jang: wavelet transform of dispersive waves 613

When two solutions of θ exist for a pair of time dif-

ferences, the source location is ambiguous. In this case, a
fourth sensor S4 (see Fig. 1) can be used to check the true
source location.

B. WT Transform

The continuous WT of a function f (t) is defined by [2]

∞  
1 t−b
(W f )(b, a) = √ f (t)Ψ dt (3)
a a
−∞

where a > 0 and the overbar indicates the complex con-

jugate. The function Ψ(t) is called the analyzing wavelet
and satisfies the admissibility condition
∞
|Ψ̂(ω)|2
dω < ∞ (4)
|ω|
Fig. 1. Experimental setup and sensor arrangement for the source −∞
location in a plate.
where Ψ̂(ω) denotes the Fourier transform of Ψ(t).
Although one can choose any analyzing wavelets that
of intersection the emission source is located. Instead of satisfy the admissibility condition (4), the Gabor wavelet
finding analytical equations for a pair of hyperbolas, To- is adopted in this study because it is known to provide the
bias [7] calculated closed-form solutions for the radius R best time frequency resolution [2]. The Gabor function is
of the source from a sensor located at a reference origin, expressed as [4]
and the angle θ from a reference axis. For a particular ar-
rangement shown in Fig. 1, where three sensors are located 1 ω0 (ω0 /γ)2 2
Ψg (t) = √ exp − t exp(iω0 t),
at S1 , S2 , and S3 with equal distances between any two 4
π γ 2
(5)
sensors, these equations reduce to
and its Fourier transform is
1
R=
, (1) √
2 C1 + E1 cos(θ − π/3) 2π γ (γ/ω0 )2
√  Ψ̂g (ω) = √ exp − (ω − ω0 )2 (6)
4
π ω0 2
−1 3(E1 − E2 )
θ = tan
E1 + E2 where
 ω0 and γ are positive constants. Here, we set γ =
  π 2/ ln 2 ≈ 5.336. The Gabor function (5) may be consid-
−1 C2 − C1
± cos  (2) ered as a Gaussian window function centered at t = 0 and
E12 − E1 E2 + E22 its Fourier transform (6) centered at ω = ω0 . The func-
tion Ψg ((t − b)/a) is then centered around t = b, and its
where Fourier transform [a exp(−ibω)Ψ̂g (aω)] is centered around
∆t1 V ∆t2 V ω = ω0 /a. The WT W f (b, a) using the Gabor wavelet thus
C1 = , C2 = , represents the time-frequency component of f (t) around
− ∆t21 V 2
d2 − ∆t22 V 2
d2
t = b and ω = ω0 /a. In this study, we set ω0 = 2π such
d d
E1 = 2 , E2 = 2 . that 1/a is equal to the usual frequency f = ω/2π.
d − ∆t21 V 2 d − ∆t22 V 2
C. Time-Frequency Analysis of Dispersive Waves
Here, d is the distance between the sensors, V is the prop-
agation velocity, and ∆t1 and ∆t2 are the arrival time
For the time-frequency analysis of dispersive waves, con-
differences between the sensors S1 and S2 and S1 and S3 ,
sider two harmonic waves of equal unit amplitude and of
respectively. Thus, the time differences between the sen-
slightly different frequencies ω1 and ω2 propagating in the
sors are required for the source location analysis. When
x-direction, i.e.,
the velocity V is dispersive, the arrival time differences of
AE waves also depends on the frequency. The single fre- u(x, t) = e−i(k1 x−ω1 t) + e−i(k2 x−ω2 t) . (7)
quency component in the output signals from each trans-
ducer can be extracted from the time-frequency analysis Introducing
of the wavelet transform, as will be seen later. From this,
the ∆t needed for the source location algorithm can be (k1 + k2 )/2 = kc ; (ω1 + ω2 )/2 = ωc
(8)
determined. (k1 − k2 )/2 = ∆k; (ω1 − ω2 ) = ∆ω,

Authorized licensed use limited to: UNIV ESTADUAL PAULISTA JULIO DE MESQUITA FILHO. Downloaded on February 17,2024 at 11:52:41 UTC from IEEE Xplore. Restrictions apply.
 614 ieee transactions on ultrasonics, ferroelectrics, and frequency control, vol. 47, no. 3, may 2000

(7) can be written as Additional gain was added by the amplifiers in the 2001,
so that 80 dB of total system gain were used. The acquired
u(x, t) = 2 cos(∆kx − ∆ωt)e−i(kc x−ωc t) . (9) waveforms were then stored on a computer for further anal-
ysis and display. Fig. 1 shows the sensor arrangement. In
It is seen that this resulting wave consists of two parts [8].
the source location experiment, the separation distance be-
The carrier wavelet represented by the exponential term
tween the two sensors was kept at 300 mm.
propagates with the phase velocity cp = ωc /kc . On the
For velocity measurements, two sensors were placed a
other hand, the modulated wave given by the cosine term
known distance apart along a line from the lead break-
travels with the group velocity cg = dω/dk in the limit of
ing source on the plate, and the arrival time difference
∆k → 0.
between the sensors was then calculated. The distance to
When the Gabor wavelet is used as the analyzing
the sensors 1 and 2 from the source was l1 = 300 mm and
wavelet, the WT of u(x, t) is given by
l2 = 600 mm, respectively. As mentioned previously, the
√ location of the peak on the (a, b) plane corresponds to the
(W u)(x, a, b) = a e−i(k1 x−ω1 b) Ψ̂g (aω1 )
arrival time b = x/cg of the wave with frequency f = 1/a.
+ e−i(k2 x−ω2 b) Ψ̂g (aω2 ) , (10) Thus, dividing the distance between the two sensors by
the arrival time difference gives the frequency-dependent
so the magnitude of the WT is obtained as group velocity
√  2  2 12 − l1
|(W u)(x, a, b)| = a Ψ̂g (aω1 ) + Ψ̂g (aω2 ) cg (f ) = . (13)
b2 (f ) − b1 (f )
1/2
+ 2Ψ̂g (aω1 )Ψ̂g (aω2 ) cos(2∆kx − 2∆ωb) . (11) The integral in (3) was evaluated numerically using the
trapezoidal rule with a time step equal to the sampling rate
If ∆ω is sufficiently small such that Ψ̂g (aω1 ) ≈ Ψ̂g (aω2 ) ≈ ∆ts . The parameters a and b were discretized as follows:
Ψ̂g (aωc ), we obtain
a = 2m/4 , b = n∆ts (14)
√  1/2
|(W u)(x, a, b)| ≈ 2a|Ψ̂g (aωc )| 1 + cos(2∆kx − 2∆ωb) .
(12) where m and n are integers.

This result indicates that the magnitude of the WT takes

its maximum value at a = ω0 /ωc and b = (∆k/∆ω)x = IV. Results and Discussion
x/cg . In other words, the location of the peak on the (a, b)
plane indicates the arrival time of the group velocity cg at A. Plate Wave Group Velocities
frequency ωc = ω0 /a, i.e., f = 1/a [4].
Typical waveforms caused by 45◦ lead break on the edge
of the aluminum plate are shown in Fig. 2. In Fig. 2(a),
III. Experiment where source-to-receiver distance = 300 mm, the exten-
sional mode can be observed at 30 µs, followed by the flexu-
A schematic of the experimental setup is shown in ral mode beginning roughly at 90 µs. To verify the mode of
Fig. 1. The source location was carried out on an aluminum these plate waves, two transducers were mounted on oppo-
plate of 6061-T6, with dimensions 1200 mm×1200 mm and site faces of the plate at the same location. A comparison of
3 mm thick. the response of the two transducers showed that they were
Lead breaks were used either on the surface or on the in phase for the initial part but were out of phase for the
edge of the plate to simulate various AE waveforms. For subsequently arriving part. The amplitude of the in-plane
the lead break at 90◦ with respect to the plane of the plate, and out-of-plane components of these modes depends on
most of the source motion is normal to the plate, which the source orientation. It is observed from Fig. 2(b), where
should produce a large flexural mode. For the case of 0◦ source-to-receiver distance = 600 mm, that the two modes
break, the break was positioned on the edge of the plate are more separated in time because the flexural mode prop-
for the source motion in the plane of the plate. For source agates much slower.
motion at a certain angle, the break was positioned on the Fig. 2 also shows the time-frequency distribution of the
inclined edge that was machined to allow the lead to be magnitude of the WT and its contour plot. The broken
broken at an angle. Gorman and Prosser [9] showed the curve in the contour plot indicates the location of the
dependence of the amplitudes of plate waves on the source peak of the magnitude for each value of frequency. As men-
orientation. tioned before, each peak represents the arrival time of a
Four broadband, piezoelectric sensors (PAC R15) were wave traveling with the group velocity. According to such
coupled to the surface of the plate by vacuum grease. A a plot, the traveling time of corresponding wave mode be-
PAC four-channel analyzer (Mistras 2001) was used for tween any two points can be obtained for each value of
waveform acquisition. The detected signals were pream- f = 1/a. Because the distance between two measuring
plified by 40 dB with PAC model 1220A preamplifiers. points is known, the group velocity can be identified at

Authorized licensed use limited to: UNIV ESTADUAL PAULISTA JULIO DE MESQUITA FILHO. Downloaded on February 17,2024 at 11:52:41 UTC from IEEE Xplore. Restrictions apply.
 jeong and jang: wavelet transform of dispersive waves 615

(a) (b)

Fig. 2. a) Waveform detected at l1 = 300 mm from the source, time-frequency distribution of the magnitude of the WT, and its contour
plot. b) Waveform detected at l2 = 600 mm from the source, time-frequency distribution of the magnitude of the WT, and its contour plot.

B. Source Location

To determine the arrival time of the simulated AE signal

at a single frequency, we used 150 kHz in the WT analysis.
This frequency was chosen because the spectral analysis
showed that this was a frequency that was typically present
in the captured waveforms.
Three different problems were considered for the loca-
tion analysis in the present study. This was done by break-
ing the pencil lead at angles with respect to the surface or
the edge of the plate. Depending on the source orienta-
tion, differences in the amplitudes of the two modes were
observed.
We first consider the case where the flexural modes are
Fig. 3. Group velocities of the extensional and flexural modes. The dominant in the waveforms detected at the three sensors.
measured velocities were calculated from the WT data of Fig. 2. Fig. 4 shows typical waveforms received at the sensors S1
and S2 when the source was placed at B in Fig. 5. Because
these waveforms were created by the 90◦ lead break on the
plate surface, the amplitude of the extensional mode was
each frequency. Fig. 3 shows the measured group velocities negligibly small. So the two plate modes were easily distin-
of the two plate modes. Also shown in Fig. 3 are the theo- guishable, and the flexural mode was used for the source lo-
retical group velocities calculated from the phase velocities cation. Also shown in Fig. 4 are the magnitudes of the WT
of the Rayleigh-Lamb equations [10] via the relationship calculated at 150 kHz. This figure represents the variation
cg = dω/dk. The material properties used in the calcula- of the spectral intensities of the 150-kHz component as a
tion were E = 70 GPa, ν = 0.33, and ρ = 2.73 g/cm3 . The function of time. The peak corresponds to the arrival time
measured velocities of the two modes agree well with the of the group velocity at 150 kHz. Thus, ∆t needed in the
theory for a wide range of frequency. This confirms that source location analysis could be easily determined. The
the Lamb waves generated are the two plate modes. measured group velocity of the flexural mode at 150 kHz

Authorized licensed use limited to: UNIV ESTADUAL PAULISTA JULIO DE MESQUITA FILHO. Downloaded on February 17,2024 at 11:52:41 UTC from IEEE Xplore. Restrictions apply.
 616 ieee transactions on ultrasonics, ferroelectrics, and frequency control, vol. 47, no. 3, may 2000

Fig. 5. Source location results of problem 1: lead breaks (+) and the
measured locations by the WT method (×).

TABLE I
Errors of the Source Locations for Problem 1.

Error (mm)
Source (x, y)
(mm) WT method Cross correlation method
A (150,180) 5.7 3.2
B (100, 80) 5.3 3.0
C (270, 30) 2.0 3.2
D (80, 130) 2.8 6.0

sors S1 and S2 when the source is placed at C in Fig. 7.

Fig. 4. Source location problem 1: a) waveform of the sensor S1 and Although the distance from the source to the sensor S1 is
its magnitude of the WT at f = 150 kHz; b) waveform of the sensor short (50 mm), the two modes are easily discerned in the
S2 and its magnitude of the WT at f = 150 kHz. time domain. This fact is more clearly reflected in the time-
frequency domain, as shown in the magnitude of the WT
at 150 kHz. We are able to observe the larger separation
was 2900 m/s (Fig. 3), and this value was used in the lo- of the two modes as the propagation distance increases to
cation algorithm. 150 mm.
Fig. 5 shows the source location results. The four lead When both modes are equally dominant, it is impor-
breaks at each position are plotted (+) along with the tant to identify each mode and to use the same mode in
measured location using the WT method (×). The data the source location. Here, the experimental location was
comparisons show good agreement. Table I summarizes performed using the extensional and the flexural modes.
these results together with errors. The source location re- The arrival time and the velocity at 150 kHz were used for
sults obtained by the cross correlation method [1] are also the extensional mode source location. The velocity of the
shown in Table I. The two techniques show similar accu- extensional mode at 150 kHz was 5400 m/sec as obtained
racy. The errors in the data are within the sensor radius from Fig. 3.
(6.3 mm). The errors here could be attributed to measure- Fig. 7 shows the results of the source location using the
ment errors in the placement of sensors and location of the WT technique. The three lead breaks at each position are
lead breaks. plotted (+) along with the measured locations using the
The second case deals with the simulated AE waveforms WT of the flexural mode (×) and the extensional mode (∗).
where both the extensional and the flexural modes are The data comparisons show good agreement with the mea-
dominant. This type of waveform can be generated using sured values. Table II summarizes these results together
the lead break at 45◦ on the machined edge of the plate. with errors. Because the correct modes were used in the
Fig. 6 shows the typical waveforms received at the sen- location algorithm, the two results show a similar accuracy.

Authorized licensed use limited to: UNIV ESTADUAL PAULISTA JULIO DE MESQUITA FILHO. Downloaded on February 17,2024 at 11:52:41 UTC from IEEE Xplore. Restrictions apply.
 jeong and jang: wavelet transform of dispersive waves 617

Fig. 7. Source location results of problem 2: lead breaks (+) and the
measured locations by using the flexural mode (×) and the exten-
sional mode (∗).

The last case considers the dominance of the flexural

mode at one sensor and the dominance of the extensional
and flexural modes at the other two sensors. Fig. 8 shows
the output waveforms from the sensors at S1 , S2 , and S3
and their magnitudes of the WT at 150 kHz. Note that the
dominant modes in the waveforms detected at the sensors
S1 and S2 are the extensional and the flexural modes. Be-
cause of the short source-to-sensor distance, the two modes
present in the waveform of S1 are almost indistinguishable
in the time domain as well as in the time-frequency do-
main at 150 kHz. However, one will be able to observe
the gradually larger separation of the two modes as the
lower frequency value is used in the time-frequency analy-
sis. This is because the group velocity of the flexural mode
becomes slower as the frequency becomes lower because of
dispersion, as shown in Fig. 3. The very weak extensional
Fig. 6. Source location problem 2: a) waveform of the sensor S1 and modes are notable in the waveform of the sensor S3 in the
its magnitude of the WT at f = 150 kHz; b) waveform of the sensor time domain, but their spectral intensities at 150 kHz are
S2 and its magnitude of the WT at f = 150 kHz. not detected in the magnitude plot of the WT.
To locate the correct source position, one should iden-
tify and use the common flexural mode present in all three
waveforms. Fig. 9 shows the result of source location when
the arrival times of the flexural modes were used. The lead
break source was plotted (+) along with the measured lo-
cation by the WT method (×). The data comparisons show
TABLE II
good agreement between the true and measured values.
Errors of the Source Locations for Problem 2.

Error (mm)
Source (x, y) V. Conclusions
(mm) Use of a0 mode Use of s0 mode
A (150, 0) 2.8 3.8
A new method for locating acoustic emission sources
B (100, 0) 3.6 4.3
C (50, 0) 5.2 5.7 in a thin plate was developed based on the WT of the
dispersive plate waves. It was found that the WT using
the Gabor wavelet was an effective tool for analyzing the
time-frequency behavior of dispersive waves. The time dif-
ferences needed in the source location could be determined

Authorized licensed use limited to: UNIV ESTADUAL PAULISTA JULIO DE MESQUITA FILHO. Downloaded on February 17,2024 at 11:52:41 UTC from IEEE Xplore. Restrictions apply.
 618 ieee transactions on ultrasonics, ferroelectrics, and frequency control, vol. 47, no. 3, may 2000

Fig. 9. Source location result of problem 3: lead break (+) and the
measured location by the WT method (×).

from the properties of the peak of the magnitude of WT in

the time-frequency distribution. The source location algo-
rithm, based on the WT data of the flexural or the exten-
sional plate mode, was found to give accurate results. The
cross correlation technique was also applied and was found
to provide similar results. When both plate modes were
present in the waveforms received at the output sensors,
it was important to identify each mode first and then use
the time-frequency data of the identical modes for accurate
source location. There may be cases where distinguishing
these two modes in the time or in the time-frequency do-
main is difficult, so an efficient way should be sought. Al-
though the source location scheme based on the WT of
dispersive waves was tested on the simple plate geometry,
we anticipate that it may be extended to the problem of
more complicated thin-walled structures.

References

[1] S. M. Ziola and M. Gorman, “Source location in thin plates

using cross-correlation,” J. Acoust. Soc. Amer., vol. 90, no. 5,
pp. 2551–2556, 1991.
[2] C. K. Chui, An Introduction to Wavelets. San Diego: Academic
Press, 1992.
[3] M. B. Ruskai, G. Belykin, I. Daubechies, Y. Meyer, R. Coifman,
S. Mallat, and L. Raphael, Eds. Wavelets and Their Applica-
tions. Boston, MA: Jones and Bartlett, 1992.
[4] K. Kishimoto, H. Inoue, M. Hamada, and T. Shibuya, “Time
frequency analysis of dispersive waves by means of wavelet trans-
form,” J. Appl. Mech., vol. 62, pp. 841–846, 1995.
[5] H. Inoue, K. Kishimoto, and T. Shibuya, “Experimental wavelet
analysis of flexural waves in beams,” Exp. Mech., vol. 36, no. 3,
pp. 212–217, 1996.
[6] H. Suzuki, T. Kinjo, Y. Hayashi, M. Takemoto, and K. Ono,
“Wavelet transform of acoustic emission signals,” J. Acoust.
Emission, vol. 14, no. 2, pp. 69–84, 1996.
[7] A. Tobias, “Acoustic emission source location in two dimensions
by an array of three sensors,” Non-Destructive Testing, vol. 9,
Fig. 8. Source location problem 3: waveforms at the output sensors no. 1, pp. 9–12, 1976.
S1 , S2 , and S3 and their magnitudes of the WT at f = 150 kHz. [8] J. D. Achenbach, Wave Propagation in Elastic Solids. New York:
North-Holland Publishing Company, 1973.

Authorized licensed use limited to: UNIV ESTADUAL PAULISTA JULIO DE MESQUITA FILHO. Downloaded on February 17,2024 at 11:52:41 UTC from IEEE Xplore. Restrictions apply.
 jeong and jang: wavelet transform of dispersive waves 619

[9] M. G. Gorman and W. H. Prosser, “AE source orientation by Young-Su Jang received the B.S. degree
plate wave analysis,” J. Acoust. Emission, vol. 9, no. 4, pp. 283– in 1998 from Pusan National University in
288, 1990. Machine Design Engineering. He is a M.S.
[10] K. F. Graff, Wave Motion in Elastic Solids. Columbus, OH: candidate and a research assistant at Pu-
Ohio State University Press, 1975. san National University. His research inter-
ests include digital signal processing and elas-
tic wave propagation in thin-walled structures
with applications in acoustic emission and
material properties characterization.
Hyunjo Jeong received the M.S. degree in
1986 from Rensselaer Polytechnic Institute
and the Ph.D. degree in 1990 from Iowa
State University in mechanical engineering
and engineering mechanics, respectively. Af-
ter a short period of employment as a post-
doctoral associate at the Center for Nonde-
structive Evaluation at Iowa State University,
Dr. Jeong joined the Agency for Defense De-
velopment, Korea, as a senior researcher. He
was in charge of design and nondestructive in-
spection of composite rocket motor cases. Dr.
Jeong has been assistant professor of mechanical engineering at Pu-
san National University, Korea, since 1996. On March 1, 2000, he
moved to the Division of Mechanical Engineering, Wonkwang Uni-
versity, Korea. His current research interests include wavelet signal
processing, ultrasonic characterization of creep-damaged materials,
and scattering model-based ultrasonic nondestructive evaluation. He
is also involved in the research of piezoelectric sensors for monitoring
strain, vibration, and fatigue crack growth. Dr. Jeong has published
about 40 papers in the area of nondestructive evaluation and char-
acterization of composite materials. He is editor of Journal of the
Korean Society for Nondestructive Testing.

Authorized licensed use limited to: UNIV ESTADUAL PAULISTA JULIO DE MESQUITA FILHO. Downloaded on February 17,2024 at 11:52:41 UTC from IEEE Xplore. Restrictions apply.