Ultrasonics 54 (2014) 1832–1841

Contents lists available at ScienceDirect

Ultrasonics
journal homepage: www.elsevier.com/locate/ultras

Corrosion and erosion monitoring in plates and pipes using constant

group velocity Lamb wave inspection
Peter B. Nagy a,b,⇑, Francesco Simonetti a,b, Geir Instanes b,c
a
Department of Aerospace Engineering and Engineering Mechanics, University of Cincinnati, Cincinnati, OH 45221, USA
b
Cincinnati NDE, Cincinnati, OH 45244, USA
c
ClampOn AS, 5162 Laksevaag, Bergen, Norway

a r t i c l e i n f o a b s t r a c t

Article history: Recent improvements in tomographic reconstruction techniques generated a renewed interest in short-
Received 26 August 2013 range ultrasonic guided wave inspection for real-time monitoring of internal corrosion and erosion in
Received in revised form 29 December 2013 pipes and other plate-like structures. Emerging evidence suggests that in most cases the fundamental
Accepted 30 January 2014
asymmetric A0 mode holds a distinct advantage over the earlier market leader fundamental symmetric
Available online 17 February 2014
S0 mode. Most existing A0 mode inspections operate at relatively low inspection frequencies where the
mode is highly dispersive therefore very sensitive to variations in wall thickness. This paper examines
Keywords:
the potential advantages of increasing the inspection frequency to the so-called constant group velocity
Lamb waves
EMATs
(CGV) point where the group velocity remains essentially constant over a wide range of wall thickness
Corrosion monitoring variation, but the phase velocity is still dispersive enough to allow accurate wall thickness assessment
from phase angle measurements. This paper shows that in the CGV region the crucial issue of tempera-
ture correction becomes especially simple, which is particularly beneficial when higher-order helical
modes are also exploited for tomography. One disadvantage of working at such relatively high inspection
frequency is that, as the slower A0 mode becomes faster and less dispersive, the competing faster S0 mode
becomes slower and more dispersive. At higher inspection frequencies these modes cannot be separated
any longer based on their vibration polarization only, which is mostly tangential for the S0 mode while
mostly normal for the A0 at low frequencies, as the two modes become more similar as the frequency
increases. Therefore, we propose a novel method for suppressing the unwanted S0 mode based on the
Poisson effect of the material by optimizing the angle of inclination of the equivalent transduction force
of the Electromagnetic Acoustic Transducers (EMATs) used for generation and detection purposes.
Ó 2014 Elsevier B.V. All rights reserved.

1. Introduction reflection measurements are rather sensitive to the presence of a

Corrosion and erosion detection and monitoring are essential
prognostic means of preserving material integrity and reducing
the life-cycle cost of industrial infrastructure, ships, aircraft,
ground vehicles, pipelines, oil installations, etc. Long-range guided
wave inspection has the potential to extend ultrasonic corrosion
measurements in pipes over very long distances [1–6]. Carefully
selected extensional, flexural, or torsional ultrasonic guided waves
in the pipe wall provide an attractive solution for long-range corro-
sion monitoring because they can be excited at one location on the
pipe and will propagate along the pipe, returning echoes indicating
the presence of corrosion or other pipe features. However,
⇑ Corresponding author at: Department of Aerospace Engineering and Engineer-
ing Mechanics, University of Cincinnati, Cincinnati, OH 45221, USA. Tel.: +1 513 556
3353; fax: +1 513 556 5038.
E-mail address: peter.nagy@uc.edu (P.B. Nagy).
distinct sharp transition between sections of different thickness.
Transmission measurements in pitch–catch mode work better
when no such localized transition exists and the wall thickness
varies in a gradual manner.
Dry-coupled piezoelectric transducer systems were shown to
detect corrosion in chemical plant pipework using cylindrical Lamb
waves in pulse-echo mode over distances approaching 50 m in
steel pipes [7] and they can propagate through multiple bends
[8]. It was also shown that low-frequency axisymmetric modes
can propagate over long distances even in buried, water-filled iron
pipes [9]. Most of such inspections are based on reflection mea-
surements in pitch–catch mode [2,6,10,11]. Carefully selected
extensional, flexural, or torsional ultrasonic guided waves in the
pipe wall provide an attractive solution for long-range corrosion
monitoring because they can be excited at one location on the pipe
and will propagate along the pipe, returning echoes indicating the

http://dx.doi.org/10.1016/j.ultras.2014.01.017
0041-624X/Ó 2014 Elsevier B.V. All rights reserved.
 P.B. Nagy et al. / Ultrasonics 54 (2014) 1832–1841
presence of corrosion or other pipe features. However, reflection
measurements are rather sensitive to the presence of a distinct
transition between sections of different thickness. Transmission
measurements in pitch–catch mode work better when no such
localized transition exists and the wall thickness varies in a gradual
manner [12]. It was shown that ultrasonic guided wave attenua-
tion measurements can be also exploited for the detection of wall
loss due to corrosion [13].
Various wave modes can be used to best detect thinning of the
pipe wall based on mode cutoff, group and phase velocity, trans-
mission coefficient or attenuation measurements. For example,
by carefully selecting the inspection frequency to match the range
of wall thickness in the pipe, one can measure the group velocity of
the S0 mode for corrosion monitoring [14]. Ultrasonic guided wave
inspection methods can be also distinguished based on the gener-
ation and detection principles they rely on as well as the different
physical principles of the transducers used. Conventional normal
and angle beam transducers exhibit very different spatial and tem-
poral frequency characteristics that can be analyzed using source
influence theory [15]. Typically, inspection is based on a single
carefully selected guided mode. However, in some cases, a multi-
mode approach is adapted, e.g., by using a linear array comb
transducer [16]. Guided waves generated by axisymmetric and
non-axisymmetric surface loading have their distinct advantages
and disadvantages [17]. Time-delay periodic ring arrays have been
used to generate axisymmetric guided wave modes in hollow
cylinders [18].
Most structural health monitoring (SHM) systems focus on cru-
cial areas that are particularly susceptible for damage, e.g., erosion
or corrosion. In such cases localized inspection strategies are pref-
erable over long-range inspection that inevitably sacrifices detec-
tion sensitivity to maximize area coverage. Recently, Cawley
et al. devised an optimal inspection strategy for designing a perma-
nently installed corrosion/erosion monitoring (CEM) system [19].
When relatively small wall thickness loss is expected more or less
uniformly distributed over the area of interest, a small number of
spot sensors should be used. When the loss tends to be severe
and concentrated at a few unpredictable locations, an averaging-
type area monitoring system is preferable. The decision is harder
when moderate loss is expected over a significant but unpredict-
able fraction of the surface [19]. Short-range ultrasonic guided
wave tomography (GWT) is especially well suited to map the wall
loss distributed over the targeted area from a limited number of
transducer locations [20–24]. In a typical GWT configuration, a pair
of transmitting and receiving ring arrays of ultrasonic transducers
surrounds the area to be monitored. Different combinations of the
array elements are used to transmit and receive guided wave sig-
nals to interrogate the area of interest from multiple directions.
Each received signal carries information about the geometrical
characteristics of the encountered defects, which is then decoded
using appropriate reconstruction algorithms.
Ultrasonic guided waves are particularly well suited for inspec-
tion of pipelines. In relatively thin-walled pipes, the guided waves
can be approximated as Lamb modes propagating along helical
paths that allow the same mode to arrive to the receiver at differ-
ent times [25]. Fig. 1 shows a schematic diagram of the three low-
est-order helical paths along and around a cylindrical pipe. The
propagation length of the nth-order helical path is.
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
‘n ¼ z2 þ ða þ pnDÞ2 ; ð1Þ
where z and a are the axial and azimuthal distances between the
transducers, respectively, D is the average diameter of the pipe,
and n is the azimuthal order.
Although higher-order helical modes are somewhat more af-
fected by the circumferential curvature of the pipe, in thin-walled
1833
n = +1
a
n=0
D
n = -1
z
Fig. 1. A schematic view of the three lowest-order helical paths in a cylindrical
pipe.
pipes their velocity is still almost the same as that of the zero-order
mode that follows the most direct path between the transmitter
and the receiver therefore all modes can be crudely approximated
by the corresponding Lamb mode in a flat plate. The problem is
complicated by the fact that a number of dispersive Lamb modes
can propagate in each direction in a pipe of given material
depending on its wall thickness and the inspection frequency
range used by the monitoring system. Because of the highly
dispersive nature of Lamb modes, fast modes following longer
helical paths (n = 1, 2, 3, . . .) can actually beat slower modes
following the shortest direct route (n = 0), therefore the observed
vibration at the location of the receiver is much more complicated
than one would assume based on direct Lamb wave propagation
only.
As an example, Fig. 2 shows the (a) phase and (b) group velocity
dispersion curves, respectively, for Lamb waves in a steel plate.
Both velocities were normalized to the shear velocity cs of the
material. In these calculations the longitudinal and shear bulk
velocities were assumed to be cd = 5900 m/s and cs = 3200 m/s. Of
particular interest in the following will be the region surrounding
the point where the group velocity of the A0 mode reaches its max-
imum, the so-called constant group velocity (CGV) point, which is
around fd  1.4 MHz mm for steel. This region is indicated by an
open circle in Fig. 2.
Even without the added complexity of higher-order helical
modes, separation of numerous dispersive Lamb modes presents
a formidable problem and renders reliable inversion all but impos-
sible. Therefore, most guided wave inspections are conducted at
frequencies well below the cut-off frequency of the first-order
asymmetric Lamb mode so that only the two fundamental, i.e., zer-
oth-order, Lamb modes are present. At low frequencies, the funda-
mental symmetric or S0 mode is a simple dilatational plate
vibration with weak dispersion and mostly in-plane vibration
while the fundamental asymmetric or A0 mode is a flexural plate
vibration with strong dispersion and mostly out-of-plane
vibration.
Generally speaking, dispersion is good since it provides sensi-
tivity to wall thickness variations, while out-of-plane displacement
is bad since it provides strong coupling to the surrounding medium
and results in strong attenuation through energy leakage. Conse-
quently, neither of the fundamental modes is particularly useful
below a certain minimum frequency, especially because the
increasing wavelength limits the spatial resolution of any inspec-
tion scheme. As the frequency increases the differences between
the S0 and A0 modes decrease and, in some respects, even reverse.
The S0 mode becomes more dispersive while the A0 mode becomes
less dispersive and the aspect ratio of their elliptically polarized
surface displacement trajectories decreases. Reversal occurs
roughly around the point when the clockwise rotation of the sur-
face particle displacement produced p by
ffiffiffi the S0 mode changes to
counterclockwise rotation (cp ¼ 2cs ) that is around
fd  2.4 MHz mm for steel. Above this frequency  thickness prod-
uct, the S0 mode exhibits lower group velocity than the A0 mode,
1834 P.B. Nagy et al. / Ultrasonics 54 (2014) 1832–1841

Normalized Group Velocity [mm/μs]

2.5 2.5

Normalized Phase Velocity [mm/μs]

symmetric modes
2 2 asymmetric modes
CGVpoint

1.5 1.5

1 1

symmetric modes
0.5 0.5
asymmetric modes
CGVpoint
0 0
0 2 4 6 8 10 0 2 4 6 8 10
Frequency ×Thickness [MHz mm] Frequency ×Thickness [MHz mm]

(a) (b)
Fig. 2. Normalized (a) phase and (b) group velocities versus frequency  thickness in steel.

but this happens outside the interest of frequency range where toneburst of f = 180 kHz carrier frequency, corresponding to
only the two fundamental modes exist. fd  1.33 MHz mm. The first arrival at each receiver is the direct
In a recent study, Huthwaite et al. presented a detailed investi- (n = 0) S0 mode followed by the direct A0 mode at Receiver #1
gation of the advantages and disadvantages of the S0 and A0 modes and by the S0 mode along the shortest helical path (n = 1 for a po-
for guided wave tomography and compared their performance sitive offset value of a) for most other receivers. The complexity of
using both numerical and experimental data [26]. As expected, the wave patterns exhibited by the later parts of these signals dem-
the sensitivity of the A0 mode to thickness variations was shown onstrates that it is essential to suppress one of the fundamental
to be superior to that of the S0 mode, while the leaky attenuation modes in order to facilitate unequivocal identification of the direct
caused by liquid loading was found much higher for the A0 mode path from the higher-order helical paths.
than for the S0. The A0 mode was shown to be less sensitive to Additional requirements in short-range GWT include the need
the presence of various surface coatings than the S0 mode. Finally, to minimize ultrasonic scattering from neighboring transducers
the authors found that both modes could achieve similar levels of and to maximize the temperature stability of not only the trans-
spatial resolution in thickness mapping. The final decision between ducers themselves but also the coupling between the transducers
the two fundamental modes depends on the issue of transduction. and the structure to be monitored. These requirements heavily fa-
Since the S0 mode is faster in the frequency region of interest for vor noncontacting Electromagnetic Acoustic Transducers (EMATs)
corrosion/erosion monitoring, it will always arrive first along the over conventional contact piezoelectric transducers, especially
direct route, so any other modes that will arrive later can be re- since, in SHM with permanently deployed transducers, the inher-
moved easily by gating. In contrast, when using the A0 mode the ently lower transduction sensitivity of EMATs can be fairly easily
transducers must be carefully optimized to suppress the excitation compensated by more excessive signal averaging. Later in this pa-
of S0 waves. When relying on later arriving helical modes in pipe per we will demonstrate that in the frequency range of interest the
inspection, the speed advantage of the S0 mode is less significant, S0 mode transduction sensitivity of EMATs can be fairly easily sup-
so if the S0 mode is used, A0 mode generation and detection must pressed without affecting the A0 sensitivity by controlling the geo-
be suppressed [26]. metrical features of the transducer while the opposite is not
In order to illustrate the necessity of single mode generation feasible. For this reason, in the following we will focus solely on
and reception, Fig. 3 shows the cascade plot composed of the re- ultrasonic guided wave monitoring of wall thickness variations
ceived signals of 16 receivers distributed along the circumference using the A0 mode.
of a D = 226 mm diameter steel pipe of d = 7.4 mm wall thickness.
The transmitter was z = 600 mm away from the plane of the 2. Constant group velocity mode
receiver array positioned at the same circumferential position as
Receiver #1. The transmitted signal was a 4-cycle windowed The fundamental asymmetric A0 mode is particularly well sui-
ted for wall thickness monitoring in the frequency range around
the previously defined CGV point (see Fig. 1). CGV inspection ex-
1 ploits the wide plateau region where the group velocity is essen-
2 tially independent of wall thickness variations, therefore the time
3
4 of arrival of the signals to be monitored does not change [27]. Addi-
5
6 tional advantages of conducting wall thickness monitoring in the
Channels

7 CGV region include relatively low susceptibility to attenuation

8
9 due to absorbing coatings as well as to leakage into the surround-
10
11 ing fluid in the case of fluid-filled and/or immersed pipes. The good
12 sensitivity of A0 guided wave inspection in the CGV regime is en-
13
14 tirely due to the significant phase dispersion exhibited by this
15
16 mode in this region. Fig. 4 shows the fundamental symmetric
and asymmetric arrivals after z = 400 mm propagation in steel
0 100 200 300 400 500 600 700
plates of different thickness. The leading S0 mode exhibits weak
Time [µs]
group and phase velocity dispersion. In comparison, the lagging
Fig. 3. Cascade plot of the received signals detected by 16 transducers distributed A0 mode exhibits very strong phase velocity dispersion, but not
along the circumference of a D = 226 mm diameter steel pipe. perceivable group velocity dispersion.
 P.B. Nagy et al. / Ultrasonics 54 (2014) 1832–1841 1835

9 mm
where sg is the group slowness. Since the phase slowness
9.5 mm
10 mm k
sp ¼ ; ð7Þ
Amplitude [a.u.]

10.5 mm
11 mm
x
and group slowness
@k @sp
sg ¼ ¼ sp þ x ð8Þ
@x @x
are both related to the wave number k and the angular frequency x,
the true phase angle can be re-written from Eq. (6) as follows
@sp
80 90 100 110 120 130 140 150 160 170 180 U ¼ x2 z : ð9Þ
@x
Time [µs]
It is advantageous to introduce a normalized frequency
Fig. 4. Fundamental symmetric and asymmetric arrivals after z = 400 mm propa-
gation in steel walls of different thickness. xd
n¼ ð10Þ
cs
and normalized phase slowness
0
136.8 µs PðnÞ ¼ sp cs : ð11Þ
-45
136.1 µs
Phase Angle [degree]

Then, the true phase angle can be written as

.

-90 135.3 µs
134.6 µs
-135 z
133.9 µs U ¼ FðnÞ ; ð12Þ
-180
d
-225 where F is a normalized sensitivity function that is dependent only
on Poisson’s ratio m through the normalized phase slowness P [29]
-270
@PðnÞ
-315 FðnÞ ¼ n2 : ð13Þ
@n
-360
0.00 0.05 0.10 0.15 0.20 Fig. 6 shows the normalized sensitivity function versus the nor-
Frequency [MHz] malized frequency for steel. Interestingly, the sensitivity function
peaks at the same value of the normalized frequency as the group
Fig. 5. Phase spectra of the fundamental flexural mode at different gate positions
velocity, i.e., at the CGV point of n0  2.85, where its value is
(tw) in a steel plate of wall thickness d = 10 mm over propagation distance
z = 400 mm. F0  0.73 radian or 42°.
The coincidence of the peak group velocity and peak sensitivity
In order to determine the sensitivity of guided wave inspection locations is of great significance from the point of view of thermal
based on phase measurement to wall thickness variation, let us as- stability [30]. From Eq. (8), the derivative of the group slowness is
sume that the fundamental flexural mode signal h(t) is gated by a !
rectangular window of tw center delay and Tw total length. The @sg 1 @P @2P
¼ 2 þn 2 : ð14Þ
magnitude H(x) and phase u(x) spectra of the gated signal can @n cs @n @n
be obtained from the complex Fourier transform as follows
From Eq. (13), the derivative of the normalized sensitivity func-
Z t w þT2w
iuðx;t w Þ ixðtt w Þ tion is
jHðx; t w Þje ¼ hðtÞe dt; ð2Þ !
t w T2w
@F @P @2P
¼ n 2 þn 2 : ð15Þ
where t and x denote time and angular frequency. The phase spec- @n @n @n
trum can be calculated from
  It can be easily seen from Eqs. (14) and (15) that they have the
uðx; tw Þ ¼ x tw  sp ðxÞz ; ð3Þ same roots, which proves the above mentioned coincidence. There-
where sp denotes the phase slowness and z is the propagation dis- fore, in a fairly wide range around the CGV point, the true phase
tance. As it is illustrated in Fig. 5, the phase spectrum is a function angle can be approximated using the peak of the normalized sen-
of the gate position even if no truncation of the signal occurs. With- sitivity function as follows
in close vicinity of the center frequency of inspection x0, this rela- 0.80
tionship is more or less linear
Normalized Sensitivity, F [rad]

CGV point
uðtw ; xÞ  u0 þ u1 ðx  x0 Þ; ð4Þ
where u0 and u1 are linear regression coefficients. 0.70
It is advantageous to introduce the so-called ‘‘true’’ phase angle
U as the interception point of the local linear regression with the
ordinate axis (x = 0) 0.60
u ¼ u0  u1 x0 ; ð5Þ
which is independent of the gate position [28]. Assuming that the
center of the gate position in Eq. (3) is equal to the group time of 0.50
0 1 2 3 4 5
arrival at the center frequency
  Normalized Frequency, ξ
U ¼ x sg ðxÞ  sp ðxÞ z; ð6Þ
Fig. 6. Normalized sensitivity function versus the normalized frequency for steel.
1836
40
30
20
Phase Angle [deg]
10
0
-10
-20
-30
-40
0 10 20 30 40 50 60
Temperature [°C]
Fig. 7. Measured absolute and true phase angle variations as functions of
temperature in a steel pipe (D = 200 mm, d = 12.5 mm, z = 330 mm and f = 120 kHz).
z flux density B0 in the conductive specimen below it. A spiral coil
U  F 0 :
d driven by an alternating current Ic is placed between the specimen
This approximation illustrates not only the easiness of inverting
the measured true phase angle into the average wall thickness, but
it also assures the thermal stability of the measurement without
external temperature compensation.
As an example, Fig. 7 shows measured absolute and true phase
angles versus temperature in a steel pipe (D = 200 mm,
d = 12.5 mm, z = 330 mm and f = 120 kHz). Over a 90 °C tempera-
ture range the total variation of the true phase angle is less than
±5°, far less than the ±30° variation of the absolute phase, which
is due to the temperature dependence of the shear wave velocity
in steel. Since the effect of temperature on the Poisson coefficient
is negligible, the sensitivity function F(n) is temperature indepen-
dent to a very good approximation. Therefore, a change in temper-
ature only causes a shift of the non-dimensional parameter n due
to its dependence on cs, which however does not result in a signif-
icant change in the value of F due to the plateau region around the
CGV point. Therefore, in the vicinity of the CGV point the true
phase angle is effectively independent of temperature.
For small wall thickness variations Dd << d, the change in the
true phase angle can be approximated as
z Dd
DU  F 0 : ð17Þ
d d
In typical applications d/z  30, so that 1% change in wall thick-
ness causes about 13° change in the true phase angle, which is
fairly easily measurable.
3. Optimized EMATs for guided wave tomography
Earlier we demonstrated that in the frequency  thickness
range of interest for monitoring purposes (fd  0.8–1.4 MHz mm),
the decreasing group velocity of the S0 mode approaches the
increasing group velocity of the A0 mode and their inherent disper-
sion makes them increasingly difficult to separate by mere time
gating as they partially overlap. Therefore, it is essential that trans-
mitters and receivers be used that are sensitive to either the S0 or
the A0 mode only, but not to both. Efficient suppression of the un-
wanted mode is particularly important since, although these two
modes can propagate without coupling in a wall of uniform thick-
ness, they become coupled when either of them scatters on a cor-
rosion/erosion defect or other structural features present in the
inspected component.
Theoretically, such mode selectivity can be achieved by various
means such as symmetric/asymmetric excitation/sensing, spatial
frequency separation depending on their significantly different
P.B. Nagy et al. / Ultrasonics 54 (2014) 1832–1841
phase velocities and therefore wavelengths, even when their group
absolute phase velocity is almost identical, or exploiting their different polariza-
true phase tion characteristics. A critical element of A0 mode GWT, especially
in the previously described CGV regime where the S0 mode be-
comes slower and more dispersive, is the use of EMATs that are
optimized for suppressing the spurious symmetric mode. The opti-
mized EMAT must exhibit a high A0/S0 sensitivity ratio to allow the
separation of the direct and, especially, the higher-order A0 arriv-
als. In order to facilitate tomographic reconstruction of transmis-
sion data in all possible directions on plate-like structures and
the use of helical modes on tube-like structures, the optimized
EMAT must also exhibit omni-directional transmitting and receiv-
70 80 90 100
ing characteristics.
The proposed configuration of the optimized EMAT is based on
the standard radially polarized normal-incidence shear wave
EMAT design shown in Fig. 8a that is routinely used for spot-by-
spot wall thickness measurements. A cylindrical permanent or
electro-magnet is used to produce an essentially normal magnetic
ð16Þ
and the magnet. The alternating magnetic field produced by this
primary coil current generates a secondary eddy current loop in
a thin surface layer of the specimen. The eddy current loop is
essentially a mirror image of the coil except that the secondary
current runs opposite to the primary one. The moving charge car-
riers experience the so-called Lorentz force that acts normal to
both the magnetic flux lines and the velocity direction of the mov-
ing electrons. This force gives a radial momentum to the electrons
that is then transferred to the solid lattice via thermal collisions
resulting in radially polarized shear wave radiation normal to the
surface.
When the magnet diameter is comparable or smaller than the
EMAT coil the bias magnetic flux density exhibits an inward radial
component tangential to the surface therefore the Lorentz force
will also have a weaker normal component that generates an addi-
tional axially polarized longitudinal wave through the thickness of
the specimen. Such dual-mode operation is considered spurious
therefore usually a large enough magnet is used to assure strictly
normal magnetic bias and thereby essentially pure shear mode
generation. Alternatively a small-diameter but tall magnet can be
used for dominantly longitudinal mode generation as shown in
Fig. 8b. The EMAT can work essentially the same wave in reception.
cylindrical spiral cylindrical spiral
magnet coil magnet coil
N N
S S
Ic Ic
B0 B0
radialy polarized axially polarized
shear wave longitudinal wave
(a) (b)
Fig. 8. Basic designs of (a) radially polarized normal-incidence shear wave EMAT
with mostly normal bias flux density and (b) axially polarized normal-incidence
longitudinal wave EMAT with mostly radial bias flux density.
 P.B. Nagy et al. / Ultrasonics 54 (2014) 1832–1841
The possibility of optimizing EMATs for single-mode operation
was analyzed by Wilcox et al. [31]. They considered the Lorentz
force transduction mechanism only and separately investigated
the modal content produced by the normal and tangential compo-
nents of a general inclined surface traction force. Of course the Lor-
entz force is a body force, but it acts only in a thin surface layer
determined by the electromagnetic skin depth. This skin depth is
typically much smaller than the acoustic wavelength, therefore
the Lorentz body force can be accurately modeled by a surface trac-
tion. Wilcox et al. were primarily interested in S0 mode guided
wave inspection, therefore they optimized their EMAT configura-
tion by suppressing the A0 mode [31], which is the opposite of
what we are interested in. Still, their analysis clearly indicates that
it is possible to enhance mode selectivity by controlling the geo-
metrical features of EMATs. Our purpose in this section is to illus-
trate that such geometrical optimization can be used for
suppressing the S0 mode in A0 mode guided wave inspection even
more efficiently due to a fortuitous consequence of the familiar
Poisson effect in the material.
It should be mentioned that in ferromagnetic materials, in addi-
tion to the above described effect of Lorentz forces, two other
transduction mechanisms, namely magnetostriction and magneti-
zation forces are also significant [32]. Recent findings indicated
that the effect of magnetostriction is generally less significant than
previously suggested [33]. Together with the usually dominant
Lorentz forces, these two other transduction mechanism should
be taken into consideration in modeling ultrasonic guided wave
generation and detection in ferromagnetic plates and pipes. How-
ever, for the immediate purposes of the current study, the impor-
tant point is that the inclination angle of the effective surface
traction can be optimized to suppress the unwanted S0 mode and
the absolute magnitude of the traction force is less significant.
While the magnetization force is a body force and magnetostric-
tion is an eigenstrain, just like the Lorentz body force, they both
act only in a thin surface layer determined by the electromagnetic
skin depth below the otherwise traction-free surface. Therefore, all
three transduction mechanisms can be combined into a combina-
tion of tangential and normal traction forces the ratio of which
controls the inclination angle of the resultant traction force relative
to the surface. We will demonstrate that the suppression of the un-
wanted S0 mode is achieved by controlling this inclination angle,
therefore for our current purposes a simplified model containing
only the Lorentz force mechanism will suffice while the other
two transduction mechanisms, namely the magnetization force
and magnetostriction, will be neglected.
3.1. Traction force optimization
The basic EMAT configuration shown in Fig. 8 essentially gener-
ates a ring of axisymmetric traction cone. The ratio between the
tangential (radial) and normal components of the traction force de-
pends on the inclination angle h of the traction cone relative to the
surface of the specimen. This configuration is identical to the one
considered by Wilcox et al. earlier [31]. When placed on a thick
plate, the alternatingly expanding/contracting ring produces two
normal-incidence ‘‘bulk’’ waves, namely a radially polarized shear
wave and a longitudinal wave, through the thickness of the speci-
men. When such an EMAT is placed on a thin plate, the same sur-
face traction distribution generates outward running cylindrically
diverging guided modes as it is illustrated schematically in Fig. 9.
Below the lowest cut-off frequency of higher-order modes, only
the two fundamental modes, i.e., the S0 and A0 modes are gener-
ated. Their amplitude ratio A0/S0 can be controlled within certain
limits by changing the angle of inclination h between the surface
traction produced by the Lorentz force.
1837
n
T
θ
S0 A0 r A0 S0
traction cone
Fig. 9. Schematic illustration of how a ring of axisymmetric tangential tractions
produced by the Lorentz force generates cylindrically diverging guided waves in a
thin plate.
As an example, Fig. 10 shows the A0/S0 ratio as a function of the
inclination angle in a steel plate for three different fre-
quency  thickness products. These results were obtained with
the COMSOL MultiphysicsÒ finite element simulation software
using a simple axisymmetric model. In this example, the outer
diameter of the traction ring was assumed to be equal to the plate
thickness d. Naturally, the absolute value of the A0/S0 ratio depends
on how these two different modes are compared to each other.
However, since the displacement distribution throughout the
thickness of the plate is constant for a given fd product, the specific
way the two modes are compared has no effect on the optimization
itself. Arguably, the most rigorous definition of the A0/S0 ratio is
based on the ratio of the generated outgoing power flow carried
by the modes. In numerical simulations, the ratio of the out-of-
plane and in-plane vibration displacement in the middle plane of
the plate can be chosen, a definition that exploits the different
symmetry properties of the A0 and S0 modes. Because the transduc-
tion mechanism considered here is limited to a very thin surface
layer, we arbitrarily chose the total surface particle displacement
produced by the two modes to express their ratio. However, this
choice does not affect the shape of the angular-dependence exhib-
ited by the A0/S0 ratio. It should be also mentioned that, in order to
characterize the overall suppression of the S0 mode, we should use
the combined sensitivities of the transmitting and receiving trans-
ducers. Assuming the same dependence on inclination angle for
both transduction mechanisms due to reciprocity, the A0/S0 ratio
will be doubled in dB.
In the case of pure tangential traction (h = 0°) both S0 and A0
modes are generated due to the resulting in-plane extension and
out-of-plane bending, respectively. The A0/S0 ratio first increases
with the inclination angle because of stronger bending caused by
the increasing normal component of the traction. Above h  20°
50
fd
[MHz mm]
40 1.0
1.2
A0/S0 Ratio [dB]
1.4
30
20
10
tangential normal
traction traction
0
0 15 30 45 60 75 90
Inclination Angle of the Traction Force, θ [deg]
Fig. 10. A0/S0 ratio as a function of the inclination angle of the Lorentz force in a
steel plate for three different frequency  thickness products (one-way
transduction).
1838 P.B. Nagy et al. / Ultrasonics 54 (2014) 1832–1841

inclination angle, depending on the frequency  thickness product, 1

Surface Displacement [a. u.]

the A0/S0 ratio goes through a peak and starts to decrease. In this 0º
range the thinner the plate the higher the A0/S0 ratio is because 90º
0.5
of the decreasing flexural stiffness of the plate compared to its less
affected in-plane stiffness.
0

3.2. S0 phase reversal

-0.5
The question arises whether the observed peak in the A0/S0 ratio
is caused by a maximum in the efficiency of the A0 mode genera- -1
tion or by a minimum in that of the S0 mode. Clearly, the first sce- 60 80 100 120 140 160 180 200 220 240
nario would be highly preferable since it would also represent a
Time [μs]
badly needed increase in the absolute sensitivity of the EMATs.
Unfortunately, that is not the case. Fig. 11 shows the surface dis- Fig. 12. Tangential surface displacement pulses produced by the S0 mode at
placement amplitudes of the A0 and S0 modes as functions of the tangential (h = 0°) and normal (h = 90°) traction force orientation.
inclination angle at the CGV point. Clearly, the A0/S0 ratio peak at
h  25° is caused by suppressing the fundamental symmetric
mode. n
What causes this minimum? Close inspection of the generated
P P
wave forms reveals that complete phase reversal occurs in the re- T T θ
gion where the S0 mode goes through its minimum. Fig. 12 shows
tangential surface displacement pulses produced by the S0 mode at Q Q
r

d
tangential (h = 0°) and normal (h = 90°) orientations of the generat-
ing traction force. The center frequency of the exciting toneburst
was chosen to coincide with the CGV point, therefore the S0 pulse 2a
shape exhibits strong dispersion, but still it is clear that the polar-
ities of the two pulses are opposite. It is interesting, that the two (a) general surface traction
pulses are not exactly the inverses of each other since the shape
of their envelopes are different. This is because at normal traction n n
orientation (h = 90°) the high-frequency components of the tone- P/2 P/2
burst are more efficient in generating the S0 mode and these Q/2 Q/2
high-frequency components exhibit stronger dispersion than the vA vA vA vA
low-frequency components of the signal. It should be emphasized r r
that this phase reversal does not occur in the output signal of the
receiver when identical transducers are used for transmitting and Q/2 Q/2 P/2 P/2
receiving purposes since the two phase reversals occur at the same (b) asymmetric mode
fd product.
The next obvious question is what causes this phase reversal? A
n n
closer look at the transduction mechanism of these low-frequency
Lamb modes indicates that the phase reversal is a subtle conse- Q/2 Q/2 P/2 P/2
quence of the familiar Poisson effect. Fig. 13a shows a schematic
uS uS uS uS
diagram of an axisymmetric traction cone T acting on the top sur- r r
face of a plate as it was previously illustrated in Fig. 9. The relative
strength of the tangential Q = cos(h) and normal P = sin(h) compo- Q/2 Q/2 P/2 P/2
nents of the surface traction are controlled by the angle of inclina-
tion h. Fig. 13b and c shows how such a general excitation can be (c) symmetric mode
decomposed into the sum of asymmetric and symmetric excita-
Fig. 13. Decomposition of (a) a general surface traction cone into (b) asymmetric
tions following the technique used by Achenbach [34]. and (c) symmetric excitations.

Let us consider asymmetric excitation first. Tangential traction

Surface Displacement Amplitude [dB]

40
f d = 1.4 MHz mm forces expand the top surface and contract the bottom surface,
which produces a bending deformation that results in positive
30 out-of-plane displacement vA of the plate. Normal traction forces
simply push the plate upwards, i.e., they also result in positive
out-of-plane displacement vA. Since both the tangential and nor-
20
mal components of the traction force produce outgoing flexural
waves with the same polarity, regardless of their ratio, they will al-
10 ways reinforce rather than weaken each other’s effect.
A0 Now, let us consider symmetric excitation. Tangential traction
S0
forces expand both the top and bottom surfaces, which produces
0 radial expansion of the plate that in turn results in a positive in-
0 15 30 45 60 75 90
plane displacement uS of the plate. Normal traction forces increase
Inclination Angle of the Traction Force, θ [deg] the separation between the top and bottom surfaces, i.e., they pro-
Fig. 11. Surface displacement amplitudes of the A0 and S0 modes as functions of the duce a positive strain in the out-of-plane n direction. However, due
inclination angle at the CGV point (one-way transduction). to the Poisson effect, this deformation causes a contraction in the
 P.B. Nagy et al. / Ultrasonics 54 (2014) 1832–1841 1839

in-plane direction and results in negative in-plane displacement uS.
Since the tangential and normal components of the traction force
produce outgoing flexural waves with opposite polarity, they will
weaken rather than reinforce each other’s effect. Consequently,
the S0 mode vanishes at a given inclination angle h of the traction
force. The exact angle when this cancellation occurs depends on
Poisson’s ratio, the ratio between the diameter of the traction cone
on the surface and the plate thickness, as well as on the fre-
quency  thickness product. We can conclude that the efficiency
of the suppression as well as the exact inclination angle where it
is most effective depends on various factors, but suppression of
the S0 mode is always possible as it is a direct consequence of
the Poisson effect.
3.3. Magnetic field shaping
In the case of EMATs, the orientation of the traction force can be
readily controlled by shaping the magnetic field distribution. As an
example, Fig. 14a shows the magnetic field distribution produced
by a cylindrical permanent magnet of equal diameter and height
in air. The magnetic force lines are superimposed to indicate the lo-
cal direction of the field. Of course, when the magnet is placed on a
ferromagnetic plate of high magnetic permeability most of the
magnetic flux is confined in the plate so that the force lines enter
the plate at almost normal angle and become more or less parallel
to the surface inside the plate. Still, for demonstration purposes,
we can assume that the conducting plate is nonmagnetic so that
the static bias field remains undistorted inside the specimen. The
effective traction vector T acting on the surface of the specimen
is proportional to the Lorentz force F acting on the material of a
to the surface. Closer to the perimeter of the magnet, the field turns
away from the axis and at around 70% of the radius the angle is
optimum (h  30°) for maximum suppression of the S0 mode.
Fig. 16a shows an example of the transmitted signal on a steel
pipe between two regular EMATs separated by z = 350 mm dis-
tance (the outer pipe diameter was Do = 200 mm and the wall
thickness was d = 12.5 mm). The center frequency of the signal
was f = 113 kHz so that the frequency  thickness product was
fd  1.4 MHz mm, i.e., at the CGV point. Under such circumstances,
the sought A0 mode is quite fast and its envelope is relatively short
because of its negligible group velocity dispersion. The spurious S0
mode has slowed down considerably relative to its low-frequency
asymptotic value, though it is still faster than the A0 mode, but its
envelope is relatively long because of its significant group velocity
dispersion. The separation of the two fundamental modes is prob-
lematic since their amplitude ratio is only A0/S0  5 or 14 dB so
that the numerous helical modes following the direct arrivals
would be irrevocably mixed. Fig. 16b shows an example of the
transmitted signal on the same steel pipe between two enhanced
EMATs separated by the same z = 350 mm at the same inspection
frequency f = 113 kHz. In this case the separation of the two funda-
mental modes is much easier since their amplitude ratio is as high
as A0/S0  60 or 36 dB. Needless to say, that the inclination angle of
the magnetic flux lines is changing over the effective width of the
90
s/R
0.1
normal
0.2
traction
Polar Angle, θ [deg]

0.4
unit length of the channel carrying the total eddy current Ie and 60
F = Ie  B0, where B0 is the bias magnetic flux density. The eddy
current is running in the azimuthal direction and the Lorentz force
is always perpendicular to the local direction of the bias magnetic
field as illustrated in Fig. 14b. This fact can be exploited to maxi- tangential
30
mize the crucial A0/S0 ratio by controlling the direction of the mag- traction
netic bias field of the magnet inside the specimen.
Fig. 15 shows the polar angle between the direction of the mag-
netic flux density vector and the surface normal, which is equal to
the inclination angle h between the surface and the Lorentz force as 0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
a function of the radial position r at three different stand-off dis-
tances s from the face of the magnet. For easier comparison, both Normalized Radial Position, r/R
r and s were normalized to the radius R of the cylindrical magnet. Fig. 15. Polar angle of the magnetic flux vector as a function of the normalized
Close to the axis of the magnet, the magnetic field is strictly axial radial position at three different normalized stand-off distances from the face of the
therefore the polar angle is zero and the Lorentz force is tangential cylindrical magnet.

cylindrical
magnet

Ic
F Ie
θ θ B0

conducting
specimen

(a) (b)
Fig. 14. Schematic illustrations of (a) the magnetic flux density distribution produced by a cylindrical magnet of equal diameter and height in air and (b) the Lorentz force F
acting normal to the local direction of the bias magnetic flux density vector B0.
1840 P.B. Nagy et al. / Ultrasonics 54 (2014) 1832–1841
1 angle to wall thickness variations peaks exactly at the CGV point.
Amplitude [a.u.]
S0
0
A0
-1
60 80 100 120 140 160
Time [μs]
(a)
1
Amplitude [a.u.]
S0
0 of the material that can be exploited by optimizing the angle of
A0
-1
60 80 100 120 140 160
Time [μs]
(b)
Fig. 16. Examples of the transmitted signals on a steel pipe between two regular (a)
and enhanced (b) EMATs separated by z = 350 mm distance and operated close to
the CGV point.
generated eddy current loop, therefore the expected improvement
in the A0/S0 ratio is necessarily lower than the peak values calcu-
lated earlier using finite element simulations based on traction
forces of well defined angle of inclination.
Finally, it should be mentioned that steering the magnetic field
for better S0 mode suppression usually has only a slight effect (1–
2 dB) on the amplitude of the A0 signal. Compared to normal orien-
tation, oblique orientation of the same bias magnetic flux density
would slightly increase the A0 amplitude (see Fig. 11). However,
depending on how steering of the bias magnetic field is achieved,
the flux density is usually reduced, therefore the overall effect on
the A0 amplitude is negligible. Steering the magnetic field also af-
fects the pulse shape of the A0 signal by slightly changing the effec-
tive diameter of the EMAT relative to the wavelength at the center
frequency, which is readily visible in Fig. 16.
4. Conclusions
A novel ultrasonic guided wave inspection technique was pro-
posed for the monitoring of corrosion and erosion in pipes and
other thin-wall structures based on the fundamental flexural Lamb
mode. The inspection frequency is chosen so that the group veloc-
ity of the fundamental flexural mode is essentially constant
throughout the wall thickness range of interest. This means that
the time of arrival of A0 pulses remains constant independent of
wall thickness variations regardless whether they follow the direct
path between the transmitter and the receiver or a longer helical
path circumnavigating the pipe once or more. Since the phase
velocity of the A0 mode is still fairly strongly dispersive in the
CGV region, the sought wall thickness information is embedded
in the true phase of the received signals. The true phase affects
the shape of the received toneburst rather than its time of arrival,
therefore can be easily distinguished from not only inevitable
amplitude variations caused, for example, by variations in external
loading, but also from absolute phase variations caused by temper-
ature variations. It was found that the sensitivity of the true phase
This means not only that it is as high as it can get, but also that
it is essentially constant, which are both advantageous from a tech-
nical point of view. As a consequence, the crucial issue of temper-
ature correction becomes especially simple in the CGV region,
which is particularly beneficial when higher-order helical modes
are also exploited for tomography. Since the group velocity of the
A0 mode is essentially the same as the bulk shear velocity of the
material, over a wide range of wall thickness, the temperature
coefficient of the group velocity is the same as that of the bulk
shear velocity.
One disadvantage of working at such a relatively high inspec-
tion frequency (fd  1.4 MHz mm) is that the faster S0 mode be-
comes slower and more dispersive, therefore also more difficult
to distinguish from the A0 mode. We proposed a novel method
for suppressing the unwanted S0 mode based on the Poisson effect
inclination of the electromagnetic traction force produced by the
EMATs used for both generation and detection. Although, for the
sake of simplicity, this method was explained in terms of the Lor-
entz force, other transduction effects, such as magnetostrictive
strain and magnetization force, present in ferromagnetic materials
could be readily included in the optimization. We showed that the
suppression of the S0 mode can be maximized by adjusting the an-
gle of inclination of the effective electromagnetic traction force to
allow a cancellation between the effects of its tangential and verti-
cal components.
References
[1] J.L. Rose, D. Jiao, J. Spanner, Ultrasonic guided wave NDE for piping, Mat. Eval.
54 (1996) 1310–1313.
[2] M.J.S. Lowe, D.N. Alleyne, P. Cawley, Defect detection in pipes using guided
waves, Ultrasonics 36 (1998) 147–154.
[3] D.N. Alleyne, B. Pavlakovic, M.J.S. Lowe, P. Cawley, Rapid long-range inspection
of chemical plant pipework using guided waves, Insight 43 (2001) 93–101.
[4] J.L. Rose, A baseline and vision of ultrasonic guided wave inspection potential,
J. Press. Ves. Techn. – Trans. ASME 124 (2002) 273–282.
[5] P. Cawley, M.J.S. Lowe, D.N. Alleyne, B. Pavlakovic, P. Wilcox, Practical long
range guided wave testing: applications to pipes and rail, Mat. Eval. 61 (2003)
66–74.
[6] D.N. Alleyne, B. Pavlakovic, M.J.S. Lowe, P. Cawley, Rapid, long range inspection
of chemical plant pipework using guided waves, Key Eng. Mat. 270 (2004)
434–441.
[7] D.N. Alleyne, P. Cawley, Long range propagation of Lamb waves in chemical
plant pipework, Mat. Eval. 55 (1997) 504–508.
[8] A. Demma, P. Cawley, M. Lowe, B. Pavlakovic, The effect of bends on the
propagation of guided waves in pipes, J. Press. Vess. Techn. 127 (2005) 328–
335.
[9] R. Long, M. Lowe, P. Cawley, Attenuation characteristics of the fundamental
modes that propagate in buried iron water pipes, Ultrasonics 41 (2003) 509–
519.
[10] D.N. Alleyne, P. Cawley, The excitation of Lamb waves in pipes using dry-
coupled piezoelectric transducers, J. Nondestr. Eval. 15 (1996) 11–20.
[11] A. Demma, P. Cawley, M. Lowe, A.G. Roosenbrand, B. Pavlakovic, The reflection
of guided waves from notches in pipes: a guide for interpreting corrosion
measurements, NDT&E Int. 37 (2004) 167–180.
[12] W. Zhu, J.L. Rose, J.N. Barshinger, V.S. Agarvala, Ultrasonic guided wave NDT for
hidden corrosion detection, Res. Nondestr. Eval. 10 (1998) 205–225.
[13] M.K. Chang, R.R. Chang, C.M. Shu, The application of ultrasonic attenuation
measurements to estimate corrosion in pipeline at support structures, Mat.
Eval. 60 (2002) 631–634.
[14] F. Jenot, M. Ouaftouh, M. Duquennoy, M. Ourak, Corrosion thickness gauging in
plates using Lamb wave group velocity measurements, Meas. Sci. Techn. 12
(2001) 1287–1293.
[15] W. Luo, J.L. Rose, Lamb wave thickness measurement potential with angle
beam and normal beam excitation, Mat. Eval. 62 (2004) 860–866.
[16] M.J. Quarry, J.L. Rose, Multimode guided wave inspection of piping using comb
transducers, Mat. Eval. 57 (1999) 1089–1090.
[17] H.J. Shin, J.L. Rose, Guided waves by axisymmetric and non-axisymmetric
surface loading on hollow cylinders, Ultrasonics 37 (1999) 355–363.
[18] W. Zhu, An FEM simulation for guided elastic wave generation and reflection
in hollow cylinders with corrosion defects, J. Press. Vess. Techn. 124 (2002)
108–117.
[19] P. Cawley, F. Cegla, M. Stone, Corrosion monitoring strategies-choice between
area and point measurements, J. Nondestr. Eval. 32 (2013) 156–163.
 P.B. Nagy et al. / Ultrasonics 54 (2014) 1832–1841
[20] D. Jansenand, D. Hutchins, Lamb wave tomography, in: IEEE 1990 Ultrasonics
Symposium, 1990, pp. 1017–1020.
[21] K.R. Leonard, M.K. Hinders, Lamb wave tomography of pipe-like structures,
Ultrasonics 43 (2005) 574–583.
[22] F. Yan, R.L. Royer, J.L. Rose, Ultrasonic guided wave imaging techniques in
structural health monitoring, J. Intel. Mat. Syst. Str. 21 (2010) 377–384.
[23] A. Volker, J. Bloom, Experimental results of guided wave travel time
tomography, Rev. Prog. Quant. Nondestruct., Eval. (30) (2011) 215–222.
[24] P. Huthwaite, F. Simonetti, High-resolution guided wave tomography, Wave
Motion 50 (2013) 979–993.
[25] J. Li, J.L. Rose, Natural beam focusing of non-axisymmetric guided waves in
large-diameter pipes, Ultrasonics 44 (2006) 35–45.
[26] P. Huthwaite, R. Ribichini, P. Cawley, M.J.S. Lowe, Mode selection for corrosion
detection in pipes and vessels via guided wave tomography, IEEE Trans. Ultras.
Ferr. Freq. Contr. 60 (2013) 1165–1177.
[27] G. Instanes, M. Toppe, B. Lakshminarayan, P.B. Nagy, Corrosion and erosion
monitoring of pipes by an ultrasonic guided wave method, Adv. Ultrason.
Methods Mater. Struct. Insp., ISTE (2006) 115–157.
1841
[28] W.L. Pilant, L. Knopoff, F. Schwab, Transmission and reflection of surface waves
at a corner, J. Geophys. Res. 69 (1964) 291–297.
[29] J.D. Achenbach, Wave Propagation in Elastic Solids, North-Holland Publishing,
1973.
[30] G. Instanes, A. Pedersen, M. Toppe, P.B. Nagy, Constant group velocity
ultrasonic guided wave inspection for corrosion and erosion monitoring in
pipes, Rev. Prog. Quant. Nondestruct., Eval. 28 (2009) 1386–1393.
[31] P.D. Wilcox, M.J.S. Lowe, P. Cawley, The excitation and detection of Lamb
waves with planar coil Electromagnetic Acoustic Transducers, IEEE Trans.
Ultras. Ferr. Freq. Contr. 52 (2005) 2370–2383.
[32] M. Hirao, H. Ogi, EMATs for Science and Industry: Noncontacting Ultrasonic
Measurements, Kluwer Academic Publishers, 2003.
[33] R. Ribichini, P.B. Nagy, H. Ogi, The impact of magnetostriction on the
transduction of normal bias field EMATs, NDT&E Int. 51 (2012) 8–15.
[34] J.D. Achenbach, Reciprocity in Elastodynamics, Cambridge University Press,
2003.