Investigationof Eddy Current Sensor Modelsfor Sharp Transitional Areasand Geometries

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 7

ASNT Fall Conference and Quality Testing Show 2009 [Columbus, OH, October 2009]: pp 279-285.

© Copyright 2009, 2011,


American Society for Nondestructive Testing, Columbus, OH.

Investigation of Eddy Current Sensor Models for Sharp Transitional


Areas and Geometries

Evgueni Iordanov Todorov

EWI
1250 Arthur E. Adams Drive, Columbus, Ohio 43221
(614) 688-5268; fax (614) 688-5001; e-mail evgueni_todorov@ewi.org

INTRODUCTION
Several methods are currently employed for nondestructive evaluation (NDE) in service and during manufacture of steel
structures and joints with sharp transitions. Magnetic particle inspection (MPI) and liquid-penetrant inspection (LPI) are the
methods mostly used for steel-welded structures and joints. Both MPI and LPI require removal of paint or other coatings.
Both methods also produce excessive number of false indications in the sharp transitional areas (STA) due to the geometry.
For many structures, radiography (RT) is not a practical alternative due to accessibility requirements.

MPI, LPI, and RT techniques provide the length of the flaw only. The flaw depth is often important in assessing structural
integrity. Ultrasonic (UT) conventional and advanced [phased-array (PA)] techniques are used extensively for length and
depth sizing. However, UT performance for small surface flaws detection and sizing may not be adequate. Coupling media is
required on the surface (e.g., water, grease) and coatings may have to be removed to conduct reliable UT.

The application of array eddy current probes and techniques has shown very promising results in the detection of small
fatigue cracks at butt welds using advanced signal presentation and processing [1]. The capabilities of eddy current
techniques to detect and size fatigue cracks and other flaws in welds under coating are well known. An extensive state-of-the-
art review was conducted. The review focused on identification of possible solutions for eddy current inspection of STA. The
main findings from the literature review were that single plus- or cross-point probes and alternative current field measurement
(ACFM) techniques were the standard for inspection of difficult to access areas (weld toe and uneven weld crown) [2-5]. The
techniques with single probes, however, are slow, counterproductive, and difficult to automate for complex shapes such as
weld crown and toe regions, jet engine slots, steering rack teeth, pistons, valves, fitting flanges, etc. Most of the eddy current
techniques using array probes were for flat areas and areas with smooth transitions made of conductive nonmagnetic or
slightly magnetic material such as austenitic stainless steel, nickel based, titanium and aluminum alloys [6-9].

Major problem for eddy current inspection with current rigid and flexible array probes is the inability to maintain constant
and minimal distance (lift off) and tilt between the probe and inspection surface during scanning of complex and variable
shape surfaces. In this project, possible probe configurations were investigated through advanced 3D finite-element computer
simulation for eddy inspection of STA. One of the objectives was to conduct optimization through computer modeling of
various promising probe designs and configurations for typical sharp transitional areas (e.g., weld toe) that could provide
increased sensitivity and immunity to noise during scanning.

RESULTS AND DISCUSSION

Transformer Probe Basic Configurations


Four configurations were simulated and analyzed at the initial stage of the study (Figure 1) at frequency of 20 kHz. A typical
transmit-receive (transformer) or reflection probe consists of one transmitter and one receiver coil as shown in Figure 1,
Configuration 1. A typical surface semicircular flaw 2 mm (length) × 1 mm (depth/height) was simulated in specimens made
of carbon steel with initial relative magnetic permeability 40 and electrical conductivity 3 mega siemens per meter (MS/m).
For Configuration 1 and 2, the direction of probe scanning was aligned and symmetrical with the flaw length. A flaw signal
(red color in Figure 1) was simulated for Configuration 1 used later for reference to assess the effect of all subsequent probe
virtual modifications.

279
conductivity 3 Mega Siemens per meter (MS/m). For Configuration 1 and 2, the direction of probe scanning was
ASNT Fall Conference and Quality Testing Show 2009 [Columbus, OH, October 2009]: pp 279-285. © Copyright 2009, 2011,
aligned and symmetrical with the flaw length. A flaw signal (red color in Figure 1) was simulated for Configuration
American Society for Nondestructive Testing, Columbus, OH.
1 used later for reference to assess the effect of all subsequent probe virtual modifications.

Configuration 1 Flaw
Transmitter Configuration 2
Flaw
Receiver

Eddy Current
Scanning Density
Distribution

Weld
Crown

Configuration 3

Configuration 4

Figure 1: Simulation of transformer probe signals at 20 kHz for four probe configurations.
Figure 1: Simulation of transformer probe signals at 20 kHz for four probe configurations.
The eddy current density distribution for a flaw position at the middle between the two coils where the flaw signal reached
maximum is shown
The eddy in Figure
current density1 distribution
for each of the for four
a flawconfigurations.
position at theThe lift off
middle (LO) signals
between the twofor each
coils configuration
where were also
the flaw signal
simulated and maximum
reached oriented from right to
is shown in left (standard
Figure phase
1 for each ofrotation
the fourcalibration)
configurations.to allow
Thefor
liftcomparison of flaw
off (LO) signals forsignal
each phases.
The effect of field concentration
configuration is shownand
were also simulated in Configuration 2. Thetosignal
oriented from right amplitude
left (standard wasrotation
phase almost 4× stronger compared
calibration) to
to allow for
Configuration 1. Further, the probe orientation during scanning was skewed with respect to the flaw length atsignal
comparison of flaw signal phases. The effect of field concentration is shown in Configuration 2. The 45 degree as
shownamplitude
in Configuration
was almost3, Figure
4 times1. The phase
stronger changed to
compared slightly and the amplitude
Configuration 1. Further,dropped
the probe significantly
orientation(~4 times)
during due to
scanning
the skew compared to Configuration 2. The last Configuration 4 (STA – e.g., weld toe) where half of the probe is above the
was skewed with respect to the flaw length at 45 degree as shown in Configuration 3, Figure 1. The phase changed
parentslightly
metal and andhalf
the is on the weld
amplitude crownsignificantly
dropped did not have(~4significant
times) dueeffect
to theonskew
the signal compared
compared to Configuration
to Configuration 3 (flat
2. The last
specimen and skewed probe). As expected, the addition of field concentrator significantly improved the sensitivity whereas
Configuration 4 (STA – e.g., weld toe) where half of the probe is above the parent metal and half is on the weld
the coil skew was a detriment. In addition, parametric (one coil acting as transmitter and receiver) probe performance was
crown did not have significant effect on the signal compared to Configuration 3 (flat specimen and skewed probe).
As expected, the addition of field concentrator significantly improved the sensitivity whereas the coil skew was a
also assessed for the same four cases. The spatial resolution of the parametric probe is expected to be better than transformer
detriment. In addition, parametric (one coil acting as transmitter and receiver) probe performance was also assessed
configuration due to the smaller probe foot print. However, the transformer configuration was selected due to better
for the same four cases. The spatial resolution of the parametric probe is expected to be better than transformer
sensitivity, lower noise
configuration dueand better
to the thermal
smaller probestability.
foot print. However, the transformer configuration was selected due to better

Magnetic Flux Density Disturbance


sensitivity, lower noise and better thermal stability.

Further improvement of probe performance was achieved by reshaping the magnetic field concentrators so that higher density
of eddy currents was produced at the STA (e.g., weld toe) area. Ferrite cores were also introduced in the coils for some
cases. One of the advantages of the general-purpose finite-element electromagnetic software is that various characteristics
of the electromagnetic field may be obtained assisting the probe optimization process. For one of the probe configurations,
the magnetic flux density disturbance components along the three coordinate axes in space were simulated. The magnetic
flux density disturbance caused by the flaw at the weld toe or STA above the metal surface was analyzed to identify the best
orientation(s) of the receiver elements of the sensors.

280
ASNT Fall Conference and Quality Testing Show 2009 [Columbus, OH, October 2009]: pp 279-285. © Copyright 2009, 2011,
American Society for Nondestructive Testing, Columbus, OH.

Two components of the magnetic flux density disturbances caused by the presence of the flaw at the STA are shown in
Figure 2. The plot in Figure 2(a) illustrates the magnetic flux density component perpendicular or normal to the parent metal
surface. It has two characteristic peaks at each of the flaw ends on the surface. The plot in Figure 2(b) shows the magnetic
flux density component parallel to the weld toe. It has three characteristic peaks – two at each end of the flaw and one
maximum peak at the middle of the flaw length. The analysis of the magnetic flux density components indicated that the
component parallel to the weld was greater by ~20% than the component perpendicular to the parent metal at 1.6 mm from
the weld toe. Flaw signals were simulated for the receiver elements (coils) sensitive to the component normal to the parent
metal. Additional sensors sensitive to the component parallel to the weld toe line or scanning direction might be used for flaw
detection in STA.
Amplitude of (By - By)

Amplitude of (Bz - Bz)

Scan distance Scan distance

(a) (b)
Figure 2: Simulation of magnetic flux density disturbance from a flaw at weld toe on weld surface at
20 kHz perpendicular to parent metal surface (a) and parallel to weld toe (b).

Eddy Current Density Maps


Eddy current density maps were simulated at various frequencies and several probe configurations with and without flaws at
the weld toe or STA to investigate the surface flaw effect on eddy current distribution. Typical surface eddy current density
plots are illustrated in Figure 3 for weld toe STA. The simulated maps without flaw (top plots) on the metal surface indicated
that the STA acted as eddy current concentrator at optimal ferrite cores length and field concentrators.

The configuration with longer ferrite cores in Figure 3(a) will be sensitive to flaws parallel and perpendicular to the STA
line because part of eddy currents are flowing perpendicular and another part almost parallel to the STA. For example,
flaws parallel and transverse to the weld toe might be detectable. The configuration in Figure 3(b) is mostly sensitive to
flaws parallel to the STA or weld toe. The presence of flaw parallel to the STA as shown in Figure 3 (bottom plots) which
is most difficult to detect with conventional probes disrupts the eddy current contours for both configurations indicating
good detectability with optimized probe design. The density of eddy currents is high at the flaw ends indicated earlier by the
presence of peaks in the magnetic flux density components shown in Figure 2.

281
flaws at the weld toe or STA to investigate the surface flaw effect on eddy current distribution. Typical surface eddy
current density plots are illustrated in Figure 3 for weld toe STA. The simulated maps without flaw (top plots) on
ASNT Fall Conference
the metal and Quality
surface indicated thatTesting
the STAShow 2009as[Columbus,
acted OH, concentrator
eddy current October 2009]:at
ppoptimal
279-285.ferrite
© Copyright
cores 2009,
length2011,
and field
American Society for Nondestructive Testing, Columbus, OH.
concentrators.

Crown

Parent Metal

Increased Eddy
Flaw at Current Density
Weld Toe

(a) Long ferrite cores. (b) Short ferrite cores.


(a) Long Ferrite Cores (b) Short Ferrite Cores

Figure 3: Simulation of eddy current density (view on arrow “A”)


Figure 3: Simulation of eddy current density (view on arrow “A”) on weld ontoe,
weldcrown,
toe, crown, and metal
and parent
parent metal surface with (bottom) and without (top) flaw at
surface with (bottom) and without (top) flaw at 5 kHz. 5 kHz.

FlawThe
Signal Simulation
configuration with longer ferrite cores in Figure 3(a) will be sensitive to flaws parallel and perpendicular to the
SignalSTAsimulation plots were
line because part ofgenerated and analyzed
eddy currents for all
are flowing optimizationand
perpendicular parameter
another combinations. A sample
part almost parallel to theisSTA.
shownForin
Figureexample,
4 illustrating
flawsthe effectand
parallel of changing
transversefrequency.
to the weld The
toeLO signals
might (not shown)The
be detectable. were simulated and
configuration oriented
in Figure from
3(b) is right
to leftmostly
(standard phase to
sensitive rotation calibration)
flaws parallel toSTA
to the alloworfor comparison
weld toe. The of flaw signal
presence phases.
of flaw Thetosensor
parallel is designed
the STA as shownininsuch a
way that differential
Figure 3 (bottom and absolute
plots) whichsignals fordifficult
is most better sensitivity, noise
to detect with reduction and
conventional detection
probes of various
disrupts the eddyflaw conditions
current are
contours
provided. Theconfigurations
for both signal plots in indicating
Figure 4 indicated that the phase
good detectability withangle between
optimized thedesign.
probe flaw signal
Theand horizontal
density of eddyaxis (LO) of
currents is the
plot decreased with the decrease of frequency from 5 to 1 kHz. Based on the simulation results (Figure 4), the frequency of
1 kHz would provide better separation between the flaw and LO signals, however, the sensitivity would also drop with the
decrease of frequency (Figure 5).

Signal from the element over the parent metal at lower frequency of 1 kHz is slightly stronger than the signal from the
identical sensor element over the weld crown. The signal difference is explained with the flaw (normal to the parent metal
at weld toe) being tilted toward the sensor element over the parent metal at 12.5 degrees with respect to the bisector of the
STA angle (225 degrees) through the metal. This indicates the possibility of flaw tilt estimation from the differential signal
generated with the optimized STA sensor.

282
detection of various flaw conditions are provided. The signal plots in Figure 4 indicated that the phase angle
between the flaw signal and horizontal axis (LO) of the plot decreased with the decrease of frequency from 5 to 1
ASNT Based
kHz. Fall Conference and Qualityresults
on the simulation Testing(Figure
Show 2009
4), [Columbus, OH, of
the frequency October
1 kHz2009]:
would ppprovide
279-285.better
© Copyright 2009, between
separation 2011,
American
the Society
flaw and LOfor Nondestructive
signals, however,Testing, Columbus,would
the sensitivity OH. also drop with the decrease of frequency (Figure 5).

(a) (b)

Figure 4. Simulated absolute and differential signals at 5 (a) and 1 kHz (b).

Signal from the element over the parent metal at lower frequency of 1 kHz is slightly stronger than the signal from
the identical sensor element over the weld crown. The signal difference is explained with the flaw (normal to the
parent metal at weld toe) being tilted towards the sensor element over the parent metal at 12.5 degrees with respect
to the bisector of the STA angle (225 degrees) through the metal. This indicates the possibility of flaw tilt
estimation from the differential signal generated with the optimized STA sensor.
(a) (b)
(a) (b)
Figure 4.
Figure 4: Simulated
Simulated absolute
absolute and
and differential
differential signals
signals at
at 55 (a)
(a) and
and 11 kHz
kHz (b).
(b).
Optimization Results
Optimization Results
Signal from the element over the parent metal at lower frequency of 1 kHz is slightly stronger than the signal from
To obtain quantitative
sensorestimates of thethe
optimization process, two common objectives were defined. The (normal
first objective
to the was to
To obtain
the quantitative
identical estimates
element over of the optimization
weld crown. process,
The signal two common
difference objectives with
is explained werethedefined.
flaw The first
maximize
parentthe ratio
at in decibel (dB) oftilted
the flaw to reference amplitude
element refereed later asmetal
relative amplitude “Arel”:
objective was
metal to maximize
weld the
toe) being ratio in decibel
towards (dB) of
the sensorthe flaw to reference amplitude
over the parent refereed
at 12.5 later
degrees as relative
with respect
amplitude
to “Arel”:
the bisector of the STA angle (225 degrees) through the metal. This indicates the possibility of flaw tilt
estimation from the differential signal generated with the optimized STA sensor.
, (1)
(1)
Optimization Results
where: AS – flaw signal amplitude simulated for each combination of parameters and ARS - amplitude of flaw
where: AS – flaw
reference signal
signal foramplitude simulated
Configuration for each
1 in Figure 1. combination of parameters and ARS - amplitude of flaw reference signal
for Configuration 1 in Figure
To obtain quantitative 1.
estimates of the optimization process, two common objectives were defined. The first
objective
The secondwas to maximize
objective the ratio
was also in decibel
to maximize the(dB)
flawoftothe
LOflaw to reference
amplitude in dB amplitude refereed
referred later later as
as “AtoL” relative
ratio:
The amplitude
second objective
“Arel”:was also to maximize the flaw to LO amplitude in dB referred later as “AtoL” ratio:

, (1)
, (2)
(2)
where: AS – flaw signal amplitude simulated for each combination of parameters and ARS - amplitude of flaw
where: ALO - amplitude of LO signal for LO from 0 to 0.025 mm. The LO was considered to be the major source of noise for
reference
where: ALO signal for Configuration
- amplitude 1 inforFigure
of LO signal 1. 0 to 0.025 mm. The LO was considered to be the major source of
LO from
noise for this application.
this application.
The second objective was also to maximize the flaw to LO amplitude in dB referred later as “AtoL” ratio:
Further, the ability of any given parameter combination to produce a phase angle between the flaw and LO signals
Further, thetoability
close of anyforgiven
90 degrees best parameter
separationcombination to produce
and noise reduction wasaalso
phase angle between the flaw and LO signals close to
considered.
90 degrees for best separation and noise reduction was also considered.
A summary of representative portion of the optimization matrix is shown in Table 1.
A summary of representative portion of the optimization matrix is shown in Table 1.
Table 1. Summary of optimization matrix.

Combi- Coils Skew Coil Ferrite Coil


nation (deg) Core Separation Frequency (kHz)
5 45 - D1 20
6 45 L D1 20
7 45 L D1 5
8 45 L D1 1
9 45 Sh D1 1
10 45 Sh D1 5
11 45 Sh D1 20
12 34.7 Sh D2 20
13 34.7 Sh D2 5
14 34.7 Sh D2 1
283
Abbreviations: L-Long ferrite, Sh-Short ferrite, D1-Distance 1, D2-Distance 2.
Further, the ability of any given parameter combination to produce a phase angle between the flaw and LO signals
ASNT
closeFall Conference
to 90 degrees and Quality
for best Testing Show
separation and 2009
noise[Columbus,
reduction OH,
wasOctober 2009]: pp 279-285. © Copyright 2009, 2011,
also considered.
American Society for Nondestructive Testing, Columbus, OH.

A summary of representative portion of the optimization matrix is shown in Table 1.


Table 1: Summary of Optimization Matrix
Table 1. Summary of optimization matrix.
Combi- Coils Skew Coil
nation (deg) Skew CoilCoil
Combi- Coils Ferrite Core
Ferrite Separation
Coil Frequency (kHz)
5nation 45 (deg) - Core D1
Separation 20
Frequency (kHz)
5 45 - D1 20
6 45 L D1 20
6 45 L D1 20
7 7 45 45 LL D1
D1 55
8 8 45 45 LL D1
D1 11
9 9 45 45 Sh
Sh D1
D1 11
10 45 Sh D1 5
10 45 Sh D1 5
11 45 Sh D1 20
11 12 45 34.7 Sh
Sh D1 D2 2020
12 13 34.734.7 Sh
Sh D2
D2 20 5
13 14 34.734.7 Sh
Sh D2
D2 5 1
14Abbreviations:
34.7 Sh ferrite, D1-Distance
L-Long ferrite, Sh-Short D2 1, D2-Distance 2.1
Abbreviations: L-Long ferrite, Sh-Short ferrite, D1-Distance 1, D2-Distance 2.

20 kHz 5 kHz 1 kHz

Figure
Figure 5:
5. Plot
Plotofofobjective
objectivefunctions
functionsfor
forcombinations
combinationsfrom
fromTable
Table11grouped
groupedby
byfrequency.
frequency.

Optimization resultsresults
Optimization for parameter combinations
for parameter from Table
combinations from 1Table
are shown in Figure
1 are shown 5. The 5.
in Figure objective functions
The objective functions Arel,and
Arel, AtoL,
phaseAtoL,
angleand
between the flaw and LO signals “Phase wrt LO” are presented for three frequencies – 20, 5 and 1 kHz.
phase angle between the flaw and LO signals “Phase wrt LO” are presented for 3 frequencies – 20, 5 and A common
trend 1iskHz.
reduction of Arel, AtoL and phase angle with reduction of frequency. For a given frequency, the highest
A common trend is reduction of Arel, AtoL and phase angle with reduction of frequency. For a given sensitivity
is achieved with long ferrites in the coils (combination 6, 7 and 8). The AtoL ratio, however, is the smallest for these three
conditions. The gain in sensitivity (Arel) is smaller than the decrease in AtoL which indicates increased sensitivity of these
configurations to surface roughness and irregularities. The configurations from 9 to 14 with short ferrites are less sensitive to
the LO variations at the expense of reduced sensitivity to flaws in STA. The change of distance between sensors in the range
from D1 to D2 for pairs of combinations 11-12, 10-13 and 9-14 at each frequency had very little or no effect on sensitivity
and phase angle. The increased frequency may seem beneficial for the increase in overall sensitivity and AtoL ratio if the flaw
signal phase with respect to LO is not considered. The reduction of frequency and associated reduction of phase angle from
approximately 140 to 100 degrees is actually improving the sensor capabilities to separate the flaw signals from the noise
produced by surface irregularities as discussed above (Figure 4). Another important benefit of frequency reduction is increased
depth of penetration. The depth of penetration obtained from the computer models for this typical steel, sensor configuration
and three frequencies is as follows: 0.3 mm for 20 kHz, 0.5 mm for 5 kHz and 0.7 mm for 1 kHz. A reliable flaw depth sizing
of surface-breaking flaws could be expected up to depths of 0.6, 1 and 1.4 mm for each frequency, respectively.

284
for this typical steel, sensor configuration and three frequencies is as follows: 0.3 mm for 20 kHz, 0.5 mm for 5 kHz
and 0.7 mm for 1 kHz. A reliable flaw depth sizing of surface-breaking flaws could be expected up to depths of 0.6,
1ASNT Fallmm
and 1.4 Conference
for eachand Quality Testing
frequency, Show 2009 [Columbus, OH, October 2009]: pp 279-285. © Copyright 2009, 2011,
respectively.
American Society for Nondestructive Testing, Columbus, OH.

Flexible Backing
Material
Weld Crown

Scan Direction

Figure 6: Artist’s concept illustrating scanning of weld STA with optimized eddy current sensor.
Figure 6. Artist’s concept illustrating scanning of weld STA with optimized eddy current sensor.
OneOneartist’s concept
artist’s concerning
concept concerningthe the
design of the
design sensor
of the andand
sensor implementation
implementation waswas
generated andand
generated is shown in Figure
is shown 6. The
in Figure
scan6.direction
The scan direction is shown with thick red arrow. The concept in Figure 6 used flexible backing material (e.g., etc.)
is shown with thick red arrow. The concept in Figure 6 used flexible backing material (e.g., rubber, foam,
to hold the foam
rubber, sensoretc.)
elements
to holdfirmly pressed
the sensor and conforming
elements to theand
firmly pressed inspection surface.
conforming to the inspection surface.

CONCLUSIONS
TheCONCLUSIONS
following conclusions can be drawn from the work on this project:
• Advanced simulation indicated that it was feasible to inspect STA.
• Novel probe design
The following was created
conclusions anddrawn
can be optimization
from thematrix wasthis
work on employed
project:to optimize the probe through computer modeling.
• Probe specifications
! Advanced were generated.
simulation indicated that it was feasible to inspect STA.
! Novel probe design was created and optimization matrix was employed to optimize the probe through computer
Future work will focus on selecting target applications, optimizing probe final configuration, fabricating a prototype probe
modeling.
and demonstrating
! the capabilities
Probe specifications for the intended application.
were generated.

ACNOWLEDGMENTS
Future work will focus on selecting target applications, optimizing probe final configuration, fabricating a prototype
This work was performed under contract with the State of Ohio with financial support from the Ohio Department of
probe and demonstrating the capabilities for the intended application.
Development. The content reflects the views of EWI and does not necessarily reflect the views of the State of Ohio,
Department of Development.
ACNOWLEDGMENTS
REFERENCES
This work was performed under contract with the State of Ohio with financial support from the Ohio Department of
1. Todorov, E.I., The
Development. W.C.content
Mohr and M.G.the
reflects Lozev,
views“Detection
of EWI andand Sizing
does of Fatigue Cracks
not necessarily in Steel
reflect the viewsWelds
of thewith Advanced
State of Ohio, Eddy
Current Techniques,” Review
Department of Development. of Progress in Quantitative Nondestructive Evaluation, Vol. 27, edited by D.O. Thompson
and D.E. Chimenti, Golden, Colorado, pp. 1058-1065, July 22-27, 2007.
2. Jessop, T.J., “Advances in Nondestructive Testing of Weldments,” Metal Progress, Vol. 123, No. 4, pp. 62-65. March
1983.
3. Calvert, J., “A Practical Guide to Eddy Current Inspection of Welded Structures,” Insight, Vol. 48, No. 8, pp. 492-499.
August 2006.
4. “The ABC’s of Eddy Current Weld Inspection,” Welding Journal, Vol. 82, No. 5, pp. 26-30. May 2003.
5. Devries, D. and J. Cox, “Inspecting Welds on Complex Tube Forms. Manual Eddy Current Options,” TPJ-The Tube &
Pipe Journal, Vol. 14, No. 3, pp. 42-44. April-May 2003.
6. Hurley, D.C., K.H. Hedengren, P.J. Howard, W.P. Kornrumpf, G.E. Sutton and J.D. Young, “An Eddy Current Array
System for Aircraft Engine Inspections,” Review of Progress in Quantitative Nondestructive Evaluation, Vol. 13A,
Brunswick, Maine; United States; 1-6 Aug. 1993, pp. 1111-1118. 1994.
7. Uesaka, M., T. Nakanishi, K. Miya, H. Komatsu, K. Aoki and K. Kasai, “Micro Eddy Current Testing by Micro Magnetic
Sensor Array,” IEEE Transactions on Magnetics, Vol. 31, No. 1, pp. 870-876, January 1995.
8. Goldfine, N. and K. Walrath, “Nondestructive Coating Characterization and Flaw Detection, Using a Conformable
Meandering Winding Magnetometer (MWM),” Fifth EPRI Steam Turbine/Generator Workshop, Lake Buena Vista,
Florida, August 1, 1997.
9. Gilles-Pascaud, C., J.M. Decitre, F. Vacher, C. Fermon, M. Pannetier and G. Cattiaux, “Eddy Current Flexible Probes for
Complex Geometries,” Review of Progress in Quantitative Nondestructive Evaluation; Volume 25A, pp. 399-406. 2006.

285

You might also like