~ UTTERW0RTH

I N E M A N N
NDT&E International, Vol. 28, No. 2, pp. 103-112, 1995
Copyright 1995 Elsevier Science Ltd
Printed in Great Britain. All rights reserved
0963-8695(94)00009-3 0963-8695/95 $10.00 + 0.00

An eddy current array i n s t r u m e n t

for application on ferritic welds
A. McNab and J. Thomson
Department of Electronic and Electrical Engineering, University of Strathclyde,
Glasgow G1 lXW, UK

Received 13 October 1994

This paper describes a multi-element, eddy current array for the local testing of
ferritic specimens including welds. A novel transient method is used to extract coil
impedance parameters, and has the advantage of fast measurement, digital
compatibility and a minimum of front-end components; a useful feature when large
element numbers are involved. The problem of displaying lift-off and flaw
information from an array is solved by using a novel 'volcano display', which provides
a simple two-dimensional image of conditions under the array. However, to produce
this display, the burden of interpreting the coil element responses must be left to
the instrument. The resident analysis software contains a number of operating modes
which provide options for single element and multi-element operation. The array
has been applied to the detection of cracks in the HAZ of a weld, where the large
surface area of the probe permits a coarser scanning pattern to be used whilst
providing a high probability of detection due to the number of elements deployed.

Keywords: eddy current testing, impedance measurement, arrays

Eddy current inspection is an attractive option for crack position of the array. The advantage of array probes in
detection due to its high sensitivity and its ability to work extending coverage when scanning for both eddy cur-
over a significant thickness of non-metallic protective rent testing and other electromagnetic methods, such
coating; an important requirement in the offshore as ACFM, has been identified by a number of
industry. The electronic nature of the data gathering researchers 11-3J. However, when deploying significant
process interfaces well with modern computer systems, numbers of sensors, the problem of both exciting and
and the results obtained can be readily displayed and receiving the required signals places a burden on the
stored for further use. There are problems, however, when electronic system. The EC instrument makes use of a
testing welded structures which arise from the variable novel measurement technique which permits rapid
nature of the material in the heat-affected zone (HAZ) of scanning of the eddy current coils, but provides a
the weld, setting up spurious indications, and from the simplified design of exciter hardware.
limited coverage obtained when scanning with a single
An additional problem, arising from the deployment of
probe.
significant numbers of probes, is the display of relevant
This paper describes an eddy current instrument for data. This may require that two parameters from each
localized manual testing which uses a two-dimensional element be represented, such as lift-off from the surface
array of eddy current sensors to improve the coverage and flaw signal amplitude. Multiple, linear, signal traces,
when scanning, and which incorporates comprehensive one for each parameter and coil can be used 131, but for
software to interpret the received eddy current responses the portable unit described here, a special 'volcano
and identify spurious responses. An additional benefit display' has been developed, which can provide
comes from enhanced flaw detection, since more than meaningful information on lift-off and flaw for a large
one sensor element will intercept the flaw at any given number of elements.

103
 A. McNab and J. Thomson

Rs 2 It consists of a charged capacitor, connected to the coil

oA through an FET switch to set up an initial decaying
1 oscillation: Mode 1 in Figure 3b. Changes in this main
Vs
oscillation due to changes in material properties can be
Rb monitored by the 'current switching' technique described
above, and since the resulting voltage spike is very
RM sensitive to these changes, very little voltage amplification
OB is required. In the application considered here, the testing
of ferritic material for cracks, two physical parameters
Figure I Transient circuit for current measurement were of interest. These were coil lift-off above the test
surface, and the presence of the crack discontinuity. By
monitoring the peak height using a fast peak detector,
and the length of another portion of this transient
Eddy current measurement system
waveform after the spike which proved sensitive to coil
The detection of cracks and other material anomalies lift-off(Mode 3 in Figure 3b), the two physical parameters
was carried out by monitoring changes in the impedance could be separated.
of the coil associated with each array element. In a
The measurement system was calibrated using a test
conventional system, the test coils are excited with a
block with slots of varying depth, and response diagrams
constant current source, and as the coil impedance
established as shown in Figure 4. In the first of these,
changes in magnitude and phase, proportional changes
Figure 4a, the measured second-mode peak voltage (Y
occur in the coil voltage. This can be measured using
axis) is plotted against the third-mode delay (X axis) at
two synchronous phase detectors in quadrature t4]. For
different lift-off heights and over different depths of
an array of coils, separate sets of phase detectors can be
artificial flaw (fine slots). In Figure 4b, the actual
used for each coil or they can be multiplexed onto one
impedance changes taking place in terms of magnitude
or more sets of multiplexers. Whilst the former
and phase are displayed. It should be noted that lift-off
configuration results in maximum scanning speed, it is
produces large phase changes on the ferritic steel, whilst
expensive in hardware when large coil-element numbers
are used.
Taking these factors into account, an alternative system (a) Typical response (no capacitance)
was devised which had fewer analogue components to t0
simplify the array front end hardware, and which had
greater compatibility with measurements taken using
digital techniques. The basis of the method relies on
establishing an initial oscillation within the coil, and then, 5
as shown in Figure 1, deflecting the coil current at a
suitable instant through a high resistive load. As the coil
tries to maintain this current, a voltage spike is produced
> 0

which is proportional to the instantaneous current. From 0

this measurement taken at two points on the waveform, 2 ,\ o ,/ ,o
together with a knowledge of the applied voltage taken Time (l~s)
at the same point across the coil, the coil impedance can
easily be calculated t~]. -5
Since the above technique relies on the accurate
(b) Typicalresponse (added capacitance)
measurement of coil current, both modelling and tO
experimental studies were conducted to assess the affects
of coil self-capacitance. Figure 2b shows the coil
capacitance causing a decaying oscillatory response
compared to the exponentially decaying spike with zero
capacitance in Figure 2a. The measurement of the first
peak output voltage of the oscillation follows the current
waveform as the switching point is varied, the effect of
increasing coil capacitance being to absorb some of the
coil energy, thereby reducing the magnitude of the peak.
;o
A simple scaling factor can be used to compensate for Time (Its) ~ lw
this effect.
-5 --
In the array probe itself, a modified form of excitation
was used. As opposed to a constant AC source, each coil Figure 2 Transient voltage waveforms from the current measuring
has its own transient excitation circuit as in Figure 3a. circuit

104
 An eddy current array instrument

+5V their construction be as simple as possible. Thus, in order

to allow high packing densities, absolute coils were used
Sb in preference to transmit/receive or differential coils.
However, to obtain a good separation between the effects
of lift-off and flaw presence on the transient measured
parameters, the coils had to be constructed to a high
standard of tolerance. For the transient measurement
system, the measured parameters of second-mode peak
I
_VI_ 0
]) IL and third-mode delay must be as independent as possible,
so that the peak is sensitive to a flaw (crack) and the
delay to lift-off. Proper coil design can achieve this, since
Firing circuit lift-off introduces a large percentage change in resistance
and hence a large phase change in the coil's complex
5V impedance. On the other hand, by varying the mutual

OV '~
i-- ---Sf~s -- - i '

No flaw - - - - - 2.0ram flaw

~ - - mode 3 . . . . 0.5mm flaw . . . . . 4.0mm flaw
5V mode l ~[[ [
- - 1.0ram flaw . . . . . . . . Constant LO
lS = 0.1mm lift-off

"" -: "', (a)

~" 5 - ..~ :,,
OV
1
o
:"

-A \%.~' X "x
",~

o 4
O 2:,< ".."MZ~- L_*
-5V 3s ~ . . - - . -'%..-..
F i r i n g p u l s e a n d i n d u c t o r voltage O
8 3
Figure 3 Array-element transient excitation and associated coil (/3

voltage waveforms I I I I I-~ 12s

0.4 0.6 0.8 1.0 1.2

T h i r d mode delay (Its)

a change in flaw depth at constant lift-offproduces mainly

a magnitude change. The resulting diagram has very 19

(b )
similar characteristics to the transient parameter plot,
where second-mode peak height and third-mode delay
o
are analogous to the impedance magnitude and phase, 17
respectively. These effects are again reflected in the 0
is ' ~ . ~ " ~ : ~ .
response diagrams of Figure 4c, which were derived using
the measured impedance data in conjunction with a 15
transient modelling program, based on Laplace trans-
I I I I I
forms, which can predict the coil response described 8O 82 84 86 88
previouslytSj. The circuit of Figure 5 is used to represent
the combination of firing circuit components, coil and
Phase()
test piece in the model. 6 -

A comparison of the measured and calculated transient ,_.~ ~ ~ . ~ (c)

response diagrams shows the general shape is similar and ,=~ 5 -- " " - ."

o
that there is good agreement in the second-mode peak
height changes. However, the third-mode delay variation 4-

due to lift-off is greater m the measured result; this arose

from uncertainties in the VMOS switch characteristics. o= 3-

Array probes m 2- I I -I ~i~'-" "

0.4 0.6 0.8 1.0 1.2
The coil e l e m e n t Third mode delay (its)

The design of the coil elements used in the array is Figure 4 (a) Plot of s e c o n d - m o d e peak versus third-mode delay
important in determining the overall performance of the for different lift-off values over artificial slots of varying depth; (b)
c o r r e s p o n d i n g coil i m p e d a n c e p l o t t e d a s m a g n i t u d e v e r s u s p h a s e
system. As the arrays were designed to deploy a u n d e r t h e s a m e c o n d i t i o n s of lift-off a n d flaw; (c) m o d e l l e d t r a n s i e n t
substantial number of elements, it was important that response parameters using the measured coil impedance values

105
 A. McNab and J. Thomson

,Js
tll I RF

1
- CL

RD ' ~ RDeff

Ren Vv LS
T Vcc

Where
Rea = FET on resistance
Reff = FET off resistance
R B = Rbi u II Ref f
Rdoa = diode on resistance
(FET drain-source diode)
Rdeff = diode off resistance
C L = coil self-capacitance
LS = coil self-inductance, LF = coupled test piece inductance
RLS = coil self-resistance, RF = coupled test piece resistance

Figure 5 Transientmodel of array coil element and associatedfiring circuit

coupling between the coil and the material through scanning pattern could be used. Two basic designs were
positioning it above the test surface, zero phase change produced; one with a fixed pattern of elements t63 and the
in the impedance can be produced when moving over a other with vertically mobile elements which could
crack tS]. These effects are exploited in the measurement conform to the contour of the test surface. In each case,
system. the number of elements deployed was 16, although the
measurement system could handle substantially more.
An experimental programme evaluated the most effective
combination of coil features related to frequency, physical In the fixed array, a four by four matrix of ferrite-cored
core dimensions, winding position on the core and the coil elements was mounted on a printed circuit board
detection area. A number of features were observed in and spaced 3.4 mm apart. This gave an area of coverage
this study which provided useful design information. It 20.4 mm by 20.4 ram. Whilst such a unit operated well
was noted that lift-off and flaw sensitivity fell off as the on substantially flat surfaces, the varying surface profile
core was shortened due to the magnetic field becoming associated with angled, welded pipe sections required a
less concentrated under the core. Winding position, too, different design. Although consideration was given to
was important as a means of varying mutual coupling; using a flexible PCB, it was felt that the degree of local
flaw sensitivity increased almost linearly the closer the surface variation in the vicinity of a capped weld dictated
coil came to the surface down to 1 mm and then decreased a wider range of vertical movement for each element than
slightly after that with an associated significant phase would be possible with this approach.
change in the impedance over a crack. The use of a
The second design, a surface-conforming array, is shown
magnetically shielded coil was also investigated and was
in Figure 6 and again comprised a four by four matrix
found to provide a larger crack response, but not to the
with a 6 mm element spacing. Each coil element is
extent that it merited the additional constructional
mounted within an individual plastic holder (Figure 7)
complexity. It was established that a satisfactory design
before being placed in the array housing, where it is held
for operation on steel had an operating frequency in the
in place by a spring. The total vertical movement
range 100 to 200 kHz, a ferrite core of length 18 mm, a
permitted is 13.5 mm. When designing the unit using this
core diameter of 1.7 mm, 60 coil-loops, and the winding
'bed-of-nails' approach, a number of factors have to be
positioned 2 mm from the end adjacent to the metal.
taken into account. These include:
(a) With 16 spring-mounted probes, attention has to be
Array designs
paid to the spring force constants. If they are too
In designing the arrays for the eddy current instrument strong then the pressure required to maintain the
(ECI), the intention was to provide a satisfactory area of array on the test piece becomes too great. In this case,
coverage under the probe head so that a fairly coarse long, extra-light springs made from phosphor-bronze

106
 An eddy current array instrument

magnetic coupling can occur through the spring of

the neighbour.
(d) The design of the array housing is important to ensure
that access can be gained when specific applications
are envisaged (Figure 8).
The final arrays produced proved to be stable in
operation and showed good consistency between results
taken at different spring compressions.

t, 64.0 mm Array instrumentation

Element pattern Cross section
The ECI was designed to control the operation of arrays
having a substantial number of elements, and at the same
Figure 6 Surface conforming array time to process the measured data to give flaw
information to the operator in as convenient a manner
as possible. The overall system is shown in Figure 9. Each
coil element has its own FET firing circuit, and the
scanning of the array proceeds in a raster fashion, with
only one element being active at a time to eliminate
cross-talk. The transient voltage output from each coil
is passed to an analogue multiplexer and then to a fast,
10ns peak-detector which measures'the second-mode
peak parameter. A zero-crossing detector coupled with
counters clocked at 100 MHz are used to measure the
third-mode delay parameter.
The instrument was designed to be PC compatible in
terms of hardware and software, and incorporates an
2- Intel 80286 processor together with a 80287 maths
coprocessor. The computer has three main tasks to
perform: (a) to control the firing of the array and the
multiplexing of the array data; (b) to analyse the array
data to extract lift-off and flaw parameters; (c) to display

il
these parameters. This latter task presented some
...... 6 i problems in that a conventional impedance plane display
was too complicated for the number of probes involved.
An LCD graphics display was therefore mounted on the
instrument (Figure 10), and used to show the condition
of the array probe. Each element was represented as a
circle, the width of whose circumference decreased with
--m . increasing lift-off, and whose centre became filled with
dots as flaw severity increased. This 'volcano display'
T
2

Figure 7 Array element holder (dimensions in mm)

were employed and their compression limited to half

their original length.
(b) The forces imposed on the array probe can cause
internal stressing of the element cores, resulting in a
measurable change in the coil impedance and
masking flaws. Careful mounting of the ferrite core
within its holder is necessary to ensure that it is
attached at one end only, thus minimizing its
constraint.
(c) Physical cross-talk between elements becomes
apparent if insufficient spacing is left between the top
of the measurement coil and an adjacent spring. When
a coil element is displaced relative to its neighbour, Figure 8 Array applied to a fusion butt-weld with weld cap

107
 A. McNab and J. Thomson

t
RS 232 Bus

Processor Board k-q Display Controller H

480 x 128 LCD
Display

ipeiin, o ector.ord Counter Board

& Crossover Digitisation
I

Figure 9 Block diagram of eddy current instrument (Eel)

relevant calibration curves which can be stored for future

reuse and comparison.
Within the ROM of the ECI unit there reside a number
of software modules which carry out a number of specific
tasks. These can be summarized as follows:
(1) Instrument initialization.
(2) PC communication.
(3) Auto calibration of array elements for variable lift-off.
The array is lifted on and off the test piece, as a
second-mode peak versus third-mode delay calibra-
tion curve is recorded.
(4) Display operations for arrays and single probes. In
addition to the volcano display, a single probe, X - Y
Figure 10 Eddy current instrument showing LCD 'volcano display of the measured parameters is useful for
display' setting up the system, since the performance of an
individual array element can be assessed.
(5) Array scanning and detection.
(6) Calibration of array firing pulses. The setting up of
presents crack-like flaws as a line of circled dots across
optimal firing pulses in relation to second-mode peak
the array aperture.
heights on and off the test piece.
(7) Code specific to the operational mode of the
Control and communication software instrument.
(8) Other utilities.
The control software is an important component of the
instrument and consists of two main bodies of code: a
communication/development program resident on a PC,
O p e r a t i o n a l m o d e s of t h e i n s t r u m e n t
and the ECI firmware resident in the instrument ROM.
The methods applied to the interpretation of parameter
Whilst the instrument can stand alone it can be connected
data obtained from the instrument have an important
to a PC for remote operation in a master/slave
influence on the reliability of flaw identification in
configuration. Commands can be sent to the ECI and
different test situations. With the array instrument, the
data returned to the PC, where a duplicate volcano
opportunities for operator-based assessment are restric-
display can be formed. Figure 11 shows a typical ECI
ted due to the very large amounts of information coming
display, formed on the PC and then passed as a screen
from each element of the array. The onus therefore lies
dump to a laser printer. The communication software
with the system software to make an intelligent analysis
within the PC consists of C code modules which instruct
of flaw characteristics and to display this information in
the ECI to change to different arrays and implement
a more readily assimilated form.
calibration procedures which set up optimum firing
pulses to measure second-mode peak and third-mode In this regard, the transient parameters of second-mode
delay parameters. It can also request the ECI to transfer peak and third-mode delay are analysed to extract the

108
 An eddy current array instrument

(a) physical conditions of lift-off from the surface and the

presence and severity of crack indications under each
array element. The analysis itself is essentially based on
the parameter plot of Figure 4a. Giveri that the diagram
remains fixed with time, the identification of the solid
lift-off curve for no-flaw establishes a base line reference,
below which a flaw vector will not occur. Thus for a
constant lift-off of 3 s say, a flaw will cause a parameter
vector to move away from the solid curves towards the
broken lines. Thus a region in the parameter plane can
be identified with flaw responses. However, the
assumption of constant response curves is not entirely
valid since material changes and long-term drift can cause
the base reference line to move in the plane. It is therefore
necessary to provide a facility for local recalibration of
the instrument.
To assist in setting up the system and for handling
different test situations a number of test or operational
modes were implemented. These can be listed as follows:

Single probe operation (Mode 1 )

This mode of operation is most suitable for development
purposes and for initial setting up procedures. It most
closely resembles a conventional vector display, except
that a parameter lift-offcurve showing second-mode peak
on the vertical axis versus third-mode delay on the
horizontal is plotted onto the LCD display rather than
impedance, Figure 12. The top left-hand corner is the
(b)
probe in intimate contact with the test piece, while the
dropping curve shows lift-off increasing until the probe
is in air beyond any level of detection. In Figure 13 this
display has been duplicated on the PC at a higher
resolution by the transfer of drawing interchange files
(DXF). Whilst the top diagram establishes the initial
lift-off reference curve, the second diagram indicates the
presence of a flaw at a lift-off of approximately 1 mm.
Subtraction of these responses removes the lift-off
response to reveal the flaw alone. Within the ECI, there
are two display modes; the first has a permanent
recording of the parameter vector with a manual clear,
whilst the second provides a continuous vector with
temporary sustain.
In the development of this mode, it became apparent that
for optimum measurement of parameters, the firing pulse
to measure the peak should be different from that used
to measure the third-mode delay. This greatly improves
performance by reducing the effects of cross-over between
the parameters due to saturation of the peak signal.
However, it does introduce a reduction in the scanning
frequency of the array elements since each must be fired
twice.

Array probe in air and analysed through a

look-up table (Mode 2)
Figure 11 Eel display transferred to PC. The array is shown testing
a butt-weld with weld cap (test piece 2) at t w o different rotations This is the simplest operating mode when using the
relative to the fatigue crack volcano display. With the probe held in air, each element

109
 A. McNab and J. Thomson

Probe double-fired for independent parameter

measurement (Mode 3)
This mode is a refinement of the previous one. In this
case, the probe is fired in air with a variable delay, to
reduce the second-mode peak almost to zero. As the
~ 1 1 1 1 1 1 1 probe moves towards the metal the peak increases
substantially. To set up the third-mode delay, the probe
is placed on the metal and the firing point adjusted until
(a) ECI Lift-off vector the peak is reduced to zero. As the probe is removed the
peak is cut off by the excitation circuit and does not
interfere with the third mode.
Each element of the array is then calibrated by lifting the

q probe from the metal to the air and generating lift-off

curves for the two parameters. These are checked using
Mode 1. Unlike the previous case, these parameters are
measured at different firing delays and the resulting
element curves used to detect flaws which appear as a
movement in parameters away from the curve in an
(b) ECI flaw vector at LO = 1.0mm upward direction. Compensation for a loss in sensitivity

~1111111
(c) ECI Lift-off vector with
(a) D X F Lift-off vector

superimposed flaw

Figure 12 Single probe operation (Mode 1) with the second-

mode peak (Y axis) plotted against third-mode delay (X axis) on
the LCD display. Display shows lift-off and flaw vectors

is balanced for a small second-mode peak, As the probe (b) D X F flaw vector at LO = 1.0mm
moves towards the material, the peak increases, and for
different levels of lift-off, the parameter pair of
second-mode peak and third-mode delay is recorded in
a look-up table. As the probe is scanned, the parameter
pair is measured, and if the second-mode peak is greater
than that stored plus a small noise margin, a flaw is
identified. It should be noted that over a ferritic weld
and its associated HAZ, the peak is slightly depressed
and hence spurious flaw indications are suppressed.
Whilst this method works well for lift-offs up to 3 mm,
the peak-response when fully on the steel is about 3.5 V.
This gives a good working range for the detection of
flaws at higher lift-off, but when the flaw depth exceeds
3 mm, the peak reaches the 5 V saturation level of the (c) D X F Lift-off vector with
A/D converter. Thus flaw depth indication beyond this superimposed flaw
point is not possible at zero lift-off. To obviate this
Figure 13 Single probe operation (Mode 1) with lift-off and flaw
problem, a double-firing technique using different vectors being displayed at high resolution through transfer of data
switching points can be used. to a PC

110
 An eddy current array instrument

NI Oimensions in mm
a flaw identification can be made without the need for
accurate calibration and recalibration.

Experimental assessment
The performance of the instrument was determined using
a set of welded test specimens both with and without a
weld cap. The parent metal was mild steel, since the
instrument was developed primarily as an array system
T for fatigue crack detection in welds made in this material.
In the first specimen (Test piece 1), the weld cap has been
Test Piece 1
removed and a fatigue crack introduced into the HAZ
of the weld as shown in Figure 14. Thus there are three
distinct material changes associated with this weld, the
NI Dimensions in mm
parent metal, the HAZ and the weld metal. The datum
lift-off response curve for the instrument was taken over
a good area of the parent metal, as shown in Figure ! 5,
with the responses being recorded using operational
Mode 3. A series of readings for the same lift-off
conditions, from 'intimate contact' to 'in air', was also
taken over the other weld zones; the values of lift used
were (0, 0.15, 0.3, 0.6, 0.9, 1.5, 'in air') mm. It can be seen
that the 5 mm flaw present within this specimen gives
rise to flaw parameter vectors (movement in flaw
T parameters away from the lift-off response curve) which
are of a substantial magnitude and have a clear angular
Test Piece 2 separation from that of the lift-off vector (movement in
measured parameters along the lift-off response curve) at
Figure 14 Mild steel, fusion-butt-weld specimens containing all values of absolute lift-off. In addition, the HAZ and
fatigue cracks
the weld metal exhibit lift-off responses which lie within
a workable noise margin from that produced over the
parent metal. The only notable divergence is for the weld
metal with zero lift-off. Here a response is produced which
due to probe lift-off can be made, so that the volcano gives a higher second-mode peak than that recorded for
display will show a filling-in of the circle for a given crack the datum response, together with an increase in
depth which is independent of lift-off. The reliability of third-mode delay. Whilst this is consistent with the
this detection scheme has been tested on welded test response expected from a very small flaw, it still remains
pieces where variable thicknesses of nonconducting shim within the noise tolerance imposed on any signal
have been used to simulate variable lift-off. It has been interpretation.
found that reliable detection can be obtained on samples
with and without the weld cap.

240 -- ~xx TM ~ Parent metal

Other modes of operation -- ~--~" ~ ~ HAZ
200 T . t "~, ~= ....... Weld
A numbcr of other modes have bcen assessed,each with { ';,'t, ,r --- ,law
its own advantages. For example, when testing on 80 - ~"~'~" / I .... Common LO
smooth, high quality surfaces at high detection
" 60 - \
sensitivities,the firing point of cach array clcmcnt can
bc adjustcd for optimum detection at a given lift-off
(Mode 4). However, under these circumstances, small
~ 20 -
variations in lift-offwill introduce substantialchanges in
detcction sensitivity. I I I I "~ IX,,
20 60 100 140 180
Another mode (Mode 5) was examined which exploits Third mode delay count (1 unit = 1 0 n s )
the dynamic nature of thc probe response as itisscanned.
W h e n a coil element encounters a crack, a plot of thc Figure 15 Transient parameter plot of lift-off and flaw responses
using operational mode 3, taken over the parent metal (lift-off),
parameter pair follows a characteristiclooping path. By heat-affected-zone (lift-off and flaw), and weld (lift-off) of test
tracking thc two parameters and recognizing this path, piece 1

111
 A. McNab and J. Thomson

typical coil element shows good agreement with

f
measured results, and provides an insight into factors
= 2o0 : '; Parent metal influencing the measurement technique. The system has
~" i" \ -.. ----Flaw particular value for an array based instrument as changes
80 I ~ "" ~.~. " Common LO in test piece parameters under each coil can be rapidly
~- ~,, 60 "~\'-, ""a measured. In addition, compensation for small variations
in coil impedance can be made by controlling the firing
,0 <>o\ I, pulse width to bring the different coil responses to a
predetermined datum. The array can therefore be scanned
at a reasonable speed over a surface, or can be applied
in a series of fixed placements.
20 60 100 140 180
Third mode delay count (1 unit = lOns) The ECI allows for independent operation or it can be
attached to a remote PC for data transfer and duplicate
Figure 16 Transient parameter plot of lift-off and flaw responses
display. A variety of probes ranging from a single element
using operational mode 3 for a weld containing a weld cap (test
piece 2) to a full array can be attached and test conditions
optimized using the most suitable mode of operation.
These operational modes range from a simple,
two-parameter display to one for multiple elements which
Test piece 2 was used to assess the performance of ECI uses a double-firing technique to assess separately the
when testing close to a weld cap. The test piece contained lift-off and the flaw responses. A further detection mode
a 4 mm deep, fatigue crack close to the toe of the weld. uses the dynamic nature of the flaw response during
Again similar responses to that from the previous sample scanning; by storing a running buffer of the measured
were observed (Figure 16), and although at higher lift-offs parameters, their local history can be viewed with the
the proximity of the weld cap tended to short out the aim of detecting the characteristic shape of the parameter
current-inducing flux at the weld toe, satisfactory results vector when traversing a flaw. This removed the need to
were obtained. Overall array performance was also make an absolute measurement and does not require
investigated using this test sample; the results obtained careful calibration.
with the surface conforming array, recorded on the PC
As to overall instrument performance, the primary
as a duplicate ECI display, are presented in Figure 11.
requirement of detecting fatigue cracks in mild steel
The array is shown oriented parallel to and at an angle
components was met with flaw indications being
to the crack, Figures 11a and 11b respectively. observed for flaws of 0.5 mm minimum depth and up to
a maximum lift-off of 4 mm. Detection in the toe of a
Conclusions weld is also possible with the array designs used to about
the same limits, although at higher lift-offs some
The eddy current instrument described in this paper
angulation of the array relative to the weld is required
provides a useful tool for the rapid local inspection of
for maximum response.
ferritic welds. Improvements over single coil operation
are obtained in scanning speed, and in the reliability of
flaw detection, because the electronically scanned array Acknowledgement
of coils permits a greater area of the material to be tested
This work was supported by the Marine Technology
and provides a 2D image of the flaw position. Although
Directorate and the SERC.
the 'volcano display' may give a simple representation of
array element responses, it is important to appreciate
that a higher degree of information processing is required References
and that a greater reliance must be placed on the flaw 1 Lugg, M.C. 'Rapid scanning with non-contacting ACFM array
interpretation software resident within the instrument. probes" in Review ~f Pro.qress 01 Quantitative N D E Vol 4A Plenum
Press, New York (1991) pp 1003-1010
Looking at the hardware component of the system, it is 2 Krampfner, Y.D. 'Flexible substrate eddy current coil arrays' in
Reviep" ~?/Proeress #1 Quantitative N D E Vol 7A Plenum Press, New
evident that a sensitive detection system has been
York (1988) pp 47t-478
developed which yields large transient voltage changes 3 Newton, K. anti Saumlerson, D.H. 'NDT research for the oil and
for small changes in complex impedance. An inherent gas industry' Brit J N D T 3 4 3 (1992) pp 123-128
signal gain is achieved which reduces the need for voltage 4 Libby, H.L Introduction to Electromagnetic Non-Destructive Test
Methods Wiley Interscience, New York (1971)
amplification with its attendant noise problems. The 5 McNab, A. and Thomson, J. 'A measurement technique for
technique is well suited to digital technology as the firing eddy-current arrays" IEEE Proc. 37 Pt A (1990) pp 147-154
pulse is numerically controlled, and the measured 6 MeNab, A. and Thomson, J. 'An eddy current instrument for fixed
position scanning' Proc. 12th World Conf. on N D T Vol 1, eds J.
parameters in the response are suited to immediate Boogard and G.M. van Dijk, Elsevier, Amsterdam (1989) pp
digitization. Modelling of the transient responses for a 358-360

112