Finite element prediction of

differential eddy current probe

signals from redO, deposits in
PWR steam generators
N. Ida, H. Hoshikawa and W. Lord

Finite element analysis techniques have been successfuly applied to the study of
eddy current NDT phenomena in axisymmetric test geometries Major applications of
such computer code have related to the design of probes and the gaining of physical
insight into just how electromagnetic fields interact with defects. This paper ~s
concerned with an additional application of the code, namely to the simulation of
practical testing situations which are too difficult and/or expensive to replicate in a
laboratory environment
Numerical experiments are described which simulate the differential eddy current
probe response to the build-up and chemical flushing of magnetite in the crevice gap
of a PWR steam generator unit The simulation results agree well with the only
experimental data available to the authors and lead to the conclusion that
conventional differential eddy current probes should be capable of characterizing
crevice gap conditions with respect to the presence of magnetite.

Keywords" eddy current testing, finite element analysis, magnetite, nuclear

reactors

Analysis of the electromagnetic fields associated with in an eddy current probe) sets up an alternating magnetic
eddy current NDT phenomena has to proceed from the field. The Maxwell-Faraday law of Equation (1) sum-
quasi-static form of Maxwelrs equations marizes many of Faraday's and Henry's observations
concerning electromagnetic induction and predicts that if
the alternating field of the eddy current probe interacts
f ff~'di = ff ~'ds (1) with a conducting test specimen` an alternating voltage
C S will be induced in the material giving rise to eddy currents.
Again by Equation (2), the eddy currents themselves set up
an alternating magnetic field which interacts with the
excitation probe field, resulting in a change of probe
c S
impedance in much the same manner as the load
impedance of a transformer is referred into the primary
fB'ds = 0 (3) winding,
$
Phenomonologically then, Equations (1) to (4) are very
and the constitutive relationship useful for explaining the basic physics of low frequency
electromagnetic field/defect interactions. Unfortunately,
one can only progress so far with them in developing
= (4) analytical expressions for, say, the probe coil impedance
given a particular defect shape. The reason for this lies in
where all the field variables E R H and ] and phasor the very nature of the non-destructive test itself Material
vectors because of the sinusoidal nature of eddy current properties, test geometries and the defects and anomalies
excitation, and/,t is a constant only if the material under in them tend not to be describable by simple (or even
test is non-ferromagnetic. complex) mathematical expressions. During the past
decade, researchers in the NDT and Applied Magnetic
The Maxwell-Ampere law of Equation (2) describes the Laboratory at Colorado State University have examined
physically observable fact that an alternating current (as the possibility of applying numerical analysis (speci-
0308-9126/85/060331-08
$3.00 1 985 Butterworth Et Co (Publishers) Ltd
NDT INTERNATIONAL. VOL 18. NO 6. DECEMBER 1985 331
fically, hmte element methods) to the naodelhng of
electromagnetic field/defect interactions m metalsl'l
Starting with acnve and resxdual magnetostatic leakage
field modelsN, the flmte element code has now been
extended to the study of both steady-state and pulsed eddy
current NDT phenomenal3-5) All of the early code was
limited to two-dtmensional or axisymmetnc test gec,
metnes, but recently the magnetostatic and steady-state
eddy current code has been converted to a full three-
& m e n s m n a l formN.
Applications of the eddy current code have centred on the
modelling of differential eddy current probe p h e n o m e n a
encountered m the testing of PWR steam generator tubing
particularly in relation to probe design and m developmg
an understanding of how probe fields interact with steam
generator components
The major purpose of this paper is to describe an
additional application of finite element m o d e l h n g
namely, the prediction of eddy current probe impedance
plane trajectories for testing situations that are too
difficult to replicate in the laboratory. Such studies are
particularly important in the nuclear industry, where the
very structure of the components themselves or the
operating condttions often preclude the acquismon of
reqmred experimental data.
One such s~tuation relates to the build-up of corrosion
products (primarily magnente, Fea04) in the crevice gaps
of PWR steam generators, as shown in Figure 1 Such
build-up can ultimately lead to tube denting reqmring the
affected tube to be plugged and resulting m an overall loss
of steam generator efficiency. Chemical flushing has been
suggested as a means to allevtate this problem, and the
Inconeltubes~ gener)r t
Mognetdem
crev,ce gop
Fig1 Schematm
d=agramofaF~VR
steamgenerator
332
quesuon naturally arises as to whether a con,~enuonal
differential eddy current probe ~s sensmve enough to
m o m t o r the process
Properties of magnetite
The numerical modelling of geometries containing
magnetite poses some unique problems. The first of these
is the complete lack of reliable data concerning
permeability and conductwity, especially tor corrosion
products that occur in PWR nuclear power plant steam
generators Thus, for example, Smith and WljnPl give a
value of 100 for the relative permabdny and 1428
~-1 C-1 m-i lor conductivity SnoeklSl gives a value ofT0
for relanve permeability and the conducnvity of magnemc
m Landott-Bornstem[ 91 is250 ~ - 1 C-1 m-1 These~alues.
all at room temperature, were obtained lbr Fe3(h crystals
( natural or synthetic)iS,91 or pressed materiallY1 and may or
may not be good esnmates for the actual material present
in steam generators. To compound the problem, corrosmn
products are a mixture of magnetite and other metals, salts
and oxides in a composition that differs from one
installation to another and ts dependent on the steam
generators constructmn and water chemistry
These reasons, and the fact that prehminary numerical
calculations indicated a lower permeability than the data
given above, prompted an experimental study of the
magnente properties
Two toroidal samples were prepared by a hot lsostanc
pressing process. One sample was made of pure magnentc
while the other had the following composition (b~y
weight)
74% Fe304
10% Cu (metalhc powder)
10% Cupric oxide
3% N1 (metallic powder)
2% Zn (metallic power)
I% NaC1 (powder)
This sample ts believed to resemble closely the actual
conditions m steam generators and the data obtained
were used in the numerical model. The pure magnetite
sample was tested for comparison with available data-
Both toro~ds were subjected to a DC soft magnenc matertal
testll01 using a Hysteresisgraph system to determine thexr
permeablhty The results are gwen in Figure 2, mdmating
a considerably lower permeabihty value for the pure
magneme sample in comparison to the literature The
higher permeabihty of the "blendeff magnetite can be
explained as the effectwe permeability o f a mixturetUl and
is mainly due to the higher permeability of the nickel,
For the measurement of conductivity, two cylindrical
samples of ldenncal materials were prepared. The
conductwity was measured using a network analyser The
blended magnetite sample produced a value of 1.6665
f~-x C-1 m-~ but no reliable measurement was possible
on the pure magnetite sample. This value, however,
indicates a conductivity two orders of magnitude lower
than that reported m the hterature for the pure
magnetite
The finite element model
The p a m a l differential equation governing eddy current
phenomena in two-dimensional geometries that include
NDT INTERNATIONAL. DECEMBER 1985
 1 The source current densityJ s and the magnetic vector
B2' potential A are phasor-vectors. Thus, harmonics are
absent in both the impressed and induced fields
2. Electrical conducnvity and magnetic permeability are
28 B
single valued in each element of the region. Thus, the
model is linear and cannot account for field depen-
dency of the permeability, although spatial variation in
24 /.t can be modelled.
3. Low frequency operanon The displacement currents
as well as volume or surface charges in Maxwelrs
equations are neglected
4. Eddy currents within the excitation coils are neglected.

i, This is equivalent to assuming that the AC resistance of

the eddy current coils is constant and equal to its DC
values.
These assumptions are certainly justified m eddy current
NDT applications where the frequency and excitation
levels are low
Equation (6) cannot be solved directly Instead, the
8 solution region is first discretized into a large number of
finite elements of some predetermined shape and the
equation formulated for the partiulcar element using any
4 of the number of general methods available for this
purposely51 The most appropriate method Is the use of an
energy-related funcnonal which is a statement of energy
I I I I ! balance in the solution region. The balance between
0 I0 20 BO 40 50 60 70 stored, dissipated and input energy can be written as
Mognetic field strength, H (Oe)
Fag. 2 Relative permeablhty curves for pure (A) and blended magnet)te (B) (7)
samples. (10e = 1000/47r A/m -1)

The solution now consists of finding a set of approxi-

conducting and magnetic materials can be written as
_
1 (xxA) = .T- 0 oO---i-
~ (5)
where /x, A , J and cr are the magnetic permeability.
magnetic vector potential, source current density and
electrical conductivity respectively. The derivation of this
equation from Maxwell's equation is given elsewhere112-'41
and is not repeated here.
In many practical eddy current NDT situations the
geometry is axisymmetric. Such is the case in the testing of
steam generator tubing where circular eddy curent coils
(absolute or differential) are used. In addition, the
excitation for these coils is sinusoidal.
Thus, writing Equation (5) in cylindrical coordinates and
performing the differentiation with respect to time for
steady-state sinusoidal excitation, a simplified equation is
obtained
1 102A 1 aA a2A .~2~
/1 \~--~-- + - + ] = ~ls +jcoo,4
- r ~ 0z = r=
In this system, both J andA are functions o f r and z only
and have a single component in the positive 0
direction.
In the derivation of the above equation, the following
assumptions were made.
mating functions A such that the energy-related
functional Is minimized. This is done at discrete points in
the solution region by solving a system of linear equations
derived in the minimization process1141.
From the magnetic vector potential calculated by the
finite element method, one can calculate any derivable
magnetic quantities such as field densities and probe
impedances.
The impedance of an eddy current probe is calculated
from the coils" geometry and the calculated magnetic
vector potential. The impedance of a circular loop of
radius r, and carrying an RMS current Is can be calculated
as
_ jco2rrr.~i
Zi - -
I, (8)
where A, is the value of the magnenc vector potential at
r,.
Integration of this equation over the cross-section of the
coil yields the impedance of the coil. The magnetic vector
potential is only known at the nodal points of the finite
element mesh, and therefore an average value A c is
calculated at the centroid of each element and assumed
(6)
constant within the element. The impedance of an
absolute eddy current probe then becomes
j~o21r~ ._~
~-_ x , (rc,Ai).~c,)
Is2
/_.,
i= 1
(9)
where rc, is the distance of the centrold of element i from

NDT I N T E R N A T I O N A L . DECEMBER 1 9 8 5 333

the s y m m e t r y h n e . J , ts the total current denstty m the coil
(J,=NI,. w h e r e N is the n u m b e r of turns m the coil), A, is
the area of e l e m e n t i a n d N. is the n u m b e r of fimte
elements m the cross-secnon of the coil
S,mdarly. the m l p e d a n c e ol a dlfferentml probe ts
,~h Na
Tube
2' -
Ic2rffs ~ " (rc,Ai)Aci --
N.
/
~ (rc,A,)Ac, (10)
Is= ~ ,='--~'
@@
where N, a n d N b are the n u m b e r of clemenls in the cross- Coils
section of lhe two coils of the probe
A plot of the c o m p l e x i m p e d a n c e o b t a m e d by E q u a n o n s
(9) or(l()) with the real part on the h o r i z o n t a l axis a n d the Fig 3 Crevme gap region m a PWR steam generator
i m a g i n a r y part on the vertical axis is called a n ' t m p e d a n c c
p h m e trajectory" a n d is a c o m m o n wa3, o f representing
but the signal o b t a i n e d whflc m o n i t o r i n g the process
eddy current test date
d e p e n d s on the way the magnetite is flushed. In the
n u m e r i c a l study, the possibility of radial or axml build-
Results up (or flushing), axial b u i l d - u p from one side ol the
support plate, a n d flushing in the presence ol tube
The model d e s c r i b e d above was a p p h e d to a detailed
denting were modelled, although because of the Imge
n u m e r i c a l study of magnetite b u i l d - u p in the crevice
n u m b e r of i m p e d a n c e plane Irajectones obt,uncd onl>
gap m PWR steam generators There is still a significant
representatp, e data are presenled m this work
uncertainty as to the magnetite b u f f & u p process ttsell~
since it is not k n o w n exactly how the m a g n e m e Thc gcometr~ c o n s i d e r e d is presented m t-tgurc 3 h
a c c u m u l a t e s m the crevice gap Similarly, in the consists of an Inconel 600 tube (22 m m m dlametm.
c h e m i c a l flushing process, the magnettle Is removed 13 m m v, all t h l c k n e s q lnstdc a 19 mm c a r b o n qccl

Table 1. Material properties used in the finite element studies

Carbon steel Inconel Magnetite

Conductwlty, o{f2 -1 m -1) 5.7 x 106 1 1 x 106 166

Relatwe permeability #R 50 1 7

Fig 4 Finite element mesh structure for the geometry

shown in Figure 3

a b

334 NDT INTERNATIONAL. DECEMBER 1985

support plate. The crevice gap between the tube and
support plate is 0.38 mm while the probe modelled is
a differential eddy current probe. The dashed lines
represent schematically the area m which denting was
modelled. The material properties used are
summarized in Table 1
Figure 4a is the finite element mesh used in the
numerical model. Because of symmetry only half the
geometry in Figure 3 is modelled. The mesh consists of
13 600 triangular elements and 7035 nodes, resulting in
a system of 7035 linear equations with a semi-
bandwidth of 37.
The region of probe movement ~s within the dense
mesh portion at the centre of the mesh. This uniformly
diwded region consists of a total of 180 layers, while
the probe is allowed to move for a maximum of 140
probe posmons. In some cases, such as the axial build-
up study, the geometry is also symmetrical about the
centre of the support plate, in which case only 70 probe
posmons are modelled. Figure 4b shows how the mesh

r=40% ( ~
r=60%

r=20%
Support
''"
r=80%~

r=lO0%
SB
r=O C(~Is

F~g5 Finiteelementpredlcttonsof differentmlprobe impedance plane trajectories for radialmagnetitebudd-up

f=80%

=
f =100%
;=0

F~g.6 Finiteelementpred=ctionsof dffferent=alprobe impedance plane trajectories for axial magnet=tebudd-up

NDT INTERNATIONAL. DECEMBER 1985 335

 20~o/of=730/f=41ii~t~eOr~~
=
t=50ff=60'/= / : 8 0 Yo

/=10 Vo /=90%

/=0 t=lO0%

Ftg 7 Fmtte element predmttons of ddferenttal probe impedance plane tralectones for axtal budd-up m the presence of denting d = 100 him

d=O d=lT d=2T d-.3T d=4T

Support
plate

m m
Cads

Fig 8 Effects of denting on probe impedance plane tralectone~ T=

25 /~m

is a d j u s t e d m the tube regmn u n d e r the s u p p o r t plate g e o m c t u revolved a n d the c o r r e s p o n d i n g m l p e d a n c c

to take into account d e n t i n g of the tube as well as
b u i l d - u p or flushing of magnetite m the crevice gap
over a dented tube
Here again, for lack of any defimte l n f o r m a t m n as to
the exact s h a p e of dents in tubes, the dented p o r t i o n is
represented by a c~rcular arc
The first part o f the study deals with radml b u i l d - u p o f
magnetlte from the s u p p o r t plate towards the tube. or
alternatively, the flushing o f magnetite from the tube
towards the s u p p o r t plate Figure 5 shows both the
plane trajectories for magnetite b r a i d - u p in layers ol
76 p.m T h e c h a n g e m s~gnal from a clean gap to onc
lull of m a g n e h t e is qmte d r a m a t i c both m shape a n d
magmtudc
The second part of the study assumes an axial buff&up
of magnetite from the centre of the support plate
outwards A l t h o u g h this is not very hkely to h a p p e n in
the b u d & u p process, it Is representatp,,e of a c h e m i c a l
flushing process where the c h e m i c a l agents attack the
magnetite from both sides Figure 6 shows this

336 NDT INTERNATIONAL, DECEMBER 1 9 8 5

 f=O b ~ c

,'=50%

k.\~\\\\.,o
~\\\\\\\\\\\\\\\\\~
BB ~$Bi E
I

Fig.9 Compartsonofpredlctedlmpedanceplanetralectorlesforvanouscrevmegapcondmonsa - - t = O , 6--t=lO%,c--t=60%,d--d=5Oixm, e--t=30%,

a= 10%. d= 100 btm

Figure 7 shows the impedance plane trajectories of

e in axial magnetite build-up (or flushing) in various
ram) amounts from a clean gap to a gap full of magnetite.
~lote hole From these plots it is clear that even small amounts of
magnetite in the gap give a clear indication.
While in the previous parts of the study there is a clear
distinction between the various stages of magnetite
build-up, monitoring of the denting process itself seems
to be more complicated, as shown in Figure 8. The output
for a small dent (25 ~m) and a large dent (100/xm) are
e in quite similar. On the other hand, the detection of
ram)
dote hole denting itself is straightforward as indicated by the two
leftmost trajectories.
In addition to the data presented above, studies were
carried out with axial build-up from one side of the
support plate (Figures 9b and 9c), axml build-up from
one side of the support plate with a portion of the gap
clean on both sides of the magnetne band (Figure 9e)
as well as with denting of the tube without the presence
in
ram) of magnetite (Figure 9d). For comparison, the clean
)lore hole crevice gap trajectory is given again in Figure 9a.

Fig 10 Experimentalresults obtainedfrom an enlargedcrevicegap
packed wJthpure magnetfte
snuation and again the changes, especially between a
clean support plate and one with magnetite or between
a full gap and partially filled gap, are quite dramaUc.
Another problem of sigmficant interest in NDT is that
of denting of tubes due to magnetite build-up.
Similarly, monitoring the flushing of magnetite in this
situation ~s more ~mportant to ensure the complete
cleaning of the magnetite in the gap
Experimental work
Any conceivable geometry (within the limitation of a
two-dimensional or axisymmetric model) can be
reproduced using the finite element code There still
remains, however, the question of comparison w~th
real, known data. The experimental measurement of a
support plate signal in the presence of a clean crevice
gap is. of course, only part of the answer Most of the
trajectories in Figures 5 to 9 cannot be reproduced
experimentally, because it is impossible to pack a
038 mm gap with magnetite, at least certainly not in
any controlled manner
Partly to answer this question, the experiment in
Figure 10 was carried out. Here a support plate was
drilled to provide a 1.5 mm gap which was then packed

NDT I N T E R N A T I O N A L . DECEMBER 1 9 8 5 337

 a b c

F~g 11 Resultsobtained during chemmal flushing tests on a model boder

axially with p o w d e r e d magnctlte at about 60% ( F i g u r e Authors

10a), 30% ( F i g u r e 10b) and 90% (Figure 10c) Th e
r e s e m b l a n c e to the c o r r e s p o n d i n g trajectories in Figure 5
ts l m m e d m t e l y evident.
More c o n v i n c i n g evidence as to the abd~ty, accurat O
and usefulness of this model ix provided m Figure 11
These i m p e d a n c e plane trajectories were obt,uncd
during the c h e m i c a l flushing of a model b o d e r Th e
data are from a tube with a 100 /.tm dent (radial) at
various stages o f flushing. Figure 1 I a was taken belore
flu s h m g and shows a clearly identifiable dent filled
with m a g n e n t e As the flushing progressed, the gap
shows various stages of c l e a n i n g Thus, tbr example, 11-1
Figure l i b , one side ol the gap is almost completel3;
cleaned as indicated by the single large uppm lobe
while the rest Is still packed with magnetite, indicating
tin uneven flushing from both sides of the support
plate Figure l lc shows a clean gap while the dent m
the tube Is visible
This particular lesk besides p r o v id in g a con,tint.rag
e x p e r i m e n t a l c o n f i r m a t i o n o f the model, also points
out the value o f such a tool in interpreting re,lhst~c
data
Conclusions
This p ap er descrfl'~es ,t u n i q u e apphcatlon ol
axisymmetrlc fimtc elernent code to the s l m u l a n o n ol
tin eddy current NDT situation too c o m p l e x to rephcate
experinaentally Results ot the m o d e l h n g study clearly
show the sensmvlty o f a conventional, differential edd3,
current p ro b e to c o n d m o n s m the crevice gaps of PWR
steam generators d u r i n g magnetite braid-up and
llushmg. Limited e x p e r i m e n t a l data seem to co n f i r m
the theoretical predictions
Acknowledgement
This work has been supported by the Electric Power
Research Institute u n d e r contract RP 1395-2
N Ida is ~lth the Um~crslty tit" Akron, lISA.
H H o s h l k a w a with N f l m n University, Japan. and ~ '
Lord with C o l o r a d o State Um,,erslty. USA.
References
I Lord, ~'. Appllt.atu)ns ol ntune~l~al modchng to clcLtm-
magneltt, methods of nondeslrucuve testing IfiEF Trans
Magnen~ MAG-19, No 6 (No', 1983)pp 2437-2442
2 Lord, W. and Huang, J.H. I)etet.t charactenz,mon trom
magnetu, leakage fields BtttJAomh,~tr Testing 19, No I (Jan t7
1977) pp 14-1S
3 Ida N., Palanisamy, IL and Lord, W. Lddy current probe design
using hmte etemern ,malysts Mater Llal. 41 No 12 (Nov 1983)
pp 1389-1394
4 Ida, N. and Lord, W. (Jraphlt_al smaulanon ol eleclromagnetk.
NI)I plobe heltts' 11-.1.1- (omputet Orapht~ Appl (May/June
1983) pp 21-28
5 ~llen,B. and Lord, W. Frolic element modehng of pulsed edd~
current phenomena m Revw. o/ Progres~ m Qttantllal/l~'
NondestruettteEvaluatmn D O lhompson and D t ( hunentl
[-ds Plenum (1984) pp 561-,6,'.;
6 Ida, N. and Lord, W. '3-D hnlte clement predt~.non ot
magrletostatl, leakdge fields ILl 1 Tran~ tVlagnett ~ M ~C,-19 No
5 (Sept 1983) pp 226(I-2265
7 Smith, J. and Wijn, H. P.J. Ferntes Wiley I nc Ne~ ~ ork (1959) p
229 and p 251
X Shock J.L. {Ed.)Are. Development.s tel Ferromagnetw Materials
Elsevier New York (1947) Ch 1
9 landott and Boernstein Zahlen und Funkttonen Vol II.
Springer Verlag, Berhn 11962)
II) ASTM - Direct ~.urrent mdgnellt, measurements lor soil
magncne rnalertals ASTM Vw~ml pubhcanon 371 sl ASTM
Philadelphia (1970)
Brown, W.F. 'MagnctostatJ~. ptmulptes m lcrromagneUsm" m
Seh'cted Topt~ s m Sohd Staw ehyst(~ Vol I, Ed E P Wohlfarth.
North-Holland Pubhshmg Co. New York
12 Anderson, O.W. ' I'ranMornaer leakage l]uK plogldm based on
the fimtc elcnlenI nlclhod l i d L 71un~ PoneJ Ippatatus am/
5t~lellls 92 (M,u-Apl 1973). pp 682-6S9
II Palanisamy, R. and Lord, W. I mile elcmcnl modchng ol
clet_tromagnetlt_ NI)T phenomen,t lt.k.k lpun~ ,,Ihegnetl~ ~
MAG-15 No 6 (1979) pp 1479-1481
14 Ida, N., Betzold, K. and Lord. W. | unto element modehng ot
,lb,,olute cdtl3, t_urrent probe signal4 ,I ,Mmde~tl [ val 3, No
(19~3) pp 147-154
Huebner, K.H. The Finite I"dement Method/or Lngmeers John
Wfle~ and Sons Nev, 'tork(1975)

Paper received 29 May 1985. Revised 15 August 1985

338 NDTINTERNATIONAL DECEMBER 1985