Sie sind auf Seite 1von 34

CH2.

The Principles of Eddy Current Testing



2.1. Introduction
The eddy current testing method is a form of electromagnetic
nondestructive testing based on the principles of electromagnetic induction.
The physicals of eddy current testing are shown in figure 2.1. The Maxwell-
Ampere law states that an alternating current sets up a time varying magnetic
field. When a primary coil excited with an alternating current is placed in
close proximity to a conducting surface, the alternating primary magnetic
fields from the probe interact with the law. This induced voltage produces
eddy currents in the material witch, by Lenz's Law ( ), set up
secondary magnetic fields that oppose the ones producing them. The eddy
currents, thus, are analogous to a mutually coupled secondary circuit in witch
the current flows in a circular direction opposite that of the current in the coil.
This mutual inductance causes a change in the impedance of the coil because
the load of the secondary circuit, now the eddy currents in the specimen, are
referred, or "reflected", back to the primary coil. For instance, the resistive
component is a measure of the energy losses within the material since the
eddy currents represent a resistive load. Hence, as a coil is brought close to a
non-ferromagnetic conductive material ( ) its resistance, the
real component of its impedance, increases.

1


Fig. 2.1. Principles of eddy current testing.

Impedance also consists of a reactive imaginary component, called the
inductive reactance. The work required to establish a magnetic flux in a coil,
on the energy stored in the magnetic field of a coil, is directly to its
inductance. Thus, when the opposing secondary field from the induced eddy
currents decrease the magnetic flux linkages, as shown in figure 2.1, the
coil's stored energy decreased which in turn reduces its inductance and

2

inductive reactance. However, when a coil with a certain magnetic field
strength, H, is placed over a ferromagnetic material( ), the lower
reluctance in the material allows the flux lines to increase in density within
the coil's windings. This, therefore, increases the coil's inductance.

Any material property or geometry variances that affect the eddy current's
distribution and the related magnetic fields will always result in a change in
the coil's impedance. These changes are monitored by either measuring them
directly with an impedance analyzer or, more commonly, by observing the
trajectories traced on the scope of a commercial eddy current instrument. (
EddyScope ) Usually, the signals are
measured as departures from a reference condition, such as that for the coil
placed on a standard specimen (
). Signals therefore
result from changes in one or more of the specimen's properties such as
conductivity, permeability, thickness, surface roughness, temperature and,
most notably, the presence of defects. A defect, for example, interrupts the
eddy currents' path, especially if it is a crack perpendicular to the flow of the
eddy currents, changing the secondary impedance, and hence the impedance
of the coil. Test setup factors such as liftoff, frequency, and electronic noise
also affect the signal. Under most operating conditions, the tests can

3

differentiate between various factors and material properties if they each
produce a different signal response. For example, commercial eddy scopes
will display relatively distinctive trajectories for different types of deviations.
This allows the instrument to be setup in a way that separates an unwanted
signal, such as that caused from liftoff, from the signal of interest, such as a
flaw. Recorded signals can also be processed later by removing the noisy
responses known to be caused by certain factors.

Commercial eddy current instruments are almost exclusively used in a
relative nature where reference standards are used to set the normal operating
point. A material similar to the standard is tested by observing the signals
that deviate from the reference condition; these indicate flaws or other
property changes. A good understanding of these signals and how an
instrument detects and displays them is necessary to develop a simple
calibration method that would make quantitative impedance measurements
possible. Consequently, commercial instruments would be able to perform
quantitative NDE (QNDE), such as is now performed with an impedance
analyzer in conjunction with an applicable eddy current theory.

4

2.2. Theory of Eddy Current Phenomenon

Electromagnetic phenomenon for time varying conditions are governed
by Maxwell's equations as follows:

t
B
E

= (2.1)

t
D
J H

+ = (2.2)

0 = B (2.3)

= D (2.4)

The constitutive relations for an isotropic, linear and homogeneous
medium are given as
H B = (2.5)
E D = (2.6)
E J = (2.7)
where

5


= the magnetic permeability in henrys per meter(H/m)
= the electric permittivity in farads per meter (F/m), and
= the electric conductivity in seimens per meter (S/m).

Since the divergence of B is zero from equation (2.3), B can be expressed
as the curl of another vector, called the magnetic vector potential (MVP) A,
given by

. A B = (2.8)

Combining this with equation (2.1) gives

t
A
E

= (2.9)

or

0 =
|
.
|

\
|

+
t
A
E (2.10)
Since an irrotational field can be replaced with the gradient of a scalar

6

potential function, the solution to equation (2.10) is

V
t
A
E =

+ (2.11)

where V is the scalar electric potential.

Now, substituting equation (2.5) into equation (2.2) yields

|
.
|

\
|

+ =
t
D
J B (2.12)

The displacement current t D is negligible when compared to the current
density, J , for the relatively low operating frequencies in eddy current testing
(below 10 MHz). Therefore, combining equations (2.7), (2.8) and (2.12) we
have

J A = (2.13)

E = (2.14)


7

Replacing E in equation (2.14) with the relation of (2.11) gives

|
.
|

\
|

=
t
A
V A (2.15)

Recognizing the vector identity

( ) ( ) A A A
2
= (2.16)

and choosing the Coulomb gauge , equation (2.15) becomes the
following partial differential equation (p.d.e.) for any varying current
0 = A

t
A
J A
s

+ =
2
(2.17)

where V J
s
= is the source current density in Amperes per square meter,
such as an infinite sheet of current over a conductive half-space. Assuming
the fields are time harmonic and vary sinusoidally in steady state, sinusoidal
eddy currents will exist within the conductor and A can be expressed as
t j
e A A

=
m
(2.18)

8

where is the angular frequency. Differentiating with respect to time, we
get

A j
t
A
=

(2.19)

Thus, substituting equation (2.19) into (2.17) results in

s
J A j A =
2
(2.20)

where the first term is a function of the density of the induced eddy currents,
J
e
, given by

A j J
e
= (2.21)

The ability to obtain analytic solutions for the magnetic vector potential from
equation (2.20) is usually limited to simple geometries such as a symmetrical
coil over an infinite half space medium. This equation also governs the skin
effect that will be discussed in the next section.


9

2.3. Skin Effect

Equation (2.17) is a diffusion equation that describes the propagation of
EM waves in a conductor. Assuming an exciting current sheet is flowing in
the x direction over a conducting half-space, the magnetic field intensity, H
y
,
in the conductor as a function of depth from the surface, z, is found to be

(

|
.
|

\
|

|
.
|

\
|

=
2 / 1 2 / 1
2 2
0

jz z
y
e e H H (2.22)

Hence, EM fields rapidly decay exponentially with depth. This effect also
applies to eddy currents, where the induced current density, J
e
, is given by

/ /
0
jz z
e
e e J J

= (2.23)

where

f
1
= (2.24)


10

is skin depth or standard depth of penetration. This is the depth at witch the
magnitude of the induced eddy current density decreases by a factor of 1/e
from its value at the surface, J
0
. This current distribution is illustrated in
figure 2.2. The eddy currents are confined to a thin layer of the conductor's
surface adjacent to the coil. And relationship (2.23) indicates that lower
frequencies are required to reach greater depths and detect or measure deeper
subsurface flaws.

Fig. 2.2. The skin effect.

11

2.4 Forward Solution for the Impedance of a Coil Over a Single-
Layered Half Space
The calculation of the impedance of a right-cylindrical air-core coil placed
above a 2-layer flat plate of metals is a canonical ( ) problem in
quantitative eddy-current inspection. Dodd and Deeds proposed a widely-
used solution for a coil above a two-conductor nonmagnetic metals. Each
layer is characterized by a uniform . We review this paper in the following.
III
VI
I
II
z=0
z=-c
z=l

2
Coil

Figure 2.3. Coil above a two-conductor plane.

12

The coil above a two-conductor plane is shown in Fig. 2.3. We have divided
the problem into four regions. The differential equation in air (regions I and II)
is
0
1
2 2
2
2
2
= + +
r
A
z
A
r
A
r
r
A

. ( ) (2.1)
(Note: =0 ) A j
i


The differential equation in a conductor (regions III and IV) is
0
1
2
2
2 2
2
= + + A j
z
A
r
A
r
A
r
r
A
i

(2.2)

Setting A(r,z)=R(r)Z(z) and dividing by R(r)Z(z) gives
0
1 1 1 1
2 2
2
2
2
= + +
i
j
r z
Z
Z r
R
rR
r
R
R

(2.3)

13

We write for the z dependence
i
j const
z
Z
Z

+ = =
2
2
2
1
, (2.4)
or
i i
j z j z
Be Ae Z
+
+ =
2 2
+
. (2.5)
We define
i i
j +
2
. (2.6)
Equation (2.3) then becomes
0
1 1 1
2
2 2
2
= + +

r
r
R
rR
r
R
R
(2.7)
The is a first-order Bessel equation and has the solutions
( ) ( r DY r CJ R
1 1
+ = ) (2.8)
Note: Bessel functions arise in solving differential equations for systems with
cylindrical symmetry.

14

Note:
Plot BesselJ 1, x , x, 0, 50

Plot BesselY 1, x , x, 0, 50

@ @ D 8 < @ @ D 8 <D
10 20 30 40 50
-0. 2
0. 2
0. 4
0. 6

10 20 30 40 50
-0. 6
-0. 4
-0. 2
0. 2
0. 4



Combining the solutions we have
| | ( ) ( ) | | r DY r CJ Be Ae RZ A
z z
i i


1 1
+ + = =

(2.9)
We now need to determine the constants A, B, C, D. They are functions of
the separation constant and are usually different for each value of . The
complete solution would be a sum of all the individual solutions, if is a

15

discrete variable, but, since is a continuous variable, the complete solutions
is an integral over the entire range of . Thus, the general solution is
| | r
D
( ) ( ) ( ) ( ) ( ) ( ) | |

d r Y D J C e B e A A
z z
i i


+ + =
0
1 1
(2.10)
In region I, as , ) should be finite. Therefore . + z ( z r A , ( ) 0 = A
In region IV, as , ) should be finite. Therefore . z ( z r A , ( ) 0 = B
In all region, divergence. Therefore .
1
Y ( ) 0 =
The solutions in each region then become
( ) ( )

=
0
1 1
) 1 (


d r J e B A
z
(2.11)
( ) ( ) | | ( )

+ =
0
1 2 2
) 2 (


d r J e B e C A
z z
(2.12)
( ) ( ) | | ( )

+ =
0
1 3 3
) 3 (


d r J e B e C A
z z
(2.13)

16

( ) ( )

=
0
1 4
) 4 (


d r J e C A
z
(2.14)
The boundary conditions between the different regions are
Between region I & II :

( ) ( )
( ) ( ) ( )

=
=
= = 0
) 2 ( ) 1 (
) 2 ( ) 1 (
, ,
, ,
r r I l r A
z
l r A
z
l r A l r A
l z l z


( )
( ) 16 . 2
15 . 2
Between region II & III :

( ) ( )
( ) ( )

=
=
= = 0
) 3 (
0
) 2 (
) 3 ( ) 2 (
, ,
0 , 0 ,
z z
z r A
z
z r A
z
r A r A


( )
( ) 18 . 2
17 . 2
Between region III & IV :

( ) ( )
( ) ( )

=
=
= = c z c z
z r A
z
z r A
z
c r A c r A
, ,
, ,
) 4 ( ) 3 (
) 4 ( ) 3 (


( )
( ) 20 . 2
19 . 2
Now, we have six equations to solve the six unknown coefficients

17

(i.e. ).
4 3 2 3 2 1
C C C B B B

From (2.15) : ( ) ( ) ( ) ( ) | | ( )



+ =
0
1 2 2
0
1 1


d r J e B e C d r J e B
l l l

(2.21)
Rewrite (2.21) as following
( )
( ) ( )
( ) ( )
2 2
l
l l
B e
C e B e


(
+


( ) ( )
F =
( ) ( )
1 1
0 0
1 1
0 0
1
J r J r rdr d
J r J r rdr d

(

(

=
(



Using the Fourier-Bessel integral:
( ) ( )
1 1
0 0
J r J r r r F d d


(2.22)
Then we have :
l l l
e
B
e
C
e
B



|
.
|

\
|

+
|
.
|

\
|

=
|
.
|

\
|

2 2 1
(2.23)
By the same way (dropping the primes, i.e. ), we have

18

(2.16) : ) (2.24) (
0 1 0 2 2 1
r J Ir e B e C e B
l l l


=

(2.17) :

3 3 2 2
B C B C
+ = + (2.25)
(2.18) :
3
1
3
1
2 2
B C B C
|
.
|

\
|

|
.
|

\
|
=

(2.26)
(2.19) :
c c c
e
C
e
B
e
C
2 1 1
4 3 3


|
.
|

\
|
=
|
.
|

\
|
+
|
.
|

\
|
(2.27)
(2.20) :
c c c
e C e B e C
2 1 1
4
2
3
1
3
1


|
.
|

\
|
=
|
.
|

\
|

|
.
|

\
|
(2.28)

( )
( )( ) ( )( )
( )( ) ( )( )
)

(
(

+ + +
+ + +
+ =
l
c
c
l
e
e
e
e r J Ir B




1
1
2
2 1 1 2 1 1
2
2 1 1 2 1 1
0 1 0 1
2
1
(2.29)
( )
( )( ) ( )( )
( )( ) ( )( )
l
c
c
e
e
e
r J Ir B





(
(

+ + +
+ + +
=
1
1
2
2 1 1 2 1 1
2
2 1 1 2 1 1
0 1 0 2
2
1
(2.30)

19

( )
( )
( )( ) ( )( )
(
(

+ + +

=

c
l
e
e
r J Ir B
1
2
2 1 1 2 1 1
2 1
0 1 0 3



(2.31)
( )
l
e r J Ir C



=
0 1 0 2
2
1
(2.32)
( )
( )
( )( ) ( )( )
(
(

+ + +
+
=
+
c
c l
e
e
r J Ir C
1
1
2
2 1 1 2 1 1
2
2 1
0 1 0 3




(2.33)
( )
( )
( )( ) ( )( )
(
(

+ + +
=
+
c
l c
e
e e
r J Ir C
1
2 1
2
2 1 1 2 1 1
1
0 1 0 4
2




(2.34)

Then the vector potential in each region is:
( ) ( ) ( ) ( )


=
0
1 0 1 0
) 1 (
2
1
,
z l
e r J r J Ir z r A


( )( ) ( )( )
( )( ) ( )( )


d
e
e
e
c
c
l

(
(

+ + +
+ + +
+
1
1
2
2 1 1 2 1 1
2
2 1 1 2 1 1
2
(2.35)
( ) ( ) ( ) ( )


=
0
1 0 1 0
) 2 (
2
1
,
l
e r J r J Ir z r A



20

( )( ) ( )( )
( )( ) ( )( )



d e
e
e
e
z
c
c
z

+ + +
+ + +
+

1
1
2
2 1 1 2 1 1
2
2 1 1 2 1 1
(2.36)
( ) ( ) ( ) ( )
( ) ( )
( )( ) ( )( )





d
e
e e e
e r J r J Ir z r A
c
z z c
l

(
(

+ + +
+ +
=
0
2
2 1 1 2 1 1
2 1
2
2 1
1 0 1 0
) 3 (
1
1 1 1
,
(2.37)
( ) ( ) ( ) ( )
( )
( )( ) ( )( )





d
e
e e
e r J r J Ir z r A
c
z c
l

(
(

+ + +
=
0
2
2 1 1 2 1 1
1
1 0 1 0
) 4 (
1
2 2 1
2
,
(2.38)



We have the equations for the vector potential produced by a single delta-function
coil. We can now approximate any coil by the superposition of a number of delta-
function coils.

21

( ) ( ) ( ) ( )


= = =
= =
2
1
2
1
0 0
1 1
, , , , , , , ,
r
r
l
l
N
i
i i
N
i
i Total
dl dr l r z r A r l z r A z r A z r A (2.39)
( ) ( ) ( ) ( )
2
1 1
(1)
0 0 1 0 1
0
,
2
r l
l z
A r z i r J r J r e



=


2
1
r l

( )( ) ( )( )
( )( ) ( )( )
dl dr d
e
e
e
c
c
l
0
2
2 1 1 2 1 1
2
2 1 1 2 1 1
2
1
1

(
(

+ + +
+ + +
+
(2.40)
Express the integral over as
0
r
( ) ( ) ( )
(
( )
2 2 2
0
1 1 1
0 0 0 0 1 2 1
2
1 0 1
2
0
2
1 1 1
,
r r r
x r
r r r
r dr r d r xJ x dx I r r J r J r

=
| | | | | |
= =
| | |
\ . \ . \ .


(2.41)
The integral over is l
( )
( ) ( ) | | | | { } k e e e e
e
dl ke e e dl k e e
l l l l
z
l
l
l l z
l
l
l z l

|
|
.
|

\
|
= + = +

+

1 2 1 2
2
1
2
1
2

(2.42)


22

Upon applying (2.41) and (2.42)
( ) ( ) ( ) ( )


|
.
|

\
|
=
0
1 1 2
3
0
) 1 (
,
1
2
1
,
z
e r J r r I i z r A


| | | |
( )( ) ( )( )
( )( ) ( )( )



d
e
e
e e e e
c
c
l l l l

+ + +
+ + +


1
1
1 2 1 2
2
2 1 1 2 1 1
2
2 1 1 2 1 1
(2.43)
( ) ( ) ( ) ( )| |



|
.
|

\
|
=
0
1 1 2
3
0
) 2 (
2 1
,
1
2
1
,
l l
e e r J r r I i z r A


( )( ) ( )( )
( )( ) ( )( )



d e
e
e
e
z
c
c
z

(
(

+ + +
+ + +
+

1
1
2
2 1 1 2 1 1
2
2 1 1 2 1 1
(2.44)
( ) ( ) ( ) ( )
1 2
(3)
0 2 1 1
3
0
1
, ,
l l
A r z i I r r J r e e


| |
(
=
|

\ .


( ) ( )
( )( ) ( )( )



d
e
e e e
c
z z c
(

+ + +
+ +


1
1 1 1
2
2 1 1 2 1 1
2 1
2
2 1
(2.45)

23

( ) ( ) ( ) ( )| |



|
.
|

\
|
=
0
1 1 2
3
0
) 4 (
2 1
,
1
,
l l
e e r J r r I i z r A


( ) | |
( )( ) ( )( )



d
e
e e
c
z c
(

+ + +

+
1
2 2 1
2
2 1 1 2 1 1
1
2
(2.46)

For a point (r, z) in region I-II, Substitute in (2.43) and in (2.44)
and add the two equations, we have
z l =
2
z l =
1
( ) ( ) ( ) ( )

|
.
|

\
|
=
0
1 1 2
3
0
) 2 , 1 (
,
1
2
1
, r J r r I i z r A


( ) ( )
( )
( )( ) ( )( )
( )( ) ( )( )



d
e
e
e e e e e
c
c
l l z l z l z

+ + +
+ + +
+

1
1
2 1 1 2
2
2 1 1 2 1 1
2
2 1 1 2 1 1
2
(2.47)


24

Once we have determined the vector potential (A), we can calculated any
physically observable electromagnetic induction phenomenon. We shall now
give the equations and perform the calculations for some of the phenomena of
interest in eddy-current testing.

Induced eddy currents:
We have, from Ohms law:
( )
,
A
J E j A r
t

= = = z
Induced voltage:
We have, for a single turn of the detector coil: ( ) z r rA j dl E V , 2
1
= =


The total voltage induced in a coil of N turns is:
( )
( )( )
( )
2 2
1 1
2 1 2 1
0
2 , 2 ,
N
i i i
i
l r
l r
N
V j r A r z j rA r z drd
l l r r

=
| |
=
|

\ .


z (2.48)
The current in a single loop is related to the applied current density, , by
0
i

25


) ( )(
1 2 1 2
0
r r l l
NI
i

= (2.49)

From (2.47), (2.48), and (2.49) , [ ] we have the coil impedance ) 48 ( ) 47 ( ) 49 (
I
V
I j
V
j L j Z = |
.
|

\
|
= =


( ) ( )
( )
| |

)
`


|
|
.
|

\
|

=

2
1
2
1
0
3
1
2
1 2
2
1 2
2
l
l
r
r
drdz d
r J I
r
I r r l l
I N j



Note:
( ) ( ) ( )
( )
5
2 1
5
1
3
1
, ,
) (
2
1
2
1



r r I
r d r J r dr r J r
r
r
r
r



( ) ( )
( ) ( )
( ) ( ) ( ) | |



d dz e e e e e
I
r r l l
N j
l
l
l l z l z l z


+
|
|
.
|

\
|

=
0 5
2
2
1 2
2
1 2
2
2
1
2 1 1 2
2
where ( )
( )( ) ( )( )
( )( ) ( )( )
|
|
.
|

\
|
+ + +
+ + +
=
c
c
e
e
1
1
2
2 1 1 2 1 1
2
2 1 1 2 1 1





26

( ) ( )
( ) ( )
( )
( ) ( ) | | { }


d A e l l r r I
r r l l
N j
l l
+ +

=

2 2 2 ,
1
1 2
1
1 2 1 2
0
2
5 2
1 2
2
1 2
2
(2.50)
where ( )
( )
( ) ( )
2
2 2
2 1 1 2 1
2
l l l l l l
e e e e

+
= + =
2
e A

Normalization :
( )
2
2 1
r r
r
+
= ,
all lengths are divided by r , all are multiplied by s r
2 2
r j
i i
+
( ) ( )
( ) ( )
( )
( ) ( ) | | { }


d A e l l r r I
r r l l
r N j
Z
l l
+ +

=

2 2 2 ,
1
1 2
1
1 2 1 2
0
2
5 2
1 2
2
1 2
2
(2.51)

Difference in impedance
layer no layer with HSP L
Z Z Z Z Z
_ _
= =

27

where Z
L
=Eq. (2.50)
and
0
1
) 50 . 2 .(
= =
=
c and
Eq

HSP
Z
( )
( )( ) ( )( )
( )( ) ( )( )
2
2
2 1
2 1
2
2 1 1 2 1 1
2
2 1 1 2 1 1
0
1
1
1





=
+

=
|
|
.
|

\
|
+ + +
+ + +
=
= =
c
c
c and
e
e

We have
( ) ( )
( )
1 2
0
2
5 2
1 2
2
1 2
2
,
1
r r I
r r l l
N j
Z
HSP


( )
( )
( )




d A e l l
l l
)
`

+ +

2
2
1
1 2
2 2 2
1 2
(2.52)
layer no layer with HSP L
Z Z Z Z Z
_ _
= =
( ) ( )
( ) ( ) ( )


d A r r I
r r l l
N j
(

2
2
1 2
0
2
6 2
1 2
2
1 2
2
,
1
(2.53)



28

2.5 Review of Moulder et al.
In the following section, we will review the paper by Moulder et al. and
show how to apply Dodd and Deeds solutions to the eddy current
measurements.
Moulder et al. describe a robust method that uses eddy-current
measurements to determine the conductivity and thickness of aluminum and
copper layers on various substrate metals and that of free-standing foils of
aluminum, and the thickness and conductivity of free-standing foils of
aluminum. The impedance was measured for air-core and ferrite-core coils
either in the presence or the absence of the layer at frequencies ranging from
1 kHz to 1 MHz. The thickness and conductivity of the metallic layers were
inferred by matching the data taken with an air-core coil to the exact solution
derived Dodd and Deeds [5] using a least-squares norm. The inference was
absolute in the sense that no calibration was required. Both the thickness and

29

conductivity can be determined accurately and simultaneously if the ratio of
the layer thickness to the coil radius is between 0.20 and 0.50. For thinner
samples, one has to know the thickness to determine the conductivity, or vice
versa.


30

Forward solution

The forward problem is to determine the impedance of the coil, given the
frequency, the layer thickness, and the permeability and conductivity of the
materials. The forward problem was solved by Dodd and Deeds who
considered a coil excited by a constant ac current of angular frequency .
They found that the impedance of the layered half space is given by

( ) ( )
( ) ( )
( )
( ) ( ) | | { }


d A e l l r r I
r r l l
N j
Z
l l
+ +

=

2 2 2 ,
1
1 2
1
1 2 1 2
0
2
5 2
1 2
2
1 2
2
. (2.54)
The experimentally measured quantity is the difference in impedance
between two measurements
HSP L
Z Z Z = . (2.55)
Where, denotes the impedance of the coil over a layer of metal on a thick
substrate. The thickness of the substrate has to be many times greater than
L
Z

31

the skin depth. In the theoretical calculation, we approximate the thick
substrate as a half space. The impedance of a coil above the substrate is used
as a reference, and is denoted by . The use of facilitates the
comparison between the theory and the experiment. is given by
HSP
Z Z
Z
( ) A r
1 2
,
(

(

( ) ( )
( ) ( )

d r I
r r l l
N j
Z

2
2
0
2
6 2
1 2
2
1 2
2
1
(2.56)


32

Inversion method

Using (2.56) to compute for a variety of layer thickness and
conductivity, Moulder et al. obtained an optimal set of parameters for which
the theoretical curve was the closest to the experimental data. The least-
squares norm was their measure of closeness. Specifically, a cost function is
defined as
Z

( )

=

N
I
theory
R R Q
1
2
exp
(2.57)

Where, the sum is carried out over N frequencies (typically N=20). Then Q
was minimized through a simplex direct-search procedure. The residual Q
depends only on the resistive component of the impedance data. We found
that the uncertainty in the coil geometry (the precision of coil wiring)

33

strongly affects the reactive (inductive) components of the impedance.
Consequently, they focused the inversion efforts on , which seemed to be
less sensitive to these model errors.
R


34

Das könnte Ihnen auch gefallen