Professional Documents
Culture Documents
d dω ω d dω ω
d dω ω d dω ω
d dω ω d dω ω
= ( + i)u
n is
measured. (left)
the horizontal extent of subsurface anomalies, while the depth from the concrete
surface or thickness of a void cannot be determined.
The purpose of this work is to propose a new quantitative NDE technique for
estimating the thickness of voids within in-situ concrete structures. This new
concrete void detection and monitoring method is called frequency dierential
electrical impedance scanning (fdEIS). In fdEIS, we place a scan probe and a
reference electrode on the opposite sides of the evaluated concrete wall as shown
in Figure 1. A sinusoidal voltage of V
0
sin t with various frequencies f =
2
in
the range of 10Hz to 100kHz is applied through the reference electrode, while
the scan probe is kept at the ground potential. This voltage dierence produces
electric current owing through the tested wall. Then the corresponding time
harmonic electric potential u
(x) = V
0
on the reference electrode
(1.1)
where = (x, ) is the conductivity distribution and = (x, ) is the permit-
tivity distribution of the concrete wall. The scan probe consists of a planar array
of circular electrodes to measure the exit current g
= ( +i)u
n, where
n is the unit normal vector to the probe.
The key idea of fdEIS is to use the frequency dierential Neumann data
d
d
g
g
+
g
[
=0
_
1
,
where :=
0
and
0
8.85410
12
[F/m] is the permittivity of the free space.
We demonstrate the performance of our method in numerical simulations.
2. Frequency differential electrical impedance scanning
Let the tested concrete wall occupy a rectangular parallelepiped region with
boundary . Let and be the portions of on which the scan probe and
the reference electrode are placed, respectively, as in Figure 1. Without loss of
generality, we let x
3
be the label of the axis normal to both and , and let
= (x
1
, x
2
, 0) :
_
x
2
1
+ x
2
2
< l and = (x
1
, x
2
, ) :
_
x
2
1
+ x
2
2
< l for some
> 0. We suppose that there is an air void D inside and that and are
constant in
D:
+ i =
_
i
0
in D
c
+ i
c
, in
D.
(2.1)
For a realistic example, we may set
c
= 10
6
[S/m],
0
= 8.8 10
12
[F/m],
c
= 15
0
[F/m].
When we apply a sinusoidal voltage of V
0
sin t with the scan probe kept
zero potential, the resulting time harmonic electric potential u
inside is the
H
1
()-solution to the following mixed boundary problem
_
_
(( + i)u
(x)) = 0 in
u
(x) = 0, x
u
(x) = V
0
, x
( + i)u
(x) n(x) = 0, x ( )
(2.2)
where n is the unit outward normal vector to the boundary . The existence
and uniqueness of the problem (2.2) are obvious [10]. Denoting the real and
imaginary parts of u
by v
= 'u
and h
= u
_
(v
) (h
) = 0 in
(v
) + (h
) = 0 in
v
= 0 and h
= 0 on
v
= V
0
and h
= 0 on
n v
= 0 and n h
= 0 on ( ).
(2.3)
Using the scan probe, we measure the Neumann data g
(x) := n (v
(x) + h
(x))
. .
real part
+i n (h
(x) v
(x))
. .
imaginary part
, x .
4 SUNGWHAN KIM , JIN KEUN SEO, AND TEAYOUNG HA
Figure 2. 1-D fdEIS Model
In fdEIS, we measure the Neumann data g
g
+
g
on . (2.4)
Now, we try to relate
d
d
g
(x
3
)
_
= 0, x
3
(, 0)
u
() = V
0
and u
(0) = 0.
(2.5)
In this 1-D model, it follows from the transmission condition that
g
(0) = (
c
+ i
c
)
du
+
dx
3
(0)
= (
c
+ i
c
)
du
+
dx
3
(a + d) (2.6)
= i
0
du
dx
3
(a + d)
= i
0
u
(a + d) u
(a)
d
,
NONDESTRUCTIVE EVALUATION METHOD OF CONCRETE VOIDS 5
where u
+
= u
[
\
D
and u
= u
[
D
. When = 0 in (2.3), h
0
= 0 in and v
0
satises
_
_
_
v
0
= 0 in
D
v
0
[
= V
0
& v
0
[
= 0
n v
= 0 on D ( ( )) .
(2.7)
Since v
0
is not dened inside D, we extend v
0
into D in the way that v
0
inside D
is the solution of the Laplace equation in D with the Dirichlet data v
0
[
D
. Let
u
0
= v
0
+ ih
0
in and g
0
=
v
0
u
0
in H
1
() as 0. Since u
(a) V
0
and u
(a + d) 0 as 0, we have
d
d
g
(0)[
=0
:= lim
0
g
(0) g
0
(0)
=
i
0
V
0
d
. (2.8)
This explains the key observation providing the explicit formula for the thickness
of D.
Theorem 2.1. [Fundamental principle in fdEIS] Assume g
is the measured
Neumann data corresponding to 1-D model (2.5). Let =
0
. Then
d
d
g
(0)[
=0
=
_
V
0
/d if D ,= .
0 if D = .
(2.9)
Note that if
2
100kHz, then =
0
8.854 10
12
10
4
.
2.2. 3-D model. The fundamental principle in 1-D fdEIS in the previous section
explains how the frequency map g
(, ) =
V
0
_
[g
[ds
.
According to Theorem 2.1, we can expect the following properties.
The distance dist
[g
.
Now, we consider a general 3-D model and try to derive a quantitative relation
between
d
d
g
and h
changes
with respect to . As far as we know, the fundamental theory of the maximum
6 SUNGWHAN KIM , JIN KEUN SEO, AND TEAYOUNG HA
0 0.2 0.4 0.6 0.8 1 1.2
0
0.05
0.1
0.15
0.2
0 0.2 0.4 0.6 0.8 1 1.2
0
0.05
0.1
0.15
0.2
(a-1) {( + i
1
)u
1
} (b-1) {( + i
2
)u
2
}
0 0.2 0.4 0.6 0.8 1 1.2
0
0.05
0.1
0.15
0.2
0 0.2 0.4 0.6 0.8 1 1.2
0
0.05
0.1
0.15
0.2
(a-2) {( + i
1
)u
1
} (b-2) {( + i
2
)u
2
}
Figure 3. The vector elds of real and imaginary parts of (a)
( + i
1
)u
1
with
1
2
= 10Hz and (b) ( + i
2
)u
2
with
2
2
= 10
5
Hz. The curved arrows in (a-1) and (b-1) mean that the
electrical current cannot pass through the void at the frequency
10Hz, while it can pass though the void at the higher frequency 10
5
Hz.
principle for the equation (2.3) is not known although the following maximum
principle should be true:
sup
x
_
[v
(x)[
2
+[h
(x)[
2
V
0
. (conjecture)
The main reason why the above maximum principle is true is that the amplitude
of an internal voltage should be less than the amplitude of the applied voltage.
Moreover, we cannot use the standard minimizing technique using the variational
principle since u
'(g
)ds = min
vH
re
max
hH
im
_
_
[v[
2
2v h [h[
2
dx (2.10)
and
V
0
_
(g
)ds = min
vH
re
max
hH
im
_
_
[v[
2
+ 2v h [h[
2
dx, (2.11)
where H
re
:= v H
1
() : v[
= 0, v[
= V
0
,
v
n
[
\()
= 0 and H
im
:=
h H
1
() : h[
= 0,
h
n
[
\()
= 0.
We try to nd a 3-D counterpart of (2.6). The following lemma is useful to
understand an interrelation between
d
d
g
and D.
NONDESTRUCTIVE EVALUATION METHOD OF CONCRETE VOIDS 7
Lemma 2.2. The complex potential u
= v
+ ih
satises
V
0
_
n
ds =
1
2
c
+
2
2
c
_
c
+
2
c
_ _
[v
[
2
+[h
[
2
_
and
V
0
_
n
ds =
1
2
c
+
2
2
c
_
(
c
c
)
_
[v
[
2
+[h
[
2
_
.
Proof. Since v
and h
[v
[
2
h
dx = V
0
_
c
v
n
c
h
n
_
ds
_
+ [h
[
2
dx = 0
and
_
[h
[
2
dx = 0
_
[v
[
2
+ h
dx = V
0
_
c
v
n
+
c
h
n
_
ds.
The above identities yield the followings
_
_
[v
[
2
+[h
[
2
_
= V
0
_
c
v
n
c
h
n
_
ds
_
_
[v
[
2
+[h
[
2
_
= V
0
_
c
v
n
+
c
h
n
_
ds
which lead to
_
c
+
2
c
_ _
[v
[
2
+[h
[
2
_
=
_
2
c
+
2
2
c
_
V
0
_
n
ds
_
(
c
c
)
_
[v
[
2
+[h
[
2
_
=
_
2
c
+
2
2
c
_
V
0
_
n
ds
This completes the proof.
The following theorem shows the -dierentiability of the solution u
of (2.2)
in and reveals the denite relation between the frequency dierential map
d
d
g
d
is the H
1
()-solution of the following mixed boundary value
problem:
_
_
_
_
( + i)
du
d
_
= i (u
) in
du
d
= 0 on
( + i) n
_
du
d
_
= 0 on ( ).
(2.12)
8 SUNGWHAN KIM , JIN KEUN SEO, AND TEAYOUNG HA
Moreover,
d
d
g
=0
satises
_
d
d
g
=0
ds
_
c
g
0
ds =
1
V
0
_
D
[v
0
[
2
dx. (2.13)
Proof. Suppose u
1
and u
2
are the solutions of (2.2) with =
1
and =
2
,
respectively. Then u
2
u
1
satises
(( + i
2
)(u
2
u
1
)) = i(
2
1
) (u
1
) in
with the boundary conditions
_
u
2
u
1
= 0 on
n (u
2
u
1
) = 0 on ( ).
Hence,
u
2
u
1
can be viewed as the weak solution of the following :
_
_
( + i
2
)
_
u
2
u
1
__
= i (u
1
) in
u
2
u
1
= 0 on
( + i
2
) n
_
u
2
u
1
_
= 0 on ( ).
(2.14)
Hence, (2.12) follows from the standard limiting process using (2.14).
Next, we consider the frequency derivative of g
. Since g
=
c
v
n
c
h
n
on , Lemma 2.2 implies
V
0
_
ds =
_
_
[v
[
2
+[h
[
2
_
dx. (2.15)
Using the boundary condition for v
0
, we have
V
0
_
c
g
0
ds =
_
\
c
[v
0
[
2
dx
From (2.15) and the above identity, we have
V
0
_
ds V
0
_
c
g
0
ds
=
_
_
[v
[
2
+[h
[
2
_
dx
_
\
c
[v
0
[
2
dx
=
_
D
0
[v
0
[
2
dx +
_
_
[v
[
2
[v
0
[
2
_
dx +
_
[h
[
2
dx.
Dividing the above identity by
0
yields
_
ds
_
c
g
0
ds =
1
V
0
_
D
[v
0
[
2
dx + Err
is
Err
=
1
V
0
_
0
_
[v
[
2
[v
0
[
2
_
dx +
1
V
0
_
0
[h
[
2
dx.
NONDESTRUCTIVE EVALUATION METHOD OF CONCRETE VOIDS 9
2l
a
c
d b D
a
c
d b D
2l
a
b
c
2l
d
D
(a) (b) (c)
Figure 4. fdEIS models with three dierent types of voids. (a)
internal rectangular void (b) internal non-rectangular void (c) sur-
face breaking void
The identity (2.13) follows from the fact that Err
0 as 0.
We try to understand the identity (2.13) in terms of the thickness of the air
void D. We consider the special case of and D in Figure 4 (a) in two dimension
and denote the width of D and by c and a, respectively. It is easy to see that
if c = a, then
1
V
0
_
D
[v
0
[
2
dx =
V
0
c
d
and
_
g
0
ds = 0.
In the case where l < c < a, we expect
1
V
0
_
D
[v
0
[
2
dx
V
0
d
since v
0
0 at the
center of the top line of the void D and v
0
V
0
at the center of the bottom line
of D. Hence,
1
V
0
_
d
d
g
[
=0
ds
1
d
. Various numerical simulations in gure 5
in the next section show that
d
d
g
(0)[
=0
V
0
d
(2.16)
The above approximation can be viewed as an extension of the one dimensional
result (2.9) in Theorem 2.1. However, we are not able to provide a rigorous
support for the above approximation due to little knowledge on the equation
(2.3). The proposed new approach is just at the beginning stage and further
studies are needed to develop theoretical underpinnings of this method.
3. Numerical simulations
In this section, we present numerical results to test the feasibility of the method
using (2.16) for three types of voids shown in Figure 4.
3.1. Internal voids. We set a rectangular model := (0.6, 0.6)(0, 0.2) m
2
,
the scan probe region := (x, 0) [ 0.3 < x < 0.3, and the reference electrode
:= (x, 0.2) [ 0.3 < x < 0.3. See Figure 4 (a) and (b). We assumed
that
D is homogeneous and
c
,
c
,
0
are constants. We set
c
= 10
6
[S/m],
10 SUNGWHAN KIM , JIN KEUN SEO, AND TEAYOUNG HA
0.6 0.4 0.2 0 0.2 0.4 0.6
100
300
500
700
d1=0.01
d3=0.03
d5=0.05
0.2 0.1 0 0.1 0.2
20
25
50
100
33.3
d1=0.01
d2=0.02
d3=0.03
d4=0.04
d5=0.05
1/d1
1/d5
1/d4
1/d3
1/d2
(a) (b)
Figure 5. The graph of the frequency dierential
d
d
g
along the
top line of shown in Figure 4 (a) with a = 1.2, b = 0.2, c = 0.8,
and ve dierent d from 0.01 to 0.05. (a) The graphs of
d
d
g
0
= 8.8 10
12
[F/m],
c
= 15
0
[F/m]. We used FEM by discretizing the
2D model with 1200 200 rectangular elements to solve the following mixed
boundary value problem:
_
_
(( + i)u
(x)) = 0 in
u
(x) = 0, x
u
(x) = 1, x
( + i)u
(x) n(x) = 0, x ( ).
(3.1)
We computed u
= (
c
+ i
c
)
u
x
3
on = (x, 0) [ 0.3 < x < 0.3. We
evaluated the frequency dierential
d
d
g
2
g
1
(
2
1
)
0
(
2
= 2 10
3
,
1
= 2 10)
Figure 5 shows the graphs of
d
d
g
on
(x, 0) [ 0.2 < x < 0.2 with with ve dierent thick-
ness d = 0.01, 0.02, 0.03, 0.04, and 0.05. The tested void is
D = (x, y) [ 0.5 < x < 0.3 and 0.05 sin
(x+0.1)
0.4
d
2
< y+0.1 <
0.05 sin
(x+0.1)
0.4
+
d
2
(Left).
correspond to d = 0.01, 0.02, 0.03, 0.04, and 0.05, respectively. The graphs show
that the quantity
d
d
g
[
=0
at the center of
1
d
Next, we performed the same numerical simulation by varying the shape and
the location of D. Figure 6 shows the corresponding graphs of the frequency dif-
ferential
d
d
g
to the void D = (x, y) [ 0.5 < x < 0.3 and 0.05 sin
(x+0.1)
0.4
d
2
< y+0.1 < 0.05 sin
(x+0.1)
0.4
+
d
2
. The graphs in Figure 6 also show the similar
behavior as in that in Figure 5.
3.2. Surface-breaking voids. Surface-breaking voids mainly occur in concrete
pavements and bridge decks. They tend to propagate downward so that it is
important to investigate how deep and how wide they grow. We applied the
fdEIS method to surface-breaking voids. In this case, the scan probe and the
reference electrode are located on either side of a surface-breaking void as
Figure 4 (c), and the Neumann data g
on is closely related to v
1
} (b-1) {( + i
2
)u
2
}
0.6 0.4 0.2 0 0.2 0.4
0
0.1
0.2
0.6 0.4 0.2 0 0.2 0.4
0
0.1
0.2
(a-2) {( + i
1
)u
1
} (b-2) {( + i
2
)u
2
}
Figure 7. The vector elds of real and imaginary parts of (a)
( + i
1
)u
1
with
1
2
= 10 Hz and (b) ( + i
2
)u
2
with
2
2
= 10
4
Hz. The curved arrows in (a-1) and (b-1) mean that the
electrical current cannot pass through the void at the frequency
10Hz, while it can pass though the void at the higher frequency 10
4
Hz.
(x, 0) [ 0.2 < x < 0.4 and the reference electrode := (x, 0) [ 0.4 < x <
0.2.
Figure 7 shows how the distance between and is related with the ow of
the current density ( +i)u
[
=0
_
1
.
Unlike in an internal void, though
d
d
g
are not close to 1/d. We are not able to explain the reason why this
happens. This work is just at the beginning stage, and further theoretical and
numerical studies are required to understand the fdEIS method.
Key words : concrete void detection, non-destructive evaluation, fdEIS AMS
Subject Classications: 35R30, 65N21
References
[1] Guide book on non-destructive testing of concrete structures, IAEA-TCS-17(2002), pp. 1-
231.
NONDESTRUCTIVE EVALUATION METHOD OF CONCRETE VOIDS 13
0.24 0.26 0.28 0.3
20
25
33.3
50
100
d1=0.01
d2=0.02
d3=0.03
d4=0.04
d5=0.05
1/d1
1/d2
1/d3
1/d4
1/d5
Figure 8. The graph of the frequency dierential
d
d
g
on
(x, 0) [ 0.23 < x < 0.3 with with ve dierent thick-
ness d = 0.01, 0.02, 0.03, 0.04, and 0.05. The tested void is
D = (
d
2
,
d
2
) (0.15, 0).
[2] C. Colla, P. Das, D. Mccann, M. Forde, Sonic, electromagnetic & impulse radar investi-
gation of stone masonry bridges , J. Non-destructive testing and evaluation International,
30(1997), no 4, pp. 249-254.
[3] A. Alexander, H. Thorton, Ultrasonic pitch-catch and pulse-echo measurements in concrete,
Non-destructive Testing of Concrete (Lew, H. S., Ed.), ACI SP-112(1989), no 21.
[4] M. Sansalone, W. Streett, Detecting Honeycombing, the Depth to Surface Opening Cracks
and Ungrouted Ducts using the Impact-Echo Method, Concrete Int., 10 (1988), no 4.
[5] W. Dover and R. Collings, Recent advances in the detection and sizing of cracks using
alternating current eld measurements (A.C.F.M.), Br. Journ. of NDT, (1980), pp. 298.
[6] N. Carino, M. Sansalone, N. Hsu, A Point Source-point Receiver, Pulse-Echo Technique for
Flaw Detection in Concrete, ACI Journal, 83 (1986), no. 2, pp. 199-208.
[7] H. Ammari, O. Kwon, J.K. Seo, and E.J. Woo, T-Scan electrical impedance imaging system
for anomaly detection, SIAM J. Appl. Math., 65(2004), pp. 252-266.
[8] L. Borcea, EIT Electrical impedance tomography, Inverse Problems, 18 (2002), no. 6, R99
R136.
[9] M. Cheney, D. Isaacson, and J. C. Newell, Electrical impedance tomography, SIAM Review,
41(1999), pp. 85101.
[10] Somersalo, E., M. Cheney and D. Isaacson. Existence and uniqueness for electrode models
for electric current computed tomography, SIAM J. Appl. Math. 52(1992) pp.1023-1040, .
[11] A. V. Cherkaev, and L. V. Gibiansky, Variational principles for complex conductivity,
viscoelasticity, and similar problems in media with complex moduli, J. Math. Phys., 35
(1994), no. 1, 127145.
[12] M.H. Choi, T.J. Kao, D. Isaacson, G.J. Saulnier, and J.C. Newell. , Simplied Model of a
Mammography Geometry for Breast Cancer Imaging with Electrical Impedance Tomography,
Proc. IEEE-EMBS Conf. 26, 2005.
14 SUNGWHAN KIM , JIN KEUN SEO, AND TEAYOUNG HA
Division of Liberal art, Hanbat National University, Korea
Department of Mathematics, Yonsei University, Korea
National Institute for Mathematical Sciences, Korea