You are on page 1of 6

J ournal of Mechanics Engineering and Automation 4 (2014) 40-45

Dynamic Detection of Reinforced Concrete Bridge


Damage by Finite Element Model Updating
Mohammed Lamine Moussaoui
1
, Mohamed Chabaat
2
and Abderrahmane Kibboua
3
1. Faculty of Mechanical and Chemical Engineering, University of Sciences and Technology Houari Boumediene, Algiers 16111,
Algeria
2. Faculty of Civil Engineering, University of Sciences and Technology Houari Boumediene, Algiers 16111, Algeria
3. National Earthquake Engineering Research Center (CGS), Hussein Dey, Algiers 16000, Algeria

Received: October 7, 2013 / Accepted: November 5, 2013 / Published: J anuary 25, 2014.

Abstract: The present work consists of dynamic detection of damages in reinforced concrete bridges by using a MMUM
(mathematical model updating method) from incomplete test data. A well suited finite element model of a repaired bridge is carried out.
The diagnosis enables us to locate and detect the damage in a reinforced concrete bridge. Thus, developments of analytical predictions
have been checked by modal testing techniques. Besides, the FTCS (finite time centered space) scheme is developed to solve the set of
equations which can easily handle finite element matrices of a bridge model. It is shown in this study that the method is applied to detect
damages as well as existing cracks in real time of a repaired bridge. To check the efficiency of the method, the repaired bridge of
OuedOumazer in Algeria has been selected. It is proven that identification methods have been able to detect the exact location of
damage areas to be corrected avoiding the inaccuracy from the finite element model for the mass, stiffness and loading.

Key words: Bridges, damage, detection, MMUM, finite elements, FTCS, correlation.

1. Introduction


Bridge structures are usually analyzed by using
dynamic equations of motion which are different from
previous works [1, 2] considering free vibrations of
spacecraft structures and neglecting structural damping.
In this research, we have developed a method by using
the FTCS scheme to solve the reduced set of equations
of the repaired multi-cracked OuedOumazers bridge
(Fig. 1). Finite element matrices are generated from a
finite element calculus code. The geometric
information of tetrahedral elements and mesh model
are obtained from Ansys software and the material
properties of reinforced concrete include law of
mixtures of concrete and steel [3]. MMUM updating
techniques are helpful to analyze large structural
complex systems of equations. The well suited finite

Corresponding author: Mohammed Lamine Moussaoui, Ph.D.,
research fields: damage detection, mathematical model updating
and mechanical engineering. E-mail: moussaouiml8@gmail.com.
element mesh of the repaired bridge is used to detect
exact locations of damage areas. The structure can
generate rigid body modes, corresponding to zero
frequencies as in Guyans transformation, added to
calculated flexible modes.
2. Stiffness, Damping and Mass Matrices
Using Barycentric coordinates of any regular
tetrahedron defined by Cartesian coordinates of its
vertices, we have: r
i
=
Vi
/V for i =1, 4, V = I

4
=1

at a P(x, y, z) point e Tetrahedron, V
1
=V
P234
, V
2
=
V
P134
, V
3
= V
P124
and V
4
= V
P123
. The degrees of
freedomare {J]
(121)
= _
u

_
(121)
and the displacements
inside this element are given with u = N

4
=1
=
r

4
=1
, : = r

4
=1
, w = r

4
=1
at the
corresponding P(x, y, z) point. Therefore, the
displacements of P become: _
u
:
w
_
(31)
= |N]
(312)
{J]
(121)

DAVID PUBLISHING
D
Dynamic Detection of Reinforced Concrete Bridge Damage by Finite Element Model Updating

41

Fig. 1 Cracked bridge before repair.

where, the matrix N of shape functions takes the
following form:
_
r
1
u u
u r
1
u
u u r
1
r
2
u u
u r
2
u
u u r
2
r
3
u u
u r
3
u
u u r
3
r
4
u u
u r
4
u
u u r
4
_
If we let _
x
y
z
1
_ = _
x
1
x
2
y
1
y
2
x
3
x
4
y
3
y
4
z
1
z
2
1 1
z
3
z
4
1 1
_ _
r
1
r
2
r
3
r
4
_ (1)
Then, we can find r
i
without handling the constraint
r
i
= 1
4
i=1
. We finally obtain:
_
x
y
z
_ = _
(x
1
-x
4
) (x
2
-x
4
) (x
3
-x
4
)
(y
1
- y
4
) (y
2
-y
4
) (y
3
-y
4
)
(z
1
-z
4
) (z
2
-z
4
) (z
3
-z
4
)
_ _
r
1
r
2
r
3
_ + _
x
4
y
4
z
4
_ (2)
Since r
4
= 1 -r
1
-r
2
-r
3
, then,
_
ux
uy
uz
_ = _
(x
1
-x
4
) (x
2
- x
4
) (x
3
- x
4
)
(y
1
-y
4
) (y
2
-y
4
) (y
3
- y
4
)
(z
1
-z
4
) (z
2
- z
4
) (z
3
- z
4
)
_ _
ur
1
ur
2
ur
3
_ =
= |[] _
ur
1
ur
2
ur
3
_ (3)
where, J is the J acobian. We have then uI =
obs| [ |ur
1
ur
2
ur
3
. Using the planes equation as
shown in Fig. 2, one can solve the four following sets
of equations:
_
x
1
y
1
x
2
y
2
z
1
1
z
2
1
x
3
y
3
x
4
y
4
z
3
1
z
4
1
_ _
o

_ = _
o
1
o
2
o
3
o
4
_ i = 1, 4 (4)
such that:
o
1
= 1 onJ o
2
= o
S
= o
4
= u or i = 1
o
2
= 1 onJ o
1
= o
S
= o
4
= u or i = 2
o
S
= 1 onJ o
1
= o
2
= o
4
= u or i = S
o
4
= 1 onJ o
1
= o
2
= o
S
= u or i = 4
.
The Gauss LU method with partial pivoting allows
to obtain the Barycentric coordinates versus Cartesian
coordinates with analytical geometry and without
inverting the matrix of the initial sets of equations. This
is a new and efficient method for determining the shape
functions of the constant strain tetrahedron. Here, the
inversion is not recommended numerically because it is
a source of roundoff errors and ill-conditioning
problems [4] for arbitrary shaped elements.
The strains inside the tetrahedron are given by:
{e] = |e
xx
, e

, e
zz
, e
z
, e
zx
, e
x
]
1
= |B]{u]
where, e
]
=
1
2
(u
,]
+ u
],
) in linear elasticity and 2e
]



Fig. 2 Barycentric coordinates of a Tetrahedron.
Dynamic Detection of Reinforced Concrete Bridge Damage by Finite Element Model Updating

42
= y
]
.
The |B]
(612)
matrix becomes:
l
l
l
l
l
l
l
o
1
u u
u b
1
u
u u c
1
o
2
u u
u b
2
u
u u c
2
u
c
1
2
b
1
2
c
1
2
u
c
1
2
b
1
2
c
1
2
u
u
c
2
2
b
2
2
c
2
2
u
c
2
2
b
2
2
c
2
2
u
o
3
u u
u b
3
u
u u c
3
o
4
u u
u b
4
u
u u c
4
u
c
3
2
b
3
2
c
3
2
u
c
3
2
b
3
2
c
3
2
u
u
c
4
2
b
4
2
c
4
2
u
c
4
2
b
4
2
c
4
2
u
1
1
1
1
1
1
1

such that

i
x
= o

= b

and

i
z
= c

.
The elementary (1212) stiffness matrix can be
written as
|K
c
] = ] |B]
1
|B]|B]

JI = |B]
1
|B]|B]I
c
(5)
where, |B(r
1
, r
2
, r
3
)] is constant and I
c
= ] uI

.
The elementary (1212) mass matrix is given as
follows:
|H
c
] = _ p|N]
1
|N]

uI =
p. obs| [ | ] |N]
1
|N]

ur
1
ur
2
ur
3
(6)
=
l
l
l
l
l
l
l
l
l
l
l
l
l
l
l
l
l
l
l
l
l
l
l
l
l
l
m
c
1u
u u
u
m
c
1u
u
u u
m
c
1u
m
c
2u
u u
u
m
c
2u
u
u u
m
c
2u
m
c
2u
u u
u
m
c
2u
u
u u
m
c
2u
m
c
1u
u u
u
m
c
1u
u
u u
m
c
1u
m
c
2u
u u
u
m
c
2u
u
u u
m
c
2u
m
c
2u
u u
u
m
c
2u
u
u u
m
c
2u
m
c
2u
u u
u
m
c
2u
u
u u
m
c
2u
m
c
2u
u u
u
m
c
2u
u
u u
m
c
2u
m
c
2u
u u
u
m
c
2u
u
u u
m
c
2u
m
c
2u
u u
u
m
c
2u
u
u u
m
c
2u
m
c
2u
u u
u
m
c
2u
u
u u
m
c
2u
m
c
2u
u u
u
m
c
2u
u
u u
m
c
2u
m
c
1u
u u
u
m
c
1u
u
u u
m
c
1u
m
c
2u
u u
u
m
c
2u
u
u u
m
c
2u
m
c
2u
u u
u
m
c
2u
u
u u
m
c
2u
m
c
1u
u u
u
m
c
1u
u
u u
m
c
1u
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1

3. Material Properties Matrix
The reinforced concrete is mainly composed by two
materials matrix-fibers such as concrete (m) and steel
(f) to form a mixture of materials. According to
mixture laws [3], their properties are related as
p = p
m
:
m
+ p
]
:
]
the density, :
m
, :
]
for volumes
fractions, E
x
= E
m
:
m
+ E
]
:
]
the Youngs modulus
which can be rewritten in the y and z directions as
E

= E
m
E
]
(E
]
:
m
+ E
m
:
]
) and E
z
= E


respectively. It is important to notice that the rule of
mixtures is a good fit to experimentation. A similar
formulation has been applied to shell elements by
Ferreira [5]. v
x
= v
m
:
m
+ v
]
:
]
for Poissons ratio,
v
yx
=v
xy
E
x
/E
y
, v
yz
=v
zy
E
z
/E
y
,v
zx
=v
xz
E
x
/E
z
,0
x
=
0
m
0
]
(0
]
:
m
+ 0
m
:
]
) for shear modulus. It is
assumed that 0
z
= 0
zx
= 0
x
. In this case, the
material is orthotropic with two planes of symmetry
with 9 constants for |] material properties matrix
given in Eq. (7):
|]
(66)
= _
|
uu
] | u ]
| u ] |
bb
]
_ (7)
Thus, the generalized Hookes law is given as
follows:
_
e
xx
e

e
zz
_ =
l
l
l
l
l
l
l
1
E
x
-
v
x
E

-
v
xz
E
z
-
v
x
E
x
1
E

-
v
z
E
z
-
v
zx
E
x
-
v
z
E

1
E
z 1
1
1
1
1
1
1
_
o
xx
o

o
zz
_
= |
uu
i
] _
o
xx
o
yy
o
zz
_ (8)
_
e
z
e
zx
e
x
_ =
l
l
l
l
l
l
l
1
20
z
u u
u
1
20
zx
u
u u
1
20
x1
1
1
1
1
1
1

_
o
z
o
zx
o
x
_ = j
bb
'
[ _
o
z
o
zx
o
x
_ (9)
such that o
]
=
]kI
e
kI
.
Here, |
uu
] and |
bb
] matrices are obtained from
an inverting process with the same algorithm Gauss LU
algorithm verifying some constraint equations as
conditions to be satisfied in order to respect the inverse
process.
Dynamic Detection of Reinforced Concrete Bridge Damage by Finite Element Model Updating

43
4. Mathematical Concepts
From Ref. [6], dynamic equilibrium equations in
forced vibrations are given as follows:
|H]|J

(t)| + |C]|J

(t)| + |K]{J(t)]
= {R(t)] (1u)
where, [M], [C] and [K] are, respectively the mass,
damping and stiffness matrices of the structure.
{u(t)] is the (n 1) time displacement vector and
n = (m+u) is the degrees of freedom number.
Besides, |C] o|K] + [|H] is the linearized
damping matrix [8]. Since critical damping ratio of
reinforced concrete is taken as
c
= S% |7], then;
o = 2(
2

2
-
1

1
)(
2
2
-
1
2
) =0.00335 and [ =
2
1

2
(
1

2
-
2

1
)(
2
2
-
1
2
) = 0.744 for two
frequencies f
1
= 2.25 Hz and f
2
= 2.50 Hz
corresponding to
1
=
2
= 0.05 [8]. It is important to
notice that the structural damping must not be
neglected for reinforced concrete structures like
bridges. Eq. (10) can be solved effectively by the FTCS
scheme [9] in which
|u

(t)| =
1
t
2
({u(t)]
(t+At)
- 2{u(t)]
(t)
+{u(t)]
(t-At)
)
+0(t
2
) (11)
|u

(t)| =
1
2t
({u(t)]
(t+At)
- {u(t)]
(t-At)
) +0(t
2
)
for the molecule:
(t - t), (t), (t + t) (12)
The set of equations is partitioned into (m) measured
and(u) unmeasured data as
_
K
mm
K
um
K
mu
K
uu
_ _
u
m
u
u
_
(t)
+ [
u
2At
_
K
mm
K
um
K
mu
K
uu
_ __
u
m
u
u
_
(t+At)
-
_
u
m
u
u
_
(t-At)
_
+[
[
2At
_
H
mm
H
um
H
mu
H
uu
_ __
u
m
u
u
_
(t+At)
-
_
u
m
u
u
_
(t-At)
_
+
1
(At)
2
_
H
mm
H
um
H
mu
H
uu
_ __
u
m
u
u
_
(t+At)
- 2 _
u
m
u
u
_
(t)
+
_
u
m
u
u
_
(t-At)
_
= _
R
m
R
u
_
(t)
(13)
The second subset in Eq. (13) gives the following
system:
|A

uu
]{u
u
]
(t+At)
= {R

] (14)
where,
|A

uu
] = [
o
2t
|K
uu
] +_
[
2t
+
1
(t)
2
] |H
uu
]
{R

] = {R
u
]
(t)
- _|K
mu
] -
2
(t)
2
|H
mu
]] {u
m
]
(t)
- _|K
uu
] -
2
(t)
2
|H
uu
]] {u
u
]
(t)
+
-_[
o
2t
|K
mu
] +_
[
2t
] |H
mu
] -
1
(t)
2
|H
mu
]_{u
m
]
(t+At)

+_[
o
2t
|K
mu
] +_
[
2t
] |H
mu
] -
1
(t)
2
|H
mu
]_{u
m
]
(t-At)

+_[
o
2t
|K
uu
] +_
[
2t
] |H
uu
] -
1
(t)
2
|H
uu
]_{J
u
]
(t-At)

which can be solved by the following algorithm:
t = u: {J]
(0)
= jJ
m
(0)
, J
u
(0)
[
1
, |J

|
(u)
known initial
conditions. Calculate |J

|
(0)
with |H]|J

|
(0)
= {R

]
(0)
=
{R]
(0)
- |K]{J]
(0)
-|C]|J

|
(0)
, t = -t: compute
then {u]
(-At)
= {u]
(0)
-(At)|u

|
(0)
+
(At)
2
2
|u

|
(0)
t =
t , 2t , . . calculate {u
u
]
(t+At)
in Eq. (14).
The data acquisition of {J
m
]
(t)
array is carried out
by the CityShark II station [10, 11] with a wide range
of ambient vibration recordings (Fig. 3). This method
enables to detect any variation of the computed
displacements between two measurement cases. It is
possible to connect the hardware to the structure and
detect in real-time the exact location of damaged areas.
Sparse matrix methods are helpful to handle the system
in Eq. (14).
5. Conclusions
In this research, we have developed a method by
using the finite time centered space scheme to solve the
reduced set of equations of the repaired multi-cracked
bridge. Finite element matrices are generated from a
finite element calculus source code. The geometric
information of tetrahedral elements and mesh model
are obtained from Ansys software (Figs. 4 and 5). It
proves in this research work that MMUM updating
techniques are helpful to analyze large structural complex
Dynamic Detection of Reinforced Concrete Bridge Damage by Finite Element Model Updating

44

Fig. 3 Seismometers connected to the CityShark II Station.


Fig. 4 Mesh of the bridge, first view.


Fig. 5 Mesh of the bridge, second view.

systems of equations and a well suited finite element
mesh of the repaired bridge is used to detect exact
locations of damage areas. Besides, unmeasured
displacements are obtained from expanded measured
displacements for which a number of DOF is limited.
Handling the dynamic displacements response is more
accurate than working initially with the eigenvectors
which have to be used in a superposition process.
Dynamic Detection of Reinforced Concrete Bridge Damage by Finite Element Model Updating

45
References
[1] B. Caesar, Analysis test correlation of dynamic
mathematical models, in: 2nd Int. Symp. Aeroelast. and
Str. Dynamics, Aachen, 1985.
[2] M. Imregun, W.J . Visser, A review of model updating
techniques, The Shock and Vibration Digest 23 (1991)
9-20.
[3] S.W. Tsai, H.T. Hahn, Introduction to Composite
Materials, Technomic Publishing Company, Pennsylvania,
USA, 1980.
[4] G.H. Golub, Matrix Computations, The J ohn Hopkins
University Press, London, 1991.
[5] A.J .M. Ferreira, J .A. Figueiras, A.T. Marques, Design of a
sofa for a passenger train, in: I.H. Marshall (Ed.),
Composite Structures 4, Elsevier Appl. Science, London,
1987.
[6] M. Geradin, Vibrations Theory: Application to Structural
Dynamics, Masson, Paris, 1993.
[7] Aseismic Rules Applicable to the Domain of Civil
Engineering Structures, MDTP 2008.
[8] R.D. Cook, Concepts and Applications of Finite Element
Analysis, J ohn Wiley, New York, 1989.
[9] C.A.J . Fletcher, Computational Techniques for Fluid
Dynamics 1, Springer Verlag, New York, 1988.
[10] J .L. Chatelain, B. Guillier, P. Guguen, J . Frchet, J .
Sarrault, Ambient vibration recording for single-station,
array and building studies made simple: CityShark II,
Intern. J . of Geosc. 3 (6A) (2012) 1168-1175.
[11] A. Kibboua, Ambient Vibration Recording on the Oued
Oumazer Bridge at Tipaza, Data Processing Report, CGS,
Algiers, 2006.

You might also like