Sunteți pe pagina 1din 20

Computers and Structures 80 (2002) 417436

www.elsevier.com/locate/compstruc

A structural defect identication approach based


on a continuum damage model
D.A. Castello, L.T. Stutz, F.A. Rochinha

Solid Mechanics Laboratory, Department of Mechanical Engineering, Federal University of Rio de Janeiro, Ilha do Fundao,
P.O. Box 68503, Rio de Janeiro, RJ 21945-970, Brazil
Received 1 July 2001; accepted 9 January 2002

Abstract
This paper introduces a structural identication technique built on nite element (FE) model updating. The FE
model is parameterized by a structural parameter that continuously describes the damage in the structure, and besides,
an evolution equation of this damage parameter is presented. The model updating is accomplished by determining the
subset of this damage parameters that minimizes a global error derived from the dynamic residue vectors, which is
obtained by introducing the experimental modal properties into the original model eigenproblem. A mode-shape
projection technique is used in order to achieve compatibility between the dimension of the experimental and analytical
models. The adjusted model maintains basic properties of the analytical model as the sparsity and the symmetry, which
plays an important role in model updating-based damage identication. The verication and assessment of the current
structural defect identication is performed on a analytically derived bidimensional truss structure and on a cantilever
bidimensional EulerBernouilli beam through a virtual test simulator. This simulator is used to realistically simulate the
corrupting eects of noise, ltering, digital sampling and truncation of the modal spectrum. The eigensystem realization
algorithm along with the common-based normalized system identication were utilized to obtain the required natural
frequencies and mode shapes. 2002 Elsevier Science Ltd. All rights reserved.
Keywords: Continuum damage model; Damage identication; Finite element model updating

1. Introduction
It is needless to emphasize the importance of damage
detection techniques and health monitoring in aerospace, civil and mechanical engineering. It is essential to
determine the safety and reliability of their systems and
structures. Based on experimental modal analysis and
signal processing techniques, monitoring and interpreting changes on structural dynamic measurements can be
considered as a quite promising approach for damage

Corresponding author. Tel.: +55-21-2562-8384; fax: +5521-2562-8383.


E-mail addresses: castello@mecsol.ufrj.br (D.A. Castello),
stutz@mecsol.ufrj.br (L.T. Stutz), faro@serv.com.ufrj.br (F.A.
Rochinha).

identication and health monitoring as one can see from


the dierent applications of vibration-based damage
identication techniques in mechanical and civil engineering [4,14,30,31,37].
Most prior work on damage detection of structures is
focused on the general framework of nite element (FE)
model updating methods. These methods are intended to
identify structural damage through determining changes
in the physical property matrices that minimize an error
function of the undamaged physical property matrices
and the modal properties of the damaged structure obtained from a modal testing. The basic idea of these
approaches is that the modal properties (frequencies,
mode shapes and modal damping) are functions of these
properties of the structure (mass, stiness and damping)
and, changes in the physical properties due to damage will be reected in the modal ones, which can be

0045-7949/02/$ - see front matter 2002 Elsevier Science Ltd. All rights reserved.
PII: S 0 0 4 5 - 7 9 4 9 ( 0 2 ) 0 0 0 1 5 - 9

418

D.A. Castello et al. / Computers and Structures 80 (2002) 417436

measured and used to infer about the damage. There are


several recent review articles [6,15,25] that show the
great interest in model updating, model correction
and health monitoring. Basically, there are three classical damage identication approaches based on FE
model updating, namely: optimal matrix update, sensitivity-based matrix update and eigenstructure assignment.
Optimal matrix update methods which aims to update the global mass and stiness matrices of the original
FE model, using measured data have been used extensively in FE model updating and damage detection.
Baruch and Itzhack Bar in [3] obtained a closed-form
solution to the minimum Frobenius norm adjustment to
the mass-weighted structural stiness matrix that takes
into consideration the frequencies and the mode shapes.
An inherent drawback associated with these methods is
the fact that the zero/non-zero sparsity pattern of the
original stiness matrix may be destroyed, what might
be followed by no physical meaning results. Still based
on the optimal matrix update approach, some algorithms were proposed aiming to preserve the original
stiness matrix pattern [18,19,33,34], keeping the load
paths the same as the original models one and aiming to
minimize the rank of the adjustment matrix instead
of its norm [20]. Rather than updating the global
matrices, the sensitivity-based matrix update approaches
update physical parameters of the structure, such as
cross-sectional areas, elastic modulus and Poisson ratio,
used in the FE model denition [9]. In this kind of approach, basic properties of the original model as the
sparsity pattern and the symmetry are implicitly preserved without requiring additional interpretations of
the resulting model. Another type of method, the eigenstructure assignment, designs a pseudo-controller
that minimizes the modal force error. The controller
gains are then interpreted in terms of structural parameter modications [21]. As in the optimal matrix
update case, the sparsity pattern and the symmetry may
only be maintained through introduction of constraints.
The damage identication technical literature is very
extensive and it presents many other approaches to the
problem of structural damage identication as, for instance, the time-domain based methods, whose most of
them were developed from control theory and process
automation [11,16,29] or the frequency response function curvature method, based only on measured data,
which shows some research eort in order to minimize
the dependence of the damage identication approach
on a prior model of the undamaged structure. Methods
based on neural networks have also been proposed
[22,36].
Here, in the present work, a new structural defect
identication approach based on updating a FE model
parameterized by a continuum damage model, which
will be referred as continuum damage identication

approach (CDIA) is introduced [5,28,35]. This approach


builds on a constrained minimization of a global residue
derived from the dynamic residue vectors, obtained by
introducing the experimental modal properties into the
original model eigenproblem. Since, in practice, due to
experimental limitations, it does not exist a compatibility between the dimension of the experimental mode
shapes and the original property matrices, a mode-shape
projection technique is needed. The connectivity and
sparsity of the original model are naturally preserved,
which plays an important role in the damage identication approaches based on model updating. Indeed, the
present method can be understood as the minimization
of the global residue upon a consistent physical parameterization of the FE model of the damaged structure.
The remainder of the paper is organized as follows.
Section 2 presents the principles of the continuum
damage model to be used and the theoretical development of the CDIA. Section 3 presents the numerical
details of: the Newtons method applied to the present
problem, the parameter selection and the mode-shape
projection. Finally, Section 4 presents some illustrative
examples to assess the main characteristics of CDIA,
where some of the simulations of the damage identication where performed on a cantilever beam through
a virtual test simulator. This simulator was used in
order to provide a more realistic character to the simulations.

2. Theoretical foundations
The present section is devoted to the construction of
the modelling for the dynamics of damaged structures,
which will provide the theoretical basis for the defect
identication method introduced in this work.
2.1. A continuum damage model
In the present modelling, besides the classical variables that characterize the kinematics of a continuum
medium (displacements and velocities of material
points), an additional scalar variable b 2 0; 1 is introduced. This variable is related with the links among
material points and can be interpreted as a measure of
the local cohesion state of the material. This eld of
cohesion states describes the current state of damage
within the mechanical system as it evolves in time and
possesses as extreme values b 0 and 1. Therefore, if at
a certain time t, after a period of evolution of the system,
b 1, all the links and the initial material properties
have been preserved. On the other hand, if b 0 a local
rupture is considered since all the links among material
points have been broken. The variable b is associated to
the damage variable D [32], by the following relation:

D.A. Castello et al. / Computers and Structures 80 (2002) 417436

b 1  D. As the degradation is an irreversible phenomenon, the rate b_ must be negative or equal to zero. A
detailed presentation of the basic principles that govern
the evolution of such kind of continuum damage can be
found in [10,23,24]. Just a summary of those principles is
presented here. For the sake of simplicity, the hypotheses of isothermal processes are adopted for the damage
evolution. Besides, it is also assumed the hypothesis of
small deformation and, consequently, the conservation
of mass principle is automatically satised.
From now on the variables in bold will be used to
denote vectors and the variables between square brackets will be used to denote tensors.
Let a body B that occupies a region X  R3 with a
suciently regular boundary C be subjected at each time
instant t 2 I to external forces g : C2  I ! R3 and
b : X  I ! R3 , to external microscopic forces p : X 
I ! R, q : C2  I ! R and T
to prescribed displacements
S
ux; t 8 x 2 C1 , where C1 C2 ; and C1 C2 C.
Therefore, the principle of virtual power can be expressed as:
pint pext 0

where the internal power pint of the inertial forces, the


internal generalized forces r (the stress tensor) and
the thermodynamical forces related to the degradation
process F and H can be written as:
Z
Z
pint  qu  ^u dX  r  r^u dX
X
ZX
^
^
2
 F b H  rb dX
X

Here, () stands for the second time derivative and


^
u : X ! R3 is an element of the set Vv of the virtual
displacements such that ^u jC1 0 and b^ : X ! R is an
element of the set Vb of the virtual variations of b. The
corresponding power pext of the external generalized
forces b, g, p and q assumes the representation:
Z
Z
Z
Z
pext
b  ^u dX
g  ^u dA pb^ dX
qb^ dA
X

C2

C2

3
where p is dened as a microscopic distance force while q
is a microscopic contact force, both in duality with b.
The microscopic forces are related to non-mechanical
actions (chemical and electromagnetic, for instance) that
can cause an evolution of the damage. So, substituting
(2) and (3) in (1) yields
Z 
Z

qu  ^u r  r^u  b  ^u dX 
g  ^u dA
X
C2
Z
Z 

H  rb^ F b^  pb^ dX 
qb^ dA 0;

8 ^u 2 Vv ; 8b^ 2 Vb

Under the hypotheses of small deformations and


isothermal processes, the free energy is supposed to be a
function of the deformation e, the temperature H, the
damage variable b and its gradient rb. In order to
summarize the presentation, the thermodynamic framework used to obtain the constitutive equations is not
presented in this paper, for further details see Mattos
and Sampaio in [24]. So, the nal constitutive relations
are the following:

bE  m
treI e
1 m 1  2m
bktreI 2le

r

 m

E
tre2 e  e
21 m 1  2m
 w kb C b_ k _
b

1
ktre2 le  e  w kb C b_ kb_
2
H krb

6
7

where E is the Young modulus, m is the Poissons ratio


and k and l are the Lame constants. The terms kb and kb_
are Lagrange multipliers associated, respectively, to the
constraints b 2 0; 1 and b_ 6 0, and they are such that
the following complementary conditions are satised:
kb 6 0, bkb 0 and b_ kb_ 0. It is also introduced the
material parameters w, C and k, which represent, respectively, the minimum energy required to start the
damage process and the viscosity and diusion constants
related to damage distribution.
Introducing the constitutive equations (5)(7) in (4),
neglecting the external microscopic forces and considering the initial conditions: bx; t 0 1 8 x 2 X, the
following mathematical problem dened over the time
interval I 0; s is obtained:
Find ux; t; bx; t, the displacement eld u : X 
I ! R3 with u jC1 
ut and the cohesion eld b : X 
I ! R with bx; t 0 1 such that:
Z
q
u^
u bk div u div ^
u 2leu  e^
u dX
X
Z
Z
u dX 
g^
u dA 0; 8 ^
u 2 Vv
8
 b^
X

C2


Z 
1
krb  rb^ dX 
kdiv u2 le  e  w b^ dX
2
X
X
Z
9
 C b_ b^ dX 0; 8 b^ 2 Vb

subjected to the following constraints:

C2

419

b 2 0; 1

and b_ 6 0

420

D.A. Castello et al. / Computers and Structures 80 (2002) 417436

It is worthwhile mentioning that the damage detection method, that will be introduced later on, can be
developed from many of the damage models that already exist [32]. Indeed, at a rst sight, the method
builds simply on the kinematic parameterization of the
FE model resulting of the use of the damage model. The
need of the evolution equations, like those just presented
above, is often due to either of the two typical issues: the
development of the damage scenario which can be used
as an initial guess for the iterative numerical solution of
the damage detection problem or the analysis of the
residual life of the structure upon the damage distribution obtained by means of the proposed technique. This
complete set of equations describing the non-linear
evolution of the damage can also be explored in situations where no control can be applied to the input
signals and, as a consequence, the usual laboratorial
conditions are not accomplished. This is typical in several industrial applications. In those cases a non-linear
identication scheme in the time domain, like the one
proposed in [13], based on a realist modelling like the
one summarized in Eqs. (8) and (9) could be used.
Another important aspect that must be remarked is
highlighted in Eqs. (8) and (9), where one can see that
only the elastic terms are aected by the damage eld.
This will imply, in the nite element context, that the
mass matrix remains unchanged during the damage
evolution.
2.2. Basic equations
One can summarize the present non-destructive
damage detection procedure by four main steps: modal
testing [7], location of potentially damaged regions, assessment of severity and the real extension of the damage, and assessment of the validity of the provided result
by means of some error measures.
Generally speaking, a linear n-DOF nite element
model of an undamaged structure is represented by the
following equation of motion
M
u Ku 0

10

where M and K are the n  n analytical mass and


stiness matrices, respectively, and u is a n  1 displacement vector. Following the standard procedures,
Eq. (10) leads to the generalized eigenproblem
K  ki M/i 0 i 1; . . . ; n

where xE;i and /E are, respectively, the ith frequency


and mode shape obtained from the experimental modal
analysis.
For the undamaged structure, the n  1 vector Ri
would be zero, unless identication imprecisions or
discretization limits, which will not be considered here.
In the presence of damage, the use of the nite element
method in (8) would lead to an equation similar to (10),
possessing as major dierence the stiness matrix, which
will now depend upon the variable b, and will be denoted as Kd . In the present work, it is assumed that the
FE model is reliable and that eventual discrepancies
found in the correlation between the analytical and experimental results are mainly due to the presence of
damage in the structure. One should notice that this
variational equation involves two continuous spatial
elds, namely b and u, which must be discretized giving
rise to the vectors of nodal variables B and U. Due to
the fact that the goal of the present work is damage
identication, it is considered that the damage does not
evolve during the dynamic tests, i.e., b_ 0. In other
words, it is supposed that the level of internal forces in
the structure during the experiment does not suce to
cause the continuation of the damage process, which
leads Eq. (9) to be automatically satised. Hence, B
might be considered as a vector of nodal physical parameters rather than a nodal variable one. Roughly
speaking, the nodal parameters of the vector B appear in
the new stiness matrix multiplying the elastic coecients. For an original discrete model, like the one
formed by a mass and a spring, this parameter will lead
to Kd bK, where Kd and K are, respectively, the
damaged and undamaged stiness of the spring. For
a FE model, the spatial parameter b will inuence all
the entries in the stiness matrix corresponding to the
neighborhood of the damaged region. In order to make
things clear, an elementary stiness matrix based on the
application of the proposed damage parameterization in
an EulerBernouilli beam context is addressed in Appendix A.
Therefore, the FE model is updated by nding
the most suitable vector of damage parameters B
fb1 ; . . . ; bnd gT that minimizes a global residue (GR) as
follows:
Find B satisfying
min GR
B

such that bj 2 0; 1;

j 1; . . . ; nd

13

11

where the eigenvalue ki is the square of the natural frequency corresponding to vibration mode /i . Now, replacing the analytical modal parameters by experimental
ones into the eigenproblem (11), a dynamic residue
vector is dened over each identied mode:


i
Ri K  x2E;i M /E i 1; . . . ; nm
12

where the GR is dened as follows


GR

nm
X

Ri Ri

14

i1

where nm is the number of modes to be used in the updating procedure and nd is the number of nodes used to
discretize the damage eld.

D.A. Castello et al. / Computers and Structures 80 (2002) 417436

The numerical approximation of the mathematical


problem presented by (13) is addressed in the next section, but it is important to emphasize that once the
nodal value vector B is obtained, the damage value at
any point of the body can be determined by using the
interpolation functions used for the discretization of the
eld b. Therefore, unlike other approaches, the present
detection method has as nal result an evaluation of the
damage state over the whole structure.

3. Numerical determination of the damage in the structure


This section presents the numerical details of the
proposed method to solve the mathematical problem
presented in Eq. (13). The proposed method ts into the
so-called sensitivity methodology [9,27], in which the
minimum of (13) is sought by means of the Newtons
method, providing the need of sensitivities computation.
The Newtons method was adopted due to its good
performance and robustness in non-linear problems even
for non-convex situations.
3.1. Newtons method
After obtaining all the required experimental data,
the FE model may be updated by nding the most
suitable set of the parameters associated with the damaged regions that minimizes the GR, as stated in Eq.
(13).
Recalling Eq. (12), the GR can be rewritten as follows
nm 
T

T
X
i
i
i
i
GR
K/E K/E x4E;i M/E M/E
i1



T
i
i
 2x2E;i M/E K/E

15

where xE;i and /E are the experimental modal properties. The set of parameters fb1 ; . . . ; bnd gT that minimizes
the problem (13) is given by the solution of the optimality condition presented below
F

oGR
0 such that bp 2 0; 1
oB

p 1; . . . ; nd
16

where the components of the vector F read as


0 "
# !T
nm
X
oK
i
i
@2
Fp
/
K/E
obp E
i1
"
# 1

T oK
i
i
/ A p 1; . . . ; nd
 2x2E;i M/E
obp E
17

421

The Newtons method was used to determine the


solution of the non-linear equation (16). So, the increment DBr of the vector Br1 at the rth iteration is
given by
Ar1 DBr Fr1

18

where the components of the tangent matrix A of the


Newtons method are as follows:
"
# !T "
#
nm
X
oK
oK
i
i
/E
/E p; q 1; . . . ; nd
2
Ap;q
ob
ob
p
q
i1
19
The constraint of the cohesion parameters is imposed
performing, at each iteration, a simple projection
Br SBr1 DBr

20

where the operator S is, essentially, the projection dened over each component of B and reads as
Sv fv if v 2 0; 1; 1 if v > 1; 0 if v < 0g

21

Once again, it is very important to emphasize that, it


is supposed that damage evolution will not occur during
the identication process. Therefore, Eq. (9) and the
hypothesis b_ 6 0 are automatically satised and it does
not play any role in the numerical algorithm.
3.2. Mode-shape projection
Unfortunately, the number of DOF at which the
mode shape is sampled from the test is typically much
smaller than the number of DOF in the FE model that
denes [K] and [M]. Therefore, to apply Eq. (12), either
the model must be reduced to the measured DOF or the
measured portion of the mode shape must be expanded
to the dimension of the analytical eigenvectors. So, in
order to achieve the compatibility between the dimension
of the experimental and FE models, the approach taken
in the present work is to expand the measured mode
shapes to the dimension of the analytical eigenvectors,
using for this an expansion built on the minimization of
the Frobenius norm of the dynamic residue [9]. To derive
the proper projection, one should partition the mode
shape /i into its measured and unmeasured components
(
)
/i
i
m
/
22
/i
o
where the superscript i means the ith mode shape and
the subscripts m and o mean measured and omitted
DOF, respectively. The subscript E meaning experimental was omitted due to the fact that only the experimental
modal parameters, frequencies and mode shapes, are
used in this expansion. Therefore, all modal parameters

422

D.A. Castello et al. / Computers and Structures 80 (2002) 417436

referred to in this section are the experimental ones. The


residue Eq. (12) may now be written as follows


Ri K  x2i M /i
(
)

 /i
m
Km Ko   x2i Mm Mo 
23
/i
o
where Km , Mm , Ko and Mo are the measured and unmeasured column sets of the stiness and mass matrices,
and xi is the ith experimental frequency. The mode
shape projection directly results from minimizing the
Frobenius norm of the dynamic residue with respect to
/i
o , assuming no change in the modal parameters, viz.
T

min Ri Ri

24

i
/o

Dening the impedance Zi of the ith mode shape as


Zi K  x2i M
can be written as
(
)
/i
m
Zi

o
/i
o

the residue R
Ri Zi
m

25

26

Combining Eqs. (26) and (24), the minimization


problem is
 T
iT iT i i
iT i i
min /i
m Zm Zm /m 2/o Zo Zm /m
i

/o

T
iT i i
/i
o Zo Zo / o

27

which has an explicit solution leading to the mode-shape


projection

 
 T
i
i
iT i
i
i


Z
Z
Z
Z
28
/i
o
m /m Poi /m
o
o
o
where meansTthe Moore Penroses inverse, inasmuch
i
as the matrix Zi
o Zo can be singular and does not have
a classic inverse. Therefore, recalling Eq. (22), one can
rewrite the full ith mode shape in terms of its matrix
projection Pi  and its measured partition /i
m as follows
(
) 

i
/m
I
i

/i
29
f/i g
m Pi /m
i
P
oi
/o
After the projection operator Pi  for the ith mode shape
is determined, the mode shape may be expanded and the
dynamic residue Ri with respect to the model DOF can
be computed.

In the present work, damage is described by means of a


eld b dened over the whole body. Thus, in a discretized model the number of required parameters bj will
generally increase as the number of the elements of the
mesh increases. Due to this fact and assuming that
damage is concentrated in some areas of the mesh, it is
quite reasonable to update only those parameters associated with the elements which belong to the potentially
damaged regions. This avoids ill conditioned problems
and spending time updating parameters which are not in
error. Therefore, rstly one may determine a set of DOF
associated to these potentially damaged regions, secondly determine their respective elements, then the parameters associated to them.
Among several procedures which may be chosen to
be used to nd out this set of DOF, one should notice
that the dynamic residue vector can deliver some information about this on its own. Recalling Eq. (12), one
sees that the dynamic residue equals to zero if the
original model stiness matrix K is replaced by the real
stiness matrix of the damaged structure KR , where the
subscript R means real, viz.


i
0 KR   x2E;i M /E i 1; . . . ; nm
30
Subtracting Eq. (30) from (12) one gets
i

K  KR /E Ri

i 1; . . . ; nm

31

Based on Eq. (31), one may conclude that the set of


DOF which exhibits the largest values will probably be
the one associated with the damaged regions, in other
words, the regions whose elemental stiness proposed by
the original model does not represent its actual value
any more. To nd out which components of the dynamic
residue vector are in error, one may dene some threshold level and compare these components with it. This
threshold level can be, for example, 10% of the largest
component, or can be dened by means of a statistical
procedure applied to the components [9].
Another interesting approach to select the DOF was
proposed by Kaouk and Zimmerman [20], where each
entry of the dynamic residue vector is dened as an inner
product of vectors. After selecting the DOF, the choice
of the set of parameters to be updated may be made
through the connectivity matrix or by the feeling of the
designer. The procedure used in the present work to
select the DOF associated with the damaged regions was
the statistical one presented by Farhat and Hemez [9]
and the connectivity matrix was used to select the set of
parameters to be updated.

3.3. Parameter selection


3.4. Algorithm
After obtaining experimentally the modal data, the
localization of the most likely damaged regions is the
rst step to be addressed in the model updating process.

Once the present approach has been properly introduced, and assuming that the experimental modal

D.A. Castello et al. / Computers and Structures 80 (2002) 417436

properties have already been obtained, one may formulate this damage identication algorithm as follows:
1. expand the measured mode shapes;
2. calculate the initial dynamic residue vectors; select
the set of parameters to be updated;
3. start the updating process assuming that all the selected parameters have initial value equal to one;
4. expand the measured modes (except for the rst iteration);
5. calculate the vector F and the tangent matrix A;
6. calculate the increment DB;
7. apply the projection operator S;
8. update the vector B and the stiness matrix Kd ;
9. kDBk
< 1 ? kGRk < 2 ?
kBk
Yes ) end
No ) go to 4.
At each iteration, the experimental mode shapes need
to be expanded (step 4) as the stiness matrix has to be
updated. Each such expansion is obtained by means of
the ill-posed problem (24), which can lead to several
solutions. Here, the one selected is given by (28), what
can represent troubles to the convergence of the algorithm.
3.5. Error measures
After using the proposed method to obtain an updated analytical model, one has to assess the reliability
of the provided results. In this work, four error measures
are used for this purpose. The relative error between the
experimental and the analytical natural frequencies Ef ,
the weighted normalized H2 norm [12] of the dierence
between the experimental frequency response functions
(FRFs) and the analytical ones, the amplitude correlation coecient (ACC) [8] and a normalized time-error
(TE) measure.
The weighted normalized H2 norm of the dierence
between the experimental FRF matrix GExp and models
FRF matrix Gm is dened as follows:
W
kDGxkH2




W x GExp x  Gm x 
H2


GExp x

32

H2

where W x is a weighting function. For the present


work, it will be considered a simple weighting function
that reads as

W x

0;
1;

jxj < XB
jxj P XB

where XB is the pass-band frequency.

33

The ACC is dened as




2HExp xH Hx
ACCx
HExp xH HExp x HxH Hx

423

34

where Hx is a vector comprised of all the available


FRFs at the frequency x. The ACC is dened between
zero and unity, and it only becomes unity if the amplitude of the FRFs coincide.
The normalized TE is dened as follows
TEt

YExp t  Ym t


YExp t

35

where Ym and YExp correspond to the system response


and to the model response to the same excitation.
Throughout this article the normalized TE will always
be calculated for a collocated sensor.
4. Illustrative examples
Here, the verication and assessment of the current
damage identication approach involve an analytically
derived modal test structure. It is based on the two
main tasks: the rst one consists on imposing some
damage to a chosen test structure and performs a virtual
modal experiment on it, ignoring the knowledge of both
the location and the magnitude of the imposed damage.
The second uses CDIA to assess damage considering as
entries both the physical properties of the undamaged
analytical test model, [M] and [K], and the dynamic response of the damaged ctitious test model.
Two completely dierent structural systems have
been chosen to be used in the CDIA assessment: a truss
and a beam. It is important to emphasize that the main
dierence between these two situations concerns the way
damage is treated, in the rst case it is treated as a
piecewise continuous eld and in the second as a continuous one.
4.1. Truss
The truss system to be used for the rst numerical
illustration is a two-dimensional oating truss structure
which is depicted in Fig. 1, as well as its measured degrees of freedom (Ux ; Uy ), which denotes its horizontal
and vertical displacements respectively. In the undamaged situation all the truss elements possess the same
physical properties, hence, the same axial stiness. The
properties of the truss have been chosen as: Young
modulus E 276 Gpa, cross-sectional area A 6:4 
104 m2 , specic mass q 5:0  103 kg/m3 and all the
vertical and horizontal bars have length L 1:0 m.
In the rst case (case 4.A), elements 35 and 50 (dashed inclined and dashed vertical elements depicted in
Fig. 1) have had their stiness decreased by 50%, what,
by cohesion parameters denition, provides a damage

424

D.A. Castello et al. / Computers and Structures 80 (2002) 417436

Fig. 1. Damaged elements of the case 4.A.

parameter D equals to 0.5 for both elements. All the


other elements kept the same original stiness. It is also
assumed that only the rst ve exible modes and the
three rigid-body ones have been measured at the DOF
shown in Fig. 1. Fig. 2 shows the result obtained by the
CDIA. The location of the elements were perfectly determined by the proposed approach as well the severity
of damage of the rst element, the severity of the damage of the second element was not perfectly obtained,
nevertheless it was a good result since it achieved approximately 80% of the correct value. Similar results
were obtained by Hemez and Farhat [9].
The second case (4.B) to be analyzed presents a situation where two elements of the truss, 29 and 39, have
been removed from the oating truss structure described

Fig. 2. Result obtained by CDIA for the case 4.A.

earlier. The removal of an element leads to a damage


parameter D equals to 1 for this element. The resulting
perturbed structure is shown in Fig. 3. It is also assumed
that the rst, the second, the eighth and the ninth exible
modes and the three rigid-body ones are available at the
DOF shown in Fig. 3. This choice of the modes is based
on the strain energy distribution such as in [9] and it has
been used here in order to have in hands an example of
the literature to be compared with. The result obtained
is depicted in Fig. 4 and once again CDIA was capable
of identifying the damaged elements and still managed
to predict that these elements had been removed from
the system.
4.2. Beam
The chosen example consists of a cantilever beam
which is discretized using 20 bi-dimensional EulerBernouilli beam elements (see Appendix A), what leads to a
FE model containing 40 DOF (20 translational, 20 rotational). The beam is depicted in Fig. 5. The properties
of the beam have been chosen as follows: Young modulus E 69 Gpa, cross-sectional area A 1:82  104
m2 , moment of inertia I 1:46  109 m4 , specic mass
q 6:8  103 kg/m3 and each nite element of the beam
model has length L 0:05 m.
In order to realistically simulate the corrupting eects
of noise, ltering, digital sampling and truncation of the
modal spectrum, a virtual simulator was utilized. The
simulator estimates the FRF for each inputoutput
pair through ensemble averaging. In order to achieve a
higher level of delity, the input signal is pre-ltered

Fig. 3. Damaged elements of the case 4.B.

D.A. Castello et al. / Computers and Structures 80 (2002) 417436

Fig. 4. Result obtained by CDIA for the case 4.B.

updating processes only the rst three modes have been


used. The authors have chosen this in order to illustrate
the fact that although not all the available spectral information is required for the updating process, the set of
not used information is relatively important in the validation of the damage identication result by inspecting
the error measures computed from the entire spectral
information.
For the computation of the weighted normalized H2
norm it will be used a pass-band frequency equals to the
mean value of the two rst identied natural frequencies and the excitation used to compute the normalized
TE is a sine chirp sweeping the band (01000 Hz).
Since in practice the data is contaminated with noise,
it has been considered the following denition of signal
to noise ratio (SNR) to be used henceforth

SNR 10 log

Fig. 5. Cantilever beam (and its nodes) used to assess CDIA.

before the analog conversion. The noise contaminated


system response is the available one to be processed, and
this signal is ltered at 80% of the Nyquist frequency
before digital sampling. As reported in [1], this procedure furnishes realism to the FRF obtaining problem.
The FRFs obtained from the virtual simulator are used
to derive the impulse response functions required for the
eigensystem realization algorithm (ERA), [17], which in
turn, provides a state-space equation of the structure.
The state-space equation derived from this realization is
used further in the CBSI algorithm [2] which provides
the required normal modal properties, namely, the
normal mode shapes at the measured DOF and the
natural frequencies.
For all the simulations it has been adopted the data
acquisition parameters shown in Table 1 and it has been
considered the following damping ratios for the rst six
modes of the structure: f1 0:058%, f2 0:508%, f3
0:309%, f4 0:378%, f5 0:500% and f6 0:400%.
The excitation point will always be at the position of the
rst sensor.
Within the band of interest (0450 Hz), it is contained six modes of the system, nevertheless, for the
Table 1
Data acquisition parameters
Sample frequency
Number of samples
Test band of interest
Number of averages
Excitation type

1200 Hz
4096
0450 Hz
10
sine chirp

425

r2s
r2n


36

where r2s and r2n correspond to the variance of the signal


s and of the noise n respectively.
To assess the proposed damage identication technique, dierent examples are presented. The rst six
cases assess the inuence of the noisy data, the reduced
measured DOF and the number of samples and averages
used in the data acquisition. Therefore, the same damage eld is considered for all the six cases. Table 2
summarizes the simulation parameters for all these six
cases. The seventh case considers a situation where two
damaged regions are present in the structure and the
damage eld is represented by piecewise non-linear
polynomials. Therefore, they cannot be exactly represented by the linear nite elements used to build up the
model. The last case assess the inuence of the FE mesh
on the identication process.
In the rst case to be considered (4.C), the imposed
damaged region is spread over the elements between the
nodes 4 and 7 in Fig. 5, and it is a linear distribution
dened by its values at the nodes 5 (0.20 m) and 6 (0.25
m), which were set to be 0.2 and 0.05 respectively, and
assuming null values at any other position. It was assumed that the rst three modes were available at every

Table 2
Simulation parameters for the rst six cases
Case

SNR
(dB)

Measured DOF
(NODES)

Number
of samples

Number of
averages

4.C

90

4096

10

4.D
4.E
4.F
4.G
4.H

90
50
50
50
20

2,4,6,8,10,12,
14,16,18,20
2,6,10,14,18
2,6,10,14,18
2,6,10,14,18
2,5,8,14,21
2,6,10,14,18

4096
4096
16,384
4096
8192

10
10
10
10
100

426

D.A. Castello et al. / Computers and Structures 80 (2002) 417436

Fig. 6. Initial dynamic residue vector for case 4.C.

other vertical DOF starting at node 2. It has been considered that the signal was contaminated with noise such
that the SNR was 90 dB (see Eq. (36)), which corresponds to a very accurate measure.
The selected parameters to be updated were based on
the components of the mean initial dynamic residue
associated to the vertical DOF, which is depicted in Fig.
6. Based on a statistical procedure [9], the parameters 4,
5, 6 and 8, which are associated with the greatest components of the mean residue vector, were selected to be
updated. It should be remarked that, in this case, this
remarkable selection of parameters is mainly due to the
great number of measured DOF. In a more realistic
situation, where one has to deal with a small set of experimental data, this selection is not so ecient.
The result provided by CDIA is depicted in Fig. 7,
where the thin line represents the imposed damage and
the thick line represents the result obtained by CDIA. It
is clear from Fig. 7, that CDIA was able to capture the

Fig. 7. Result obtained by CDIA for the case 4.C.

essence of the damage eld. However, it should be remarked that the validity of the provided result must be
based only on the error measures computed from the
available data. Hence, Fig. 8 shows the error indicators
associated to the case (4.C).
The relative frequency error indicator Ef shows that
all the model natural frequencies, except the second one,
have got closer to the experimental natural frequencies.
The ACC measure clearly shows an improvement of the
FRFs from the second natural frequency on, and that
the frequency band around the rst natural frequency has reached worst approximation. Nevertheless,
it should be remarked that, within this band, the poor
FRFs curve tting, due to precision errors in frequency
associated to the discrete fourier transform may be responsible to make the ACC error measure worse within
this band. The normalized TE also shows a better
agreement between the response of the damaged structure and that of the updated model than the response of
the original one. As a last check, one clearly sees that the
normalized dynamic residue matrix has been minimized.
Hence, all the error measures are positive with respect to
the response provided by the CDIA, i.e., all of them
shows that the damage identication performed by
CDIA have led to a better agreement between the updated model and the actual damaged structure, which
implies that the result can be considered satisfactory,
fact that agrees with Fig. 7.
The second situation (4.D) is exactly equal the previous one (4.C), except that here one has half the sensors
such that the measured DOF are the vertical ones of the
following nodes: 2, 6, 10, 14 and 18. The result provided
by CDIA is shown in Fig. 9 and the error measures for
this case are shown in Fig. 10.
It is clear from Fig. 9 that the present result does not
describe the damage eld as well as it has been described
in the previous example, for the same algorithm stopcriterium. The weighted normalized H2 before and after
the updating process are respectively: 0.90889 and
0.76580, what means a positive result. Despite the fact
that one can clearly see that the result presented in Fig. 7
is more accurate than the one presented in Fig. 9 it is
hardly noticed by comparing Figs. 8 and 10. Therefore,
it is really dicult to assure that, in this situation, one
solution is more consistent than the other one. This
fact may be considered as predictable as long as it is
well known that, in general, the dynamic behavior of
a structure is slightly changed due to damage. According to the set of error measures depicted in Fig. 10, the
result shown in Fig. 9 would be perfectly accepted as
a satisfactory solution for the damage identication
problem.
The third case (4.E) to be analyzed is equal to the case
(4.D) except that the SNR has been set equal to 50 dB.
The result provided by CDIA is depicted in Fig. 11 and
its respective error measures are shown in Fig. 12. After

D.A. Castello et al. / Computers and Structures 80 (2002) 417436

427

Fig. 8. Error measures for case 4.C.

analyzing some damage identication results the authors


reached to the conclusion that the lower the number of
measured DOF the greater the errors associated to the
dynamic residue inasmuch as its calculation demands a
mode expansion technique, which in turn is based on the
minimization of the own dynamic residue. Therefore, in
practical situations the authors believe that the dynamic
residue does not provide a straightforward criterium for

evaluating the reliability of the solutions and its graphics


will be omitted henceforth.
The weighted normalized H2 norm before and after
the updating process are respectively: 0.90892 and
0.77870. Hence, all the error measures are positive with
respect to the provided result. The situation here is
similar to the one which appeared in the previous case, it
is dicult to say if the solution shown in Fig. 9 is better

428

D.A. Castello et al. / Computers and Structures 80 (2002) 417436

Fig. 9. Result obtained by CDIA for the case 4.D.

than the one shown in Fig. 11. Nevertheless, the comparison of the error measures for the situations (4.C)
and (4.E) slightly indicates that the solution in Fig. 7 is
better than the one presented in Fig. 11.
The fourth case (4.F) entirely corresponds to the
previous one (4.E), except that the FRFs were obtained
with 16,384 samples in time (see Table 1), fact that led to
a higher denition of the FRFs for the same band of
interest (0450 Hz). Since the normal mode shapes and
natural frequencies required for the present approach
come from the ERA algorithm, whose the required data
are the impulse response functions obtained out of the
measured FRFs, the mode shapes and natural frequencies can be dierent from the ones obtained in case (4.E).
The result obtained by CDIA is graphed in Fig. 13 and
its error measures are depicted in Fig. 14.
The weighted normalized H2 norm before and after
the updating process are respectively: 0.76766 and
0.29434. As it was expected, since changes in the dynamic behavior of the structure due to damage are slight
and the set of available experimental data (FRFs) is
more precise in this case than in the foregoing one, the
result provided in Fig. 13 is a better approximation of
the damage eld than the solution depicted in Fig. 11,
which is noted when comparing the Ef error measure in
Figs. 12 and 14 and it is even more clear when comparing the ACC measure or the weighted normalized H2
norm.
It is well known that after performing an experiment
if one decides to do a new one, such as keeping everything the same but the position of sensors, can be really
simple or almost not practical. However, despite the
diculties inherent to this process, a last assessment of
the identied damage eld can be obtained performing a
new experiment, if it is possible. Therefore, the damage
identication result depicted in Fig. 11 will be assessed
by a new experiment (4.G), where the sensors are relocated at more suitable positions, chosen such as to en-

compass the damaged region pinpointed in Fig. 11. The


vertical DOF of the following nodes: 2, 5, 8, 14 and 21
are assumed to be measured, with the data acquisition
parameters as those shown in Table 1.
The result provided by CDIA for the case (4.G)
is graphed in Fig. 15 and its respective error measures
in Fig. 16. The weighted normalized H2 norm before
and after the updating process are respectively: 0.89519
and 0.77635. Therefore, analyzing this new set of
error measures one may conclude that once again the
updating process has furnished a positive result, corroborating the result obtained from the experiment in
case (4.E).
The problem of damage identication in the presence
of more severe noisy data is illustrated in case (4.H). In
this case, the SNR has been set equal to 20 dB, which
means that the rate between the variances of the signal
and the noise attains 10%. In this case, for the data
acquisition parameters as in Table 2, the result provided
by CDIA indicates the presence of two damaged regions, as indicated by the picture on the left in Fig. 17.
One indicated damaged region encompasses the imposed
damage eld and the other one resides near the free end
of the beam, where damage does not really exist. However, when analyzing the error measures for this case,
the validity of the result is questioned by the normalized
TE, although the other measures are somehow positive.
Therefore, the strategy adopted was to post-processing
the results by performing the damage identication twice
again, making use of the same experimental data in such
a way that, at each time, only the parameters associated
with each previous indicated damaged region were allowed to be updated.
The damage identication provided by CDIA for the
region near the xed end is depicted in the picture on the
right in Fig. 17. All the associated error measures from
this estimation indicates a better agreement between the
updated model and the experimental data than that
obtained in the damage identication performed on the
second region. Therefore, the result provided in the
picture on the right in Fig. 17 has been adopted as
the one that describes the current state of damage in
the structure.
The seventh case (4.I) to be considered corresponds
to a situation where two damaged regions are present in
the structure. Besides, the linear interpolation functions
used to approximate the damage eld are not able to
exactly describe the imposed damage. Each damaged
region have been constructed joining the following two
polynomials together:
b1x 1  0:5227 x=L3 0:7920 x=L2
 0:4253 x=L; x 2 x0 ; x0 L
b2x 1 0:1307 x=L2  0:2247 x=L
 0:1560;

x 2 x0 L; x0 3:25L

37

D.A. Castello et al. / Computers and Structures 80 (2002) 417436

429

Fig. 10. Error measures for case 4.D.

where x0 and L are the initial point of the damage region


and the length of the the beam element, respectively.
Here every other vertical DOF were measured, with the
data acquisition parameters as those shown in Table 1
and SNR 40 dB. The result provided by CDIA is
shown in Fig. 18 and its respective error measures are
shown in Fig. 19.

The weighted normalized H2 norm before and after


the updating process are respectively: 1.06626 and
0.74904. The provided result depicted in Fig. 18 shows
that CDIA was able to capture the essence of the
damage and all the error measures, in general, agree
with it. It should be remarked that in none of the analyzed examples, the second natural frequency of the

430

D.A. Castello et al. / Computers and Structures 80 (2002) 417436

Fig. 11. Result obtained by CDIA for the case 4.E.

Fig. 13. Result obtained by CDIA for the case 4.F.

updated model gets closer to the experimental one,


which is easily veried in the relative frequency error Ef
graphs. The same behavior is also observed in the rst
natural frequency in the case (4.I). The sensitivity analysis of the eigenfrequencies with respect to the damage

parameters [26] turns out that the sensitivity of the


second natural frequency with respect to the imposed
damaged in the region between nodes 4 and 7, is very
small. Therefore, one may conclude that this fact along
with the poor frequency resolution of the FRFs in the

Fig. 12. Error measures for case 4.E.

D.A. Castello et al. / Computers and Structures 80 (2002) 417436

431

Fig. 14. Error measures for case 4.F.

Fig. 15. Result obtained by CDIA for the case 4.G.

As a nal example (4.J), the inuence of the FE mesh


on the damage identication result of CDIA will be
assessed. Therefore, the same experimental data of the
damaged beam will be used for the updating of three
dierent FE meshes. The damaged region is spread over
the region (0.1,0.3) m, (Fig. 20).
We have considered that the vertical DOF at the
positions 0.11.0 m have been measured. It has been
analyzed nite element models having 10, 20 and 40
identical elements. The imposed damage eld is such
that, for each model, the damage eld discretization is
capable of perfectly describe it. The measured DOF are
located at positions which coincide with some nodes of
the FE models. The results obtained by CDIA for these
situations, with the same algorithm stop-criterium, are
shown in Figs. 2123.

5. Concluding remarks
range encompassing the second frequency may contribute to this result. The same conclusion may be drawn
with respect to the rst frequency in the case (4.I).

An approach for detecting damage based on a continuum damage model and using partial experimental

432

D.A. Castello et al. / Computers and Structures 80 (2002) 417436

Fig. 16. Error measures for case 4.G.

Fig. 17. Result obtained by CDIA for the case 4.H.

modal parameters has been presented. It is built on a


constrained minimization of a global residue and utilizes
a mode-shape projection to match the dimension of the
experimental and FE models. The approach has been

assessed on two dierent examples, a truss and a beam.


The simulations have been performed on real-like situations, where the corrupting eects of noise, ltering,
digital sampling, truncation of the modal spectrum and

D.A. Castello et al. / Computers and Structures 80 (2002) 417436

Fig. 18. Result obtained by CDIA for the case 4.I.

433

Fig. 20. Damage eld for the case 4.J.

Fig. 19. Error measures for case 4.I.

the limited number of sensors have been considered. The


ERA along with the common-based normalized system

identication (CBSI) were utilized to obtain the required


natural frequencies and mode shapes.

434

D.A. Castello et al. / Computers and Structures 80 (2002) 417436

Fig. 21. Result obtained by CDIA for the case 4.J with 20 elements.

Fig. 22. Result obtained by CDIA for the case 4.J with 40 elements.

The validity of the provided results have been based on a


set of error measures. This set of error measures is capable of qualifying whether a result is coherent or not,
nevertheless, it has not been shown to be a straightforward measure in comparing dierent results for the same
problem.
One should remark that due to the way the problem
has been formulated, the present approach has an applicability beyond damage identications point of view.
Once one has assumed the existence of a eld capable
of describing the damage over the whole system, the
damage evolution may be simulated if the external loads
are known. Furthermore, the present approach may still
be used as an alternative parameter selection estimation,
inasmuch as the damage evolution simulation can provide a set of the most likely damaged regions. Therefore,
as a direct result of joining these two areas, damage
identication and damage modelling, it has appeared a
new damage identication approach which has proved
to be ecient and robust. Furthermore, this methodology can still be used with other existent damage identication approaches in order to get results more and
more eective.
Finally, one should realize that the proposed approach can be extended by either taking a rened,
compared to the nite element mesh, damage eld discretization or using special tailored functions. Hence, in
the authors belief, an adaptive strategy, like those traditionally used in the nite element technology, can be
built in order to improve the present methods performance.

Acknowledgements
Support for this research was partially provided by
Brazilian Spacial Agency (AEB) under contract 011/98
and National Council of Scientic and Technological
Development (CNPq).

Appendix A. Elemental stiness matrix for a bi-dimensional EulerBernouilli beam based on the continuum
damage model
Fig. 24 depicts a well known bi-dimensional Euler
Bernouilli beam element where the DOF are ordered as
follows
ue fvi hi vj hj gT
Fig. 23. Result obtained by CDIA for the case 4.J with 10 elements.

The method has been shown to be ecient for estimating damage parameters for the presented situations.

A:1

The elemental stiness matrix for this element can be


obtained by the standard straindisplacement matrix H
[25], using the following equation
Z
ke 
EI H2 T H2  dX
A:2
Xe

D.A. Castello et al. / Computers and Structures 80 (2002) 417436

Fig. 24. Beam element.

where the subscript 2 is associated with the transversal/


rotational displacements and all the physical and geometrical properties were considered constant throughout the element. From the denition of the cohesion
parameter b, one may formulate the elemental stiness
matrix which takes damage into consideration, what
leads to
Z
 e
kd
bxEI H2 T H2  dX
A:3
Xe

Assuming, for instance, that bx is interpolated with


classical Lagrangean piecewise linear shape functions as
follows
X
bx
br Nr x
A:4
r

the nal result 2L=Eked  obtained after performing all


the required computations is given by
0

12I
bi bj
L2
B 4IL 2bi bj
B 12I
@  2 b b
i
j
L
4I
bi 2bj
L

435

[2] Alvin KF, Park KC. A second order structural identication procedure via state-space-based system identication.
AIAA J 1994;32:397406.
[3] Baruch M, Bar Itzhack IY. Optimum weighted orthogonalization of measured modes. AIAA J 1978;16(4):34651.
[4] Brinker R, Kirkegaard PH, Anderson P. Damage detection
in an oshore structure. In: Proceedings of the 13th IMAC,
1995. p. 6617.
[5] Castello DA. Structural damage identication via stiness
matrix updating based on a continuum damage model.
MSc Dissertation, Federal University of Rio de Janeiro,
Brasil, 1999 [in Portuguese].
[6] Doebling SW, Farrar CR, Prime MB, Sheritz DW.
Damage identication and health monitoring of structural
and mechanical systems from changes in their vibration
characteristics: a literature review. Los Alamos National
Lab., Report LA-13070-MS, Los Alamos, NM, May 1996.
[7] Ewins DJ. Modal testing: theory and practice. Taunton,
Somerset, England: Research Studies Press Ltd; 1989.
[8] Grafe H. Model updating of large structural dynamics
models using measured response functions. PhD Dissertation, Imperial College of Science, Technology and Medicine, University of London. 1998.
[9] Farhat C, Hemez FM. Updating nite element models
using an element-by-element sensitivity methodology.
AIAA J 1993;31(9):170211.
[10] Fremond M, Nedjar B. Damage, gradient of damage and
principle of virtual power. Int J Solids Struct 1996;33(8):
1083103.
[11] Frank PM. Fault diagnosis in dynamic systems using
analytical and knowledge-based redundancea survey and
some new results. Automatica 1990;26:45974.

1
4I
2bi bj
 12I
bi bj
bi 2bj
L
L2
2I3bi bj  4IL 2bi bj
2Ibi bj C
C
12I
4I
 4IL 2bi bj
b


bi 2bj A
i
j
L
L2
2Ibi bj
 4IL bi 2bj 2Ibi 3bj
4I
L

where bi and bj are the nodal cohesion parameters associated to the nodes i and j of the element depicted in
Fig. 24. One should notice that this result is a particular
one due to the fact that one could have chosen any kind
of shape functions such as quadratic, trigonometric and
etc. Therefore, the user has enough freedom to choose
the most suitable discretization of the damage eld as
well as the shape functions for some specic problem,
what makes the proposed approach very attractive inasmuch as this choice, in principle, does not have any
connection with the discretization of the FE model of
the original system.
References
[1] Alvin KF. Method for determining minimum-order mass
and stiness matrices from modal test data. AIAA J
1995;33:12835.

[12] Gawronski WK. Dynamics and control of structures: a


modal approach. In: Mechanical engineering series. New
York: Springer-Verlag; 1998.
[13] Hemez FM, Doebling SW. Review and assessment of
model updating for non-linear, transient dynamics. Mech
Syst Signal Process 2001;15(1):4574.
[14] Hu N, Wang X, Fukunaga H, Yao ZH, Zhang HX, Wu
ZS. Damage assessment of structures using modal test
data. Int J Solids Struct 2001;38:311126.
[15] Ibrahim SR, Saafan AA. Correlation of analysis and test
in modeling of structures, assessment and review. In: Proceedings of the Fifth International Modal Analysis Conference (London). Bethel, CT: Society for Experimental
Mechanics; 1987. p. 165160.
[16] Cattarius J, Inman DJ. Time-domain analysis for damage
detection in smart structures. Mech Syst Signal Process
1997;11(3):40923.
[17] Juang J-N, Pappa RS. An eigensystem realization algorithm for modal parameter identication and model
reduction. J Guidance Cont Dyn 1985;8(5):6207.

436

D.A. Castello et al. / Computers and Structures 80 (2002) 417436

[18] Kabe A. Stiness matrix adjustment using mode data.


AIAA J 1985;23(9):14316.
[19] Kammer D. Optimum approximation for residual stiness
in linear system identication. AIAA J 1998;26(1):10412.
[20] Kaouk M, Zimmerman DC. Structural damage assessment
using a generalized minimum rank perturbation. AIAA J
1994;32(4):83642.
[21] Lim TW, Kashangaki TA-L. Structural damage detection
of space truss structures using best achievable eigenvector.
AIAA J 1994;32(5):104957.
[22] Marwala T. Damage identication using committee of
neural networks. J Engng MechASCE 2000;126(1):
4350.
[23] Mattos HSC, Domingues SMP, Rochinha FA. Modelling
of nonlinear damage on elastic brittle materials. Mech Res
Commun 1998;25(2):14753.
[24] Mattos HSC, Sampaio R. Analysis of the fracture of brittle
elastic materials using a continuum damage model. Struct
Engng Mech 1995;3(5):41128.
[25] Mottershead JE, Friswell MI. Model updating in structural
dynamics. J Sound Vib 1993;167(2):34775.
[26] Nelson RB. Simplied calculation of eigenvector derivatives. AIAA J 1976;14(9):12015.
[27] Ricles JM, Kosmatka JB. Damage detection in elastic
structures using vibratory residual forces and weighted
sensitivity. AIAA J 1992;30(9):23106.
[28] Stutz LT, Castello DA, Rochinha FA. Structural damage
identication based on a nite element updating. 6th US
National Congress on Computational Mechanics. Dearborn, Michigan, 14 August 2001.
[29] Kranock SJ, Peterson LD. Real-time structural health
monitoring using model-based observers. unabridged doc-

[30]

[31]

[32]

[33]

[34]

[35]

[36]

[37]

toral dissertation. Department of Aerospace Engineering Sciences, University of Colorado, Boulder, February
2000.
Sampaio RP, Maia NM, Silva JM. Damage detection
using the frequency response function curvature method.
J Sound Vib 1999;226(5):102942.
Santos JV, Soares CM, Soares CA, Pina HL. Development
of a numerical model for the damage identication on
composite plates. Compos Struct 2000;48:5965.
Skrzypek J, Ganczarski A. Modeling of material damage
and failure of structures: theory and applications. Berlin:
Springer-Verlag; 1999.
Smith S. Iterative use of direct matrix updates: connectivity
and convergence. In: Proceedings of the AIAA 33rd
Structures, Structural Dynamics, and Materials Conference (Dallas, TX). Washington, DC: AIAA; 1992. p. 1797
806.
Smith S, Beatie C. Simultaneous expansion and orthogonalization of measured modes for structural identication.
In: Proceedings of the AIAA Dynamic Specialist Conference (Long Beach CA). Washington, DC: AIAA; 1993.
p. 26170.
Stutz LT. Damage identication approach via exibility
matrix adjustment based on a continuum damage model.
MSc Dissertation, Federal University of Rio de Janeiro,
Brasil, 1999 [in Portuguese].
Tsou P, Shen M-HH. Structural damage detection and
identication using neural networks. AIAA J 1994;32(1):
17683.
Zou Z, Link M. Identication of delamination in sandwhich plates using vibration test data. Proceedings of the
Identication in Engineering Systems, 1996. p. 5866.

S-ar putea să vă placă și