You are on page 1of 13

International Journal of Mechanical Sciences 62 (2012) 120132

Contents lists available at SciVerse ScienceDirect

International Journal of Mechanical Sciences


journal homepage: www.elsevier.com/locate/ijmecsci

Vibration based damage detection in composite beams under temperature


variations using Poincare maps
Emil Manoach n,1, Sylwester Samborski, Andrzej Mitura, Jerzy Warminski
Department of Applied Mechanics, Lublin University of Technology, Lublin 20-618, Poland

a r t i c l e i n f o

a b s t r a c t

Article history:
Received 31 October 2011
Received in revised form
8 May 2012
Accepted 11 June 2012
Available online 18 June 2012

In this work numerical and experimental study of the vibration of laminated beams with damage,
subjected to dynamic loading and temperature variations is presented.
The goal of the study is to check the applicability of the damage detection technique based on an
analysis of the Poincare maps of the beam response. The geometrically nonlinear version of the
Timoshenko beam theory is used to model the beam behavior. The damage is represented as a
reduction of the effective elastic modulus of the beam material in a small area of the structure.
The beams are subjected to a harmonic loading, leading to large amplitude vibrations and to
temperature changes. The main results are focused on establishing the inuence of the damage on
the vibration response of heated or unheated structures and the change in the time-history diagrams
and the Poincare maps, caused by a damage and elevated temperature. The damage detection criterion
formulated earlier for nonheated plates, based on analysing of the Poincare maps of the damaged and
healthy plate, is modied and tested for the case of beams additionally subjected to elevated
temperatures. The importance of taking into account the actual temperature in the process of damage
detection is shown. Performed experimental tests (by an optical method) conrm the applicability and
sensitivity of the proposed method.
& 2012 Elsevier Ltd. All rights reserved.

Keywords:
Composite beam
Nonlinear vibrations
Damage detection
Poincare map

1. Introduction
Vibration-based structural health monitoring (VSHM) methods
are quite popular and are of extensive use for monitoring
structures and fault identication. They are based on the fact that
a damage will alter the stiffness, mass or energy dissipation
properties of the structure which in turn will alter its measured
vibration response. The vibration based damage detection is an
effective method due to its simplicity of implementation and
ability of acquiring both the global and the local information of
the structure.
Many of the previous efforts of researchers on VSHM were
directed towards methods based on changes in the dynamic
characteristics of the structures such as natural frequencies,
vibration mode, and modal curvature [17]. The simplest way
to perform a fault detection is the usage of the rst several natural
frequencies of a structure which can be determined from a
conventional experiment. However, it seems that the lower order
resonance frequencies are not always very sensitive to a damage,

n
Corresponding author at: Institute of Mechanics, Bulgarian Academy of Sciences,
Acad. G. Bonchev Street, bl.4, Soa 1113, Bulgaria.
E-mail address: e.manoach@imbm.bas.bg (E. Manoach).
1
On leave of absence from the Institue of Mechanics, Bulgarian Academy of
Sciences, Soa 1113, Bulgaria.

0020-7403/$ - see front matter & 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.ijmecsci.2012.06.006

except in cases with a very large damage [3,69]. Mode shapes


are in general more sensitive to damage but they are difcult to
measure and estimate from measured quantities. Some researchers looked at the loss factor as an indicator of damage in
composites (e.g. Lee et al. [10]). It could be a promising indicator
for the approaches for health monitoring, since some of the used
tools in VSHM, like Poincare maps and phase plots, are very
sensitive to energy dissipation.
Another problem with the VSHM methods is that often they are
based on linear models of the structure. For systems which behave
as nonlinear, and the nonlinearities of the system are not taken into
account in the model, VSHM methods may give false alarms due to a
discrepancy between the measured and the expected responses.
To address some of the above mentioned problems, a lot of
authors started recently to study the possibility to employ
measured time series of the structure response and the nonlinear
dynamics theory [1114]. The idea of using nonlinear dynamics
tools for damage detection purposes stems from the fact that a
damage (or a fault) introduces an additional nonlinearity, which
affects and changes the dynamics response of the system under
consideration. In general, damage detection can be performed by
extracting and monitoring certain parameters (features) that
change with the introduction and the growth of the damage.
Most of the studies in this eld are devoted to the extraction of
such features from the structural vibration response. In [11] the

E. Manoach et al. / International Journal of Mechanical Sciences 62 (2012) 120132

authors use the beating phenomenon for damage detection (DD)


purposes. In [12,13] new attractor-based metrics are introduced
as damage sensitive features. In [1416] it is veried that due to
the inuence of the nonlinearity of typical cracks, superharmonic
vibration regimes appear in the system response. Many authors
use statistical methods to detect and localize the damage [1721].
In the authors previous works [21,22] a numerical approach to
study the geometrically nonlinear vibrations of rectangular plates
with and without a damage is developed. A damage index and a
method for DD and location, based on the Poincare map of the
response, have been proposed. The suggested damage detection
method shows good capability to detect and localize a damage in
plates.
Temperature changes can and do affect substantially the
vibration response of a structure. Thermal loads introduce stresses due to thermal expansion, which lead to changes in the modal
properties. Thermal loads can also cause buckling and in some
cases even lead to chaotic behavior [2326]. Thus, on a lot of
occasions the presence of a temperature eld can either mask the
effect of the damage or increase it, which will render a VSHM
method ineffectiveit might give no alarm when a fault is
present or might give a false alarm. This is why it is crucial to
take into account the temperature changes when developing
VSHM procedures.
The main objectives of this study are the following:
1. To study the inuence of defects, elevated temperatures and
their combination on the dynamic characteristics of composite
beams and on their geometrically nonlinear dynamic response.
2. To test numerically the Poincare maps based criteria for
identication of a fault in beams taking into account the
elevated temperature and to study its sensitivity.
3. To check experimentally (by high speed camera tests) the
applicability of the developed method for a damage detection
in composite beams.

mid-axis of the beam, the displacements in the x and z directions


are denoted by u and w, respectively. cx(x,t) is the angle of
rotation of the normal of the cross-section to the beam mid-axes.
2.1. Geometrical relationships
The strain and curvature-displacement relationships associated with the mid-axes of the beam which consider large
displacements and shear can be expressed as
 
@u 1 @w 2
@w
@c
c, k0 
1
e0x
, e0xz
@x 2 @x
@x
@x
0

2.2. Constitutive equations


For each ith layer the relations between stresses and strains
are
i
i i
i i
si
si
x E ex aT TT 0 ,
xz G exz

(i)

(i)

where E is the Youngs modulus and G is the shear modulus for


each layer.
The generalized forces are obtained by integrating the stresses
at each layer and summarizing the integrals (see [27,28])
"
#
 2
@u
@w
N A11
0:5
aT DT
@x
@x
M D11 k0


@w
s
c
Q k A55
@x

where
Nl
X

Ek zk zk1 b

k1

A55 b

Nl
X

Nl
X

Ek h

k1

Gk zk zk1 b

k1

The object of investigation is a laminated beam with length l


and width b, having Nl number of layers, symmetrically disposed
around the mid-axis (Fig. 1).
The thickness of kth layer is h(k), its Youngs modulus E(k),
Poisons ratio v(k) and the coefcients of thermal expansion a(k)
T .
The beam is subjected to a transverse load p(x,t) and temperature
variation DT (with respect to a reference temperature) leading to
large amplitude vibrations. In general, the distribution of DT can
be assumed as nonuniform along the beams thickness. The
geometrically nonlinear version of the Timoshenko beam theory
is used to model the beam behavior, so that the shear deformation
and rotary inertia are taken into account. At each point of the

The strain vector is given by e fe0x zkx ,f ze0xz g , where f(z) is


a function describing the distribution of the shear strain along the
beam thickness.

A11 b
2. Theoretical model

121

Nl
X

Gk h

k1

Nl
3
3
bX
Ek zk zk1 ,
3k1

aT

Nl
X

ak
T

k1

In Eqs. (3) N is the longitudinal force, M is the bending moment


and Q is the shear force, h(k) is the thickness of each layer and ks
characterizes the distribution of the shear stresses along the beam
cross-section (shear correction factor).
It should be noted that for laminates with nonsymmetric layups additional terms considering coupling between in-axis and
out-of-axis behavior of laminates appear in the constitutive
Eqs. (3) (see [27,28]). In this work the considerations are limited
to symmetric laminated beams and the coupling stiffness is equal
to zero.
2.3. Equations of motion
The equilibrium equations may be deduced by considering the
translational equilibrium in x and z directions and the rotational
equilibrium about x-axis.
The equations of motion of large amplitude vibrations of a
laminated Timoshenko beam with layers/plies symmetrically
arranged around the mid-axis can be written in the following
form:

Fig. 1. Geometry of the beam. x1 and x2 denote the beginning and the end of the
damaged part of the beam.

@N
@u
@2 u
c1
RH 2 0,
@x
@t
@t

h=2 r z rh=2

5a

122

E. Manoach et al. / International Journal of Mechanical Sciences 62 (2012) 120132

@M
@c
@2 c
Q c2
RI 2 0, 0 ox o l, t 4 0
@x
@t
@t
!
@Q
@2 w
@N @w
@w
@2 w
N
c1
RH 2 px,t

@x
@x @x
@x
@x2
@t


5b

5c

RI b=3

ri zi3 zi13

Nl
X

6a

ri hi

6b

and with c1 and c2 are denoted the damping coefcients, which


are assumed to be proportional to the mass terms.
Substituting Eqs. (3) into Eqs. (5) and neglecting the longitudinal inertia effects the following system of governing equations is obtained:
(
"
#)
 
@
@u 1 @w 2
A11

aT DT
0
7a
@x
@x 2 @x





@
@c
@w
@2 c
s
D11
c RI 2 0
k A55
@x
@x
@x
@t

7b

"
  #



@
@w
@u 1 @w 2 @2 w
A55
c A11

@x
@x
@x 2 @x
@x2
A11 aT DT

@2 w
@2 w
RH 2 p
2
@x
@t

7c

It is assumed that the defect manifests itself in a reduction of


the effective elastic modulus of the beam at the corresponding part
of its cross-section. Ideas of how delamination could be taken into
account in the determining of the effective properties of the beam
could be found for example in [29,30]. It is clear that this approach
for modeling a defect in a composite structure is quite simple and
does not consider the concomitant complicated phenomena as the
changes in a stress eld, additional damping, sliding between
layers, opening and closing modes of the defect (e.g. delamination),
etc. Reduction of the rigidity is only one of the manifestations of
the damage in composite beam. On the other hand, the other
phenomena going along with the damage usually enforce its
inuence on the response. The present study is focused on the
damage detection approach and due to this reason it is acceptable
to use the model of the reduced rigidity for a DD purpose.
In the theoretical model described above, the variable rigidity
of the beam due to the defect is considered.
2.4. Boundary and initial conditions
In the present work the numerical model is applied for fully
clamped and in-axis xed beams. Thus, the boundary conditions are
u0,t ul,t w0,t wl,t 0,

c0,t cl,t 0

The inuence of the temperature variation is more essential


for such beams due to the thermal expansion.
The initial conditions are accepted in the following general form:
wx,0 w0

9a

_
_0
wx,0
w

9b

cx,0 c0

9c

c_ x,0 c_

N
X

Ei =Nl ,

i1

i1

u u=l,

x x=l,

z z=h,

t tc=l,

c2 E=r

10

where

i1

RH b

In numerical solution of the problem the following set of


dimensionless variables is used:
w w=l,

Here, the following denotations are used:


Nl
X

3. Numerical solution of the problem

9d

N
X

ri =N l

11

i1

By omitting the bars in the dimensionless variables, the


governing Eqs. (7) describing the transverse vibrations of the
beam can be written in the following form:
(
"
#)
 
@
@u 1 @w 2
A11

aT DT
0
12a
@x
@x 2 @x




@
@c
@c
E @2 c
s
2 @w
D11
c d2
k A55 l
RI
0,
@x
@x
@x
@t
r @t2
0 ox o l, t 4 0



@w
@w
E @2 w
s @
A55
c d1
RH
pl GL2 GT2
k
@x
@x
@t
r @t2

12b
12c

where by GL2 and GT2 the component of the following, the so-called
pseudo-load vectors (see [20,28]) due to the geometrical nonlinearities and thermal changes are denoted
" 
 2 # 2
@u
@w
@ w
GL2 A11
0:5
13a
@x
@x
@x2
GT2 A11 aT DT

@2 w
@x2

13b

Here d1 c1lc, d2 c2lc.


It is easy to notice that from Eq. (12a) follows:
 
@u
C=A11 Guw GuT
@x
where C is a constant and
 
1 @w 2
, GuT aT DT
Guw 
2 @x

15

The solution of the above equation regarding u is


Z x
Z x
Z x
u
A1
d
x

G
d
x

GuT dx C 2
ux C
11
w
0

14

16

Considering the boundary conditions and taking into account


that the part of the beam between x1 and x2 (see Fig. 1) has
different mechanical properties it is easy to obtain that C2 0 and
!

Z l
x1
x2 x1 lx2
Guw GuT dx

C
17
A111
A211
A111
0
Here A111 and A211 denote elastic constants dened by Eq. (4) for a
healthy and damaged part of the beam, correspondingly.
Thus, if one considers Guw andGuT as known functions, the sum of
the components of pseudo-load vectors due to the large displacements and the temperature changes, can be expressed using Eqs.
(13)(15) and (17) as
" 
  #
@u
1 @w 2 @2 w
@2 w
@2 w

GL2 GT2 A11


A11 aT DT 2 C 2
2
@x
2 @x
@x
@x
@x
18
The left hand side of Eqs. (12b) and (12c) is a linear form. If one
considers GL2 and GT2 as components of the known force vectors
GL and GT the mode superposition method can be applied for the
solution of these equations. As the normal modes of the free

E. Manoach et al. / International Journal of Mechanical Sciences 62 (2012) 120132

vibrations of the beam do not correspond to those of the nonlinear system, these modes are called pseudo-normal or quasinormal modes [31,32]. Thus, the generalized displacements
vector v {c,w}T is expanded as a sum of the product of
pseudo-normal modes vn and the time dependent functions
qn(t) as
v

Nf
X

vn xqn t

19

4. Damage detection technique


The damage identication technique is based on analysing the
Poincare maps of the responses of the inspected structure. This
technique is suggested in [22,21] for damage detections in plates.
In these works damage was modeled as a part of the structure
with a reduced thickness. According to these works the following
damage index was suggested:

n1

Substituting Eq. (19) into Eqs. (12b) and (12c), multiplying by


vm(x), integrating the product over the beam length, invoking the
orthogonality condition, and assuming proportional damping in
R1
2
the sense 0 d1 cn d2 w2n dx 2xn on the equations for qn(t) are
uncoupled in the form:
q n 2xn on q_ n o2n qn F n t

21
The initial conditions dened by Eqs. (9) are transformed also
in terms of qn 0 and q_ n 0. Eqs. (20) are strongly coupled due to
the nonlinear terms on the right hand sides of the equations
(which depend on the solution).
Let the total time interval in which the dynamic response is
desired be divided into a sequence of time increments [ti,ti 1].
Using the methodology developed by Kukreti and Issa [31], the
pseudo-load vector {P G} is interpolated by a quadratic polynomial of time, i.e.
0 r t r Lt

22

where Lt ti 1 ti represents the time increment and the variable t, dened as t t  ti, identies a new time origin for each
time increment.
Denoting
P0 x Px,0,

0oxo1

23a

0 o x o1

P1 x Px,mLt ,

23b

P2 x Px,Lt ,

0 ox o 1

23c

G0 x Gx,0,

0 o m o1

23d

G1 x Gx,mLt ,
G2 x Gx,Lt ,

0 o m o1

23e

0 o mo 1

23f

the expressions for constants A, B and C can be developed in


terms of Pi, Gi (i 0/2).
Here, t mLt represents an intermediate point in the time
interval Lt.
The general solution of Eq. (20) is
qn t E1n q0n E2n q_ 0n F 1n an F 2n bn F 3n cn

24

where E1n, E2n, F1n, F2n, F3n denote complicated mathematical


expressions containing on, xn and t (for details see [24,32,33])
and
Z 1
Z 1
Z 1
an b
vTn A dx, bn b
vTn B dx, cn b
vTn C dx
25
0

Idi

9Sui Sdi 9
Sui

26

where
Sui

NX
p 1 q
_ ui,j 1 w
_ ui,j 2
wui,j 1 wui,j 2 w
j1

27a

Sdi

NX
p 1 q
_ di,j 1 w
_ di,j 2
wdi,j 1 wdi,j 2 w
j1

27b

20

Here on are natural frequencies of the (undamped) Timoshenko


beam, xn are modal damping coefcients and
Z 1
F n t
vTn xPx,t Gx,t dx, Px,t 0,pT , G GL GT

Px, t Gx, t Ax BxUt CxUt2 ,

123

The iteration procedure applied to solve the above equations is


very similar to the ones applied in [24,32,33] and that is why it is
not presented here.

In the above equations i 1, 2, y, N, where N is the number of


nodes, Np is the number of points in the Poincare map for each
_ uij and wdij , w
_ dij denote the jth point on Poincare
node, and wuij , w
maps generated for undamaged (healthy) and damaged states,
respectively.
The Poincare map is a standard tool for inspecting dynamical
systems, since it preserves many properties of periodic and quasiperiodic orbits of the original system and has a lower dimension.
The Poincare maps contain data for the displacements and the
velocities of the structure in a compact form and since these two
parameters are expected to be sensitive to damage, these diagrams can be used to detect damage. The suggested damage index
represents the relative difference between the lengths of the
curves formed by connecting the dots on the Poincare maps for
the nondamaged and the damaged plate for ith node, respectively.
This difference is accepted by us as a measure for the global
change (during the total period of vibration) in the dynamic
behavior of the damaged structure in comparison with the
undamaged one.
The damage index described above depends on the location of
the point on the beams x-axis and consequently it is a function of
the beam coordinate x.
The expectations are that if the function Id(x) has a maximum
and it is strongly concave in the vicinity of the maximum, the
beam has damage and the nodes close to the maximal value of the
function will represent the damaged area.
The damage criterion based on this index presumes setting a
threshold value Td for the damage index:
Id x, DT 4T d

28

This threshold depends on the specicity of the problem as


well as on the sampling. It is important to notice that the
temperature changes should also be taken into account. For this
reason the damage index dened by Eqs. (26) and (27) is
calculated at equal values of DT for the healthy and
damaged beams.

5. The case study


The numerical study focuses on damage detection and localization of composite beams. This is level II of damage detection
according to the Rytters classication of damage detection [34].
As the beams with different levels of damage are studied in this
work and the inuence of the level of the damage on the damage
index is shown, one can conclude that the study includes also

124

E. Manoach et al. / International Journal of Mechanical Sciences 62 (2012) 120132

elements of level III (assessment of damage) according to the


Rytters classication.
The considered structure is a symmetric cross-ply laminated
beam composed of 10 orthotropic layers, each 0.25 mm thick. The
two layers located above and under the neutral surface have 01
orientation with respect to x-axis. The whole beam sequence of
layers (plies) is [(90/0)2/0]s. The length of the beam is l 80 mm
and the width b 5 mm. Material characteristics of the single ply
are as follows:
1.
2.
3.
4.

Youngs moduli: E1 56 GPa, E2 16 GPa,


Poissons ratio: u 0.269,
thermal expansion coefcient: aT 13.2  10  6 K  1,
mass density r 2052 kg/m3.

These characteristics correspond to a glass-epoxy composite


material.
Using Eqs. (4) one can obtain D11 1.6375  102 Nm and
A55 3.9401  107 N/m. Concerning the rigidity (stiffness) properties of the beam these coefcients correspond to an equivalent
homogenous beam with effective properties Eef 41.92 GPa,
nef 0.32991.
In the rst considered case of a damage (Case A), a beam fault
is modeled by prescribing to a small part of the beam reduced
rigidity. The defect is located at x A (0.56 m, 0.64 m) (10% of the
beam length, c.f. Fig. 1) and has the following stiffness characteristic: Edef 0:5Eef 20:96 GPa. The beam is discretized by 40 linear
beam nite elements. This reduction of the Youngs modulus is
quite high and we may expect the inuence of the damage to be
well expressed. The goal is also to study the inuence of the
temperature changes on the DD process when damage is
essential.
Then two other cases are consideredCase B with the Youngs
modulus reduction Edef 0:8Eef and Case C with the Youngs
modulus Edef 0:9Eef .
These different cases of a damage are analysed in order to
clarify the sensitivity of the method to detect damages with
different sizes. It is known that the material discontinuities, such
as cracks, pores or delaminations (single or multiple) can be a
source of enlarged compliance (reduced stiffness). Some cracks or
delaminations in thick beams can provoke a very small reduction
of the stiffness of the beam. That is why we consider a beam with
10% and 20% reduction of the Youngs modulus. For the considered set of composite samples used in our experimental study
(see Section 7) a special 3 point bending test was performed, in
which the span of the bottom rollers was reduced to the length of
the delamination (approx. 30 mm) and the force-deection curve
was obtained with the Zwick Z100/SN3A universal testing
machine and the appropriate TestExpert software. The plot was
subsequently compared to a similar one, obtained for the healthy

structure at the same span of the supporting rolls. Next, inclination angles of experimental curves in their straight parts were
compared. The ratio of the inclination angles tangencies was
assumed to be the equivalent of bending-moduli ratio in case of
the delaminated and the healthy beam, respectively. According to
the measured results the value of the ratio was 0.62. These tests
dened for us that the upper limit of the Youngs modulus
reduction in our numerical calculation to be 50% (Case A).
The location of the damage for Cases B and C is the same as for
Case A.
The aim of the following numerical examples is to test the
procedure for DD at the presence of temperature changes, to
detect and localize damage (delamination), to check the sensitivity of the method and to estimate the temperature inuence on
the process of damage detection.

6. Results from the numerical simulations and discussion


In this section results of the detection and localization of the
defects described in the above-considered beams are discussed.
The inuences of the defect and the temperature changes on the
system response are also commented.
First of all, the sensitivity of the rst seven natural frequencies
of the beams (calculated by FE method) to damage are tested (see
Table 1). For the beam with a considerable fault the change in the
rst natural frequency is only 0.45%. The change in the third and
the second frequencies is more essential (about 5%) and could be
eventually an indicator for damage (without showing its location). The changes in the natural frequency in Case B and Case C
are so small (less than 1%) that they cannot be used as indicators
for damage.
Then the forced response of the beam subjected to a harmonic
loading is tested. All considered beams are subjected to two kinds
of loadings: (a) excitation with frequency equal (or very close) to
the rst natural frequency and (b) excitation with a frequency
equal to a half of the rst natural frequency. The beams are
additionally subjected to different temperature changes.
The beam from Case A is subjected to thermal loads:
DT 10 K, DT 20 K and DT 30 K. In all cases time history
diagrams and Poincare maps are plotted for intact (healthy) and
damaged beams. Let us rst have a look at the time histories of
the beams responses. In the cases when the excitation frequency
is close to the rst natural frequency of the beam a beating
phenomenon is observed (c.f. [22,24]).
The inuence of damage on the time history diagrams of the
beam subjected to a loading with excitation frequency
oe 5665 rad/s (  o1/2) at different temperature changes can
be seen in Fig. 2ac. In Fig. 3ac are shown time history diagrams
at excitation frequency almost equal to the rst natural frequency

Table 1
Comparison of the natural frequencies of the healthy and damaged beams.
Mode no.

Healthy beams natural


frequencies (rad/s)

Damaged beams
Edef 0.5Eef

1
2
3
4
5
6
7

11,330
30,996
60,171
98,265
144,702
198,818
259,887

Edef 0.8Eef

Edef 0.9Eef

Frequencies
(rad/s)

Relative differences Frequencies


(%)
(rad/s)

Relative differences Frequencies


(%)
(rad/s)

Relative
differences (%)

11,278
29,365
57,293
96,239
140,329
190,244
249,922

0.46
5.26
4.78
2.06
3.02
4.31
3.83

0.05
1.05
1.31
0.52
0.38
1.04
1.16

0.09
0.84
1.13
0.70
0.58
0.74
0.79

11,324
30,671
59,381
97,756
144,145
196,753
256,881

11,320
30,737
59,494
97,580
143,857
197,348
257,840

E. Manoach et al. / International Journal of Mechanical Sciences 62 (2012) 120132

0.0050

0.002

0.0025

w/l

w/l

0.004

125

0.000

-0.002
16000

0.0000

20000

24000

28000

32000

-0.0025

10000
20000
dimensionless time

dimensionless time

30000

0.0050

w/l

0.0025

0.0000

-0.0025
10000

15000

20000
dimensionless time

25000

30000

Fig. 2. Time histories of the beam center for different temperatures. p 50 N, oe 5665 rad/s (Case A). (a) DT 0; (b) DT 10; (c) DT 30. Black line, healthy beam;
red line, damaged beam. (For interpretation of the references to color in this gure caption, the reader is referred to the web version of this article.)

of the healthy beam (oe 11,330 rad/s). From these time-history


diagrams it can be concluded that the considered damage leads to
small changes in the amplitude of responses but the time
histories undergo signicant changes in the period of beating. It
can be seen that at the very beginning (t 0) the responses
almost coincide. Then the phase shifts and the differences
between the responses increase (see zooms inserted in the main
gures with time histories shown for a very short period of time).
The changes are more essential when oe o1. It is reasonable to
expect that small changes in the rst natural frequency due to
damage are more important when a beam is excited in the most
sensitive frequency region. The computations conrm these
expectations.
Figs. 4ac and 5ac show that the considered damage does not
change the type of the Poincare map but it only slightly inuences
the size of the radius of the circle formed by the dots. Nevertheless, in all cases of temperature loading the presence of

damage and its location is very well predicted by the damage


criterion based on the damage index (Eq. (26) ) (see Figs. 6 and 7).
It is important to notice that when oe E o1/2 the level of the
damage index increases with the temperature elevation, i.e. the
elevated temperature strengthens the inuence of damage. However, when oe E o1, the relation between the damage index and
the temperature is oppositethe increased temperature
decreases the damage index in the delaminated area. The explanation of this phenomenon is simple. The increased temperature
leads to a change of the natural frequencies of the beam and in
this way takes away the response of the beam from the most
sensitive region around the resonance. This leads to reduced
deections in comparison with case DT 0 and the absolute
values of the corresponding damage indexes for the heated beams
are a little smaller than the damage index for the unheated beam.
This observation could be important when an excitation frequency must be chosen for a damage detection purposes. In all

126

E. Manoach et al. / International Journal of Mechanical Sciences 62 (2012) 120132

0.050

0.050

0.025

w/l

w/l

0.025

0.000

0.000

-0.025

-0.025
0

10000

20000

30000

10000

dimensionless time

20000

30000

dimensionless time

0.06

w/l

0.03

0.00

-0.03
0

10000

20000

30000

dimensionless time
Fig. 3. Time history diagrams of the beam center at different temperatures. (a) DT 0; (b) DT 10; and (c) DT 30. p 50 N, oe 11,330 rad/s (Case A). Black line,
healthy beam; red line, damaged beam. (For interpretation of the references to color in this gure caption, the reader is referred to the web version of this article.)

cases the damage criterion based on the damage index suggested


here shows very good capability to predict the damage location.
The next examples aim to show the applicability of the
damage detection criterion in a case of beams having very small
faults leading to reduction of Youngs modulus by 20 or 10% only.
Usually such faults are not easy to be registered by the traditional
methods and in some cases are not dangerous for the structure.
Only those cases when the excitation frequency is practically
equal to the rst natural frequency are shown here because at this
excitation the structures are more sensitive to any changes, as can
be seen from the levels of the damage indexes plotted in
Figs. 6 and 7. As can be expected the damage does not inuence
essentially the Poincare map of the response (not shown here).

Nevertheless, if one computes the damage index by Eq. (26)


the damage location is very well expressed as the nodes with the
maximal values of the function Id(x)see Fig. 8. Similarly to the
case shown in Fig. 7 the elevated temperature decreases the value
of the damage index. However, if one tries to construct a damage
index using data for the unheated healthy beam and for the
heated damaged beam the damage index is high (which means
that there are large differences between the beams responses) but
the damage location cannot be predicted as can be seen in Fig. 9.
Moreover, the nature of the curve Id(x) in Fig. 9 could not be
considered as an indication of intact condition of the beam
because the curve Id(x) does not have a well-expressed maximum.
This example shows that it is very important to measure the data

E. Manoach et al. / International Journal of Mechanical Sciences 62 (2012) 120132

0.0002

dimensionless velocity

0.0002

dimensionless velocity

127

0.0000

-0.0002
-0.002

0.000
w/l

0.0000

-0.0002
-0.002

0.002

0.000
w/l

0.002

dimensionless velocity

0.00025

0.00000

-0.00025
-0.002

0.000
w/l

0.002

Fig. 4. Poincare maps for the response of the beam center at different temperatures. p 50 N, oe 5665 rad/s (Case A). (a) DT 0; (b) DT 10; and (c) DT 30. Black
dots, undamaged beam; red dots, damaged beam. (For interpretation of the references to color in this gure caption, the reader is referred to the web version of this
article.)

for the inspected damaged structure at the same temperature at


which data are collected for the reference one.
The smallest damage considered in Case C (Ed/E 0.9) leads to
very small changes in the time history diagrams and in Poincare
maps, which for DT 15 are smaller than for DT 0 (not shown
here). Nevertheless, the computed Damage index for this case
again shows quite precisely the location of the damagesee
Fig. 10.
It is notable that the maximum of the curve, which the damage
index forms, is not in the center of the defect but at the edge of
the defect (in the direction of the increasing values of the
deection).

7. Experimental tests
7.1. Specimens and experimental set-up
The goal of the paper is to validate the proposed method for
DD and to check its applicability by the series of experimental
laboratory tests. The specimens are composite beams prepared
from 10-layer glass-epoxy composite. The laminates were fabricated according to the prepreg technology in a form of plates.
For this purpose the TVR380M12/26%/R glass-epoxy roving tape

was used and the laminate was cured with an autoclaving


technique. A package of 10 correctly oriented composite layers
was put into a hermetic foil and a sub-atmospheric pressure of
 80 kPa was established with a vacuum pump. Next, the whole
package was placed in an autoclave under 450 kPa of overpressure, what kept the laminas in a correct position while curing.
The total curing time was more than 4 h with 2.5 h at a constant
135 1C temperature providing the high quality of polymerization
process.
The manufactured plates were subsequently cut into stripes
the beams of two different widths: 12.5 7 0.5 mm and
25 7 0.5 mm. The cutting process was realized with a cutting
machine equipped with a diamond saw. Each plys thickness was
0.255 mm. The ply sequence was [(7451)201]s. what gave
2.55 mm of the beam thickness in total. During the manufacturing
process in the mid-plane, i.e. between the two 01-plies in chosen
places, pieces of 0.02 mm thick polytetrauoroethylene (PTFE) foil
were introduced in order to imitate a damage in the form of
delamination. Such a procedure is a standard one used for making
an articial delamination in layered composites.
The experiments have been performed for cantilever beam
samples with and without delamination and two different geometries. In the rst series of the specimens (Case 1) the beam
has a 93 mm length measured from the clamp and 25 mm width.

128

E. Manoach et al. / International Journal of Mechanical Sciences 62 (2012) 120132

0.004

dimensionless velocity

dimensionless velocity

0.004

0.000

-0.004
-0.02

0.00
w/l

0.02

0.000

-0.004
-0.02

0.00
w/l

0.02

dimensionless velocity

0.005

0.000

-0.005
-0.03

0.00
w/l

0.03

Fig. 5. Poincare maps for the response of the beam center at different temperatures. p 50 N, oe 11,330 rad/s (Case A). (a) DT 0; (b) DT 10; and (c) DT 30. Black
dots, undamaged beam; red dots, damaged beam. (For interpretation of the references to color in this gure caption, the reader is referred to the web version of this
article.)

The delamination in the damaged beams has a 10 mm length and


it starts at 4 mm from the clamped edge.
The second set of specimens (Case 2) is of length 100 mm and
width 12 7 0.5 mm. The delamination for the damaged beams
for this series was located between 35 and 45 mm, measuring
from the clamp. Sketches of the samples can be seen in Fig. 11a
and b.
These gures present the 10 markers which are plotted for
each of the beams. As can be seen in the case of the longer beam
the markers are not equidistance but they are condensed in the
delaminated areatwo markers at the edges of the delamination
and one in the middle.
A special clamp has been designed in order to clamp two
beams, damaged and undamaged, simultaneously.
The experimental setup has two sub-systems. The rst excitation sub-system consists of an electrodynamic shaker with a
dynamic range 650 N and a controller. Two composite beams
with and without a defect are placed in parallel in the handle
which is xed to the shaker armature. To control the motion of
the shakers armature (kinematic excitation) a closed loop controller is applied. A desired kinematic excitation follows a sine
rule with 10 g amplitude (where g is the gravity acceleration) and
a frequency of excitation close to the rst natural frequency of the
bending mode. The control algorithm takes a signal from an
acceleration sensor which is mounted on the clamp. The second

a visional sub-system is designed to record the movements of


both beams in the series of images. The high speed camera that is
used has a maximal resolution of 1632  1200 pixels and a
corresponding speed of 1016 frames per second (fps). In the
presented example the camera resolution of 912  304 pixels
and a speed of 6250 fps have been used. The experiment lasted
8468 ms with 52,926 recorded pictures. The position of the
camera was calibrated so that its optical axis was perpendicular
to the plane of the beam motion. This allowed us to use a 2D
analysis method. The images have been processed in TEMA
software by applying the best tracking algorithm, Quad [35].
The Quad algorithm uses markers with a clearly dened shape.
The accuracy of this tracking method is about 0.1 pixels. The nal
results from the analysis of the images are time series of velocities
and displacements of the observed nodes.
The experimental setup is presented in Fig. 12a and b.
7.2. Experimental results
Series of experiments have been performed with the beams
with the two different geometries. A large number of tests with
specially selected combination of the samples have been performed in order to avoid any inuence of the accidental differences in the beams geometry or material. One health beam has
been excited consecutively with three different damaged beams,

E. Manoach et al. / International Journal of Mechanical Sciences 62 (2012) 120132

129

Fig. 8. Damage index for a beam (Case B) subjected to harmonic mechanical


loading with an amplitude p 10 N, excitation frequency oe 11,330 rad/s at
two thermal loads.
Fig. 6. Damage index for a beam subjected to harmonic mechanical loading with
an amplitude p 50 N, excitation frequency oe 5665 rad/s and different
thermal loads.

Fig. 9. Damage index for a beam subjected to harmonic mechanical loading with
an amplitude p 10 N, excitation frequency oe 11,330 rad/s computed taking
the response for unheated healthy beam and heated (DT 15) damaged beam.

Fig. 7. Damage index for a beam subjected to harmonic mechanical loading with
an amplitude p 50 N, excitation frequency oe 11,330 rad/s and different
thermal loads.

then another healthy beam has been again tested together with
three damaged beams, etc.
Case 1. For Case 1 the beams are tested with two slightly different
excitation frequencies oe 140.625 Hz and oe 140 Hz.
The results from all performed experiments are very similar. In
some of them some errors arise in the displacements and velocity
in node 10 and for these experiments node 10 has been excluded
from the computations of the damage index. Due to the limited

space of the paper it is not possible to present all results received


from the experiment. Therefore, only some representative results
are presented. In Fig. 13 typical time-history diagrams for node 5
(center of delamination) are shown. The calculated damage
indexes from the responses of the beams excited with two
different frequencies are shown in Fig. 14. As can be seen the
curve of DI has a strongly expressed maximum in node 6 which
neighbors node 5 presenting the center of the delaminated area.
In order to clarify whether this maximum of DI is connected with
the delamination, the delaminated zone was marked with three
markers in the next experimentCase 2.
Case 2. For this set of beams a simple modal analysis was
performed. Using a modal hammer with piezoelectric sensor
(to measure the input signal) and an optic sensor (to measure
the resulting displacements) the rst natural frequencies of the

130

E. Manoach et al. / International Journal of Mechanical Sciences 62 (2012) 120132

healthy and damaged beams have been determined. They are:


ou1 125:6 Hz for the healthy beam and od1 125:1 Hz for the
beam with delamination. The relative difference in the rst
natural frequencies is 0.4%.

0.06

T=0

Damage index

0.04

T=15
0.02

For the second set of beams, by changing the beams which are
simultaneously clamped, 10 tests have been performed. Three
healthy and three damaged beams have been used in various
combinations. The healthy and damaged beams have been excited
with a frequency of 125 Hz and acceleration equal to 10 g.
Actually, in 9 of all 10 experiments were obtained identical
results.
Phaseplot diagrams with Poincare maps on them for node
6 for the healthy and damaged beams are shown in Fig. 15.
A typical plot of the measured damage index obtained for the
second sets of beams (Case 2) is shown in Fig. 16. The experimentally obtained damage index has its maximum in node 6 that
is the edge of the delamination in a direction of increasing
displacements. This result has been obtained in 9 of 10 experimental series. The magnitude of the damage index varies slightly
in each experiment. This value depends on the starting time of the
sampling of the trajectories which the nodes describe in the phase
space. Only in one experiment the maximum has been obtained
in node 7.

1.5
1

2.5

0.5
0

0
1

11

21
Node numbers

31

41

-0.5

1.5

-1.5

-1

Fig. 10. Damage index for a beam (Case C) subjected to harmonic mechanical
loading with an amplitude p 10 N, excitation frequency oe 11,330 rad/s at
two thermal loads.

0.002 0.004 0.006 0.008

w, mm

1
0.5
0
-0.5
-1
-1.5

0.4

0.8

1.2

1.6

time, s

Fig. 11. Scheme of the tested beams with the markers. (a) Beam Case 1 and
(b) beam Case 2.

Fig. 13. Time history diagram for point 6. Healthy (black color) and damaged (red
color) beams. Excitation frequency oe 140 Hz. (For interpretation of the
references to color in this gure caption, the reader is referred to the web version
of this article.)

Fig. 12. Experimental set-up.

E. Manoach et al. / International Journal of Mechanical Sciences 62 (2012) 120132

131

4
Damage index

Damage index

3
2
1

0
1

Node numbers

4
6
Node numbers

10

Fig. 14. Damage indexes (Case 1). (a) oe 140 Hz and (b) oe 140.625 Hz.

1500

5
Poincare maps

1000

Damage index

velocity, mm/s

500

-500

-1000
0
-1500

-2

-1.5

-1

-0.5

0
0.5
w, mm

1.5

Fig. 15. Phase diagram of the response of healthy (black lines) and damaged (red
line) beams in Node 6. The Poincare map of the damaged beam is ploted with
black dots and the Poincare map of the healthy beam with red dots
oe 125 Hz. (For interpretation of the references to color in this gure caption,
the reader is referred to the web version of this article.)

All experimental results allow us to conclude that the proposed damage index is quite sensitive to the presence of delamination and it shows exactly the edge of the delamination in the
direction of the increasing displacements. These experimental
results correspond to the results obtained numerically (see
Section 5). Obviously, the end of the defect is the location where
a sharp change of the properties of the material occurs, which
lead to the biggest disturbance in the stress and strain elds of the
beam. These changes affect the damage index.
It should be noted that the experimental study was performed
only for beams at room temperature. With these experimental
tests we proved only the capability of the proposed method for
damage detection to predict a damage and its location. The
sensitivity of the damage index to the temperature inuence is
not studied experimentally due to lack of laboratory equipment.

8. Conclusions
A numerical approach has been applied to study the geometrically nonlinear vibration of thermally loaded composite beams

5
6
Node numbers

10

Fig. 16. Damage index for the beams excited with oe 125 Hz (Case 2).

with or without damage. Computed time-domain responses have


been used to analyse the behavior of either intact or damaged
beams. Based on these analyses a concept of damage index
developed previously has been adapted and applied for a damage
detection and location. It has been demonstrated that damage can
inuence substantially the time-domain response of a beam
despite its very small inuence on the beam natural frequencies.
It has turned out that the inuence of the temperature changes is
essential and able to change substantially the nonlinear dynamic
response of the beam. For this reason temperature changes
should be taken into account when developing damage assessment procedures.
A vision based measurement technique has been applied to
test the suggested vibration based damage detection method
experimentally. Laminated beams in intact condition have been
excited simultaneously with beams having delamination and new
damage detection criteria have been tested.
The experimental results conrmed perfectly the applicability
of the method. The proposed approach for a damage detection can
be applied without a priori information of the presence of a
damage and it does not need any knowledge of the model of the
structure. These facts and the simplicity of the method make the
proposed approach for a damage detection potentially very useful
for inspection of structures in variety of industrial applications.
The authors have submitted a patent pending application for the
method to the Polish and European patent ofce.

132

E. Manoach et al. / International Journal of Mechanical Sciences 62 (2012) 120132

Acknowledgments
The research leading to these results has received funding
from the European Union Seventh Framework Programme (FP7/
2007-2013), FP7-REGPOT-2009-1, under Grant agreement no.
245479. The support by Polish Ministry of Science and Higher
EducationGrant no 1471-1/7.PR UE/2010/7 is also acknowledged. The rst author wishes to acknowledge the partial support
received through Bulgarian NSF Grant DCVP 02/1.
References
[1] Zou Y, Tong L, Steven GP. Vibration based model-dependent damage
(delamination) identication and health monitoring for composite
structuresa review. J Sound Vib 2000;230(2):35778.
[2] Rizos PF, Aspragathos N, Dimarogonas AD. Identication of crack location and
magnitude in a cantilevered beam from the vibration modes. J Sound Vib
1990;138:3818.
[3] Banks HT, Inman DJ, Leo DJ, Wang Y. An experimentally validated damage
detection theory in smart structures. J Sound Vib 1996;191:85980.
[4] Verboven P, Parloo E, Guillaume P, Overmeire MVan. Autonomous structural
health monitoringPart I: modal parameter estimation and tracking. Mech
Syst Signal Process 2002;16:63757.
[5] Parloo E, Verboven P, Guillaume P, Van Overmeire M. Autonomous structural
health monitoringPart II: vibration-based in-operation damage assessment.
Mech Syst Signal Process 2002;16:65975.
[6] Andreaus U, Casini P, Vestroni F. Frequency reduction in elastic beams due to
a stable crack: numerical results compared with measured test data. Eng
Trans 2003;51(1):116.
[7] Doebling SW, Farrar CR, Prime MB, Shevitz DW. Damage identication and
health monitoring of structural systems from changes in their vibration
characteristics: a literature review. Report LA-12767-MS. NM, USA: Los
Alamos National Laboratory; 1996.
[8] Andreaus U, Baragatti P. Fatigue crack growth, free vibrations and breathing
crack detection of aluminium alloy and steel beams. J. Strain Anal Eng Des
2009;44(7):595608.
[9] Reddy AD, Reheld LW, Haag RS. Inuence of prescribed delaminations of
stiffness-controlled behaviour of composite laminates, effects of defects in
composite materials. ASTM STP 836. Am Soc Test Mater 1984:7183.
[10] Lee BT, Sun CT, Liu D. An assessment of damping measurements in the evaluation
of integrity of composite beams. J Reinf Plast Compos 1987;6:11425.
[11] Cattarius J, Inman DJ. Time domain analysis for damage detection in smart
structures. Mech Syst Signal Process 1997;11:40923.
[12] Todd. M, Nichols JM, Pecora LM, Virgin L. Vibration-based damage assessment utilizing state space geometry changes: local attractor variance ratio.
Smart Mater Struct 2001;10:10008.
[13] Moniz L, Nichols JM, Nichols CJ, Seaver M, Trickey ST, Todd MD, et al.
A multivariate, attractor-based approach to structural health monitoring.
J Sound Vib 2005;283:295310.

[14] Epureanu BI, Tang LS, Padoussis


MP. Exploiting chaotic dynamics for
detecting parametric variations in aeroelastic systems. AIAA J 2004;42(4):
728735.
[15] Andreaus U, Casini P, Vestroni F. Nonlinear dynamics of a cracked cantilever
beam under harmonic excitation. Int J Non Linear Mech 2007;42(3):56675.
[16] Andreaus U, Baragatti P. Cracked beam identication by numerically analysing the nonlinear behaviour of the harmonically forced response. J Sound Vib
2011;330(4):72142.
[17] Tsyfansky SL, Beresnevich VI. Non-linear vibration method for detection of
fatigue cracks in aircraft wings. J Sound Vib 2000;236:4960.
[18] Sohn H, Czarnecki JA, Farrar CR. Structural health monitoring using statistical
process control. J Struct Eng 2000;1261356 2000;126.
[19] Nair KK, Kiremidjian AS, Law KH. Time series-based damage detection and
localization algorithm with application to the ASCE benchmark structure. J
Vib Shock 2006;291:34968.
[20] Hu XT, Qin ZY, Chu FL. Damage detection in plate structures based on spacetime autoregressive moving average processes. In: 9th international conference on damage assessment of structures (DAMAS2011). IOP Publishing J
Phys Conf Ser 2011;305:012119.
[21] Trendalova I, Manoach E. Vibration based damage detection in plates by
using time series analysis. Mech Syst Signal Process 2008;22:1092106.
[22] Manoach E, Trendalova I. Large amplitude vibrations and damage detection
of rectangular plates. J Sound Vib 2008;315(3):591606.
[23] Amabili M, Carra S. Thermal effects on geometrically nonlinear vibrations of
rectangular plates with xed edges. J Sound Vib 2009;321:93654.
[24] Manoach E, Ribeiro P. Coupled, thermoelastic, large amplitude vibrations of
Timoshenko beams. Int J Mech Sci 2004;46:1589606.
[25] Ribeiro P, Manoach E. The effect of temperature on the large amplitude
vibrations of curved beams. J Sound Vib 2005;285:1093107.
[26] Ribeiro P. Thermally induced transitions to chaos in plate vibrations. J Sound
Vib 2007;299:31430.
[27] Reddy JN. Mechanics of laminated composite plates and shells: theory and
analysis. second editionCRC Press; 2003.
[28] Altenbach H, Altenbach J, Kissing W. Structural analysis of laminate and
sandwich beams and plates. An introduction into the mechanics of composites. Lublin: Lubelskie Towarzystwo Naukowe; 2001.
[29] Wee YC, Boay CG. Analytical and numerical studies on the buckling of
delaminated composite beams. Compos Struct 2007;80:30719.
[30] Jofe R, Varna J. Analytical modelling of stiffness reduction in symmetric and
balanced laminates due to cracks in 901 layers. Compos Sci Technol
1999;59:164152.
[31] Kukreti AR, Issa HI. Dynamic analysis of nonlinear structures by pseudonormal mode superposition method. Compos Struct 1984;19:65363.
[32] Manoach E. Dynamic large deection analysis of elastic-plastic Mindlin
circular plates. Int J Non Linear Mech 1994;29:72335.
[33] Manoach E, Karagiozova D. Dynamic response of thick elasticplastic beams.
Int J Mech Sci 1993;35:90919.
[34] Rytter, A. Vibration based inspection of civil engineering structures. In:

Fracture and dynamics. PhD thesis. University of Aalborg,


Denmark; 1993.
p. 44.
[35] /www.imagesystems.tvS.

You might also like