You are on page 1of 9

Thin-Walled Structures 49 (2011) 11231131

Contents lists available at ScienceDirect

Thin-Walled Structures
journal homepage: www.elsevier.com/locate/tws

On the behavior of thin walled composite beams with stochastic properties


under matrix cracking damage
Prashant M. Pawar 
Shri Vithal Education and Research Institutes College of Engineering, Pandharpur, Solapur 413304, India

a r t i c l e i n f o

abstract

Article history:
Received 19 November 2010
Received in revised form
14 April 2011
Accepted 19 April 2011
Available online 10 May 2011

Thin walled composite beam structures are prone to damage which results in change in the
performance of these structures. The change in the performance due to damage may get confused
with the variation in the performance due uncertainties in the properties of these structures. Here, the
performances of the thin walled composite beam under matrix cracking damage having material
uncertainties are studied. The cross-sectional stiffness properties are obtained using thin walled beam
formulation, which is based on a mixed force and displacement method. The stochastic behaviors of
material properties are obtained from previous experimental and analytical studies. The effects of
matrix cracking are introduced through the changes in the extension, extensionbending and bending
matrices of composites. The effects of matrix cracking on out-of-plane bending, inplane bending and
torsion cross-sectional properties are studied at different crack densities for stochastic material
properties. Further, the effects of matrix cracking and uncertainties on measurable properties such as
deections and frequencies are studied. Results show that the beam responses at different crack
densities get mixed due to the material uncertainties. The estimates of variance obtained for observable
system properties due to uncertainty can be used for developing more robust damage detection
algorithms.
& 2011 Elsevier Ltd. All rights reserved.

Keywords:
Matrix cracking
Uncertainty
Thin walled composite beams
Damage prediction

1. Introduction
Thin walled composite beam structures are used in several
engineering applications such as helicopter blades, wings, trusses
in space structures, antenna legs, submarine hulls, cooling tower
shafts, medical tubing, connecting shafts, transmission poles, tail
boom of helicopter, tube like structures in missiles and launch
vehicles. The composite materials, despite of their signicant
advantages, prone to complicated damage mechanisms are compared to metals due to their heterogeneous composition and
directional properties. The key damage modes in composites are
matrix cracking, ber failure, bermatrix debonding and delamination. For both the monotonic and the fatigue loading conditions [1,2], matrix cracking is the rst failure mode which leads to
the severe damage modes such as debonding, delamination and
ber breakage. Therefore, matrix crack detection is a useful
approach to monitor structural health and identify the point
where more dangerous damage mechanisms begin. However,
matrix crack detection problem becomes complicated in reality
due to involvement of various uncertainties in the composite
structures. The uncertainties range from the statistical nature of

 Tel.: 91 2186 225083; fax: 91 2186 225082.

E-mail address: pawarpm@gmail.com


0263-8231/$ - see front matter & 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.tws.2011.04.004

the material properties of constituent ingredients (e.g., bers and


resin) to the randomness in the fabrication (e.g., layup and curing)
and manufacturing processes.
Earlier matrix crack modeling approaches were just focused on
cross-ply laminates. These approaches were the ply discount
method [3], shear lag model [1,4], self-consistent scheme [5],
variational approach [6], approximate elasticity approach [7],
internal variable method [8] and continuum damage approach [9].
Most of the composite structures use arbitrary lamina composites
and matrix crack modeling in such a laminate becomes a
challenging task. Gudmundson and his co-workers initiated work
stlund [10]
for the arbitrary laminates. Gudmundson and O
derived high accuracy closed form asymptotic expressions for
laminates of arbitrary layups with dilute and innite matrix crack
densities. Gudmundson and Zang [11] and Adolfsson and
Gudmundson [12] developed and further modied an analytical
model for thermoelastic properties of composite laminates
containing transverse matrix cracks. Recently, Singh and Talreja
[13] presented a synergistic damage mechanics approach for
composite laminates with matrix cracks in multiple orientations.
The interaction between the 7 y-cracks and the 901-cracks is
analyzed in terms of the crack surface displacements using a
three-dimensional nite element (FE) model and found to be
signicant only for crack orientations close to 901. However, this
model is not yet demonstrated for any realistic applications.

1124

P.M. Pawar / Thin-Walled Structures 49 (2011) 11231131

Damage modeling in composite beams was initiated with


approximate modeling for matrix cracking [14] and delamination
[1517] modes. Detailed modeling of damage in thin walled
composite structures was initiated by Pawar and Ganguli [18]
by modeling matrix cracking in a hollow circular composite beam
with 7 0m =90n s family of composites. The effects of variation in
crack density on the fundamental frequency, for various combinations of 7 0m =90n s composite was studied. However, a realistic helicopter rotor blades are made of the arbitrary lamina
composites. The effects of matrix cracking on the thin walled
composite were investigated using a box beam and a two-cell
aerofoil section-beam [19]. The effects of matrix cracking on
cross-sectional stiffness properties, bending slope and twist of
composite box and two-cell aerofoil cantilever beam were studied
for matrix cracking in a single lamina group, two lamina groups
and in the complete laminate. Next, Pawar and Ganguli [20]
considered progressive damage growth in the composite blade
which includes the key damage modes in the composite materials
such as matrix cracking, debonding/delamination and ber breakage. The variation of beam cross-section stiffnesses, bending
slopes and twist was studied with the progress in damage. It
was found that the torsion stiffness shows much larger change
due to damage compared to the bending stiffnesses due to
composite damage growth. Therefore, the assumption of an equal
reduction in elastic stiffness for the bending and torsion modes
which is often made to model damage is not accurate.
Several studies have been conducted on uncertainty analysis of
composite structures [21]. Murugan et al. [22] initiated uncertainty analysis for thin walled composite beams for helicopter
rotor blade applications where the effects of uncertainties of
composite materials on the cross-sectional stiffness properties,
natural frequencies, and aeroelastic responses of composite
helicopter rotor blades were investigated. In another study by
You et al. [23] on thin walled composite beams where an
assessment is made to quantify the inuence of random material
properties and fabrication/manufacturing uncertainties on the
aeroelastic response and hub vibratory loads of composite rotor
blades. Recently, Pawar et al. [24] developed an analytical
approach for the estimation of the inuence of material uncertainties on cross-sectional stiffness properties of thin walled
composite beams. Fuzzy arithmetic operators were used to
modify a thin walled beam formulation which was based on a
mixed force and displacement method for the analysis of thin
walled composite beams with elastic couplings. The uncertainties
are introduced using the membership functions of the material
properties for obtaining the membership functions of crosssectional stiffness properties. Further, the propagation of uncertainties in the beam responses are demonstrated for the beams
with and without couplings.
In this paper, the behavior of damaged thin walled composite
beam having material uncertainties is studied. The matrix crack
damage is modeled through the extension, bending and extensionbending stiffness matrices. The material uncertainties are
introduced using Monte Carlo simulation using the stochastic
properties obtained from previous experimental work. The crosssectional stiffness properties are obtained using mixed beam
approach. The histograms of the cross-sectional stiffness properties at different crack densities are obtained using thin walled
beam theory, matrix cracking model and Monte Carlo simulation.
Further, the effects of the material uncertainties and matrix
cracking damage on the bending slopes and twist are studied.
The effects of matrix cracking under uncertainty are also studied
on the rst three modes of frequencies. Assuming that while
growing the damage the uncertainty remains same, the histograms of delta values of cross-sectional stiffness properties and
frequencies are also obtained.

2. Thin walled composite beam theory


This work is focused on investigating the effect of matrix
cracking on the thin walled composite beam structures having
uncertainty in the material properties. As the cross-sectional
dimensions are much smaller compared to its length, these beams
are typically modeled as a one-dimensional beam. One-dimensional approximations are associated with the assumption of local
deformations in terms of four generalized deformations namely
extension U, torsion f, out-of-plane bending by and inplane
bending bz . Through proper modeling of these beams, several
non-classical behaviors, which are exhibited by thin walled
composite structures, can be included. This theory is based on
the combination of force and displacement methods for obtaining
the cross-sectional stiffness properties of the beams [25]. This
theory includes the inuence of thickness of the shell wall which
affects bending and shear deformations. The resulting rst-order
(Timoshenko) shear deformation theory describes the beam
kinematics in terms of the axial, out-of-plane and inplane bending, out-of-plane and inplane shear, torsion, and torsionwarping
deformations.
The geometric relations shown in Fig. 1 between the global
deformations of the beam U, V and W along the x-, y- and z-axes,
and f twist about the x-axis and the local shell deformations are
u, vt and vn along the x, s and n directions, respectively, are used
to obtain the shell strain in terms of global displacements as

exx U,x zby,x ybz,x o f,xx


gxs gxy y,s gxz z,s u0,s V,x y,s W,x z,s rf,x
kxx bz,x z,s by,x y,s qf,xx
kxs 2f,x bz y,s by z,s r f,x =a

ess kss 0

Further using
9 2
8
A11
Nxx >
>
>
>
>
>
> 6
>
>
Nss >
>
>
6 A12
>
>
>
>
> 6
>
<
Nxs = 6
6 A16
6
> 6 B11
>
Mxx >
>
>
>
> 6
>
>
> 6
> Mss >
B
>
>
> 4 12
>
>
;
:M >
B
xs

16

the general constitutive relations for shell


9
38
A12 A16 B11 B12 B16 >
exx >
>
>
>
>
7>
>
A22 A26 B12 B22 B26 7>
ess >
>
>
>
>
>
7>
>
>
<
7
A26 A66 B16 B26 B66 7 gxs =
7
B12 B16 D11 D12 D16 7>
k >
>
> xx >
7>
>
>
> kss >
>
B22 B26 D12 D22 D26 7
>
>
5>
>
>
>
>
:
B26 B66 D16 D26 D66
kxs ;

Fig. 1. The denition of beam and section variables.

P.M. Pawar / Thin-Walled Structures 49 (2011) 11231131

with
(
)
Nsn
Nxn

"

A44

A45

A45

A55

Aij ,Bij ,Dij

XZ
m

Aij

XZ
m

zm 1
zm

#(

zm 1
zm

gsn
gxn

ki kj Qijm dz

These stiffness matrices reduce with increasing crack density.


k
The dimensionless crack density f for the ply k is dened by

)
3

Qijm 1,z,z2 dz

1125

fk

tk
sk

11

Here, sk is the average crack spacing and tk is the thickness of


ply k. The changes in stiffness matrices can be obtained by

i,j 1,2,6

i,j 1,2,6

Material properties

Mean

COV

Distribution

E1
E2
G12

141.9 GPa
9.78 GPa
6.13 GPa
0.42

3.39
4.27
4.27
3.65

Normal
Normal
Normal
Normal

n12

4
4.8 x 10

4.7
4.6
4.5
EIy

The material and ply orientation uncertainties get introduced


to the thin walled beam constitute relation through Qijm , stiffness
coefcients of mth layer in the shell wall sections whereas ply
thickness uncertainty gets introduced through zm, the distance
from the midplane to the lower bottom surface of the mth layer.
With the assumption that the hoop stress ow Nss and the
shear ow Nsn are negligible, the constitutive relation reduces and
treating that the strain measures exx , kxx, and kxs as known and
derive expressions for the shear ow Nxs and the hoop moment
Mss in terms of these quantities using the equations of equilibrium of an element of the shell wall. Modied form of Reissners
semicomplimentary energy function FR is used to obtain the
stiffness matrix relating beam forces to beam displacements as

Table 1
Stochastic material properties of graphite/epoxy.

FR 12Cne e2xx 2Cnk kxx exx Cnf kxs exx 2Cng Nxs exx Cnt Mss exx
Cmk k2xx 2Cmf kxx kxs 2Cmg kxx Nxs Cmt kxx Mss

4.4
4.3
4.2

2
Cff k2xs 2Cfg kxs Nxs 2Cft kxs Mss Cxn g2xn Cgg Nxs

4.1

2
2Cgt Nxs Mss Ctt Mss


4
3.9

The Reissner functional is used to obtain the stiffness matrix


relating beam forces to beam displacements
Z lI
d
6
fFR gxs Nxsa kss Mss g ds dx 0

fFg Kfqg

2
2.5
Crack Density

3.5

0.5

1.5

2
2.5
Crack Density

3.5

0.5

1.5
2
2.5
Crack Density

3.5

7.4

EIz

7.2
7
6.8

6.6
6.4
6.2
0

where F is the generalized beam force vector given as


fFg N My Mz Ts Mo T

1.5

7.6

where q is the generalized beam deformation vector given as


fqg U,x by,x bz,x f,x f,xx T

5
7.8 x 10

Using the equilibrium conditions of shell walls obtained from


the semicomplimentary energy function and the constraint conditions, the generalized beam forces and moments are obtained in
terms of shell wall forces. Finally, using the shell strainsbeam
displacements relation and generalized constitutive relation, the
force displacement relation for thin walled beam is obtained as

0.5

4
2.3 x 10

2.2
3. Matrix crack model
The matrix cracking in the composite is inserted through the
extension (A), extensionbending (B) and bending (D) stiffness
matrices [11]. The stiffness matrices for the presence of matrix
cracks A(c), B(c) and D(c) are obtained by subtracting damage
matrices DA, DB and DD from the stiffness matrices A, B and D
of the virgin laminate

2
1.9
1.8
1.7

Ac ADA

Bc BDB
Dc DDD

GJ

2.1

10

Fig. 2. Reduction in deterministic cross-sectional stiffnesses with increase in crack


density.

1126

P.M. Pawar / Thin-Walled Structures 49 (2011) 11231131

relating the strain increment produced by an array of cracks to


the local face displacement and written as [11]

DB

N X
N q
X
k
l
t k f t l f zl Ckl
EE

13

k1l1

DA

N X
N q
X
k
l
t k f t l f Ckl
EE

12

k1l1

DD


N X
N q
X
t k t l kl
k
l
CBB
t k f t l f zk zl Ckl
EE
4
k1l1

14

Probability

Probability

where N is the number of lamina and the matrix C takes account


of the elastic properties and crack orientation relative to applied
stress and can be written as

0.1
0.09
0.08
0.07
0.06
0.05
0.04
0.03
0.02
0.01
0

0.1
0.09
0.08
0.07
0.06
0.05
0.04
0.03
0.02
0.01
0

k
k kl l l
Ckl
m Q N bm N Q ,

C.D. = 0.0

0.9

0.95

1
EIy/EIy0

1.05

C.D. = 1.0

4. Numerical results

0.82 0.84 0.86 0.88 0.9 0.92 0.94 0.96 0.98


EIy/EIy0

C.D.=2.0

Probability

0.08
0.06
0.04
0.02
0
0.78

0.8

0.82

0.84

0.86 0.88
EIy/EIy0

0.9

0.92

The matrix cracking is introduced using Gudmonsons model


and material uncertainties are introduced using the Monte Carlo
simulation (MCS). The stochastic behaviors of material properties
E1, E2, G12 and m12 are obtained from the earlier experimental
studies [2629]. These studies show that the stochastic behaviors
of the material properties of composites can be reasonably
represented as the normal (Gaussian) distribution with a deviation to characterize most of the randomness. Table 1 shows the
mean, standard deviation (SD) and statistical distribution of the
random variables used in the material uncertainty.
Thin walled composite beam is modeled as a single-cell box
beam with outer width203.2 mm and outer depth 38.1 mm,
having 28 plies with ply thickness 0.127 mm and a balanced
layup as 04 =15=153 =30=302 s in all the walls. The beam
length is considered as 5 m whereas mass per unit length is
6.46 kg/m. These properties are of a typical model of thin walled
box beam equivalent to stiff-inplane helicopter rotor blade [30].
The effects of matrix cracking in the thin walled composite are
added through the extension (A), extensionbending (B) and

0.94
0.12

C.D. = 3.0

0.1
Probability

Probability

15

Thus, damage matrices are proportional to the crack density


[12]. The results of the
and crack displacement vector bkm
m
components connected to pure extension were obtained from
Gudmundson and Zang [11] and the components required to be
added to take bending into account are obtained from Adolfsson
and Gudmundson [12] using the least square t to the results
from numerical integration.

1.1

0.1

0.1
0.09
0.08
0.07
0.06
0.05
0.04
0.03
0.02
0.01
0

m EE, BB

where EE and BB denote pure extension and bending, respectively.


The matrix Nk is dened from the constant unit normal vectors nk
for crack surfaces of ply k can be written as
" k
#
n1 0 nk2
16
Nk
0 nk2 nk1

0.08

C.D=0.0
C.D=1.0
C.D=2.0
C.D=3.0

0.06
0.04
0.02

0.76

0.78

0.8

0.82

0.84 0.86
EIy/EIy0

0.88

0.9

0.92

Fig. 3. Histograms of out-of-plane bending cross-sectional stiffness at different


crack densities.

0
0.75

0.8

0.85

0.9
0.95
EIy/EIy0

1.05

1.1

Fig. 4. Outline of histograms of out-of-plane bending cross-sectional stiffness at


different crack densities.

P.M. Pawar / Thin-Walled Structures 49 (2011) 11231131

Probability
Probability
Probability

0.06
0.04

0
0.75

0.8

0.85

0.9
0.95
EIz/EIz0

1.05

1.1

Fig. 6. Outline of histograms of inplane bending cross-sectional stiffness at


different crack densities.

C.D.=1.0

0.82 0.84 0.86 0.88 0.9

0.92 0.94

0.96 0.98

EIz/EIz0
0.92 0.94

0.1
0.09
0.08
0.07
0.06
0.05
0.04
0.03
0.02
0.01
0
0.78

0.8

0.82 0.84 0.86 0.88


EIz/EIz0

C.D.=2.0

0.9

0.92 0.94

0.1
Probability

0.08

C.D=0.0
C.D=1.0
C.D=2.0
C.D=3.0

C.D.=0.0

0.9 0.92 0.94 0.96 0.98 1 1.02 1.04 1.06 1.08 1.1
EIz/EIz0

C.D.=3.0

0.08
0.06
0.04
0.02
0

0.1

0.02

0.1
0.09
0.08
0.07
0.06
0.05
0.04
0.03
0.02
0.01
0

0.1
0.09
0.08
0.07
0.06
0.05
0.04
0.03
0.02
0.01
0

0.12

Probability

bending (D) stiffness matrices, as given in Eqs. (10). Fig. 2 shows


the reduction in the stiffnesses EIy, EIz and GJ due to matrix
cracking for deterministic material (mean) properties for crack
densities from 0.0 to 4.0 in all the plies. From these plots, it can be
seen that at matrix crack density of about 3.0, the curve becomes
almost at which indicates crack density matrix crack saturation
after which other damage modes will take place. Therefore, in this
study the behaviors of the thin walled composite beam having

1127

0.76 0.78

0.8

0.82

0.84 0.86 0.88


EIz/EIz0

0.9

0.92

Fig. 5. Histograms of inplane bending cross-sectional stiffness at different crack


densities.

stochastic material properties at damage levels of crack densities


of 0, 1.0, 2.0 and 3.0 are studied.
The stochastic behaviors of composite material properties are
modeled using 6000 samples which are decided based on the
number of samples required for the convergence of coefcients of
variation (COV) of cross-sectional stiffness properties [22,23]. The
cross-sectional stiffness properties for the 6000 samples with
damage levels of matrix crack densities of 0, 1.0, 2.0 and 3.0 are
obtained using thin walled beam analysis and matrix crack
model. Fig. 3 shows the histograms of out-of-plane bending
cross-sectional stiffness properties of the beam with damage level
with crack densities of 0, 1.0, 2.0 and 3.0. These histograms are
obtained by normalizing the stiffness values with respect to the
values of the baseline beam at zero damage. From the gure, it is
observed that the material uncertainties of laminas get transmitted to the cross-sectional stiffnesses, resulting in a normal
distribution. The outlines of these histograms are shown in Fig. 4.
Comparison of the histograms of the out-of-plane bending crosssectional stiffnesses at different crack densities shows that the
histograms of the beams with crack densities of 3.0 is clearly
separable from the histograms of undamaged beam. The histograms of crack densities of 1.0 and 2.0 get slightly overlapped
with histograms of crack density of 1.0. However, the separation
of the damage level of crack density of 2.0 and 3.0 becomes
difcult using these cross-sectional stiffness properties.
The histograms of inplane bending cross-sectional stiffness
properties are shown in Fig. 5 and the outlines of these histograms are shown in Fig. 6. The histograms of the inplane bending
cross-sectional stiffness properties show the similar trends that of
the out-of-plane bending cross-sectional stiffness properties.
Fig. 7 shows the histograms of torsion cross-sectional stiffness
properties of the beam and outlines of these histograms are
drawn in Fig. 8. It is observed that the histograms of the torsion
stiffnesses show well separation as compared to that of the
bending stiffnesses. In case of torsion stiffnesses, the histograms
crack density of 1.0 are also separated from the histograms of
undamaged beams. The histograms of stiffnesses having a crack
density of 1.0 can be separated from the histograms of stiffnesses
having a crack density of 3.0. However, separation of histograms
of stiffnesses at crack densities of 2.0 and 3.0 becomes difcult.
4.1. Beam response
While developing the beam damage detection system, beam
stiffness cannot be measured directly. Instead, this system can be
developed by comparing the beam responses of undamaged

P.M. Pawar / Thin-Walled Structures 49 (2011) 11231131

0.1
0.09
0.08
0.07
0.06
0.05
0.04
0.03
0.02
0.01
0

Probabiity

C.D.= 0.0

0.9

0.92 0.94 0.96 0.98

C.D.=1.0

0.75

0.8

0.85

0.9
GJ/GJ0

0.95

1.05

1.1

Fig. 8. Outline of histograms of torsion cross-sectional stiffness at different crack


densities.

0.08
1.4

0.06
0.04
0.02
0
0.76

Probability

C.D=0.0
C.D=1.0
C.D=2.0
C.D=3.0

0.7

0.1
0.09
0.08
0.07
0.06
0.05
0.04
0.03
0.02
0.01
0
0.72

0.78

0.8

0.82

0.84 0.86
GJ/GJ0

0.88

0.9

0.92

Normalised Flap Slope

Probability

0.1

C.D.= 2.0

min CD=0.0
max CD=0.0
min CD=1.0
max CD=1.0
min CD=2.0
max CD=2.0
min CD=3.0
max CD=3.0

1.2
1
0.8
0.6
0.4
0.2
0
0

0.5

1.5

2
2.5
3
Beam Span (m)

3.5

4.5

1.5

2
2.5
3
Beam Span (m)

3.5

4.5

1.5

2
2.5
3
Beam Span (m)

3.5

4.5

1.4

0.74

0.76

0.78 0.8
GJ/GJ0

0.82

0.84

0.86

0.1
C.D.=3.0
0.08

min CD=0.0
max CD=0.0
min CD=1.0
max CD=1.0
min CD=2.0
max CD=2.0
min CD=3.0
max CD=3.0

1.2
1
0.8
0.6
0.4
0.2
0
0

0.5

0.06
1.5

0.04
0.02
0
0.68

0.7

0.72 0.74 0.76 0.78


GJ/GJ0

0.8

0.82 0.84 0.86

Fig. 7. Histograms of torsion cross-sectional stiffness at different crack densities.

Normalised Twist

Probability

0.1
0.09
0.08
0.07
0.06
0.05
0.04
0.03
0.02
0.01
0

1
1.02 1.04 1.06 1.08 1.1
GJ/GJ0

Normalised Lag Slope

Probability

1128

min CD=0.0
max CD=0.0
min CD=1.0
max CD=1.0
min CD=2.0
max CD=2.0
min CD=3.0
max CD=3.0

0.5

0
beams with the responses of the damaged beams. The effect of
matrix cracking on the slopes of the beams for unit bending and
torsion loading with material uncertainty is studied here. Fig. 9
shows the plots of lower (min) and upper (max) limits of the
beam bending slope and twist responses of the beams at different
crack densities. These responses are normalized using tip
responses of the baseline beam. In case of the bending responses,
the band of responses at crack density of 1.0 overlaps almost half
part of the band of responses of the zero crack density. The band

0.5

Fig. 9. Upper and lower limits of beam responses at different crack densities.

of responses at crack densities of 2.0 and 3.0 slightly overlaps


with the band of responses of zero crack density.
From the plots of torsion responses, it observed that the band
of responses at crack densities of 2.0 and 3.0 are well separated

P.M. Pawar / Thin-Walled Structures 49 (2011) 11231131

from the band responses of zero crack density. However, the band
of responses at crack density of 1.0 slightly overlaps with the
band of responses at zero crack density. Therefore, it can be noted
that the torsion responses are more useful for separating the
damaged and undamaged beams under material uncertainties. As
there is mixing of band of responses at crack densities of 1.0,
2.0 and 3.0, prediction of damage level using these responses
becomes difcult.

4.2. Beam frequencies


Most of the studies have used the vibration based signals such
as natural frequencies, mode shapes, etc. The effect of matrix
cracking on the rst three modes of frequencies of the thin walled
beam having uncertainty in material properties are studied here.
These natural frequencies are obtained by considering the thin
walled one-dimensional cantilever beam which is modeled using
EulerBernoulli cantilever beam with the governing equation,


@2
@2 w
@2 w
EIx 2 mx 2 0
2
@x
@x
dt

Kg F o2 Mg F

This equation is solved numerically as a general eigenvalue


problem to obtain the natural frequencies o and the eigenvectors
F of the beam. By considering 20 elements, the baseline values of
the rst three mode frequencies for out-of-plane bending are
14.85, 93.32 and 263.50 Hz whereas for inplane bending these
values are 59.27, 372.42 and 1051.5 Hz.
Fig. 10 shows the outlines of the histograms of the rst three
frequencies of out-of-plane and inplane bending. Similar to the
bending cross-sectional stiffnesses and responses, the frequency
responses also show that the matrix cracking with crack densities
2.0 and 3.0 can be predicted clearly. However, the responses of
crack density of 1.0 get slightly overlapped with the responses of
the crack density of 0.

Until now, the stochastic analysis of the beam cross-sectional


stiffnesses and responses is carried out for total number of
samples. For a single sample the material properties become
deterministic. For example, if we consider the ith sample, its
material properties will be deterministic. If we continue to
measure the behavior of that beam with increase in crack
density at crack densities of 1.0, 2.0 and 3.0 and compare it
with its behavior at zero crack density, that change in value is

0.12

0.08
0.06
0.04

C.D.=0.0
C.D.=1.0
C.D.=2.0
C.D.=3.0

0.1
Probability

C.D.=0.0
C.D.=1.0
C.D.=2.0
C.D.=3.0

0.1

0.08
0.06
0.04
0.02

0.02
13

13.5

14

14.5

15

15.5

0
12.5

16

13

First Flap Mode Frequency

Probability

14

14.5

15

15.5

16

0.08
0.06
0.04

C.D.=0.0
C.D.=1.0
C.D.=2.0
C.D.=3.0

0.1
Probability

C.D.=0.0
C.D.=1.0
C.D.=2.0
C.D.=3.0

0.1

0.08
0.06
0.04
0.02

0.02

0
80 82 84 86 88 90 92 94 96 98 100
Second Flap Mode Frequency

80 82 84 86 88 90 92 94 96 98 100
Second Lag Mode Frequency
0.12

0.12
0.08
0.06
0.04

C.D.=0.0
C.D.=1.0
C.D.=2.0
C.D.=3.0

0.1
Probability

C.D.=0.0
C.D.=1.0
C.D.=2.0
C.D.=3.0

0.1
Probability

13.5

First Lag Mode Freqency


0.12

0.12

0.08
0.06
0.04
0.02

0.02
0
220

18

4.3. Change in stiffness due to matrix cracking

0.12
Probability

free vibration, we have the eigenvalue problem as

17

Here EI(x) is the element exural stiffness which is EIy in case


of out-of-plane bending and EIz in case of inplane bending, m(x) is
the mass per unit length, w(x,t) is the displacement. Based on this
governing equation nite element modeling is done to calculate
the natural frequencies. Details of such method are given in
standard textbooks [31]. By nite element solution, we get the
global stiffness matrix Kg and the global mass matrix Mg. Then, for

0
12.5

1129

230

240

250

260

Third Flap Mode Frequency

270

280

0
220

230

240

250

260

270

280

Third Lag Mode Frequency

Fig. 10. Outlines of the histograms of rst three out-of-plane and inplane bending frequencies at different crack densities.

1130

P.M. Pawar / Thin-Walled Structures 49 (2011) 11231131

5. Conclusions

termed as D value:

DEIy

at C:D: 1:0 i

EIy

at C:D: 0:0 EI y at C:D: 1:0

EI y

19

at C:D: 0:0

If we obtain similar delta values for 6000 samples, the


uncertainties in the delta values can be understood. Fig. 11 shows
the outlines of the histograms of normalized bending and torsion
stiffness delta values. From the gure, it is observed that the delta
values of bending and torsion cross-sectional stiffnesses are well
separated for each crack density. As compared to that of the delta
values for torsion cross-sectional stiffness, the delta values for
bending cross-sectional stiffnesses give better separation of the
outlines of the histograms at different crack densities. This shows
that the delta values give better separation of the damage levels.
Similarly, the delta values of the frequency performances can be
obtained. Fig. 12 shows the outlines of the delta values of the
frequency response of rst harmonics. It can be observed that the
delta values of the frequency responses are also well separated at
crack density values of 1.0, 2.0 and 3.0.

0.12
C.D.=1.0
C.D.=2.0
C.D.=3.0

Probability

0.1
0.08
0.06
0.04
0.02
0
0.08 0.09 0.1

0.11 0.12 0.13 0.14 0.15 0.16 0.17 0.18


EIy

0.1

C.D.=1.0
C.D.=2.0
C.D.=3.0

Probability

0.08
0.06
0.04
0.02
0
0.12

0.14

0.16

0.18
0.2
GJ

0.22

0.24

0.26

Fig. 11. Outlines of histograms of delta values of out-of-plane bending and torsion
stiffnesses at different crack densities.

0.12
Probability

0.1
0.08

C.D.=1.0
C.D.=2.0
C.D.=3.0

0.06
0.04
0.02
0
0.045 0.05 0.055 0.06 0.065 0.07 0.075 0.08 0.085 0.09 0.095
First mode frequency

Fig. 12. Outlines of histograms of delta values of rst out-of-plane bending mode
frequency at different crack densities.

In this work, the performance of thin walled composite beam


with stochastic material properties under matrix cracking is
studied. Using the Monte Carlo simulation and the stochastic
behaviors of material properties obtained from previous experimental studies, the variation in responses of thin walled composite beam due to matrix cracking are studied. Effects of matrix
cracking on out-of-plane bending, inplane bending and torsion
cross-sectional stiffness properties, bending slopes, twist and
bending frequencies are studied. Focus is given on mixing of
beam responses at different crack densities due to uncertainties.
The following conclusions are drawn from the numerical results
which are useful for the formulation of design of matrix crack
detection system:
1. Bending cross-sectional stiffness properties are useful for
distinguishing the damage of crack density of 3.0 from undamaged beam. While separation of damage level of crack
densities of 1.0 and 2.0, slight mixing of data may be possible
due to uncertainties. However, the separation of the damage
level of crack densities 2.0 and 3.0 becomes difcult using
these cross-sectional stiffness properties.
2. Histograms of the torsion stiffnesses show well separation as
compared to that of the bending stiffnesses. Torsion stiffness
indicates that beams with crack density of 1.0 can be well
separated from the undamaged beams and also beams with
crack density of 3.0 can be well separated from beams
with crack density of 1.0. However, separation between beams
with crack densities of 2.0 and of 3.0 becomes difcult.
3. The band of bending responses at crack density of 1.0 overlaps
almost half part of the band of responses of undamaged beams
whereas the band of responses at crack densities of 2.0 and
3.0 slightly overlaps with the band of responses of zero crack
density. The band of torsion responses at crack densities of
2.0 and 3.0 are well separated from the band of responses
of undamaged beams. The band of responses at crack density
of 1.0 slightly overlaps with the band of undamaged beams.
The bending and torsion responses can be jointly used for
predicting crack density of 1.0 which indicates crack initiation
and crack density of 3.0 which indicates initiation of higher
damage levels.
4. For the beams whose responses at undamaged state are known
while developing damage prediction algorithm, delta values
can be obtained as variation of responses from the response of
undamaged state. The delta values of deection and frequency
responses give better separation of the damage levels. The
delta values of the bending deection and frequency responses
clearly separate damage levels of crack densities of 1.0,
2.0 and 3.0.

References
[1] Highsmith AL, Reifsnider KL. Stiffness reduction mechanics in composite
laminates. In: Damage in composite materials, vol. 775. New York: American
Society for Testing and Materials; 1982. p. 10317.
[2] Reifsnider KL. Damage and damage mechanics. In: Reifsnider KL, editor.
Fatigue of composite materials. New York: Elsevier; 1991. p. 1177.
[3] Sendeckyj GP, Richardson MD, Pappas JE. Fracture behavior of thornel
300/5208 graphiteepoxy laminatespart I: unnotched laminates. In: Composite reliability, vol. ASTM 580; 1974. p. 52846.
[4] Flaggs DL, Laws N. Prediction of tensile matrix failure in composite laminates.
Journal of Composite Materials 1985;19:2950.
[5] Dvorak GJ, Laws N, Hejazi M. Analysis of progressive matrix cracking in
composite laminatesI: thermoelastic properties of a ply with cracks.
Journal of Composite Materials 1985;19:21634.
[6] Hashin Z. Analysis of cracked laminates under tension. Journal of Applied
Mechanics 1987;25:8729.

P.M. Pawar / Thin-Walled Structures 49 (2011) 11231131

[7] Talreja R. Stiffness properties of composite laminates with matrix cracking


and internal delaminations. Engineering Fracture Mechanics 1986;25:
75162.
[8] Allen DH, Lee JW. Matrix cracking in laminated composites under monotonic
and cyclic loadings. Microcracking Induced Damage in Composites 1990;111:
6575.
[9] Adolfsson E, Gudmundson P. Matrix crack induced stiffness reduction in
0m =90n = yp =yq s m composite laminates. Composite Engineering 1995;5:
10723.
stlund S. Prediction of thermoelastic properties of compo[10] Gudmundson P, O
site laminates with matrix cracks. Composite Science and Technology
1992;44:283308.
[11] Gudmundson P, Zang W. An analytic model for thermoelastic properties of
composite laminates containing transverse matrix cracks. International
Journal of Solids and Structures 1993;30:321131.
[12] Adolfsson E, Gudmundson P. Thermoelastic properties in combined bending
and extension of thin composite laminates with transverse matrix cracks.
International Journal of Solids and Structures 1997;34:203560.
[13] Singh CV, Talreja R. Evolution of ply cracks in multidirectional composite
laminates. International Journal of Solids and Structures 2010;47:133849.
[14] Lakshmanan KA, Pines DJ. Damage identication of chordwise crack size and
location in uncoupled composite rotorcraft exbeams. Journal of Intelligent
Material Systems and Structures 1998;9:14655.
[15] Murri GB, OBrien TK, Rousseau CQ. Fatigue life methodology for tapered
composite exbeam laminates. Journal of the American Helicopter Society
1998;43:14655.
[16] Purekar AS, Pines DJ. Detecting damage in non-uniform beams using the
dereverberated transfer function response. Smart Materials and Structures
2000;9:42944.
[17] Morozov EV, Sylantiev SA, Evseev EG. Impact damage tolerance of laminated
composite helicopter blades. Composite Structures 2003;62:36771.
[18] Pawar PM, Ganguli R. Matrix crack detection in thin-walled composite beam
using genetic fuzzy system. Journal of Intelligent Material Systems and
Structures 2005;16:381468.

1131

[19] Pawar PM, Ganguli R. Modeling multi-layer matrix cracking in composite


helicopter rotor blades. Journal of the American Helicopter Society
2005;50:35466.
[20] Pawar PM, Ganguli R. Modeling progressive damage accumulation in thin
walled composite beams for rotor blade applications. Composites Science
and Technology 2006;66:233749.
[21] Petit CL. Uncertainty quantication in aeroelasticity: recent results and
research challenges. Journal of Aircraft 2004;43:121729.
[22] Murgan S, Ganguli R, Harursampath D. Aeroelastic response of composite
helicopter rotor with random material properties. Journal of Aircraft
2008;45:30622.
[23] You YH, Jung SN, Pawar PM, Kim CJ. Effect of uncertainty on hub vibration
response of composite helicopter rotor blades. Journal of Aircraft
2010;47:15160.
[24] Pawar PM, Jung SN, Ronge BP. Fuzzy approach for uncertainty analysis of thin
walled composite beams. In: 36th European rotorcraft forum, Paris; 7th9th
September 2010.
[25] Jung SN, Nagaraj VT, Chopra I. Rened Structural dynamics model for
composite rotor blades. AIAA Journal 2001;39:33948.
[26] Antonio CC, Hoffbauer LN. From local to global importance measures of
uncertainty propagation in composite structures. Composite Structures
2008;85:21325.
[27] Shaker A, Abdelrahman WG, Tawk A, Sadek E. Stochastic nite element
analysis of the free vibration of laminated composite plates. Computational
Mechanics 2008;41:493501.
[28] Onkar AK, Yadav D. Forced nonlinear vibration of laminated composite plates
with random material properties. Composite Structures 2005;70:33442.
[29] Vinckenroy GV, De Wilde WP. The use of Monte Carlo techniques in
statistical nite element methods for the determination of the structural
behavior of composite materials structural components. Composite Structures 1995;32:24753.
[30] Smith EC. Vibration and utter of stiff-inplane elastically tailored composite
rotor blades. Mathematical and Computer Modelling 1994;19:2745.
[31] Chandrupatla T, Belegundu A. Introduction to nite elements in engineering.
Prentice Hall-Gale; 1998.

You might also like