Escolar Documentos
Profissional Documentos
Cultura Documentos
Received 9 November 1998; received in revised form 2 March 1999; accepted 9 April 1999
Abstract
Identication of elastic properties of unidirectional glass/epoxy laminates from the measured eigenfrequencies has been performed. The stiness of the laminates has been investigated by a mixed numerical/experimental method employing the vibration
test of plates. Elastic constants of laminates have been determined by using an identication procedure based on experiment design,
the nite-element method and the response-surface approach. Elastic properties of laminates with two dierent bre-surface
treatments have been compared. It was found that only for the transverse elastic modulus is there a statistically signicant dierence
between the composites with good and poor bre/matrix adhesion. # 1999 Elsevier Science Ltd. All rights reserved.
Keywords: Composites; B. Surface treatments; C. Laminates; C. Elastic properties
1. Introduction
The transverse properties of unidirectional composite
laminates such as stiness, strength, delamination fracture toughness, etc., depend on the bre/matrix adhesion quality. The inuence of interfacial eects on the
mechanical properties have been studied in Refs. [16]
and others. Dierent experimental, analytical and
numerical methods have been used in order to model
the interface behaviour. Among the experimental methods mainly are used such as bre pull-out, fragmentation, transverse tension and shear and other destructive
methods. Some non-destructive methods are discussed
in [5] and other papers. One non-destructive method for
studying the elastic properties of laminates is the vibration test [4,7], which is based on identication of elastic
constants from the measured natural frequencies.
During recent years investigations for developing a
new technique for material identication, the so-called
mixed numericalexperimental technique, have started
[819]. The determination of stiness parameters for
complex materials such as bre-reinforced composites is
much more complicated than for isotropic materials
since composites are anisotropic and non-homogeneous.
* Corresponding author. Tel.: +371-708-9264; fax: +371-782-0094.
E-mail address: rikards@latnet.lv (R. Rikards)
Conventional methods for determining stiness parameters of the composite materials are based on direct
measurement of strain elds. Boundary eects, samplesize dependencies and diculties in obtaining homogeneous stress and strain elds are some of the most
serious problems. Because of this, indirect methods have
recently received increasing attention. One such indirect
method is based on measurements of the structure
response and application of the numericalexperimental
identication technique.
Mixed numericalexperimental methods are sensitive
for model errors because the numerical model is always
based on a series of hypotheses. If the real structure
does not satisfy one or more of these hypotheses, the
model of the structure is evidently not appropriate.
Since the development of mixed numericalexperimental techniques for material identication is aimed at
obtaining a practical method which yields quick and
reliable results, much research has been done in order to
minimise these model errors [8,20,21].
In the meantime many dierent approaches were
produced for identication of the physical parameters
directly characterising structural behaviour (i.e.
Young's modulus and density of the material). In [11]
appropriate comparisons were made between actual
eigenfrequencies of an existing structure and those
obtained through the nite-element analysis. This led to
0266-3538/99/$ - see front matter # 1999 Elsevier Science Ltd. All rights reserved.
PII: S0266-3538(99)00059-7
2016
2. Experimental
2.1. Material
Unidirectionally reinforced laminates were produced
from E-glass bres (12 mm, 63 tex) with two dierent
Sample
a (m)
b (m)
h 10 (m)
(kg/m )
PE01
PE02
PE03
PE04
PE05
PE06
Mean
0.1399
0.1401
0.1400
0.1400
0.1400
0.1401
0.1400
0.1401
0.1401
0.1399
0.1401
0.1404
0.1403
0.1401
2.095
2.072
2.060
2.070
2.010
2.060
2.061
1884
1889
1890
1900
1890
1900
1892
2017
Table 2
Experimental plate exural frequencies fi (Hz) for the PE composite
Mode
Mode shape
Specimens
Mean
m, n
PE01
PE02
PE03
PE04
PE05
PE06
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
1.1
2.0
2.1
0.2
1.2
3.0
2.2
3.1
3.2
0.3
4.0
4.1
2.3
4.2
3.3
5.0
164
270
415
493
587
742
891
1470
1600
1750
2050
2240
2400
162
266
413
487
584
725
886
1470
1600
1735
2036
2395
167
270
420
485
590
740
920
1340
1470
1570
2040
2200
2320
170
270
430
480
590
750
915
1340
1480
1610
1760
2100
2440
173
276
419
491
587
747
910
1320
1480
1630
1760
2450
170
270
410
470
570
720
910
1290
1440
1560
2310
168
270
418
484
585
737
905
1340
1317
1468
1595
1751
2057
2220
2386
2018
E2
21 23
0
E1
E2
E1
4 3 G23 25 0:58 2 33 4
;
8 22 E1
80
number of
measured eigenfrequencies fi i !~ i 2fi . The corresponding numerical eigenfrequencies fi !~ i 2fi for
the set of material parameters i are represented by
!~ 1 ; !~ 2 ; :::; !~ I . Let us consider the scaling parameter C
which is chosen according to the relation [8]
C
!21
!~ 21 E01
x
h
i2
I
!2i C!~ i x2
X
!4i
i2
Subject to constraints
g 1 x 4 2 > 0
g 2 x
E2 4 2 G12 8 2 33 4
;
;
E1
4
E1
160
12
where
0 1 212
or
E2
>0
E1
8 2 33 4
# > 0 or
4 3 2 4 2
16 1
8 22
4
"
G12
>0
E1
25 0:58 2 33 4
g 3 x "
# > 0 or
4 3 2 4 2
8 1
8 22
4
10
G23
>0
E1
r
4 3
4
> 0 or
g4 x
4 2
8 22
r
E1
j12 j > 0
E2
11
max
;
min
i 4i 4i
12
i 2; 3; 4; 5
13
14
15
hence
E1 CE01
16
2019
17
xj i xmin
j
1 max
xj xmin
Bij 1
j
k1
18
2020
Further, in these reference points (18) the nite-element solution of the eigenvalue problem (13) is performed and at least 16 rst numerical frequencies of the
PE composite are obtained for each reference point. A
similar procedure was repeated also for the other plates
(PE01PE06).
6. Approximation of the response surface
Techniques from experiment design and responsesurface methodology [35] are used to build the approximate models from the data in the reference points.
Information on the behaviour of an object can be
obtained from the computer solution in the reference
points of the experiment design. The information can be
represented as a data table, where the response function
yx of the object is to be in relationship with the variables
x. In our case there are four identication variables
representing the elastic constants of the material. The goal
is, by using the data only in the reference points [in our
case these data are obtained by the nite-element solution of the eigenvalue problem (13) in the reference
points], to obtain the relation yx in the mathematical
form or the so-called response surface. Here such
mathematical models (response surfaces) have been
obtained for the rst 16 eigenfrequencies of the laminated plate.
The existing methods of regression analysis are based
on the principle that the equation form is known a
priori, and the problem is to nd coecients of the
equation. However, in most cases the form of the equation must also be determined. Such a method was proposed earlier [2427] to obtain a simple mathematical
model for the structural optimisation problems. This
method was briey outlined and applied to the identication problems in [7].
Further, the procedure to get the approximation yx
for the rst natural frequency f1 of the PE composite
plate is shown. In order to obtain the equation of
regression, the resint program [24] was employed. In
this program the perspective functions are selected by
using the least-squares estimation. Then a step-by-step
reduction procedure of the number of terms in the
model is applied. The diagram of reduction of terms in
the model for the rst frequency is shown in Fig. 2. It is
seen that the rst break in the diagram corresponds to
the regression equation with eight terms. The second
break corresponds to the expression with four terms.
This means that the model with eight terms should be
chosen. On eliminating the eighth term, the correlation
with the data of numerical experiment decreases more in
comparison with the previous step of reduction. In the
expression with four terms, only three parameters of
identication are presented. Thus, the simple model
(response surface) for the rst frequency of the PE
19
z2 5:84 8:67x2 ;
z4 3:39 3:83x4
20
2021
f FEM
x f exp
i
i
100
f exp
i
21
ms
Fs h
Table 3
Elastic constants of the PE composite obtained by identication
Property
PE01
PE02
PE03
PE04
PE05
PE06
1 s1
2
E1 (GPa)
E2 (GPa)
G12 (GPa)
G23 (GPa)
12
38.69
11.80
4.65
4.90
0.263
40.12
11.80
4.67
7.19
0.213
39.80
12.16
5.01
1.31
0.137
38.75
12.30
5.19
2.78
0.315
39.57
12.67
5.73
7.71
0.358
36.16
11.99
5.26
0.246
38.841.44
12.120.33
5.090.41
4.782.76
0.2550.077
39.15
12.16
5.11
2.06
0.269
2022
Table 4
Vibration modes used for identication
Plate
PE01
PE02
PE03
PE04
PE05
PE06
Mean
Mode i
1
10
11
12
13
14
15
16
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
Table 5
Flexural frequencies fi (Hz) (i 1; 2; :::; 16) and residuals i for PE
composite
Mode i
FEM
Exp.
i (%)
Used in identication
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
168
270.5
418
477
585
737
905
1340
1317
1468
1595
1751
2057
2220
2386
167.9
266.4
430.7
484.5
588.3
740.7
887.6
919.6
1382
1320
1449
1588
1737
2073
2232
2337
0.06
1.37
3.04
0.10
0.56
0.50
1.31
3.13
0.23
1.29
0.44
0.80
0.78
0.54
2.06
+
+
+
+
+
+
+
+
+
+
+
Table 6
Elastic constants for Pe composite obtained by independent tests
Property
Vibration test
E1 (GPa)
E2 (GPa)
39.501.18
8.471.76
38.841.44
12.120.33
n1 n2 2
n1 n2
< t=2; n1 n2 2
22
2023
Table 7
Comparison of dierent properties
Property
Test
EP composite
PE composite
t-test (22)
E1 (GPa)
E1 (GPa)
E2 (GPa)
E2 (GPa)
G12 (GPa)
Vibration
Tension [4]
Vibration
Tension [4]
Vibration
39.031.14
38.451.62
12.690.20
10.350.69
5.330.23
38.841.44
39.501.18
12.120.33
8.471.76
5.090.41
Pass
Pass
Fail
Fail
Pass
material with good bre/matrix adhesion (EP composite) the stress/strain curve is nearly linear till ultimate
stress (see Fig. 4) and the value of E2 can be measured
in the static test [4]. For the EP composite the E2 modulus obtained in the vibration test [7] is also about 20%
higher than modulus obtained in the static tensile test.
This could be due to the inuence of viscoelasticity in
the dynamic test.
8. Comparison of elastic properties for PE and EP
composites
In order to compare the laminates with good bre/
matrix adhesion (EP composite) and poor bre/matrix
adhesion (PE composite), dierent material properties
can be used. For example, in [4] the transverse strength
and in [37] the interlaminar fracture toughness properties
have been compared for the same composites considered here. In the present paper elastic properties are
2024
References
[1] Mahiou H, Beakou A. Modelling of interfacial eects on the
mechanical properties of bre-reinforced composites. Composites
Part A 1998;29A:103548.
[2] Steen M, Valles JL. Interfacial bond conditions and stress distribution in a two-dimensionally reinforced brittle-matrix composite. Composites Science and Technology 1998;58:31330.
[3] Wacker G, Bledzki AK, Chate A. Eect of interphase on the
transverse Young's modulus of glass/epoxy composites. Composites Part A 1998;29A:61926.
[4] Kessler A, Bledzki, AK. Does the interface control the transverse
modulus of GFRP? Composite Interfaces, in press.
[5] Mai K, Mader E, Mhule M. Interphase characterization in composites with new non-destructive methods. Composites Part A
1998;29A:11119.
[6] Keusch S, Queck H, Gliesche K. Inuence of glass bre/epoxy
resin interface on static mechanical properties of unidirectional
composites and on fatigue performance of cross ply composites.
Composites Part A 1998;29A:7015.
[7] Rikards R, Chate A, Steinchen W, Kessler A, Bledzki AK.
Method for identication of elastic properties of laminates based
on experiment design. Composites Part B 1999;30B:27989.
[8] Mota Soares CM, Moreira de Freitas M, Araujo AL. Identication of material properties of composite plate specimens. Composite Structures 1993;25:27785.
[9] Sol H, De Visscher J, De Wilde WP. Identication of the viscoelastic material properties of orthotropic plates using a mixed
numerical/experimental technique. In: Brebbia CA, Carlomagno
GM, editors. Computational methods and experimental measurements VI, vol. 2: stress analysis. London/New York: Elsevier
Applied Science, 1993. p. 13142.
[10] Sol H. Identication of anisotropic plate rigidities using free
vibration data. PhD thesis, Free University of Brussels, 1986.
[11] Bolognini L, Riccio F, Bettianli F. A modal technique for the
identication of stiness and mass parameters in large structures.
In: Brebbia CA, Carlomagno GM, editors. Computational methods and experimental measurements VI, vol. 2: stress analysis.
London/New York: Elsevier Applied Science, 1993. p. 33752.
[12] Frederiksen PS. Identication of temperature dependence for
orthotropic material moduli. Mechanics of Materials 1992;13:7990.
[13] De Wilde WP. Identication of the rigidities of composite systems by mixed numerical/experimental methods. In: Vautrin A,
Sol H, editors. Numerical identication of composites. London/
New York: Elsevier Applied Science, 1991. p. 115.
[14] Grediac M, Vautrin A. Mechanical characterization of anisotropic plates: experiments and results. Eur J Mech A/Solids
1993;12(6):81938.
[15] Link M, Zou Zh. A two steps procedure to identify physical parameters of composite material structures using vibration test data.
In: Proc. Int. Conf. on Vibration Engineering, ICVE 94, Beijing.
Beijing: International Academic Publishers, 1994. p. 21724.
[16] Moussu F, Nivoit M. Determination of elastic constants of
orthotropic plates by a modal analysis method of superposition. J
Sound and Vibrations 1993;165(1):14963.
[17] Pedersen P. Optimization method applied to identication of
material parameters. In: Eschenauer HA, Thierauf G, editors.
Discretization methods and structural optimizationprocedures
and applications. Berlin: Springer Verlag, 1989. p. 27783.
[18] Frederiksen PS. Identication of material parameters in anisotropic platesa combined numericalexperimental method. PhD
thesis, Department of Solid Mechanics, The Technical University
of Denmark, 1992.