Вы находитесь на странице: 1из 171

Structural

Analysis
of
Advanced
Materials

Edited by
Moussa Karama

Structural Analysis of
Advanced Materials

Selected, peer reviewed papers from the International Conference on


Structural Analysis of Advanced Materials (ICSAAM 2009),
September 7-10, 2009, Tarbes, France

Edited by

Moussa Karama

TRANS TECH PUBLICATIONS LTD


Switzerland UK USA

Copyright 2010 Trans Tech Publications Ltd, Switzerland

All rights reserved. No part of the contents of this publication may be


reproduced or transmitted in any form or by any means without the written
permission of the publisher.
Trans Tech Publications Ltd
Laubisrutistr. 24
CH-8712 Stafa-Zurich
Switzerland
http://www.ttp.net
Volume of
Key Engineering Materials
ISSN 1013-9826
Full text available online at http://www.scientific.net

Distributed worldwide by

and in the Americas by

Trans Tech Publications Ltd.


Laubisrutistr. 24
CH-8712 Stafa-Zurich
Switzerland

Trans Tech Publications Inc.


PO Box 699, May Street
Enfield, NH 03748
USA

Fax: +41 (44) 922 10 33


e-mail: sales@ttp.net

Phone: +1 (603) 632-7377


Fax: +1 (603) 632-5611
e-mail: sales-usa@ttp.net

Preface
International Conference on Structural Analysis of Advanced Materials (ICSAAM 2009)
The increased use of advanced materials in high efficiency structures electronic devices, medical equipment,
planes and vehicles requires improved reliability, resistance to degradation, failure and life-span forecasts
under a wide variety of loading conditions. The development of materials with advanced structural properties is
becoming a key factor in industrial and technological progress. The aim of the conference is to provide a forum
for engineers, and researchers, scientists and industrial experts, to present their work and discuss the present
situation with regard to advanced and associated technologies, experimental techniques, numerical analyses and
recent developments in the field of advanced materials. ICSAAM 2009 will include conferences by internationally
renowned researchers as well as oral and poster presentations covering aspects of research and advanced
material technology.
Congress website: http://www.enit.fr/ICSAAM2009/
Our conference establishes a link between process, materials and structure specialists. There is a strong
connection between the various fields of research and their industrial applications. It enables companies and
research organisations to discuss new means already used or usable in industry and to present their innovations.
The conference tries to attract younger researchers and scientists by introducing a system of reduced
participation costs and presentation of their work during plenary sessions; 20% of the participants are PhD
students.
Our Scientific and Technical committee includes representatives from many countries: Europe, USA, Canada,
Israel, Algeria, Tunisia, Morocco, Japan, Latin America and many regions of France - in particular Midi Pyrnes
and Aquitaine. Eminent colleagues include Professor George Papanicolaou (University of Patras), Professors
Costica Atanasiu and Gheorghiu Horia (Polytechnic University of Bucharest) with whom we decided to organise
the 2009 ICSAAM congress in Tarbes (September 7-10, 2009) with the Laboratoire Gnie de Production at the
Ecole Nationale dIngnieurs in Tarbes.
The success of previous congresses (2005 in Roumania and 2007 in Greece) is without doubt related not only to
the topics suggested and the efforts made to promote closer links between Research and Industry, but also to the
effective support provided by the many varied complementary organizations.

Professor Moussa Karama


Chairman ICSAAM 2009 http://www.enit.fr/ICSAAM2009/index.html
Guest Editor Special issue Advanced Materials Research

Table of Contents
Preface
Detectors in Barrier Structures of Metal-Lamellar Semiconductors
M. Mehrabova
The Influence of Some Technological Parameters on the Fracture Toughness of Ceramic
Materials
V. Goanta and M. Mares
Manufacturing of Carbonaceous Materials Based on Olive Stones Biomass for
Electrochemical Applications
A. Pikasi, P. Georgiou and J. Simitzis
Physicochemical and Electro-Rheological Characterization of Kaolinite / CMS / Silicone Oil
Fluid
T. Reiss, S. Laribi, J. Fleureau and J.F. Tassin
The Effect of the Cure Temperature on the Thermomechanical Characteristics of an
Adhesive
M. Ali, B. Lorrain, M. Karama and B. PUEL
Study of DGEBA and Novolac Adhesives Hydrothermal Ageing for a Ceramic/ Steel
Substrates Bonding Assembly
C. Gaussens, V. Nassiet and B. Hassoune-Rhabbour
Influence of Strain Rate on the Yielding Behavior and on the Self Heating of Thermoplastic
Polymers Loaded under Tension
I. Hadriche, E. Ghorbel, N. Masmoudi and F. El Halouani
Dynamic Response of Symmetric and Asymmetric E-Glass / Epoxy Laminates at High
Strain Rates
M. Tarfaoui, S. Choukri and A. Neme
Influence of Process and Material Parameters on Impact Response in Composite Structure:
Methodology Using Design of Experiments
C. Garnier, S. Mistou and O. Pantal
Approximate Solution of the Structural Problems Using Probabilistic Transformation
S. Ouhimmou, A. El Hami, R. Ellaia and M. Tkiouat
Structural Shape Optimization Using an Adaptive Simulated Annealing
W. El Alem, A. El Hami and R. Ellaia
Graft Interpenetrating Continuous Epoxy-Polysiloxane Polymeric Network
T. Djilali, V. Nassiet and B. Hassoune-Rhabbour
Experimental and Numerical Modelling of LRI Process
R. Brault, M. Niquet and S. Mistou
BEM Simulation of 3D Updated Resin Front for LCM Processes
R. Gantois, A. Cantarel, J.N. Flices, N. Pirc and F. Schmidt
FEA of Dynamic Behavior of Top Hat Bonded Stiffened Composite Panel
G. Mohamad, M. Tarfaoui and V. Bertram
Experimental Characterization of Behavior Laws for Titanium Alloys: Application to
Ti5553
V. Wagner, M. Baili, G. Dessein and D. Lallement
Effect of Ductile Damage Evolution in Sheet Metal Forming: Experimental and Numerical
Investigations
F. Abbassi, O. Pantal, S. Mistou, A. Zghal and R. Rakotomalala

1
11
23
33
43
51
63
73
83
91
101
111
121
131
137
147
157

Key Engineering Materials Vol. 446 (2010) pp 1-10


(2010) Trans Tech Publications, Switzerland
doi:10.4028/www.scientific.net/KEM.446.1

DETECTORS IN BARRIER STRUCTURES OF


METAL- LAMELLAR SEMICONDUCTORS
Matanat Mehrabova
Institute of Radiation Problems, Azerbaijan National Academy of Sciences
AZ 1143, Azerbaijan, Baku, 9 F.Agayev str.
metanet-mehrabova@rambler.ru
Keywords: semiconductor, detector, Shottki barrier, photoconductivity, irradiation, absorption
coefficient, voltage.

Abstract. Recently peaceful use of the nuclear energy and radioactive nuclides has increased the
radiation pollution factor in the world and radiation safety problems have become actual ones. The
development of the nuclear power engineering, protection and processing of radioactive wastes of
nuclear reactors, the use of radioactive isotopes in national economy, nuclear explosion, industrial
wastes and etc. may cause the radioactive pollution of the environment. In the case of such pollution
the environment as well as living organisms are exposed to radioactive radiation (-, -, - etc.).
Radioactive substances emit radioactive rays and as their decay time varies for natural and artificial
radioactive substances the time of their environmental effect ranges from some years to million
years.
In consequence of radiation effect new specific properties ecological, psycological,
biological and so on are observed. Its necessary to determine the harmful effect of radioactive
pollution areas on the environment as well as human life and radiation dose in order to protect the
area, foods and people from this effect. It is of great importance to make dosimeters capable of
registering -radiation dose in a wide energy range and with high effectiveness and radiationresistant devices to reveal the singularities of radioactive radiation and assess its safety risks in
order to protect environment and living organisms from these effects [1-4]. The dosimeters made on
this purpose can be used during environmetal control, radioecological service at sanitary-andepidemiologic stations, detection of radioactive materials and plants at customs service, emergency
cases and military dangers. The pollution areas, radiation type and nature, equivalent dose of -rays
and exposure dose rate is determined by dosimeters and the prospecting for radiactive materials
and ionizing sources are realized.
A scintillation method of the registration and the spectrometry of ionizing radiation is a more
developed one among the different methods of detection. This is conditioned to a high extent by the
fact that scintillation materials differ for their high registration efficiency, sufficient temporal and
energy resolution. The demand for scintillation detectors rises for their sensitivity and registration
efficiency of ionizing radiation, space resolution and high-performance, lack of hygroscopicity,
capacity to be applied under extreme climatic conditions (temperature stability, high radiation field)
due to the development and creation of the new generation of high-sensitive and high-performance
radiation equipment (including nuclear power engineering, space researches under extreme
conditions, geophysical instrument engineering).
In comparison with other crystals A3B6 semiconductor crystals are much more radiationresistant, have high sensitivity and anisotropic properties in ultraviolet (UV) spectrum region.
Scintillation semiconductor detectors (SD) on the base of metal-36 lamellar semiconductors are
perspective materials for electron, x- and -irradiation. High photoconductivity in UV-spectral
region exceeding marginal region is observed in these semiconductors[5].
One of the main parameters of electron, x- and -irradiation SDs is a collection efficiency of
charge carriers. In the given work the collection of charge carriers in the detectors on the base of
metal-GaSe, GaS and InSe lamellar semiconductors has been investigated. The mode of SDs
functioning is based on the generation of electron-hole pairs in contact to a metal-semiconductor and

Structural Analysis of Advanced Materials

their collection at applying reverse bias voltage. The optimal values of sample thickness, the voltage
of electric field applied on contacts and photon energy in incident light have been theoretically
determined which supplies effective functioning of the detector. The photoconductivity (PC) has
been calculated depending on electric field voltage and light absorption.
A spatial distribution of non-equilibrium electrons n() along the crystal thickness during
their excitation at several values of the absorption coefficient: =1cm-1, =10 cm-1, =100cm-1 has
been theoretically investigated. During the generation of the electrons at = 1cm-1 non-equilibrium
electrons are distributed at zero bias (V=0 ) more or less uniformly along the crystal thickness. At
imposing external field photoexcited electrons may pass into the crystal depth. The bias voltage is
applied at so high values that all excited electrons reach back electrode (the opposite one) for
efficient functioning of the detector.
It has been theoretically determined that PC in the region of low energy values on negatively
charged irradiated surface during rise in voltage increases but at an opposite polarity of the applied
voltage decreases. PC decreases up to a limiting value in the region of high energy values during rise
in voltage, after which the marginal PC increases in four orders. Besides, during rise in voltage the
margin of PC moves to a short-wave region. By this way, the optimal values of the parameters have
been determined at which a PC maximum is observed in the given structure.
The obtained theoretical results agree with the experimental data. The investigated n-InSe
samples with 106Ohmcm specific resistance at room temperature have been grown by the
Bridgman method.
The surface-barrier structure has been formed by the vacuum deposition of Au film with 22.5mcm thickness on a freshly cleaved InSe surface. Ohmic contacts have been pasted on the
opposite surface of the crystal by a silver paste.
In a wide range of direct (U=0.05-100V) and reverse bias voltages the current of the
investigated structure follows the IUn power law. Besides, n power exponent for direct and reverse
currents in the voltage region up to U5V turns out to be close to the unit which is usually connected
with tunneling of the charge carriers or is explained by the current restricted by a space charge.
The typical spectral dependence of the obtained structures PC at 300K temperature has been
studied. During the illumination of the structure by the side of Au, PC is observed in 0.36-1.8 mcm
spectral region. At the values of U=0.05-1V reverse bias photosensitivity increases which can be
explained by an exciton decay at h=1.6eV quantum energies. The dependence of a PC maximum on
the electric field has been formulated.
On the base of the conducted investigations we have come into conclusion that x- and irradiation detectors can be constructed on the base of metal GaSe, GaS or InSe semiconductor
which will have high sensitivity and differ according its speed.
Introduction
SDs are the most perspective materials in optoelectronics because of the small size of an active
area and the possibility of providing high sensitivity by means of a highly active area creation [6].
To find ways of optimization conditions for the operation of SDs is of great interest. The
optimization of SDs parameters depends on both internal structures of an active element and external
conditions. Scintillation SDs on the base of metal 36 lamellar semiconductor are perspective
materials for electron, x- and -irradiation. One of their main parameters is efficiency of a charge
carrier collection [6, 7]. The study of a charge carrier collection problems in A3B6 based x- and
-ray detectors is now actual as the lifetime of the carriers in these detectors is much less than in
other semiconductors and in order to collect carriers at small thickness (1-2 mm) of the crystal it is
important to put high voltage (100-300 V). These semiconductors are interesting as high PC in UVspectral region exceeding marginal region [8, 9] is observed in these semiconductors.
In the present work charge carrier collections in x- and - ray detectors on the basis of Au
GaS (GaSe, InSe) is investigated. The operation principle of SD is based on the generation of
electron-hole pairs in contact to a metal-semiconductor and their collection at the enclosed reverse

Key Engineering Materials Vol. 446

bias voltage [7]. The optimal values of the sample thickness, the intensity of the electric field
enclosed to the contacts and the energy of falling light photons at which it is possible to provide
effective operation of the detector have been defined. The calculation of the PC depending on the
intensity of an electric field and light absorption has been carried out.

Mathematical description
The operation principle of SD based on the generation of electron-hole pairs on the contact of a
metal-semiconductor and their collection at the enclosed pressure of an inverse displacement has
been determined. The simulated semiconductor detector represents planar structure of metal-GaS
(GaSe, InSe). A crystal with 2 mm thickness is considered. We shall choose an axis of coordinates
on a normal surface of the detector, the beginning of the coordinates =0 we shall fix on the border
of the semiconductor contact with a metal electrode. The electric field is directed along the
distribution of falling light beam on the sample (Fig.1).

Fig.1 The scheme of PC excitation (metal - GaS, GaSe, InSe).


The mobility of the holes in the layered semiconductors is much less in comparison with the
electrons and consequently a hole component of PC can be neglected in comparison with
nn electron (e- is electron charge, and n are excess concentration accordingly holes and
electrons). The concentration distribution for the electrons is determined according to continuity
equation:
1 dj n n

0 exp( x ) 0 .
(1)
e dx
n
Where jn density of an electronic current, n electrons life time, (-) velocity of
electron photocarriers generation, number of the photons falling on a unit area in unit of time,
optical absorption coefficient. By considering the drift and diffusion components of current we
shall obtain:

j n eFn eD n

dn
.
dx

(2)

Given (Eq.2) in (Eq.1) we shall obtain the equation of a special solution:


n (x ) C exp( x ) .

Where:

0
2
1 Ldr 2 L n
Ldr=nFn

(3)

Structural Analysis of Advanced Materials

Ldr drift length the average distance through which an electron passes along the direction of an
electric field during the electron life time n (n = 10-6s). The intensity of an electric field is
determined by an enclosed external voltage F=V/d. The electron diffusion length is:
Ln = (Dnn)1/2
Where, Dn=0.207 diffusion coefficient of electrons. According to the known distribution of
superfluous electrons on the thickness of the sample and the dark conductivity of the semiconductor
we reveal the resistance of a crystal at irradiation:
R( ,V )

e
0

dx
n n 0

(4)

The density of photocurrent is equal to the difference of current at irradiation and darkness:

I V (

1 0

)
R d

(5)

Here 0 dark conductivity of the semiconductor. PC calculated by the formula:

I
.
VS

(6)

S contact area.
The results of the calculation of the electron distribution n () at their excitation at = 1 cm-1,
= 10 cm-1 and = 100 cm-1 are presented on Fig.2a, 2b, 2c, 2d.

a)

b)

c)

d)

Fig.2 Space electron distribution along the sample thickness at different voltages for GaS with an
absorption coefficient: a) =1cm-1, b) =10cm-1, c) =100cm-1, d) =1cm-1 at an opposite polarity
of the electrons for GaS.

Key Engineering Materials Vol. 446

At the generation of the electrons with a small absorption coefficient (=1cm-1)


nonequilibrium electrons are distributed at zero reverse bias voltage (V=0) more or less in regular
intervals on the crystal thickness with recession near to both surfaces because of a superficial
recombination (Fig.2a).
In spite of the fact that the sample thickness d>>Ln at imposing external pulling field
photoexcited electrons can come into the crystal depth. Except for an effective detector operation the
pressure voltage of displacement is so high that all excited electrons reach the back (opposite one)
electrode. At the appendix the reverse bias voltage irrespective of polarity electrons are pushed aside
by a field from a frontal surface. It means that if is so small, that 1/>> d, electron hole pairs are
generated in regular intervals in volume, and under the action of external electric field electrons are
delayed from a negatively charged electrode, PC decreases (Fig. 2a). If the optical absorption
coefficient is big enough and the Ln is less than the crystal thickness d (1/ << d Lnd ), the
absorption of photons occurs in the thin layer adjoining to the irradiated surface. At the appendix to a
crystal of a negative pressure voltage excited electrons will be delayed from a surface in volume and
PC of the sample as a whole increases (Fig.2b, 2c). At an opposite polarity electrons will be pushed
out by an electric field from a crystal and PC of the sample will decrease (Fig.2d).
In Fig.3a, 3b, 4a, 4b, 4c the curves of PC calculated at various polarities and various sizes of
reverse bias voltage are presented by a formula (Eq.6). As it is evident from Fig.3a, 3b PC in the
range of low energy values at a negatively charged irradiated surface voltage increases and at an
opposite polarity of the enclosed voltage decreases.

a)
b)
Fig.3 Spectral characteristics of PC in a low energy range: ) for GaS at various polarities and
various sizes of reverse bias voltage, b) for 1-GaSe, 2- InSe, 3- GaS.
In the region of high energy values (Fig.4a, 4b, 4c) at increase of the pressure PC decreases up
to a limiting value then there occurs an increase edge PC on four orders. Besides, with increase of
voltage the edge of PC moves into a short-wave range.

Structural Analysis of Advanced Materials

a)

b)

c)
Fig.4 Spectral characteristics of PC in a high energy range a) for GaS, b) for 1-GaS, 2- GaSe, 3-InSe,
c) for GaSe at V= -0.01 V voltage.
In Fig.5 the voltage dependences of PC are presented at various values of an absorption
coefficient.

Fig.5 Dependence of photoconductivity on a pressure voltage for GaS


The obtained results will be agreed by the experimental data [5, 8, 9]. Recession of PC in 36
crystals in the high energy region is explained by the photoexcitation of excitons of intermediate
type (EIT). The enclosed electric field of a Shottki barrier on the basis of these crystals ionizing EIT
eliminates the given recession.

Key Engineering Materials Vol. 446

Experiment and results


The obtained theoretical results agree with the experimental data. The investigated samples of
n-InSe with 106 Ohmcm specific resistance at room temperature have been grown by a
Bridgman method.
The surface-barrier structures have been formed by Au film vacuum deposition with 22.5mcm thickness on a freshly cleaved InSe surface. Ohmic contacts have been pasted on an
opposite surface of the crystal by a silver paste.
In a wide range of direct (V=0.05-100V) and reverse bias voltages the current of the
investigated structures follows the IUn power law. Besides, n power exponent for direct and reverse
currents in the voltage region up to V5V turns out to be close to unit which is usually connected
with tunneling of charge carriers or is explained by the current restricted by a space charge (Fig.6).
The typical spectral dependence of the PC of the obtained structures at 300K temperature has
been studied. During the illumination of the structure by the side of Au, PC is observed in 0.36-1.8
mcm spectral region (Fig.7). At the values of a reverse bias V=0.05-1V the photosensitivity
increases which can be explained by an exciton decay at h=1.6eV quantum energies. The
dependence of a PC maximum on an electric field was formulated (Fig.8).

Fig.6 Current-voltage characteristic of Au-InSe at 300 : 1) reverse, 2) direct bias

Structural Analysis of Advanced Materials

a)

b)

c)
Fig.7 Spectral distribution of PC Au-InSe at room temperature: a) illumination from the
semiconductor side, b) = 4.4 . 104 V/cm2, c) E= 2.8 . 104V/cm: 1) illumination from the metal side,
2) illumination from the semiconductor side

Fig.8 Dependence of a PC maximum on an electric field.

Conclusion
The calculation of PC for a planar Au--GS structure depending on the intensity of an electric
field and the light absorption for a configuration when the external voltage is directed along the
distribution of exciting light beam was spent, i.e. under the conditions of division of electron-hole
pairs and collecting of carriers similarly to how it happens in the x- and -ray detector. At such
scheme the value of PC and the form of the obtained spectral curves strongly depend on the polarity
and size of reverse bias voltage. The optimum values of these parameters at which a PC maximum of
the given structure is observed are certain.
It has been determined that PC in the region of low energy values on a negatively charged
irradiated surface during the rise in voltage increases but at the opposite polarity of the applied
voltage decreases. PC decreases up to a limiting value in the region of high energy values during the
rise in voltage, after which a marginal PC increases in four orders. Besides, during the rise in a

Key Engineering Materials Vol. 446

voltage margin of PC moves to a short-wave region. By this way the optimal values of the
parameters have been determined at which a PC maximum is observed in the given structure.
The photon detector on the base of Au InSe was developed, obtained and tested in the
laboratory Radiation physics of semiconductors of Institute of Radiation Problems of ANAS
(Fig.9), on which it is possible to create a scintillation detector of -radiation [10].

Fig.9 Photon detectors on the base of Au - InSe


The suggested detector-avalanche photodiodes of a new generation allows recording the light
of minute intensity. They differ from their analogues according to their small sizes, high sensitivity
(106-107 V/Vt) and operation speed.
The areas of application are in medicine physiotherapy, blood autotransfusion, human
solarization; in agriculture greenhouse and hothouse agrotechnology; in biotechnology synthesis
of D2, D3 vitamins; in disinfection of water, air, clothes, instruments and food products during longterm storage and epidemics; in astronavigation and UV location; in astronomy - data accessing
about physical processes in non-terrestrial objects capable to irradiate UV radiation; in material
science determination of substance composition and electron structure of elements; in ecology
the problem of ozone hole, detection of environmental pollution; in nuclear physics and power
engineering recording nuclear particles with the help of scintillators; in defectoscopy,
criminalities [11].
On the base of the conducted investigations we come into conclusion that x-ray and irradiation detectors can be constructed on the base of metal-GaSe, GaS and InSe semiconductor
which will have high sensitivity and differ according to its speed.
Summary
PC has been calculated depending on an electric field voltage and light absorption. Spatial
distribution of non-equilibrium electrons along the crystal thickness during their excitation at several
values of absorption coefficient has been investigated. It has been determined that PC in the region
of low energy values on a negatively charged irradiated surface during rise in voltage increases but
at an opposite polarity of the applied voltage decreases. PC decreases up to a limiting value in the
region of high energy values during rise in voltage, after which marginal PC increases. During rise in
voltage margin of PC moves to a short-wave region.
References
[1] S.S.Berdenosov, Y.A.Sapojnikov: Ionizing radiation and environment. Soros Educational
Journal, v.7, 2 (2001), p.40-46
[2] A.Kastalsky, S.Luryi, B.Spivak: Semiconductor high-energy radiation scintillation detector.
Nucl. Instrum. Methods Phys. Res., A 565 (2006), p 650-656
[3] F. Knoll Glenn: Radiation Detection and Measurement. (2000), 816 p.

10

Structural Analysis of Advanced Materials

[4] R.S.Madatov, M.A.Mehrabova, F.R.Abbasov, S.S Suleymanov: Rapid detectors for - quanta
on the base of Au-Si. The IV Euroasian Conference on Nuclear Science and its Application.
2006, Baku, Azerbaijan, p.236-238.
[5] O.Z.Alekperov, M.O.Godjaev, M.Z.Zarbaliyev, R.A.Suleymanov: Interband photoconductivity
in layer semiconductors GaSe, InSe and GaS. Solid State Communication, v.77, 1 (1991)
p.65-67
[6] L.A.Bakaleynikov, E.J.Flegontova, K.J.Pogrebitskiy, I.V.Yeryomin: Theoretical Principles of an
Operation of a Semiconductor Detector Based on p-n Junction. Journal of Technical Physics
(Russian), 9 (2004) p.74-85
[7] L.A.Kosyachenko, E.L.Maslyanchuk, I.M.Rarenko, ..Sklyarchuk: Study of Carriers
Accumulation in CdZnTe- Detectors of x-Ray and -Radiation by a Photoelectric Method.
Physics of a Solid State(Russian), 38, 8 (2004) p.1018-1023
[8] R.S.Madatov, M.A.Mehrabova, T.B.Tagiyev, N.I.Huseynov, F.P.Abbasov: x-ray, electron and
-irradiation detectors in barrier structures of metal-lamellar semiconductor. Int. Conf. on Role
of Nuclear Energy in the Energy Supply of the Future, Baku, 2008, p.60-61
[9] M.A.Mehrabova: Photoelectric processes in x-ray and -irradiation detectors on the Au-A3B6
base. Transitions of Azerbaijan National Academy of Sciences, 3 (2009), p.103-108
[10] Information on http:// www.google.com Photon detectors of new generation. Mehrabova
Matanat, Nuclear energy and safety.
[11] T.V.Blank, Y.A.Goldberg: Semiconductor photoelectrosensor for ultraviolet spectral range.
Semiconductors of Physics and Technique, v.37, 9 (2003), p.1025-1055

Abbreviations

Explanations

UV
SD
PC
EIT

ultraviolet
semiconductor detector
photoconductivity
excitons of intermediate type

Metal-A3B6-is a barrier structure which is formed by a metal deposition on a lamellar crystal

Key Engineering Materials Vol. 446 (2010) pp 11-21


(2010) Trans Tech Publications, Switzerland
doi:10.4028/www.scientific.net/KEM.446.11

The influence of some technological parameters


on the fracture toughness of ceramic materials
Viorel Goanta1, a, Marian Mares2, b,
1,2

Technical University Gh. Asachi from Iasi, Bd. Mangeron, 67, Romania
a

email: vgoanta@tuiasi.ro, bemail: mmares@tuiasi.ro

Keywords: fracture toughness , Vickers micro-indentation, finite element analysis, technological


parameters.

Abstract. In order to determine the fracture toughness of the materials presenting high hardness
values in the superficial layers, the Vickers micro-indentation was imposed as a reliable procedure.
That method became attractive because of the relative simplicity of the experimental technique and
because of its low cost. There are several calculus relationships that could be applied using the data
provided by that method, in order to determine the material fracture toughness. The determination
of fracture toughness using the Vickers indentation method is based on the analysis of radial cracks
propagation, from the corners of the indentation trace. The length of these cracks is connected with
the material fracture toughness, on the basis of some semi-empirical calculus relations that are
taking into account the indentation load and some physical characteristics of the test material, as
Youngs modulus and Poissons coefficient.
In the present paper, fracture toughness was determined on a series of ceramic samples, made of
the same material, but with different geometrical shapes and obtained by applying different
technological procedures. The influence of some technological parameters on the fracture toughness
was evaluated. The material fracture toughness was determined, into the vicinity of the propagated
cracks (in a sample that could be a final product), on an area with a specified geometric contour.
As a preliminary stage, a step by step FEM analysis was made, into the Vickers indentation
material region, for different values of indentation load. In this manner, it was proved that the
maximum stress value, on the perpendicular direction, as related to the crack diagonal plane, is
always located at the peak of the indentation trace, and that is the effective start-point of cracking,
for this type of indentation.
Introduction
A major goal for the current ceramic research and development is to produce some tough and strong
ceramic materials, with a reliable performance. Nevertheless, the fracture toughness of ceramics
remains poor, when comparing to that of metals and composites [1]. Precise design and modeling
methodologies are therefore necessary, in order to predict the performance of ceramics. For such
methods of analysis to be applied, statistical valid data are requested about the studied material,
regarding some of its properties as fracture toughness, wear resistance, strength and hardness.
Fracture toughness values are extensively used for characterizing the fracture resistance of
ceramics and other brittle materials [2]. The fracture of brittle ceramics is usually controlled by the
mode I fracture toughness. A simple dimensional analysis of a body containing a crack with a
length of 2a and subject to an applied stress shows that the stress-intensity factor at the crack tip
KI is:
KI = Y a

(1)

where Y is a dimensionless constant that depends on sample geometry and crack configuration.
Fracture toughness of brittle materials is considered as a material parameter. It is admitted that
fracture will occur at a critical stress intensity level, KIc. Most of the fracture toughness studies for
metallic materials are typically using the Chevron notch technique, based on compact specimens

12

Structural Analysis of Advanced Materials

and round notched tensile specimens [3]. These methods require considerable time for sample
preparation and for notch geometry control. Moreover, only a few of them could be applied for
ceramics, because of their very high brittleness. Although numerous testing techniques are available
[1, 2], there is no standard specimen type for determining fracture toughness of engineering
ceramics. The choice of a specific technique is determined by the type of information that is needed.
Specimen geometry, preparation and manufacture history are critical to correlate the test specimen
behavior and the actual component fracture toughness.
A method that is frequently used for determining fracture toughness is the so-called indentation
fracture (IF) method [4]. The importance of that method comes from its simplicity and the small
material volume that is requested for conducting the KIc measurements. A Vickers microindentation is implanted onto a flat ceramic surface, and consequently some cracks develop around
the indentation, their lengths being inverse proportional to the material toughness. By measuring the
crack lengths, it is possible to estimate KIc for the studied material. One can say that the indentation
technique has long been considered an attractive method for assessing the toughness of ceramic
materials, due to the ease and low cost of its application. The results of indentation toughness
measurements critically depend on the assumption of crack type (the Palmqvist type or the
median/radial one) (see Fig. 1), on the equations that are used for fracture toughness calculation,
and on some material-dependent and material-independent constants [3].

Fig. 1. The crack formation by Vickers indentation


The measured crack lengths are then connected with the material toughness KIc using a semiempirical calculus relation [5]:
K Ic =

E P

H c3 2

(2)

where P is the applied load, E the test material Youngs modulus, H - Vickers hardness number, c
- the radial crack length, measured from the indentation center, and - an empirically determined
calibration constant that is usually taken to be 0.016 0.004.
The Palmqvist crack model equation could also be used for determining fracture toughness, but
only when shallow cracks are observed. According to Niihara [6] the following equation is valid, if
c/a<3.5:
E
K Ic = 0.018 H a1/ 2
H

0.4

a
1
c

1 2

(3)

It must be emphasized that the major difficulties for obtaining some reliable and consistent fracture
toughness values are the cracks type determination, together with the precise measurement of crack
lengths [7].

Key Engineering Materials Vol. 446

13

The Boussinesq stress field


After the Vickers indentation, a small rhombic trace is obtained on the ceramic specimen surface.
That situation has as a suitable theoretical description the Boussinesq solution regarding the
deformation status of the half-space (Fig. 2a): a solid with large dimensions, but having a
limitation-plane, and carrying a concentrated load that is oriented perpendicularly to the respective
plane.

a. coordinate system
b. the stress state for the point M
Fig. 2. The coordinate system for the Boussinesq solution, as applied for a Vickers indentation
For the point M of (rOz) plane, one can consider the following stress-state components (Fig. 2b):
- the normal stress t, perpendicular to the (rOz) plane;
- the normal stress r, perpendicular to the (tOz) plane;
- the normal stress z, perpendicular to the (rOt) plane;

- the shear stresses, having equal values, rz and zr.


It is important to observe the absence of shear stresses for the (rOz) plane, indicating that plane as a
principal one, for the stress state that is studied. The problem is to determine all the stress values,
for any point M(r,z) of the solid, expressed as a function of the point coordinates, r and z.
Two possible relations for calculating the u and w displacements of an arbitrary chosen point
M(r,z) were indicated by Boussinesq, as follows:

u=A

rz

w = A[

z2

1
+ (3 4 ) ]

(4)

or
u=B

r
1
w=B
( + z)

(5)

where = r 2 + z 2 is the distance of M to the origin of coordinate system (Fig. 2a), and is the
Poissons coefficient for the studied material. For determining the values of A and B constants one
can firstly consider the static equivalence of the external load P to the resulting effect of axial
stresses z, acting on the plane that is defined by z=const. The second condition is based on the fact
that the surface of the above-cited half-space is not charged with any tangential forces, so the shear
stresses are zero at z=0. Using the resulting A and B values, one can find the final calculus relations
for the stress components in M(r,z) (Fig. 2b), as follows:

14

Structural Analysis of Advanced Materials

- radial stress:

r =

3zr 3
P 1 2
[
5 ]
2 ( + z )

(6)

- circumferential stress:

t =

P
z
1
(1 2 )[ 3
]
2

( + z )

(7)

- axial stress:

z =

- shear stress:

P 3rz 2
=
2 5

P 3z 3
2

(8)

(9)

One can observe that at =0 (the origin of coordinate system and the point of load application) all
the stress components tend to infinite values (Fig. 3b).

a. the crack growth under the action of load P;


b. the stress components on the surface
Fig. 3. Load application and stress components for the Boussinesq solution
The stress component that is given by the above-presented relation (Eq. 7) is recommended to be
used for calculating the fracture toughness value for the studied material, having in view that the
crack opening (corresponding to Mode I) is determined by the circumferential stress component. As
a result one can obtain the following calculus relation:

K Ic =

(1 2 )P a
2(a ) 2

a2 + c2
ln
c2
2

(10)

FEM analysis
It is well known that, for ceramic materials, it is quite difficult to establish the geometry of cracks
that are determined by a Vickers indentation. Mainly for low levels of indentation load (50 daN)
that were used for the above-presented experiments, the cracks are not easy to be observed. On the
other hand, as a consequence of the particular manufacturing process that was used for the studied
ceramic specimens, one could obtain different values for the diagonal dimensions of cracks, but also
different crack geometrical aspects, when regarding in its depth direction. As a result, it is difficult
to establish the specific type (Palmqvist or median-radial, see Fig. 1) of induced cracks (by Vickers

Key Engineering Materials Vol. 446

15

micro-indentation) cracks, just on the basis of a simple superficial or even sectional visual analysis
of a ceramic specimen. Those practical aspects could lead to the idea of using a numerical method,
as FEM, for example, in order to obtain an image of the stress distribution into the region of a
Vickers indentation trace. Such an identification of the maximum stress values could facilitate a
more precise analysis of the diagonal cracking process.
The following-presented FEM analysis was step-by-step made, considering indentation loads of
50N, 100N, 200N, and respectively 300N. The experimental data on the mean values of the halfdiagonal (a) of indentation trace were used. Seven measurements were conducted for each sample,
neglecting the minimum and the maximum values, and the other five results were averaged to
obtain a mean value, as presented in Table 1:
Table 1: Dimensions of indentation traces, for different values of indentation load
P
a
c
h
q
[N]
[mm]
[mm]
[mm]
[N/mm2]
50
0.123
0.210
0.0008
1.652
100
0.171
0.292
0.0011
1.710
200
0.258
0.440
0.0016
1.502
300
0.277
0.473
0.0017
1.955
The Table 1 is also presenting the two other dimensions of indentation traces (see Fig. 4a), that
could be calculated as follows (on the basis of the effective value 136 of pyramid peak angle):
b=

a
2

, h = b tan(22 0 )

(11)

The last column of Table 1 is referring to the assumed uniform distributed pressure (q), which is
applied on the faces of pyramidal indentation trace (Fig. 4a). Using the above-presented expression
(11), one can obtain the pressure value as follows:
q=

P 
2a 2 mm 2

(12)

The values from Table 1 were used as parameters into the FEM analysis (on the bases of Von Mises
criterion), in order to obtain an image of stress distribution in the vicinity of indentation trace. As an
example, the map from Fig. 4a is corresponding to 300N indentation load.

V o n M i s e s s tr e s s [M P a ]

3500
3300
3100
2900
2700
2500
50

100

150

200
Force [N]

a)
b)
Fig. 4. a) Stress distribution map and b) the maximum stress variation,
as a result of Vickers indentation.

250

300

16

Structural Analysis of Advanced Materials

On the basis of the symmetry that is characterizing both the loading and the trace geometry, the
analysis was made just for one quarter of pyramidal indentation trace, imposing as boundary
conditions two plane of symmetry ((XOZ) and (YOZ)).
One can observe the symmetry of above-illustrated stress distribution, when reporting to the
diagonal of indentation trace, and that fact could be assumed as a confirmation of accuracy for the
conducted FEM model and analysis. The maximum stress value, for each step of that FEM
analysis, was located at the peak of Vickers indentation trace (its variation with the magnitude of
indentation force is presented in Fig. 4b), indicating that region as the starting-point of induced
crack (by Vickers indentation) crack. As a consequence, one can assume the median-radial type of
that crack. One can observe a minimum value of above-discussed maximum stress, corresponding
to 200N indentation load. The low value of (q) pressure (see again Table 1) which was obtained for
the respective indentation case could explain that situation.
On the other hand, the experimental observations lead to the conclusion that the crack is
developing following the diagonal direction of indentation trace. According to the brittle character
of ceramics, one can assume the crack developing is caused by some stresses that are oriented
perpendicular to that diagonal direction (the radial stresses (t) from Fig. 4a).
Fig. 5 is presenting the variation of these stress components, into a diagonal plane that includes
the peak of pyramidal indentation trace.

a) map of the von Mises stresses


b) spatial graph of the t stress distribution
Fig. 5. Radial stress (t) distribution, in the vicinity of Vickers indentation trace
One can observe, once again, the fact that the radial stress component (t) reaches its maximum
value at the peak of pyramidal indentation trace. Also, important values of (t) are obtained on the
diagonal line. Finally, one can assume that, for ceramic materials, the crack which is determined by
a Vickers indentation is usually starting at the peak of pyramidal indentation trace, and is
developing on the diagonal direction of that trace. Moreover, the specific type of crack is medianradial, and that character must be used as a basis for choosing the calculus relation, in order to
establish the fracture toughness of the studied material.

Experimental procedure
The analysis that is presented here is focused on a ceramic material, based on alumina and having
some mechanical characteristics as listed below, in Table 2. The Vickers microhardness tests were
made on the upper surfaces of some sintering manufactured specimens (Fig. 6), with different
shapes and dimensions. The Vickers hardness (Hv) was measured on polished surfaces, using an
EMCOTEST MC10 hardness tester with varying indenting loads of 50, 100, 200 and 300 N.
The fracture toughness (KIc) values at 300 N load were calculated from the measurement of radial

Key Engineering Materials Vol. 446

17

cracks around the Vickers indentations, according to the relations above-numbered with (2), (3) and
(10). Seven different measurements were made, on a radial direction (see Fig. 6a and Fig. 6b), for
every ceramic specimen.
Table 2. Mechanical properties and composition of the studied material
Material
Alumina
ceramic
99%
Alumina

Density

Hardness
HV30

Flexural
strength
(3 point)

[g/cm3]
3.67

[daN/mm2]
800-1000

[MPa]
375

Tensile
strength

Fracture
toughness
KIc

Youngs
modulus

Poisson
ratio

[MPa]
205

[MPa m ]
3-6

[GPa]
30

0.26

a. the specimen configuration;


b. the traces of Vickers indentations
Fig. 6: The ceramic specimen and the Vickers microhardness indentations
The Youngs modulus value for the studied material was experimentally determined using two
methods. Firstly, the technique indicated by the ISO Standard 14577:1 (Instrumented indentation
test for hardness and materials parameters) was applied (see Fig. 7a). Secondly, a standard tension
test was used, with strain gauge rosettes instrumentation on some prismatic specimens (10x20x200
mm3). For that method, the modulus value was established, for each specimen, as the slope of the
stress-strain dependence curve.

a)
b)
Fig. 7. a) Determination of Youngs modulus and b) the image of a Vickers indentation trace

18

Structural Analysis of Advanced Materials

On the other hand, a technique based on penetrating liquids was used for measuring the length of
cracks that were propagated from the corners of indentation traces (Fig. 6b). For that situation, the
electronic image that could be obtained with an EMCOTEST apparatus is not sufficiently clear. As
a consequence, a direct microscopic observation of the specimen surface was preferred.

Results and discussion


The measure of the diagonals of indentation traces (see Fig. 7b) was calculated, for every load level,
as an average value from at least seven pyramidal traces. For IFT calculation, it is recommended to
take into account the true (load-independent) hardness of the studied material [3, 4]. Actually, one
can observe (Fig. 8b) an increase of experimentally determined hardness number, as a consequence
of load increasing, from 100N to 300N (HV10, HV20, and respectively HV30). Moreover,
corresponding to those load levels, a similar variation of hardness number with the position r of the
indentation point (see Fig. 6a) was found.
It is clearly shown, in Fig. 8a, the fact that hardness numbers are bigger, for points 1 and 7, than
for the rest of indentation points, while the minimum values are corresponding to point 5.

2c

a. hardness variation with radius r;


b. hardness variation with load increasing
Fig. 8 The dependence of Vickers hardness number on the indentation point and load level
On the other hand, a different variation was obtained for an indentation load of 50N (HV5 in Fig.
8a), with lower values of hardness number for the marginal points (1 and 7), when comparing with
the internal ones.
One can admit that the hardness number variation with the load level could be explained by the
increase of the amount of ceramic material that must be dislocated, when the indentation load is
increasing. Moreover, it seems that even the opposition of ceramic material to the indentation is
increasing, on the measure of the enlargement of indentation depth.
The fracture toughness of the studied ceramics was established using the experimental data
corresponding to the load level of 300N, at which a crack growth, from the corners of indentation
traces (Fig. 6b), was observed. The cracks were emphasized using some penetrating liquids, and
then measured using an electronic microscope.
From Fig. 9, one can observe a variation of crack length with the position of indentation point,
and also significant differences between the crack length values, when measuring in a radial (Cr),
and respectively in a circumferential direction (Ct), on the ceramic specimen surface. Those
differences could be an expression of a certain degree of anisotropy, regarding the crack resistance
of the studied ceramic specimen.

Key Engineering Materials Vol. 446

19

Fig. 9. Crack length variation with radius r


The graphs from Fig. 10 are illustrating the variation of material fracture toughness, as a function
of the position of the indentation point. The calculus was made using the crack lengths values that
were measured on the radial (KIc-Cr), and respectively on the circumferential direction (KIc-Ct), but
also using their average value (KIc-Cmed), for every indentation point. One can observe that the
graph which was obtained from the average values can be approximated by a horizontal line, when
using just the internal indentation points (4 to 6). As a consequence, one can consider that the value
KIc=4.1 MPa m could be assumed as illustrative for the studied ceramic material.

Fig. 10. The dependence of fracture toughness on the position of the indentation point
On the other hand, the graphs from Fig. 11 were obtained using the values of material fracture
toughness, calculated (for each of the indentation points) with the above-presented relations that
were numbered with (2), (3) and respectively (10). One can observe a similar variation for the
graphs obtained using the relations (2) and (3), together with a slightly difference from point to
point, for both cases. Moreover, the respective fracture toughness values are close to those that are
usually presented in literature (see again Table 2). On the contrary, the values that were calculated
from the relation (10) are significantly different from the other ones, with an important variation
from an indentation point to another. When comparing to the values that are indicated by the
literature, one can assume that the results which were obtained from the relation (2) are the closest
to them. As a consequence, that relation seems to be suitable for the calculus of fracture toughness
values, into the case of the studied ceramic material.

20

Structural Analysis of Advanced Materials

Fig. 11. Comparison between fracture toughness values based on different calculus relations
Some specimens of the shape from Fig. 6a, with different values of internal radius (Ri = 12mm,
21.5mm, and respectively 41.5mm) were used, in order to determine (using the above-described
experimental procedure) the fracture toughness (KIc) of the studied material. For any (r) radius
value, seven different indentations were made and, from the corresponding values of (KIc), the
maximum and the minimum ones have been neglected. As a consequence, an average value was
calculated using the remaining five results.
The variation of those average values, as a function of (r) radius (position of the indentation
point) is presented in Fig. 12, for the three specimen dimensional categories.
Firstly, one can observe an unexpected increase of fracture toughness values, when increasing
the specimen dimension. That fact could be explained by the possible existence of some residual
stresses that were bigger for the small specimen than for the large ones. Those stresses could lead to
some longer cracks to develop, and consequently to some lower values of (KIc) for the small
specimens. On the other hand, one can say that, also for the specimens of small dimension, a less
important variation from point to point was obtained, when comparing to the results that correspond
to the bigger specimens.

Fig. 12. Comparison between fracture toughness values, for specimen with different internal radius

Key Engineering Materials Vol. 446

21

Conclusions
An experimental study was made, in order to establish the fracture toughness value of a certain
ceramic material. It was observed that the use of a loading force of 300N was the suitable condition
for obtaining reliable experimental data, because of a good visibility of crack growth, at the corners
of indentation traces. The Vickers hardness numbers were determined by directly measuring the
diagonals of traces, on the electronic images, resulted from the acquisition made by the
EMCOTEST MC10 hardness tester. On the other hand, those images could not be used for crack
growth identifying. As a consequence, the cracks were emphasized using penetrating liquids, and
then measured using an electronic microscope.
A step FEM analysis was used, in order to establish the stress distribution for many different
magnitudes of indentation load, into the region of a Vickers indentation trace. On that basis, it was
concluded that, for ceramic materials, the crack is usually developing in a median-radial way (see
Fig. 1). An increase of Vickers hardness numbers was observed, when increasing indentation load,
and it could be explained by a different capability of various superficial material layers, regarding
the resistance to indentation.
Some significant differences were observed between the crack lengths that were measured in a
radial, and respective in a circumferential direction, on the ceramic specimen surface. As a result,
similar differences were obtained for the corresponding calculated fracture toughness values.
Nevertheless, the graph based on the average experimental data was approximately horizontal for
the internal indentation points, and one could assume the respective value of fracture toughness as
illustrative for the studied ceramic material.
Finally, one can assume that the calculus relation, based on the Bousinesq solution, is not
suitable for that kind of material, and probably also for other types of ceramics.

Acknowledgements
The present work was supported by the National Council of Scientific Research CNCSIS
Romania under the Grant No ID_597/2007, for the study of technical materials that are used in the
high technology.

References
[1] F. Sergejev, M. Antonov, Comparative study on indentation fracture toughness measurements
of cemented carbides, Proc. Estonian Acad. Sci. Eng., 12, 4 (2006), p. 388-398.
[2] N. Nayeb-Hashemi, Microhardness Indentation Application and Limitation in Fracture
Toughness Evaluation of Ceramics, In Mechanical Engineering, no. 12, (1993), p. 79-84.
[3] K. Strecker, S. Ribeiro, M-J Hoffman, Fracture toughness measurements of LPS-SiC: a
comparison of the indentation technique and the SEVB method, Mat. Research, 8, 2 (2005).
[4] J. Kruzic, R. Ritchie, Determining the Toughness of Ceramics from Vickers Indentations Using
the Crack-Opening Displacements: An Experimental Study, J. Am. Ceram. Soc., 86, 8 (2003) p.
143336.
[5] G.R. Antis, P. Chantikul, B.R. Lawn, D.B. Marshall, J. Am. Ceram. Soc., 64(9), (1981), p. 533538.

[6] K. Niihara, A fracture mechanics analysis of indentation-induced Palmqvist crack in ceramics,


J. Mater. Sci. Lett., 2 (1983), p. 221223.
[7] K. Hagihara, S. Maeda, T. Nakano, Y. Umakoshi, Indentation fracture behavior of
(Mo0.85b0.15)Si2 crystals with C40 single-phase and MoSi2(C11b)/bSi2(C40) duplex-phase
with oriented lamellae, Sci. & Techn. of Adv. Mat., 5 (2004), p. 1117.

Key Engineering Materials Vol. 446 (2010) pp 23-31


(2010) Trans Tech Publications, Switzerland
doi:10.4028/www.scientific.net/KEM.446.23

Manufacturing of carbonaceous materials based on olive stones


biomass for electrochemical applications
Anastasia Pikasi, Pantelitsa Georgiou and Johannis Simitzis*
National Technical University of Athens, School of Chemical Engineering, Department III,
Materials Science and Engineering, Laboratory Unit Advanced and Composite Materials", 9
Heroon Polytechniou str., Zografou Campus, 157 73 Athens, Greece
*

Corresponding author: simj@chemeng.ntua.gr, Fax. +30 210 7723252, Tel. +30 210 7723178

Keywords: Carbonaceous materials, carbon fibres, Pt electrodeposition, cyclic voltammetry


Abstract. Carbonaceous materials have been obtained by pyrolysis of composites based on olive
stones biomass, novolac resin as binding agent, with or without an aromatic compound
(naphthalene). The pyrolysis residue at 1000 C is 40 w% and its electrical conductivity, , is 0.13
S/cm. Small cylindrical specimens have been manufactured and pyrolyzed at 1000 C in order to be
used as electrodes. Platinum was electrodeposited by cyclic voltammetry on these specimens using
them as working electrodes or on commercial carbon fibres, respectively, for correlating purposes.
The morphology of both carbon materials, used as electrodes, was characterized by SEM images
and the presence of Pt was determined based on EDS analysis. The crystallographic planes of Pt
carbon of Pt deposited on carbon materials were characterized with XRD. The oxidation of ethanol
from a proper solution using the carbonaceous specimen and the carbon fibres as working electrodes
was examined by cyclic voltammetry.
Introduction
Carbon represents a very attractive material because of its different allotropic forms, the various
microtextures due to the degree of graphitization, the ability to exist under different macroscopic
forms (powders, foams, fibres, fabrics and composites), etc. [1]. Carbon having different
combinations of properties can be prepared by using proper precursor materials under appropriate
thermal treatment. The variation in properties is a result, primarily, of the variation in carbon
crystallite size, orientation and extent of cross linking between trigionally bonded regions [2].
The conversion of an organic material into carbon is a complex process involving a wide variety
of reactions and a great number of intermediates. Model aromatic compounds, such as naphthalene
have been investigated in order to gain useful data for the reaction mechanisms of the carbonization
process [3]. An important requirement for the products in nowadays is the use of cheap raw
materials. Such raw materials are certainly some agricultural byproducts, from which other useful
final products after appropriate treatment can be produced. As biomass precursors have been used
peanut shells, starch and oak, lignin, cotton, wool, rice husk and sugar etc. [4]. Furthermore,
carbonaceous adsorptive media have been produced by combination of an agricultural byproduct (as
olive stones) or typical raw material for production of activated carbon (as lignite) with polymer
resins (as phenol formaldehyde resin) [5].
Carbonaceous materials have attracted great interest and are used as negative electrodes for
consumer lithium ion batteries because they are more advantageous than lithium metal negative
electrodes in terms of cycle performance and safety [6]. Carbon materials can be utilized as
electrode support and as catalysts support as well. These materials are widely used as electrodes in
electrochemical reactors for different electrochemical applications [7]. The use of carbonaceous
materials as catalyst supports for precious metals rapidly increased in the last years, due to the
continuous advancing development of fuel cells [8]. Low operation temperature is one of the most

24

Structural Analysis of Advanced Materials

important features of the fuel cells. In direct alcohol fuel cells (DAFCs), ethanol is the most
promising one among other alcohols, due to considerable energy capacity (~8 kWh/kg), nontoxicity,
biological renewability, easy handling and storing unlike hydrogen fuel [9].
Platinum is one the best anodic materials that exhibit catalytic properties for the oxidation
process of alcohols to proceed at a sufficient rate in fuel cells. One of the ways to reduce the cost of
fuel cells anodes is the deposition of platinum on the less expensive material such as carbon
materials [10]. Electrochemical deposition is a versatile technique by which a thin desired metallic
coating can be obtained on to the surface of another material by simple electrolysis of an aqueous
solution containing the desired metal ion or its complex. Moreover, electrodeposition is the only
technique by which metals with high melting points (e.g. platinum) can be deposited.
Electrodeposits have fine structure and have valuable physical properties. A great advantage of
electrochemical deposition is that the thickness of the layer can be controlled. Electrochemical
deposition of platinum is practiced from chloride, ammine, sulfate nitrite and hydroxy complexes.
The chloride based electrolytes contain platinum complexed in the +4 state (usually H2PtCl6) [11].
For synthesizing novel carbon material as a viable product, it is necessary to select a precursor,
which is not derived from fossil fuels or related to petroleum products [12]. Hence, olive stones
biomass is a valuable source, a natural product from which carbonaceous materials can be
synthesized by pyrolysis.
The aim of this work is the manufacturing of carbonaceous materials based on olive stones
biomass in combination with novolac resin as binding agent, with or without an aromatic compound
(naphthalene) and the investigation of these products for electrochemical applications, after
electrodeposition of platinum on them.
Experimental
Composites were fabricated by mixing olive stones biomass (as grains with diameter less than 300
m) in different proportions (80% and 40% w/w) without or with the naphthalene, as an aromatic
precursor (40% w/w) with AlCl3 as catalyst (1/1 mol/mol or 0.95/1 w/w) and novolac (20% w/w)
with hexamethylenetetramine (hexa) (7/2 w/w) as curing agent and the proper mixture was placed
into small cylindrical moulds. The curing was performed by heating at 170 C for 30 min and the
specimens were pyrolyzed under N2 atmosphere with a heating rate of 10 C/min up to 1000 C.
Small cylindrical specimens had been manufactured and pyrolyzed at 1000 C in order to be used as
electrodes. The latter had the following characteristics: average diameter: 6 mm, average length
immersed in the electrolyte solution : 9 mm, corresponding immersed surface: 198 mm2 = 1.98 cm2.
The electrical conductivity, , of the carbonaceous materials was also measured.
Platinum was electodeposited by cyclic voltammetry on these specimens using them as working
electrodes or on commercial carbon fibres, respectively, for correlating purposes. The carbon fibres
used have the following characteristics : filament diameter: 7 m, number of filaments: 24000, cross
section of the total fibre: 0.90 mm2, average length immersed in the electrolyte solution: 15 mm,
corresponding immersed surface (assuming that all monofilaments are separate, without common
surface) : 7913 mm2 = 79 cm2. The electrochemical measurements were carried out in a
conventional three-compartment glass cell at room temperature. The counter electrode was a
platinum sheet. The saturated calomel electrode (SCE) (+0,241 V vs. NHE) is used as a reference
electrode, connected with the cell through a Luggin capillary system. All the potentials mentioned in
this work are referred to this electrode. The electrodeposition of Pt on this carbonaceous electrode
or on the carbon fibres was performed by cyclic voltammetry with the following series of solutions:
a) 0.1 M HClO4, b) 0.002 M H2PtCl6 + 0.1 M HClO4, c) 0.1 M HClO4. The experiments in the
solutions a) and c) were performed in order to study the performance of the electrode prior and after
electrodeposition of Pt. All experiments were recorded in the potential range from +1.2 V to -0.4 V
and back to +1.2 V, using a scan rate of 50 mV/s, for 20 cycles. The resulting material is
characterized as Pt-carbonaceous electrode. Then, the oxidation of ethanol from an aqueous solution

Key Engineering Materials Vol. 446

25

of 0.15 M ethanol mixed with 0.2 M H2SO4 was performed by cyclic voltammetry, in the potential
range from -0.22 V to 0.96 V and back to 0.22 V for the Pt-carbonaceous electrode and from +0 V
to +1.2 V and back to +0 V for the carbon fibre electrode, using a scan rate of 50 mV/s, for one or
more cycles.
The morphology of the electrodes was characterized by SEM images and the presence of Pt was
determined based on Energy Dispersive Spestroscopy (EDS), using a FEI Quanta 200 scanning
electron microscope. Additionally, the crystallographic planes of Ptcarbon of Pt deposited on the
carbonaceous materials and on the carbon fibres were characterized with X-ray diffraction (XRD),
using a Bruker D8 Focus X-Ray Diffractometer.
Results and Discussion
Table 1 presents the raw materials and the pyrolysis parameters used for the production of
carbonaceous materials, and their electrical conductivity.
Table 1. Raw materials and pyrolysis parameters for the production of carbonaceous materials and
Code of
materials

Raw materials (w/w)


'ovolac
'

'aphthalene
A

Pyrolysis

Electrical
conductivity

Olive
Residue
Maximum
stones
temperature
(w/w)
biomass
B
(S/cm)
(%)
(%)
(%)
(C )
(%)
20
80
800
33.6
1.10-3
'B-800
20
40
40
800
43.7
3.62 x10-2
'AB-800
20
40
40
1000
40
0.13
'AB-1000
20
40
40
1000
39.3
0.10
'AB-1000-E *
* E: indicates a formed electrode as small cylinder with the ratio of length to diameter equal to 1.2
Pt-carbonaceous electrode is the material NAB-1000-E after electrodeposition of Pt by cyclic
voltammetry

their electrical conductivity


The electrical conductivity of the materials is increased by increasing the pyrolysis temperature.
The weight losses during the pyrolysis are high up to 800 C and then the residue yield remained
almost constant. On the other hand, the pyrolyzed materials at 1000 C have the highest electrical
conductivity.
Fig. 1a presents the cyclic voltammograms for the electrodeposition of Pt on carbon fibres used
as the working electrode. The peaks observed can be interpreted based on the literature [13]. During
the first scan from +1.2 V to -0.4 V there is a broad cathodic peak from 0.1 V up to -0.05 V, which
is characteristic for Pt(IV) reduction. The peaks at more negative potentials (-0.3 V) are due to the
reduction of H+ and adsorption of hydrogen atoms on the Pt clusters formed. During the reverse
scan, the anodic peak at -0.14 V is attributed to the oxidation of adsorbed hydrogen atoms
(desorption) from the platinum surface [10]. The latter peak is increased with increasing number of
cycles. The peaks at around +0.9 V for the formation of platinum oxides and at +0.4 V for the
reduction of platinum oxides which are mentioned in the literature for Pt-carbon ceramic electrodes
[10] are not clearly defined in the Fig. 1a.

26

Structural Analysis of Advanced Materials

20

(a)

(b)

-0,6 -0,4 -0,2

10

0
0,0

0,2

0,4

0,6

0,8

1,0

1,2

1,4

-1

I (mA)

I (mA)

-0,6 -0,4 -0,2

0
0,0

0,2

0,4

0,6

0,8

1,0

1,2

1,4

-10

-2
-20

-3

E vs. SCE (V)


st

nd

1 cycle
th

20

E vs. SCE (V)

3rd cycle

cycle

4th cycle

cycle

Figure 1a. Cyclic voltammograms for the


electrodeposition of Pt on carbon fibres used as the
working electrode, with the solution of 0.002 M H2PtCl6
+ 0.1 M HClO4, after previous treatment with 0.1 M
HClO4

st

1 cycle

th

cycle

20th cycle

Figure 1b. Cyclic voltammograms for the


electrodeposition of Pt on carbonaceous materials, used
as the working electrode, with the solution of 0.002 M
H2PtCl6 + 0.1 M HClO4, after previous treatment with
0.1 M HClO4
2

1,0

(a)

(b)

0,8
1

0,6

I (mA)

I (mA)

0,4
0,2
0,0
-0,6 -0,4 -0,2 0,0
-0,2

0,2

0,4

0,6

0,8

1,0

1,2

1,4

-0,6 -0,4 -0,2

0
0,0

0,2

0,4

0,6

0,8

1,0

1,2

1,4

-1

-0,4
-0,6

-2

-0,8

E vs. SCE (V)

-1,0

E vs. SCE (V)


carbon fibres electrode
Pt-carbon fibres electrode
Figure 2a. Cyclic voltammograms of carbon fibres
electrode before (dashed line) and after (solid line) Pt
deposition, in 0.1 HClO4

carbonaceous electrode
Pt-carbonaceous electrode
Figure 2b. Cyclic voltammograms of carbonaceous
electrode before (dashed line) and after (solid line) Pt
deposition, in 0.1 HClO4

Key Engineering Materials Vol. 446

27

100
1,5

(b)

(a)

50

I (mA)

I (mA)

1,0
0,5

-0,4

-0,2

0
0,0

0,2

0,4

0,6

0,8

1,0

-50
0,0
-0,2
0,0

0,2

0,4

0,6

0,8

1,0

1,2

1,4

-100

-0,5

-150

-1,0

E vs. SCE (V)


Pt-carbon fibres electrode
carbon fibres electrode
Figure 3a. Cyclic voltammogram of ethanol oxidation
using the Pt-carbon fibres electrode (solid line) and
voltammogram of carbon fibres electrode (dashed line),
using aqueous solution of 0.15 M ethanol mixed with
0.2 M H2SO4

E vs. SCE (V)


Pt-carbonaceous electrode
carbonaceous electrode
Figure 3b. Cyclic voltammogram of ethanol oxidation
using the Pt-carbonaceous electrode (solid line) and
voltammogram of carbonaceous electrode (dashed line),
using aqueous solution of 0.15 M ethanol mixed with 0.2
M H2SO4

Fig. 1b presents the cyclic voltammograms for the electrodeposition of Pt on the carbonaceous
electrode. During the forward scan (+1.2 V to -0.4 V), there is an insignificant change in the slope
of the cathodic curve, which is not enough evident in order to be attributed exclusively to the Pt(IV)
reduction. Currents corresponding to the reduction of oxygen groups will superpose the current for
Pt deposition. The oxygen groups like OH, >C=O, -COOH are formed on the surface of the
carbonaceous material during the pyrolysis, mainly due to the content of olive stone biomass
(lignin-cellulosic origin) in the initial composites. Contrary to the carbon fibres electrode, the %
(w/w) of oxygen element present in the carbonaceous electrode is about 4.5 times higher (see EDX
analysis, Table 2). Furthermore, due to the high porosity of the carbonaceous material (see SEM
Figure 4), the electrocatalytic activity is not uniform and it is difficult to obtain distinct peaks or
limiting currents. The characteristic peaks for adsorption (-0.3 V) and desorption (-0.14 V) of
hydrogen atoms in the voltammograms of carbon fibres electrode (Fig.1a) are not observed in the
voltammograms of the carbonaceous electrode.
Fig. 2a presents the cyclic voltammograms of the carbon fibers electrode (before and after Pt
electrodeposition, respectively) in 0.1 HClO4. The first treatment (one cycle) activates the surface
of the electrode and then follows the Pt electrodeposition (Fig. 1a). After this Pt electrodeposition,
the voltammogram obtained by the next treatment with the same solution (Fig. 2a) does not indicate
distinct peaks which could be expected according to the literature for treatment with HClO4 [13] or
H2SO4 [14]. However, by intensive electrochemical oxidation of the carbon fibres electrodes prior
to Pt electrodeposition (i.e. under electrolyte decomposition forming 2 / 2 and 10 or 20 cycles of
electrochemical treatment), the corresponding voltammograms obtained after Pt electrodeposition
showed distinct anodic and cathodic peaks.
Fig. 2b presents the cyclic voltammograms of the carbonaceous electrode (before and after Pt
electrodeposition, respectively) in 0.1 HClO4. After this Pt electrodeposition, the voltammogram
obtained by the next treatment with the same solution (Fig. 2b) does not indicate distinct peaks. The
voltammograms obtained under the same conditions are reproducible.
Fig. 3a presents the cyclic voltammogram of ethanol oxidation using the Pt-carbon fibres
electrode and the carbon fibres electrode (i.e. without Pt deposition). The peaks observed can be
interpreted based on the literature [10]. Based on this literature, the oxidation of ethanol commences
at about 0.1 V, and two oxidation peaks, at 0.63 V and 1.05 V, are observed during positive

28

Structural Analysis of Advanced Materials

potential scanning. The first oxidation peak can be ascribed to CO2 production and the second
appears due to acetaldehyde formation. In the voltammogram of Fig. 3a for the Pt-carbon fibres
electrode, during the forward scan from +0 V to +1.2 V the ethanol oxidation begins at about 0.1 V.
One broad oxidation peak is observed from +0.58 V up to +0.90 V, which can be attributed to CO2
production and acetaldehyde formation peaks which are overlapped. During the reverse scan (from
+1.2 V to +0 V), there is an anodic peak at +0.3 V which corresponds to reoxidation of ethanol and
intermediate species (e.g. PtOCH2CH3, PtCHOHCH3, (Pt)2=COHCH3, PtCOCH3 and PtCO)
that are formed from incomplete oxidation of ethanol in forward scan [10,15]. The small peak at
+0.5 V may correspond to the formation of ethylacetate [16]. On the other hand, regarding the
carbon fibres electrode (without Pt electrodeposition), there is no anodic peak attributed to CO2
production from +0.58 V up to +0.90 V, or other changes, indicating that ethanol is not oxidized.
Fig. 3b presents the cyclic voltammograms of ethanol oxidation using the Pt-carbonaceous
electrode and the carbonaceous electrode (i.e. without Pt electrodeposition). During the forward
scan from -0.22 V to +0.9 V (regarding Pt-carbonaceous electrode) there is a broad anodic peak at
+0.58 V, which is attributed to CO oxidation (i.e. CO2 formation). The peak at around 0.9 V is
attributed to acetaldehyde formation. The small peak at +0.5 V at the reverse scan may correspond
to the formation of ethylacetate [15]. On the other hand, regarding the carbonaceous electrode
(without Pt electrodeposition), there is no anodic peak attributed to CO2 production from +0.58 V
up to +0.90 V, or other changes, indicating that ethanol is not oxidized.
The separate steps for the oxidation of ethanol on Pt have been extensively studied [9,10,17].
The ethanol is oxidized to (I) CO2, (II) acetaldehyde and (III) CH3COOH. Acetaldehyde can be
further oxidized to CO2 or CH3COOH. The major achievement in ethanol oxidation is the formation
of CO2. The first stage of the electrochemical oxidation of ethanol concerns its adsorption on Pt
bound on them by the O or C atom [10,17]. Many intermediate reactions take place with the
intermediate species adsorbed on the Pt. Additionally ethylacetate can also be formed by a followon chemical reaction of the electrochemically formed acetic acid with ethanol [16].
The SEM images of the Pt-carbonaceous electrode and Pt-carbon fibres electrode are presented
in Fig. 4 (a), (b) and (c), respectively. According to the SEM images, light-white spots can be
observed which were identified as Pt deposits by EDS analysis.
The EDS analysis of the Pt-carbonaceous electrode and the Pt-carbon fibres electrode are listed
in Table 2. The presence of carbon in the spectrum is derived from the novolac, the biomass and the
naphthalene, while the oxygen is derived from novolac and biomass. The presence of aluminum and
chloride are due to the catalyst used. Traces of potassium, calcium and silicon (0.74 % w/w) are due
to the biomass. Iron is also detected from the sample carrier during pyrolysis. The electrodeposition
of platinum is about 1 % w/w.

Key Engineering Materials Vol. 446

29

(a)

(b)

(c)
Figure 4. SEM images of the Pt-carbonaceous electrode (a) x50 and (b) x8000 magnifications and the Ptcarbon fibres electrode (c) x1500 magnification

Table 2. Results of spot EDS analysis of the Pt-carbonaceous electrode and the Pt-carbon fibres
electrode
ELEME'TS

(w/w)

C
(%)

'i

Al.

Cu

Pt

Cl

Ca

Fe

(%)

(%)

(%)

(%)

(%)

(%)

(%)

(%)

(%)

19.09

9.71

0.93

0.81

1.23

0.90

0.81

4.28

0.77

1.96

13.62

MATERIALS

Ptcarbonaceous 65.79
electrode
Pt-carbon
fibres
electrode

79.36

The X ray diffraction patterns of materials (a) Pt-carbonaceous electrode and (b) Pt-carbon fibres
electrode are presented in Fig. 5. Additionally, the XRD diffractogram of the material Pt-carbon
fibres electrode beginning the measurement from 2 = 35 is also given (c) in order to acquire better

30

Structural Analysis of Advanced Materials

resolution of the small diffraction peaks. The arrows show the corresponding axes for both
materials. The X ray diffraction peaks are characterized according to literature [10,18,19]. The
following diffraction peaks for the diffractogram of the Pt-carbonaceous electrode (a) are observed :
i) at 2 = approx. 25 corresponding to graphite as Gr(002) and ii) at 2 = approx. 40, 46 and 67
corresponding to platinum crystallographic planes (111), (200) and (220), respectively. The major
diffraction peak of Gr (002) is also observed for the diffractogram of the material Pt-carbon fibres
electrode (b). In the diffractogram (c), there is the diffraction peak of Gr(100) as a wide peak, which
might overlaps the Pt(200) and the diffraction peaks Gr(004) and Pt(311) as minor peaks. Therefore,
the presence of Pt on the carbon fibres electrode and on the carbonaceous electrode confirmed its
deposition on them during the corresponding cyclic voltammetry treatment.
Gr (002)

50

500

40

Counts

Pt(111)

Counts

Pt(200)

(a)

30

Gr
(100)

Gr (004)

400
300

Pt(220)
Pt (311)

200

20

(c)

(b)

100

10

10

20

30

40

50

60

70

80

90

100

110

120

2theta (degrees)
Figure 5. X ray diffraction patterns of Pt-carbonaceous electrode (a) and Pt-carbon fibres (b) and
(c), beginning the measurement from 2 equal to approx. 25, in order to omit the highest peak at
25
Conclusions
The carbonaceous materials originated by the mixture of olive stones biomass, naphthalene, novolac
and hexamethylenetetramine after pyrolysis at 1000 C showed the highest electrical conductivity,
compared to the other pyrolyzed composites. This material as well as commercial carbon fibres for
correlating purposes, were used as electrodes for the electrodeposition of Pt with cyclic
voltammetry. The Pt-carbonaceous electrode showed the Pt crystallographic planes (200) and (220),
according to XRD; light-white spots were observed (by SEM) which were identified as Pt deposits
(~1% w/w) on this electrode, also containing ~19% w/w oxygen element, determined by EDS
analysis. The Pt-carbon fibres electrode contained ~ 4% w/w oxygen element and ~13.6% w/w Pt.
Both electrodes efficiently decompose ethanol by cyclic voltammetry.
Acknowledgements
The authors would like to thank the Assistant Professor E. Chatzitheodoridis for some of the XRD
measurements.

Key Engineering Materials Vol. 446

31

References
[1] E. Frackowiak and F. Bguin: Carbon Vol. 39 (2001), p. 937.
[2] P.L.Jr. Walker: Carbon Vol. 10 (1972), p. 369.
[3] R.A. Greinke and I.C. Lewis: Carbon Vol. 22 (1984), p. 305.
[4] Y.J. Hwang, S.K. Jeong, J.S. Shin, K.S. Nahm and A.M. Stephan: Journal of Alloys and
Compounds Vol. 448 (2008), p.141.
[5] J. Simitzis and J. Sfyrakis: Journal of Applied Polymer Science Vol. 54 (1994), p. 2091.
[6] Y. Kida, K. Yanagida, A. Funahashi, T. Nohma and I. Yonezu: Journal of Power Sources Vol.
94 (2001), p. 74.
[7] M.M.E. Duarte, A.S. Pilla, J.M. Sieben and C.E. Mayer: Electrochemistry Communications Vol.
8 (2006), p.159.
[8] E. Antolini: Applied Catalysis B: Environmental Vol. 88 (2009), p.1.
[9] F. Vigier, C. Coutanceau, F. Hahn, E.M. Belgsir and C. Lamy: Journal of Electroanalytical
Chemistry Vol. 563 (2004), p. 81.
[10] H. Razmi, Es. Habibi and H. Heidari: Electrochimica Acta Vol. 53 (2008), p. 8178.
[11] C.R.K. Rao, D.C. Trivedi: Coordination Chemistry Reviews Vol. 249 (2005), p. 613.
[12] X. Zeng, D. Wu, R. Fu and H. Lai: Materials Chemistry and Physics Vol. 112 (2008), p. 1074.
[13] D. Stoychev, A. Papoutsis, A. Kelaidopoulou, G. Kokkinidis and A. Milchev: Materials
Chemistry and Physics Vol. 72 (2001), p. 360.
[14] F. Montilla, E. Morallon, J.L. Vasquez, J. Alcaniz-Monge, D. Cazorla-Amoros, A. LinaresSolano: Carbon Vol. 40 (2002), p. 2193.
[15] T. Iwasita and E. Pastor: Electrochimica Acta Vol. 39 (1994), p. 531.
[16] H. Wang, Z. Jusys and R.J. Behm: Journal of Power Sources Vol. 154 (2006), p. 351
[17] T. Iwasita, B. Rasch, E. Cattaneo and W. Vielstich: Electrochimica Acta Vol. 34 (1989), p.
1073
[18] T. Hyde: Platinum Metals Review Vol. 52, (2008), p. 129
[19] L. Jiang, G. Sun, S. Wang, G. Wang, Q. Xin, Z. Zhou, B. Zhou: Electrochemistry
Communications Vol. 7 (2005), p. 663

Key Engineering Materials Vol. 446 (2010) pp 33-41


(2010) Trans Tech Publications, Switzerland
doi:10.4028/www.scientific.net/KEM.446.33

Physicochemical and electro-rheological characterization of kaolinite /


CMS / silicone oil fluid
Thomas Reiss1,a, Saoussen Laribi2,b, Jean-Marie Fleureau1,c
and Jean-Franois Tassin3,d
1

Ecole Centrale Paris & CNRS UMR 8579, grande voie des vignes, 92295 Chtenay-Malabry,
France,
2
Ecole Nationale Suprieure des Mines de Paris, 60, Boulevard Saint-Michel 75272 Paris, France
3
Universit du Maine, Avenue Olivier Messiaen - 72085 Le Mans, France
a
thomas.reiss@ecp.fr,bsaoussen.laribi@ensmp.fr,cjean-marie.fleureau@ecp.fr,
d
jean-francois.tassin@univ-lemans.fr
Keywords: electrorheological fluid, kaolinite, intercalation, DMSO, CMS, synthesis

Abstract. The aim of this study is to elaborate electro-rheological fluids based on kaolinite. the
scientific characterization made it possible the identification of the composite nature and the checking of the
intercalation of the polymer among the clay particles The rheological behaviour of the fluid depends on

the electric field. A yield stress of the suspensions is observed, which increases with the applied
electric field.
An interpretation based on the different modes of association between the clay particles is proposed
to account qualitatively for the observed behaviour.
Introduction
Electro-rheological fluids (ERF) are materials that contain nanometric particles strongly polarized
by an electric field, dispersed in dielectric liquid.
These fluids have recently attracted extensive interest. The viscoelastic properties of an electrorheological fluid can be adjusted by modulating the applied electric field through an external
solicitation to make damping devices.
The aim of this study is to explore new ways to prepare electro-rheological fluids based on clay
nano-particles. For this purpose, a kaolinite clay and a polymer having interesting electrorheological properties and good affinity with clay have been chosen.
This study is divided into two parts. The first one deals with the preparation and characterization of
the clay and clay/polymer nanocomposite. The results of the characterization by X-ray diffraction,
thermal analysis and infrared spectroscopy have confirmed the intercalation of polymers in the clay
particles. The second part is a study of the influence of the electric field on the rheological
properties.
Experimental Materials
The research was carried out on a kaolinite clay material that comes from France. It is referred
under Kp300.
Clay minerals are hydrous silicates or aluminosilicates. Geotechnical engineers call these minerals
clays if their particle size is below 2 m. Clays belong to the family of phyllosilicates, which means
they present layered structures. Each layer is composed of more or less continuous two dimensional
tetrahedral sheets (T) composed of (SiO4)4- and octahedral sheets (O) made of (AlO6)3- which can
also contain either Al3+, Fe3+ or Mg2+ cations. In this study we are interested in kaolinite clays that
his structure consists of a two-sheets phyllosilicate where the T:O ratio = 1. Its ideal formula is
Si4 Al 4 O10 (OH )8 [1].

34

Structural Analysis of Advanced Materials

Methods.
Purification. The difficulty in the study of clay minerals is the heterogeneity of natural minerals.
For this reason, we have elaborated pure materials from natural clays using the sodium exchange
method [2] which permits to eliminate impurities.
To carry out the purification, 1 g of raw clay is mixed with 100 ml of sodium chloride solution
(1N). After approximately 12 hours agitation of the prepared mixture, the suspension is centrifuged
during 10 minutes at a speed of 3000 rpm. The agitation-centrifugation cycle is repeated several
times in order to ensure that the possible cation exchange by sodium is completed. Finally the
material is freeze-dried in order to eliminate all the water from its structure. The final product, a
kind of foam, will be called Kp after purification.
Intercalation of DMSO. Kaolinites are known to be composed of very rigid particles in which
the sheets are strongly linked with each other and do not allow polymer molecules to penetrate.
For this reason and in order to permit the intercalation of the polymer between kaolinite layers, an
increase in the thickness of the interlayer space by intercalating organic molecules was used. The
dimethylsulfoxide (DMSO) organic molecules was chosen because of its marked proton acceptor
[3,4], its small molecular size [5] and its high dipolar moment: 3.6 D.
The DMSO-kaolinite complex was prepared by mixing 10 g of kaolinite with 200 ml of DMSO
solution and stirring them at 80C. After 24 h, the resulting material (kaolinite-DMSO) was filtered
and dried in a vacuum oven at 80C for 3 h to eliminate the DMSO adsorbed on the surface of
kaolinite.
Preparation of carbo-methyl Starch (CMS). 10 g corn starch and 40 ml ethanol were placed in
a 100 ml vessel and stirred for 1 h. Then potassium hydroxide (1.5 g) was added at 30C and
reacted for 0.5 h. After that, 1.5 g potassium hydroxide and 5 g chloroacetic acid were added to the
vessel at 50C and the mixture was stirred for 3 h. The product was filtered off and washed three
times with ethanol, then dried under vacuum at 70C for 1 h.
Synthesis of nanocomposite Kaolinite/CMS. In a 500 ml vessel, 8 g kaolinite-DMSO and 150
ml of distilled water were mixed and stirred at 30C for 2 h. Then 8 g CMS was dissolved in 100 ml
distilled water and dripped into the vessel at 60C. After the addition was completed, the
temperature of vessel was increased to 80C and the mixture stirred for 8 h. the product was filtered
off and washed three times with ethanol, then dried under vacuum at 60C for 3 h and at 80C for 2
h. The obtained kaolinite-CMS nanocomposite was then ground 5 h in a ball-mill and dried under
vacuum at 80C for 2h.
Silicone oil 47V20 (20 cSt) was first dried at 100C for 2 h, and then the electro-rheological fluid
(ERF) was prepared by dispersing the nanocomposite particles in the silicone oil (volume percent
25%).
Experimental techniques. The physicochemical, textural and electro-rheological characterization
was determined by several methods:
- The specific surface area (ST) was measured using the ethylene glycol monoethyl ether method
[6].
- The cation exchange capacity CEC was determined using the methylene blue absorption technique
[7].
- Grain size distribution analysis was carried out using a Micrometrics Sedigraph 5100.
- Atterberg limits were measured with a Casagrande apparatus following AFNOR (1993) relating
to the Casagrande dish [8].
- The scanning electron microscopy (SEM) was carried out on a JEOL JSM-840A.
- X ray diffraction: The X-ray diffractograms were obtained using a Philips PW1710 diffractometer
with a Cu anticathode (CuK = 0.154040 nm 0.154438 nm) at 40 kV and 20 mA.
- Infrared spectroscopy: The Fourier-transform infrared (FTIR) spectra were recorded with a 1000
spectrophotometer Perkin-Elmer FTIR spectrum and were normalized to the spectrum of the KBr
pellet. The IR spectrum ranges from 350 to 4000 cm-1.

Key Engineering Materials Vol. 446

35

- Thermal and thermogravimetric analyses (DTA-TGA): The thermograms were performed on a


SETARAM TG-TDA92. The heating rate is 2C/min, the used gas is argon.
- Electro-rheology: Rheological properties were measured with a Physica MCR301 rheometer with
a Couette geometry. In such geometry, the tested material is subjected to an approximately uniform
shear rate provided that the gap between the cylinders is sufficiently small compared to the diameter
of the inner cylinder. On the other hand the gap has to be much larger than the larger heterogeneity
length scale in the material in order to provide average measurements. In our case, the inner
(mobile) cylinder was 25 mm diameter and the outer (fixed) cylinder 29 mm. The gap was thus of 2
mm, which satisfies the above-mentioned condition in the case of the pastes considered here. The
coaxial geometry is linked to a high tension generator (FUG HCL 14-12500) supplying tensions up
to 12.5 kV with intensity under 1 mA.
Results and discussion.
Physicochemical and textural characterization. Table 1 shows the physicochemical
characterization of the kaolinite. The grain size distribution shows the clay fraction (< 2 m) is
about 60% ; the relationship between the plasticity index (20%) and the clay fraction (60%)
indicates that the clay have a medium swelling potential if plotted using the classification of
Williams and Donaldson [9, 10].
Table 1 Physicochemical properties of the kaolinite clay
2

Specific surface area St (m /g)


Cation exchange capacity CEC (meq/100g)
Liquid Limit wL (%)
Plastic Limit wP (%)
Plasticity Index Ip (%)
< 2 m (%)
Swelling potential

Kp
12
5.3
40
20
20
60
medium

Fig. 1 shows the SEM micrograph of Kp and indicates morphology of the clay particles, the photo
reveals a stack of rigid platelets with lateral extensions of several hundreds of nanometer and the
thickness around of a the order few tens of nm [11]. These crystallites are formed by association of
several sheets of hexagonal and pseudo-hexagonal shaped plates characteristic of kaolinite [12] this
shape is the result of the nature of kaolinite layers and the strong dipole-dipole and H bond
interactions between the octahedral and the tetrahedral sheets of adjacent layers. This structure
explains the necessity of intercalation of DMSO in to the interlayer before the polymer.

36

Structural Analysis of Advanced Materials

Figure 1 SEM micrograph of purified kaolinite: hexagonal and pseudo-hexagonal shaped plates
characteristic of kaolinite.
Synthesis of the final composite. The X-ray diffractograms of Kp, Kp/DMSO complex
(Kp+DMSO) and of the composite Kp/CMS (Kp+CMS) are shown on figure 2; they demonstrate:
- The intercalation of DMSO between the kaolinite layers. In fact the reflection at d = 0.7 nm of Kp,
typical of the basal spacing (d001) of kaolinites, moves after intercalation of DMSO to 1.1 nm, this
value is typical of the basal spacing d001 of DMSO/ kaolinite and confirms the expansion of the
interlayer space of the kaolinite. These results are in agreement with those found by Frost et al. [13].
- The evaporation of DMSO from the final product leading to the Kp/CMS composite. This
conclusion is proved by the decrease of the interlayer space from 1.1 nm to 0.7 nm after the
substitution of DMSO by CMS, which confirms the disappearance of DMSO by evaporation [14].
- The combination between kaolinite particles and CMS rather than an intercalation of CMS in the
interlayer space of kaolinite. In fact the reflection at 0.7 nm shown on the Kp+ CMS is typical of
kaolinite basal spacing, which means that the basal spacing is not affected by the presence of CMS,
This interpretation is proposed by Wang [15] explaining the increasing stability of the composite by
substitution of DMSO by CMS. The same observations have been made by Benchabane [16] and Le
Pluart et al. [17] on other (bentonite / CMC) composites. Le Pluart explains this phenomenon by the
fact that CMC does not penetrate into clay layers but reacts on the clay particles surface.
The broading of the X-ray peak centred at 0.7 nm could correspond to a non parallel orientation of
the polymer molecule with respect to the host lattice [17].

Key Engineering Materials Vol. 446

37

Figure 2 X-ray diffractograms of pure kaolinite(Kp), kaolinite/DMSO complex (Kp+DMSO) and


kaolinite/CMS composite (Kp+CMS)

In order to confirm these hypotheses, an IR characterization was carried out.


The structure of kaolinite clay consists of a stacking of 1:1 layers. Each layer contains a pseudohexagonal silica sheet of corner-shared SiO4 units linked to a dioctahedral sheet of edge sharing
AlO2(OH)4 octahedra. The evolution of the OH stret ching and bending vibration modes
characterizing the hydroxyl groups present on the interlayer surfaces (between the dioctahedral and
tetrahedral sheets, pointing horizontally toward the center of the ditrigonal cavity), the inner surface
OH group (which are located at the top of the dioctahedral sheet, are hydrogen-bonded with the
basal plane of oxygen atoms of the next layer), gives information on the structural modifications
taking place in the interlayer space. The intercalation of a polymer in to clay will result a new H
bond network resulting from interaction of some inner surface OH groups with the polymer by
breaking some hydrogen bonds between Al and Si sheets.
Fig. 3 illustrates the infra-red spectra of pure kaolinite (Kp), kaolinite/DMSO complex
(Kp+DMSO) and kaolinite/CMS composite (Kp+CMS).
The spectrum of Kp+DMSO shows the deformation bands characteristic of DMSO at 1428, 1410,
1395 and 1310 cm-1 and also two vibration bands at 3500 et 3540 cm-1.
The band at 3695cm in kaolinite and DMSO remains observed at a close position (3695 cm) which
is in agreement with the attribution of this band to the stretching of the interlamellar hydroxyl
groups pointing perpendicularly to the c-direction [17]. After the intercalation of DMSO, the band
at 3653, 3669 and 3695 cm-1 are significantly modified which can be explained by that part of the
hydroxyl groups pointing perpendicularly to the tetrahedral sheet become free after separation to the
layer sheet by DMSO [18, 19]. These observations confirm the intercalation of DMSO on kaolinite
clay.
Bands at 2940 and 2880 cm-1 on the spectrum of Kp+CMS are characteristic of the C-H valence
vibration bands, which confirms the presence of polymer in the composite. Also this spectrum
shows the missing of both deformation and vibration bands of DMSO, which confirms the
evaporation of DMSO after the substitution of DMSO by CMS and consequently confirms the
previous conclusions from XRD.

38

Structural Analysis of Advanced Materials

1,2

1395 1310

1,0

Kp+DMSO
3695

0,8

3500

1428

Kp+CMS
0,6

1410

2880

3695

2940

3540

Kp

0,4

3652

0,2

3619

3695
3669
36523619

0,0
4000

3500

3000

2000

1500

1000

500

-1

wavenumber (cm )
Figure 3 IR spectra of pure kaolinite (Kp), kaolinite/DMSO complex (Kp+DMSO) and
kaolinite/CMS composite (Kp+CMS)
Fig. 4 and Fig. 5 illustrate the Kp and Kp+CMS thermograms. The Kp thermogram shows
endothermic peaks at 90C accompanied by a weight loss shown on the TGA curve, corresponding
to the removal of externally adsorbed water [20,21]. The second loss and associated with an
endothermic effect at 510C, corresponds to the desydroxylation process of kaolinite during the
transformation of kaolinite to metakaolinite [22].
Kp/CMS thermogram differs from that of pure kaolinite by another loss of weight around 200 and
300 C accompanied by an exothermic reaction. This weight loss is attributed to the partial thermal
removal of the organic material corresponding to a decarboxylation and degradation of the polymer.
These observations confirm the formation of the (Kp + CMS) composite.
0
st

TGA
DTA

1 loss of weight

-10

-10

510C

Exo

Exo

-40

90C

-20

290C

-30
d

S loss of weight

DTA

-50

-40

510C

TGA

1st loss of loss


of weight

90C

-50
rd

Th loss of weight

-60
100

200

300

400

500

TC

Figure 4 Thermogram of pure kaolinite

600

100

200

300

400

500

-60
600

T C

Figure 5 Thermogram of kaolinite/CMS composite

Loss of weight (%)

-30

Loss of weight (%)

-20

sd loss of weight

Key Engineering Materials Vol. 446

39

Influence of the electrical field on the rheological behaviour of the ER fluid. The results of the
study of the influence of the electric field on the rheological behaviour of Kp+CMS / silicone oil are
presented on Fig. 6 (Kp+CMS / silicone oil 10% of weight) and on Fig.7 (Kp+CMS / silicone oil
25% of weight).
Figure 6 shows a typical electro-rheological behaviour in fact the suspension exhibit a Bingham
fluid behaviour with a yield stress under an applied electric field [23, 24].
Figure 6 and Figure 7 reveal a linear increase in the yield stress with the electrical field, and
consequently in the viscosity because of the electro-rheological effect.
Results also show that in the absence of any electrical field, the fluid has a yield stress, which could
be attributed to two effects: (i) van der Walls forces which are responsible for the formation of
aggregates and consequently the resistance to flow [25], and (ii) the fact that the percentage of clay
in the composite is so high that the number of clay particles and the repulsion of double layers
increase, which may lead to an ordered and stable structure and therefore a continuous rigid
network and consequently to a yield stress.
Note that the structure in a concentrated suspension in general under an applied electric field is
sufficiently rigid for the material to withstand a certain level of stress without flowing the maximum
stress
that
can
be
sustained
without
flow
is
the
yield
stress.
The increase in the yield stress as a function of the electric field could be explained by the fact that,
when applying an electric field between the two cylinders, the kaolinite particles are aligned parallel
to the electric field, forming a more rigid structure that requires to apply a higher torque on the
inner cylinder to break the bounds between the particles and obtain the fluid flow [26].

Figure 6 Influence of the electric field on the rheological behaviour of Kp+CMS / silicone oil (10%
of weight)

40

Structural Analysis of Advanced Materials

1000

kaolinite/CMS

Yiel stress (Pa)

800

600

400

200

0
0,0

0,5

1,0

1,5

2,0

2,5

3,0

3,5

4,0

Electric field (kV/mm)

Figure 7 Variation of yield stress of Kp+CMS / silicone oil (25 % of weight) as a function of the
applied electric field
Conclusion
The aim of this study was to elaborate, to characterize and to test a clay based electro-rheological
fluid. The principle of such a material is to allow the orientation of the particles under an electrical
field in order to significantly modify the flow properties.
The obtained results are encouraging. In fact the physicochemical characterization allowed on one
hand to prove the synthesis of the composite and on the other hand to observe electro-rheological
effects similar to those of prior studies.
References
[1] G. Baudet: Lindustrie Cramique N 753 (1981), p. 627
[2] H. Van Olphen: Introduction to clay colloid chemistry (Wiley and Sons 2nd eds., New York
1977).
[3] C. Agami: Bull. Soc. Chim. de France N 4 (1965), p. 1021
[4] J. Courtot-Coupez and M. Le Demeyet: Bull. Soc. Chim. de France N 3 (1969), p. 1033
[5] R. Thomas, C.B. Shoemaker and E. Eriks: Acta cryst. N 21 (1966), p. 12
[6] D.L. Carter, M.D. Heilman and C.L. Gonzalez: Ethylene glycol monoethyl ether for determining
surface area of silicate minerals. Soil Sci 100 (1965), p. 356
[7] R.K. Taylor: Cation exchange in clays and mud rocks by methylene blue. Journal of Chemical
Technology and Biotechnology. Vol. 35 A (1985), p. 195-207
[8] AFNOR Sols. Reconnaissance et Essais Dtermination des limites dAtterberg Limite de
liquidit la coupelle Limite de plasticit au rouleau. Association Franaise de Normalisation
(1993)
[9] A.H. Weir and R. Greene-kelly: Beidellite. Am Miner Vol. 47 (1962), p. 137
[10] A.B. Williams and G.W. Donaldson: Building on expansive soils in South Africa: 1973-1980,
Proc of 4th int conf on expansive soils (1980)
[11] S. Laribi, B. Jouffrey and J.M. Fleureau: Eur. Phys. J. 39 (2007), p. 257
[12] D. Tessier: Etude exprimentale de lorganisation des matriaux argileux. Doctoral Thesis.
University of Paris 6, France (1984)

Key Engineering Materials Vol. 446

41

[13] R.L. Frost, E. Mako and J. Kristof: J. Colloid Interface Sci. No. 239 (2001), p. 458
[14] R.L. Frost: Journal of Physical Chemistry (1999)
[15] B.X. Wang: Journal of Solid State Chemistry Vol. 179, N 3 (2006), p. 949
[16] A. Benchabane: Doctoral Thesis, University of Strasbourg I, France (2006)
[17] L. Le Pluart, J. Duchet, H. Sautereau and P. Halley: Applied Clay Science Vol. 25, N 3-4
(2004), p. 207
[18] S. Letaief and C. Detellier: J. Mater. Chem., N 17 (2005), p. 1476
[19] J. Tunney and C. Detellier: Chem Mater., N 8 (1996), p. 927
[20] J. Tunney and C. Detellier: Chem Mater., N 5 (1993), p. 747
[21] J.B. Dixon and S.B. Weed: Soil Sci Soc Am. (1977)
[22] S. Caillre, S. Henin and M. Rautureau: Minralogie des argiles Vol. 1-2 (Masson (eds.), Paris
1982)
[23] J.W. Kim, H. Noh, J. Choi, D.C. Lee and S. Jhon: Polymer 41 (2000), p. 1229
[24] I.S. Lee, J.Y Lee, J.H. Sung, H.J Choi: Synthetic metals. 152 (2005), p. 173
[25] J.S. Reeds: John Wiley & sounds (eds.) (1998), p. 227
[26] C. Boissy, P. Atten and J.N. Foulc: J. Phys. III France. N 5 (1995), p. 677

Key Engineering Materials Vol. 446 (2010) pp 43-49


(2010) Trans Tech Publications, Switzerland
doi:10.4028/www.scientific.net/KEM.446.43

The effect of the cure temperature on the thermomechanical


characteristics of an adhesive
M. ALI1,a, B. LORRAIN1,b, M. KARAMA1,c and B. PUEL2,d
1

Laboratoire Gnie de Production Ecole Nationale dIngnieurs de Tarbes


Avenue dAzereix BP 1629 65016 Tarbes - FRANCE
2

ALSTOM Transport, 50, rue Docteur Guinier - 65600 Smac - FRANCE


a

mohamad.ali@utc.fr, bbernard.lorrain@enit.fr,cmoussa.karama@enit.fr,
d

bernard.puel@transport.alstom.com

Keywords: adhesive joint, thermomechanical behaviour, optimization of polymerization,


hygrothermal ageing, Differential Scanning Calorimetry (DSC), cohesive and adhesive failure.

Abstract. The origin of this work takes place in the framework of an industrial collaboration in
order to study the behavior of a mechanical joint, used in the industry of railway structures. One
problem in using the structural adhesive in the industry is hygrothermal ageing. In these papers; the
effect of the adhesive cure temperature on the mechanical properties of a commercial bi-component
epoxy bonding joint will be studied. Usually the bi-component epoxy is cured at room temperature.
The industrial goal in this study is to increase the mechanical performances of adhesively bonded
structures. To achieve this goal the curing process of the adhesive will be optimized.
Introduction
The cubicles of traction developed by the ALSTOM Transport Company [1] are composed of
different modules, power, electronic and other modules used in railway traction. This modularity
requires a structural assembly with relatively important size. Currently, the metal parts of the
structure are assembled by mechanical connections and in particular riveted and screwed. These
technologies of assemblies require a great number of fasteners accompanied by as many installation
operations, also the adaptation of tools to reach all the places of the cubicles. Moreover, the
assembly must ensure the sealing for protecting the electric components from the environmental
constraints. The structural bonding represents an alternative solution, thanks to the industrial
evolution of the structural adhesives.
The adhesive joints present a homogeneous stresses distribution, high corrosion resistance, easy
application, and low weight. Nevertheless, the adhesion durability depends on many and opposite
parameters, such as mechanical, thermodynamical and chemical adhesion [2, 3].
The commercial adhesive used in this study is the bi-component epoxy scotch weld 7240 B/A
produced by 3M. It is a structural adhesive and it may cure at room temperature. A pneumatic gun
and a mixer to mix the two parts of the adhesive (Part A: hardener, Part B: epoxy) are used.
Experimental
To study the effect of cure temperature on the mechanical and thermo mechanical behaviour of
the adhesive, the adhesive cured at room temperature and at 70C in 2 hours is tested.
The choice of the temperature of 70C was based on the exothermic Differential Scanning
Calorimetry (DSC) of the adhesive, which shows the reaction between the two parts (Fig. 1) [4].
Several networks can be formed during the adhesive curing [5]. With a rate of temperature rise
below 2C/min, two exotherms can first be observed wich may correspond to the formation of two

44

Structural Analysis of Advanced Materials

networks. The first network begins to form around the 70 C. When increasing the heating rate, the
exothermic signal is lost, it is clear that the potential detection of different exotherms due to the
presence of several epoxies and hardeners is very sensitive to heating rate.
1.5
2C/min

3C/min

5C/min

Energy Flow (mW/g)

1
0.5
0
-0.5
-1
-1.5
Temperature (C)

-2
20

40

60

80

100

120

140

160

Fig. 1: exothermic DSC of the adhesive


The main characteristics of the peaks of crosslinking obtained and the reaction heat (Hr) and
glass transition temperature (Tg) for each rate of temperature rise are presented in Table 1. Hr is
lower at 5 and 3 C /min, which means that the adhesive curing is not properly completed. The Tg
is greater when the rate of temperature rise its value is greater than 2 C /min compared to other rate
of temperature rise.
5C/min

3C/min

2C/m in
First peak

Second peak

T beginning (C)

80 3

77 4

54 3

83 2

T peak (C)

107 2

102 3

72 4

97 3

T end (C)

134 3

125 2

81 4

115 3

Hr (J/g)

60,4 5,2

63,5 4,8

120,4 6,1

Tg (C)

103 3

105 3

115 10

Table 1: Characteristics of peaks of crosslinking, reaction heat (Hr), and glass transition
temperatures for the adhesive 7240

Key Engineering Materials Vol. 446

45

The choice of the 2 hours is based on the DSC results for the adhesive cured at 70C during 30
minutes, 1, 2, 3 and 4 hours (Fig.2). The objective is to select the time curing required by heating
the adhesive cured with a ramp of temperature rise of 10 C/ min, and then check the residual
exothermic peaks and the Tg for each sample [6]. In the DSC, the identification of the Tg is hardly
detectable, but the polymerization begins to stabilize from two hours.
2.5

30 minutes
2h
4h

Energy Flow (mW/g)

2
1.5

1h
3h

1
0.5
0
0

20

40

60

80

100

120

140

160

180

200

-0.5
-1

Temperature (C)
-1.5

Fig.2: DSC for the adhesive cured at 70C during 30 minutes, 1, 2, 3 and 4 hours
So the adhesive cycle optimized is 2 hours at 70 C with a rate of temperature rise of 2 C/min,
this cycle will be compared with the adhesive cured at room temperature to evaluate the effect of
curing temperature on the adhesive. Two types of tests were performed: the rheometer for the
adhesive structural characteristics and the tensile test for the mechanical characteristics of bonded
joint.
The rheometer: the device used is the ARES rheometer, operated by RSI orchestrator software
produced by Rheometric society. The applied behaviour on the adhesive specimen is the triangular
torsion. An engine requires a sinusoidal angular deformation. The adhesive specimen sends a
sinusoidal torsion torque measured by a sensor. The results stress is phase shifted with an angle ,
called the loss angle, for the deformation applied. During the test, the torsion torque and are
measured. The shear complex modulus G*, is obtained by the following [7]:
G = G + iG" =

G"
tan =
G

Te
3

ab 1 0,63 a

(1)

(2)

a and b represent the width and thickness of the rectangular adhesive specimen, is the torsion
angle, G' is storage modulus characterizing the energy stored as elastic and G" is loss modulus
representing the energy dissipated by the material.

Tensile test: this type of test has been used to characterize the mechanical strength of the bonded
joint. This test was performed on a tensile machine INSTRON 5500R, equipped with a 100 kN load

46

Structural Analysis of Advanced Materials

cell. The regulation was made in displacement with a speed of 1 mm/min. This test is necessary to
validate the improvement of the mechanical strength of bonded joints with an adhesive cured at
70C compared to adhesive cured at room temperature. The type of the bonded joint used is the
single lap joint with two plates of galfan (steel covered zinc (D X 52D + ZA)). The mechanical
characteristics data of this material are presented in table 2 [1].
E

Re : yield stress

Rm : rupture limit

(MPa)

(MPa)

200

270 420

(GPa)
200

0. 3

Table 2: Mechanical characteristics of the material


The adhesive 7240 B/A is charged with the grains of glass of 200 micrometer diameter to gauge
the thickness of adhesive layer. The preparation of surfaces is done by a simple degreasing. To
study the effect of high curing temperature on the mechanical behavior of the bonded joint, taking
into account the hygrothermal ageing, tensile tests were performed on joints in hygrothermal aged
with a humidity RH = 95% and a temperature of 80 C.
Results and Discussion
In rheometer, the storage and loss modulus G' and G" were obtained as a function of
temperature. The storage modulus G' and thereafter the shear complex modulus G* for the adhesive
cured at 70C is greater than the modulus obtained at room temperature. The loss angle can also
be detected. It allows the determination of the temperature T associated with the main mechanical
relaxation which is the mechanical response associated with the glass transition.
The following Figure (Fig. 3) shows the storage modulus G as a function of the temperature.
1'000
room temperature
70C-2H

G'(MPa)

100

10
Temperature (C)
1
0

20

40

60

80

100

120

140

Fig. 3: G= f(T)
The following Figure (Fig. 4) shows the loss modulus G" as a function of the temperature

Key Engineering Materials Vol. 446

100

47

room temperature

G"(MPa)

70C-2H
10

Temperature (C)

0
0

20

40

60

80

100

120

140

Fig. 4: G" = f(T)


The following Figure (Fig. 5) shows the loss angle as a function of the temperature and it is so
possible to find the temperature T which is 82C for a polymerization of 2 Hours at 70C, and
53C for a polymerization at room temperature.
0.80
0.70

room temperature

tan

0.60

70C-2H

0.50
0.40
0.30
0.20
0.10
0.00
0

10

20

30

40

50

60

70

80

90 100 110 120 130


Temperature (C)

Fig. 5: Tan =f(T)


The Table 2 presents G max. and T obtained with the two types of polymerization, 2 hours in
70 C and room temperature. These results issued from the above figures show the interest of
performing the polymerization at a controlled temperature. In effect, all the inflexions visible
correspond to the temperature T. This temperature is greater when the polymerization is optimized.
*

Curing cycle

room temperature

2 hours in 70 C

G*max. [MPa]

564 5

817 5

T [C]

53 2

82 1

Table 2: G*max. and T for two types of polymerization


In tensile test, the result of the failure effort of the bonded joints was obtained. The tests were
performed with an imposed displacement for the two types of polymerization (Fig. 6).

Structural Analysis of Advanced Materials

Effort (kN)

48

45
40
35
30
25
20
15
10
5
0

room temperature
70C-2H
Displacement (mm)
0

10

Fig. 6: Effort-Displacement curves for the single lap joint (adhesive cured at room temperature and
at 70C in 2 hours)
As shown on the above Fig.6, the tensile tests performed after hygrothermal ageing of 50 days
show improvement of the mechanical strength of joints cured at 70 C in 2 hours [8]. The
degradation of the failure effort is 40% for the joints cured at room temperature and less than 12%
for the joints cured at 70 C in 2 hours.

Cohesive failure

Adhesive failure

Fig. 7: Failure topographies of an adhesive joint


The above Fig.7 presents the topography of the joint after failure on the left without ageing, on
the right after 50 days hygrothermal ageing. It is remarkable to note that the non aged specimen
exhibits a totally cohesive failure, what proves the quality of bonding. The aged one presents a
combination between cohesive and adhesive failure, what shows the effects of ageing on bonding.
To improve the cross linking of the adhesive is not a protection of ageing but it produces a
limitation of the harmful effects of hygrothermal ageing while limiting the area of adhesive failure.

Key Engineering Materials Vol. 446

49

Conclusion
The first conclusion of these results is that the polymerization at a higher temperature from
room temperature (23 C) has a positive effect on the mechanical strength and the environmental
degradation of bonded joints. The polymerization of the adhesive at high temperature can ensure
good cross linking and consequently strengthen the tridimensional network of adhesive, which
increases the shear complex modulus G* and the temperature T.
Optimization cycle of polymerization of the adhesive is very important to have an adhesive
with better thermo-mechanical resistance. This optimization can lead to cure the adhesive at
temperatures above 70 C to increase as T above 80 C. It appears real that optimizing the
parameters of polymerization leads to better behaviour of the adhesive joint and to higher strength
and resistance to ageing. It seems very important to check the level of the cross linking of the
adhesive to ensure a better life for the joint. In effect the joint will be better because this mechanical
behaviour will be more cohesive than adhesive. Also it can be concluded that the polymerization at
a higher temperature from room temperature has a positive effect on the mechanical strength and
the environmental degradation of bonded joints.
The future prospects of this work is to confirm the mechanical performance of bonded joints
in hygrothermal ageing with a larger study to identify the endurance thermal limit of the adhesive.
References
[1] M. Ali, M. Karama, B. Lorrain, B. Puel, Fatigue behaviour of the screwed/riveted joints and the
bonded joints, International conference on the respective input of the numerical simulation and the
experimental approach in fatigue design 21-22 Novembre 2007, Senlis, france.
[2] Bernd R. Burchardt and Peter W. Merz (2006), Elastic Bonding and Sealing in Industry.
Handbook of Adhesives and Sealants Volume 2, chapter 6, p. 356-375.
[3] J. K. Gillham, J. A. Benci, (1974) J. polym. Sci, 46, 279-284.
[4] G. Possart, M. Presser, S. Passlack, P.L. Geib, M. Kopnarski, A. Brodyanski, and P. Steinmann
(2009) Micromacro characterisation of DGEBA-based epoxies as a preliminary to polymer
interphase modeling. International Journal of Adhesion & Adhesives N29 p.478-487.
[5] Material data Sheet Safety, SCOTCH-WELD 7240 B/A FR, 3M Company, FDS N:09-6884-2,
http://www. Quickfds.com
[6] Y. Grohensb, B. Georgea, F. Touyerasa and J. Vebrela, (1998), Analysis of curing mode and
mechanical properties of an anaerobic adhesive, European Polymer Journal Volume 34, Issues 3-4,
P. 399-404.
[7] J.P. Habas, (1995), thse de doctorat, universit de Pau et des Pays de lAdours, N252. Pau,
France.
[8] Y.C. Lin, Xu Chen, H.J. Zhang, Z.P. Wang, (2006), Effects of hygrothermal aging on
epoxy-based anisotropic conductive film, Materials Letters N60 p. 2958-2963

Key Engineering Materials Vol. 446 (2010) pp 51-61


(2010) Trans Tech Publications, Switzerland
doi:10.4028/www.scientific.net/KEM.446.51

Study of DGEBA and Novolac adhesives hydrothermal ageing for a


ceramic/ steel substrates bonding assembly.
Cllia Gaussens1, a, Valrie Nassiet1, b, Bouchra Hassoune-Rhabbour1, c
1

Laboratoire Gnie de Production, Equipe Interfaces et Matriaux Fonctionnels, Ecole Nationale


dIngnieurs de Tarbes,
B.P. 1629, 65016 Tarbes cedex, France
a

email: Clelia.Gaussens@enit.fr, b email : Valerie.Nassiet@enit.fr,


c

email:Bouchra.Hassoune@enit.fr

Keywords: thermosetting resins, adhesives formulations, hydrothermal ageing, sorption models.

Abstract. The industrial application (not described for confidential reasons) needs the assembly of
a ceramic piece and a steel one. This PhD research work aims to provide a reliable industrial
bonding between ceramic and steel substrate using structural adhesive. This industrial joint shall
withstand a wide range of temperature with pretty brutal thermal changes and also severe humidity
conditions. As prerequisite to this bonding assembly qualification, the purpose of this work is the
study of the durability, in a hot-wet environment. Depending on the formulations, we observe
different behaviours in term of weight uptake, morphological, thermomechanical and chemical
modifications of the epoxy network.
Introduction
The brittleness of the ceramic and the need of a total transfer of the load require a structural joining
as solution. From an industrial point of view, to include structural bonding solutions is difficult
because of the polymer degradation during life service involving adhesive thermomechanical
properties and adherence changes.
Thermosetting can resist to significant loads, so they are the only ones able to be used as structural
adhesives. Epoxy resins are often used as adhesives, coatings, encapsulate, casting materials,
potting compounds, and binders for a large range of temperatures from -55 to 200C. They are
formulated to generate specific chemical and thermomechanical properties.
Organic polymers are prone to the adsorption of surrounding media either of a liquid or gaseous
nature with mechanical properties reduction as consequence. There are several modes of humid
ageing: by plasticization [1, 2], by differential swelling related to concentration gradients [3, 4],
polymer degradation by hydrolysis [5, 6], damage localised at the filler/ polymer interface. Even
epoxy resins regarded to be water resistant have their mechanical properties decreased [7-10].
Classically short time exposure can lead to reversible plasticization, while longer time exposure
results by crack occurrence and growth. All the previous physical phenomena have two main
origins. I) the affinity of the matrix for water called hydrophilicity, II) the penetration rate of water
in the polymer (also called diffusivity). The hydrophilicity is related to the existence of specific
interactions between the water molecules and polar groups of the polymer [11-15]. The diffusivity
has been linked to morphology [16], to molecular mobility [17] and free volume [11, 12, 18].
Hydrophylicity and diffusivity are two inseparable concepts. Indeed Polymer interactions with
water play a significant role in the diffusion.
Three basic classes of behaviours related to diffusivity can be established [19].
- Fickian diffusion in which the rate of diffusion is much less than that of relaxation due to
mechanical, structural modes of polymer/water system.
- Case II [20] or Super Case II in which the diffusion rate is very quick compared to the
polymer molecular relaxation during diffusivity.
- Non-fickian or anomalous diffusion [21] which occurs when the diffusion and relaxation
rates are similar.

52

Structural Analysis of Advanced Materials

Carter and Kibler [22] have developed a model so-called Langmuirtype diffusion model (because
of its similarity to Langmuir dual sorption model). This model involves bound water and mobile
(free) water molecules by assuming a probability for bound molecules (due to specific interactions
between water molecules and polar groups of the polymer) to become free molecules and inversely.
Indeed H2O molecules establish hydrogen bonds relatively strong (12Kcal/mol) with the polar
groups of the epoxy chain [23-24]. The model graphic shape corresponds to a mass increase firstly
proportional to the square root of ageing time, then transforming in a linear part of lower gradient
but not zero as in a fickian model.
At long exposure periods or in more severe environments, thermosetting hydrolysis occurs and
leads to the macromolecular chains failure. Particularly the polar groups as hydroxyls (-OH) and
ether-(-C-O-C-) or (-CONH; COOH; COO-), which dont participate during the cure of the network
are converted by hydrolysis [25]. The first part of this article deals with gravimetry tests for
different formulations in a humid environment. Then thermomechanical tests by rheometry have
been performed on aged samples to explain the water diffusion in the resin network. The last
purpose of this study is the chemical identification of network structure changes.
Materials
Studied adhesives are based on DGEBA and Novolac epoxy resins mixed to an amine hardener
which basic molecular structures are shown on Fig. 1, 2 and 3.
CH3

O
CH CH2 O

H2C

CH3

OH
C O CH2 CH

CH2 O

O
O CH2 HC

CH2

CH3

CH3

Fig.1: Diglycidylether of bisphenol A formula


H2C

H2C
O

H2C
O

HC

HC

HC

CH2

CH2

CH2

CH2

CH2
n

Fig.2: Novolac formula


Novolac functionality varies with polymerization degree as shown in Fig.2 since oxirane group is in
the repeated monomer and it ranges between 1 and 7.
The curing agent Dicyandiamide is used with a catalyst in order to decrease the initial curing
reaction temperature, commonly higher than 150C. The curing reaction mechanism is represented
on Fig.3 for a secondary amine with an oxyran group.
C CH2
O

C CH2
O

HN

HO

Fig. 3: Cure mechanism

OH
CH2 N

OH
CH2 O

Key Engineering Materials Vol. 446

53

The three formulations named adhesives A, C and D for industrial confidential reasons vary by the
two resins proportions. All the data are collected in Table 1.
Adhesive name
A
C
D

Resin nature
DGEBA
DGEBA / Novolac 1
Novolac 1/ Novolac 2

Table 1: Adhesive characteristics


As evidence, Novolac viscosity increases with functionality. Novolac 1 has functionality higher
than Novolac 2 in order to decrease the blend viscosity making the cure process easier. DGEBA is a
bifunctional prepolymer.
Experimental procedures
Calorimetry. During curing optimization and after ageing (at 40C and 95%HR in a climatic
oven), calorimetric analyses were carried out with a Differential Scanning Calorimeter (Shimadzu
50 from Shimadzu Corporation) to determine the optimized cure cycle.
About 12 mg of the adhesive is set in a sealed aluminium pan placed in a measurement heating
cell. An empty pan is used as reference. All experiments are realized under nitrogen atmosphere in a
20ml/min flow. The glass transition temperature determination is performed at 10C/min from 20C
to 180C for A and from 20C to 210C for C and D.
Hydrothermal Ageing. After hydrothermal ageing (40C and 95% HR imposed by industrial
requirements) for a given period, the mass gain experiments are performed on three parallepipedic
samples surface (2*10*10mm3) dried and weighed using a Sartorius R18 DF1 balance, whose
sensitivity is 10 mg. The samples are obtained through the process described in Table 2. Aging
times are ranging from 0 to 9000 h.
Infrared Spectroscopy. The infrared spectra during ageing were obtained using the ATR FTIR
spectrometer of Perkin Elmer. FTIR is a common technique to investigate chemical structures of
polymer, and ATR method is used to, the material surface laid on the total reflection element
(crystal of diamond coupled with the ZnSe); while the other surface faced the contacting device controlled through careful torque. Each spectrum was scanned 4 times from 600 to 4000 cm1. The
spectra were taken at a resolution of 4 cm1 and the penetration depth is around 0.5 m according to
the spectroscope specifications.
Scanning electron microscopy (SEM). After ageing some samples are observed by Scanning
Electron Microscopy (Jeol 7000F), which permits to highlight very fine details (high resolution) but
also to have a good focus over a wide range of surfaces (large depth of field). The micrographs are
realized under a 2 kV voltage which gives an imaging resolution around 3nm. For a better
observation, a very fine palladium metallization is done by plasma spray on the analyzed surface.
Dynamic mechanical analysis. The thermomechanical characterizations of the cured adhesives
samples (40*10*2) mm3 are performed in a strain-imposed dynamic rheometer (ARES from
Rheometrics Scientific) with rectangular torsion geometry. The storage shear modulus G, the loss
shear modulus G and the loss angle tan () (G/G) are recorded in the thermal range from 25 to
200C. The angular frequency is fixed at 1 rad/s, the temperature ramp is equal to 3C/min and the
strain fixed at 0,3% to perform experiments in linear viscoelasticity domain.

54

Structural Analysis of Advanced Materials

Results and discussion


It is evoked, in the introduction, that the durability of adhesive bonded joints is considerably
reduced in the presence of an aggressive medium associating heat and moisture. The hydrothermal
degradation of joints then results either from a decrease of the mechanical properties of materials
(generally those of the adhesive) or from a reduction of adhesion properties, or even from both
phenomena.
Moreover the first step is to optimize the vulcanization process as a guaranty of a better durability.
Time and temperature are essential parameters of the reaction. It has been widely shown that an
incomplete reaction would diminish reliability of adhesive-bonded assemblies.
The curing processes for the three adhesives are given in the following Table 2. Before curing,
adhesive are degassed at 70C to evacuate the reaction inhibitor and the air bubbles. Then cured
samples stay 24 hours in a desiccant before performing the rheological tests.
Curing process of bulk
Glass transition temperature
Conversion rate
adhesive
(Tg)*
100C 1h, 120C 1h, +140C
147 3C
97 1
A
1h, +160C/1h
C
185 3C
96 2
100C 1h, 135C 1h +160C
1h +185C 1/2h
175 3C
95 2
D
DSC experiments
Table 2: Curing process for bulk samples
The curing steps are chosen in the exothermal temperature range defined by DSC. This process is
explained by the Gillhams and Bencis theory [26] expressing that glass transition temperature
cannot overtake thermosetting maximal internal temperature during cure.
In order to minimize the weight effect on cure exothermal peaks leading to the thermoset oxidation,
a first step at 100C is maintained during one hour for the three formulations.
The conversion rate is calculated from the Beer Lambert law (equation 1). The main curing reaction
is the aperture of oxyran groups (915 cm-1 band) by the hydrogen of the hardener amine. Since the
ring stretch vibration at 1507cm-1 band is not modified; the law is applied by comparing the
variation of 915 cm-1 band with 1507cm-1 band [27, 28, 29]:
Adhesive

(1)

abs (915cm-1)t and abs(915cm-1)t=0 are respectively the area value under the oxyran group
absorbance band for the cured and uncured network. The same definition is applied for
abs (1507cm-1) band.
As regards the conversion rate values (Table 2), we can conclude that the cure is optimized.

Water Sorption. Fig. 4 shows the sorption behaviour of adhesives A, C and D aged in a 40C and
95%HR climatic oven.

Key Engineering Materials Vol. 446

55

Fig.4: Water weight uptake versus time root normalized by sample thickness
In the first time the water diffusion is quiet similar in the three adhesives. Initially, the mass
increases linearly with the square root of time but without mass stabilization as in a Fickian way
[19]. So water doesnt behave as a single free phase driven by the concentration gradient. On
contrary the mass always increases with a weaker slope. No stabilization appears after more than
9000 hours of adhesives ageing. Moreover the increase is higher for Novolac based adhesives.
How to explain this weak increase in a second period? What about balance between bound and free
water molecules, possible hydrolysis and /or swelling between surface and core?
Because the polymer rheological behaviour is governed by the molecular dynamics, we have
performed thermomechanical tests on aged samples to quantify the molecular relaxation changes
which should be involved by water molecules.
Thermomechanical and Morphological Analyses. Fig.5a depicts the thermomechanical behaviors
of adhesives A, C and D before ageing. The so-called alpha temperature (T) associated to the
maximum of the relaxation peak on the loss modulus G curve is due to the polymer transition
from its glass state to its rubbery state. So T is the rheological signature of the glass transition
temperature Tg. Its value is 148 3C for A, 182 3C for C, 169 3C for D and is in accordance
with the Tg values gathered in Table 2. The broadening of G relaxation in glass transition
temperature range for C and D adhesives may be explained by the large polymolecularity of the
thermosetting network - blend of chemically different molecules (DGEBA and Novolac) for
adhesive C, and blend of two Novolac resins with different functionalities, therefore different
molecular weights for adhesive D.
As to visualize better the different mechanical relaxations associated to water diffusion, we choose
to represent the loss factor tan for aged samples instead of G variation. Because all the adhesives
are modified in the same way, Fig.5b only gathers the variations of the elastic modulus G and the
loss factor tan of A adhesive after several ageing times in a 40C and 95%HR climatic oven.

56

Structural Analysis of Advanced Materials

Fig.5a: Thermomechanical behavior of the three Fig.5b: Thermomechanical behavior


unaged adhesives
adhesive at several ageing times

of

The tan maximum can also be related to the glass transition temperature. We shall speak about
T instead of T so as to differ the two values. Subsequently, only the T variation according to
ageing time is commented. One can see a Tdecrease with ageing time. The alpha relaxation
becomes broadened with ageing time with occurrence of one new relaxation which appears as a
shoulder on the relaxation. On the thermomechanical curve of a one week aged sample, the two
relaxations are around 130C and 145C and are shifted towards lower temperatures at longer
ageing times. Fig. 5b too shows the variations of G and the loss factor tan for two samples
previously aged for one week and one month then subjected to air drying in an oven at 120C
during 48 hours. The aim of this drying was to observe the possible reversibility of the water uptake
by the adhesive. After desorption of the week aged sample, the shoulder at 130C is removed but
the tan maximum, signature of the glass transition temperature, is unmodified. For longer ageing
times the shoulder remains and T is unchanged after desorption. It seems that chemical
irreversible modifications occur in epoxy network, leading to an unmodified tan maximum value
after desorption. Infra-red spectroscopic analyses have been performed and are commented in
Chemical ageing of the adhesive section. In addition, according to Zhou [30], water molecules are
bonded with polymer threshold by hydrogen links. After desorption he proposed an explanation of
residual water, invoking two kinds of bonded water. The first one, easily removed by desorption, is
characterized by a single hydrogen bond between water molecules and the polymeric network. This
behaviour seems to correspond to samplesone previously aged for one week then subjected to air
drying. The second type involves the creation for a water molecule of more than one hydrogen bond
via multiple interaction sites. In this latter case, higher activation energy requires a high temperature
for totally drying the sample. So, for the month aged and desorbed sample, the second type of
hydrogen bonds seems to be the dominant form of the total amount sorbed water which is not
removed after a 48 hours drying. This explanation corroborates the Carter and Kibler model on the
reciprocal conversion between free and bound water molecules, these later possessing a higher
activation energy compared to free water molecules; it is correspondly harder to remove from the
resin.
As to support our interpretation the water mass variation Mt as a function of time square root
normalized by sample thickness has been fitting by Carter and Kibler model according to the
procedure detailed by Popineau and al [31]:
For short exposure times
when Kt0.7

(2)

Key Engineering Materials Vol. 446

57

For long exposure times


when Kt1
with K=

(3)

. D is a coefficient of diffusion, M the saturation mass and e the sample thickness.

and is the probability for bonded water molecules to become free and inversely.
At a given time t, n molecules are bonded among the n free molecules and N become free among
the N bonded molecules with at saturation

(a)

(b)

(c)

Fig.6: Gravimetric data for water absorption for adhesive A (a), adhesive C (b), adhesive D (c): ()
experimental, (-) model
Fig. 6a, 6b and 6c illustrate the good agreement enough between the experimental and simulation
results for the three adhesives.
The diffusion coefficients are (9,2 0,2) 10-13 m2/s for A; (13,1 2)10-13 m2/s for C and
(9 0,3)10-13 m2/s. These values are comparable to these found in literature, [6]; [31]; [32]; [33]
and similar for the three formulations.
Then the reciprocal conversion between free and bound water molecules is verified.
Some authors [34-35] previously reported a similar deviation, on the sorption curve (Fig.4), for a
glass epoxy composite immersed in water and in glycol solution at different temperatures. They
attributed it, after microscopic examinations, to the creation of microcracks and voids within the
matrix and at the matrix fibre interface. The mass uptake thus becomes a signature of the both water
sorption within the epoxy/amine network and the filling of these voids and microcracks.
So scanning electronic microscopy analyses (Fig.7 and 8) have been performed at 2kV tension, on
A and D adhesive samples aged of 50 hours and 5000 hours.

58

Structural Analysis of Advanced Materials

Fig.7a:
SEM Fig.7b: Adhesive
micrographs
aged 50h(x5000)
Unaged adhesive B
(x5000)

A Fig.7c: Adhesive A aged


5000h (x5000)

x10

Fig.8a:
SEM Fig.8b:Adhesive
micrographs
aged 50h (x5000)
Unaged adhesive D
x5000

D Fig.8c:Adhesive D aged 5000h


x50000)

(x5000 and

The D adhesive surface aged during 50 hours (Fig.8b) shows first microcracks and a little porous
form to become totally damaged with a spongy morphology. Numerous microcavities and a few
microcracks appear at 5000 hours of immersion. The spongy morphology appears weaker for
the A adhesive than for the D adhesive, for long times as 5000 hours. The more severe damage of
the D adhesive can explain the stronger positive deviation in the second part of sorption curve
compared to the DGEBA based adhesives. We remember that the Novolac functionality varies with
polymerization degree as shown in Fig. 2 since oxyran group is in the repeated monomer. Moreover
Novolac structure is essentially formed by phenyl groups As a consequence, the Novolac average
intercrosslink molecular weight is very weak (115 g/mol) compared to the DGEBAs one (344
g/mol), then making this resin very stiff as confirmed by curves stress/strain of the two adhesives
(Fig.9). Indeed it is noticeable that adhesive D is more rigid (the highest young modulus) and brittle
(the lower break elongation) compared to A. The water diffusion involves concentration gradient in
the sample thickness from shell to core of the sample, leading to more and more compressive areas
from surface inwards sample core. As evidence, stiffner is the sample, greater is the differential
swelling into the sample hence microcracks occurrence.

Key Engineering Materials Vol. 446

59

Young
modulus
Adhesive [GPa]

Failure
strain [%]

2,30,02

10,20,1

2,90,3

3,90,4

Fig.9: Curves stress/strain for unaged A and D adhesives.


To perform the structure chemical evolution, Infra-red spectroscopic spectra have been performed.
Chemical ageing of the adhesive. Because all the adhesives are modified in the same way, Fig.10
just shows the A adhesive chemical evolution during ageing.

Fig.10: Infrared spectra obtained for unaged, aged and desorbed samples
After one week in a 40C and 95 % HR oven, the infrared spectrum is modified as following:
1- Increase of the OH band intensity between (3200-3600 cm-1) ;
2- Increase of the intensity at the wavelength between 1710 and 1780cm-1
3- Decrease of the intensity between 1680 and 1710cm-1.
These tendencies are magnified after one month exposure and stays after desorption
Normalization in the spectra represented Fig. 10 is carried out on the 1507cm-1 band which is
attributed to the ring stretch vibration which is not modified during ageing. The modification of the
OH band is due to water sorption. A desorption cannot fully erase this modification as shown by the
OH band variation. So the water has partly reacted with the matrix by hydrogen bonds as evidenced
by rheological approach and/or chemically reacted by hydrolysis with the thermoset matrix [5-6].
In a previous article [36], for an unaged sample, the carbonyl band (C=O) which appears between

60

Structural Analysis of Advanced Materials

1710 and 1780cm-1 has been attributed to the Gilbert mechanism [37] describing the oxazolidinone
(containing a carboxylic acid group) formation by a chemical reaction between secondary amine
groups and oxyran groups. The bands between 1680 and 1710cm-1 have been attributed to amide
groups formation. Indeed, according to the Saunders mechanism [38], reaction between
dicyandiamide and hydroxyl groups finally creates a substituted guanylurea containing amide
groups.
After ageing and ageing more desorption, the irreversible modifications of FTIR bands between
1710 and 1780cm-1 and 1630 and 1710cm-1 can be attributed to the amide (1680-1710cm-1)
hydrolysis (decrease of the associated band) producing carboxylic acids (1710-1780cm-1) according
the following mechanism in humid medium:

Fig.11: Hydrolysis mechanism


Conclusions
The hydrothermal ageing study of epoxy adhesives shows different behaviours depending on their
formulation parameters. Firstly, the diffusion seems follow the Langmuir model. This hypothesis is
confirmed by bonded water presence after desorption and good concordance between gravimetry
reponses and Carter and Kibler model. However this model does not take into account the cracks
occurrence and/or chemical modifications experimentally highlighted.
Indeed a same irreversible chemical modifications of FTIR bands are observed on the three
adhesives. These chemical modifications can be attributed to the amide hydrolysis.
Finally Novolac based adhesives, essentially formed by phenyl groups, is stiffer than the DGEBA
based adhesive. So, at long ageing time, water diffusion creates a great differential swelling leading
to microcracks creation, signature of physical modifications

References
[1] D. Colombini, J. J. Martinez-Vega, and G. Merle. Polymer, 43, 16, (2002).4479
[2] C. L. Soles and A. F. Yee. J. Polym. Sci. Pol Phys, 38, 5, (2000), 792
[3] M. R. Vanlandingham, R. F. Eduljee, J. W. Gillespie Jr., J. Appl. Polym. Sci., 71, 5, (1999), 787
[4] I. Merdas, F. Thominette, A. Tcharkhtchi, J. Verdu. Compos Sci Tech, 62(4), (2002), 487
[5] B. De Neve, Applied Surface Science, 134, (1998), 202
[6] G.Z Xiao and M. E. R. Shanahan Polymer, 39, 14, (1998), 3253[7] C. M. Shen and G. S.
Springer, J. Comp. Master, 11, (1977),2
[8] C. M. Shen and G. S. Springer, J. Comp. Master, 11, (1977), 250
[9] B. DeNve and M.E.R. Shanahan, Polymer, 34, (1993), 5099
[10] P. Montois, V. Nassiet, J. A. Petit, Inter J Adhesion and Adhesives, 27, (2007), 145
[11] M.J. Adamson, J. Mater. Sci. 15, (1980), 1736
[12] P. Johncok, G.T. Tudgey Br .Polym. J. , 18, (1986)242.
[13] E. Morel, V. Bellenger, J. Verdu, polymer , 26, (1985), 1719
[14] V. Bellenger, J. Verdu, E. Morel, J. Mater. Sci. , 24, (1989), 63

Key Engineering Materials Vol. 446

61

[15] A. Tcharkhtchi, P.Y. Bronnec, J. Verdu, Polymer, 41, (2000), 5777


[16] V.B. Gupta, L.T. Drzal, M.J. Rich, J. Appl. Polym. Sci., 30, (1985) 4467
[17] S. Cukierman, J.L. Halary, L. Monnerie, Polym. Eng. Sci., 35, (1992), 534
[18] J.S. Vrentas, J.L. Duda, H.C. Ling, Macromolecules, 21, (1988), 1470
[19] H. L. Frisch. Polymer Eng. Sci., 20, 1, (1980), 2
[20] T. Alfrey, E. F. Gurnee, W. G. Lloyd, J. Polym. Sci., 12, (1966), 249
[21] C. H. M. Jaques, H. B. Hopfenberg, T. Stannett, Permeability of Plastic Films and Coatings,
(1974), 73
[22] H. G. Carter and K. G. Kibler, J. Comp. Mater, 12, (1978), 118
[23] J. Verdu, Matriaux et tchniques, 9-10, 24 (1992)
[24] P.J. Flory, J. Rehner, Polym. Eng. Phys, 11, 11 (1943)
[25] M.A. Boyle, C.J. Martin, J.D. Neuner ; Epoxy Resins ASM Handbook 21, (2001), 78
[26] J. K. Gillham,J. A. Benci, J. Polym. Sci, 46, (1974), 279
[27] S.G. Hong, C.S. Wu, Thermochim. Acta, 316, (1998), 167
[28] S.G. Hong, C.S. Wu, J. Therm. Anal. Calorim., 59, (2000), 711
[29] P. Bartolomeo J.F. Chailan, J.L. Vernet, Eur. Polymer. J., 37, (2001), 659
[30] J. Zhou, J.P. Lucas, Polym, 40, (1999), 5505
[31] S. Popineau, C. Rondeau-Mouro, C. Sulpice-Gaillet, M.E.R. Shanahan. Polymer, 46, 24,
(2005), 10733
[32] E. Legghe, E. Aragon, L. Blec, A. Margaillan, D. Melot. Progr. Org. Coating., 66, 3, (2009),
276
[33] Y. Ding, M. Liu, S. Li, S. Zhang, W. Zhou, B. Wang, Macromol. Chem. Phys., 202, 13,
(2001), 2681
[34] J. Evieux, Study of bonded plasma-treated, The Journal of Adhesion, 80: 263_290, 2004.
[35] A. Chateauminois, L. Vincent, B. Chabert, J.P. Soulie, Polymer 35 (22), 4766_4774 (1994).
[36] C. Gaussens, V. Nassiet, B. Hassoune Rhabbour, submitted to International Journal Advanced
Materials Research
[37] M.D. Gilbert and N.S. Schneider, Macromolecules, 24, (1991), 360
[38] T.F. Saunders, M.F. Levy, J.F. Serino, J. Appl. Polym. Sci. Pol. Chem., 5, (1967), 1609

Key Engineering Materials Vol. 446 (2010) pp 63-72


(2010) Trans Tech Publications, Switzerland
doi:10.4028/www.scientific.net/KEM.446.63

Influence of strain rate on the yielding behavior and on the self


heating of thermoplastic polymers loaded under tension
Ismail Hadriche 1,2,a, Elhem Ghorbel 1,b, Neila Masmoudi 2,c
and Foued El Halouani 2,d
1

L2MGC-EA4114, University of Cergy Pontoise, Rue dEragny, Neuville sur Oise,


95031 Cergy Pontoise Cedex, France

LASEM-LR99ES36, National Engineering School of Sfax, B.P. W1173, Sfax, Tunisia


a

ismail_hadriche@yahoo.fr, b elhem.ghorbel@u-cergy.fr, c neilakhabou@yahoo.fr,


d
foued.elhalouani@enis.rnu.tn

Keywords: thermoplastics, yielding behavior, self heating, infrared thermographs, strain rate
effects

Abstract. The effect of strain rate on the mechanical behavior of thermoplastic polymers
(Polymethyl methacrylate, Polycarbonate and Polyamide 66) has been studied. Deformation tests in
tension were conducted over the range of strain rate varying between 2.6 10-4s-1 to 1.3 10-1s-1. The
Youngs Modulus E and Yield stress ST evolutions have been identified and modelled as a function
of the strain rate. It has been established that, in the range of the considered strain rates, the yielding
behavior of PMMA and PC is well described by the Eyring theory while for PA66 the Ree-Eyring
theory is successfully used to illustrate the yielding behavior. During tensile tests the specimen
surface temperatures were monitored using an infrared camera. Results reveal a significant
temperature rise at large deformations for PA66 and PC. As the strain rate increases the temperature
is steadily increased with deformation due to plastic work. Hence, for PC and PA66, a significant
thermal softening is observed after yielding which affects the stress-strain behavior. Thermomechanical coupling during polymer deformation can be considered in the modeling of the
mechanical behavior of polymers. No self-heating has been detected for PMMA.
Introduction
The research of the behaviors laws dedicated to polymeric materials is generally based on
experimental tests. The developed models depend on a great number of parameters that must be
identified using, generally, a restricted number of experiments [1,2]. For polymers, several authors
have shown that the constitutive models dedicated to simulate the mechanical behavior reproduce
well the tension and compression behavior. Most of them are established by considering isothermal
deformation conditions. Although heating related to the plastic strain of some polymers is evoked in
the literature [3,4] it is not introduced for most cases in the constitutive equations dedicated to the
modeling of deformation behavior of polymers. Indeed, it is supposed that the tests remain
isothermal for selected operating conditions.
Aruda et al. [4] showed that strain hardening observed for Polymethyl methacrylate (PMMA) tested
under compression is due to the temperature rise during deformation. The surface temperature was
measured using a Mikron M676S infrared detector. They established that increasing the strain rate
(-0.1s-1, -0.01s-1 and -0.001s-1) leads to an increase of the heating of PMMA during inelastic
straining and therefore produces a thermo-mechanically coupled behavior at large strains.
Li et al [5] showed that PMMA subjected to compressive dynamic loading exhibits a minor
temperature increase while a thermal softening overtakes strain hardening for Polycarbonate (PC).
Temperature is recorded using IR detector array. These results confirm those of Rittel et al. [6]
obtained by means of thermocouple technique.
The aim of this investigation is to quantify the change in temperature which occurs during quasistatic tensile deformation of glassy and semi-crystalline polymers during loading.
To achieve this target, experiments were performed over the range of strain rate varying between

64

Structural Analysis of Advanced Materials

2.6 10-4s-1 to 1.3 10-1s-1 using three thermoplastics. Temperature rise is recorded during deformation
using an infrared camera, which allows a rapid control without contact [7].
Materials and experimental methods
Three industrial thermoplastics are used: Polyamide PA66 (Thechnyl), Polycarbonate PC
(Makrolon) and Polymethyl methacrylate PMMA (Altuglas). To determine the thermal properties of
these polymers, Differential Scanning Calorimetry (DSC) tests have been realized using DSC Q100
(TA Instruments) with a heating rate of 10C/min with 25mL/min Helium protection. The weight
fraction crystallinity can be assessed based on Eq. 1.
Xc =

H f
.
H

(1)

Where Hf is the enthalpy of fusion of samples and H is the extrapolated value of the enthalpy
corresponding to the melting of 100% perfect infinite crystalline sample. For PA66, an average
value of 188J/g has been chosen for H [8,9] while Hf =47.76J/g. The obtained properties (the
glass transition temperature Tg, the melting temperature Tf and the weight fraction crystallinity) are
summarized in Table1.
Table 1 Thermal properties of polymers studied
Thermal properties PC PA66 PMMA
Tg [C]
149.7 58.8 127.7
Tf [C]
* 261.4
*
*
25
*
Xc [%]
* No Tf for amorphous polymers
Cylindrical samples were machined and polished from extruded bars respecting the dimensions
1
2
shown in Fig. 1. Tensile tests were conducted at 230.5 C and a relative humidity of 505 %. Each test
is repeated at least three times. I STRO 5567 machine assisted by a computer with a Dynamic
Strain Gauge Extensometer I STRO 2620-601 is used (Fig. 2). To collect the surface temperature
versus time data during tension a FLUKE ThermoView Ti25 infrared camera was used (Fig. 2).
Such camera allows the detection of heat loss radiation in the infrared range of the electromagnetic
spectrum and the production of images of that radiation, called thermograms.
The radiative heat loss energy is expressed as:
E r = 0 ( T 4 T04 ) .

(2)

Where is the emissivity, 0 is the Stefan-Boltzmann constant, T is the kept temperature of the
polymer and T0 surrounding (room) temperature. This operation requires the definition of the
emissivity of each polymer to calibrate its radiant energy. In order to determine the emissivity
values of the tested polymers, thermocouples are embedded at the surface of the specimens heated
in a furnace to a considered temperature (23C <T<80C). Each temperature measured using K-type
thermocouple is compared to the one obtained using FLUKE ThermoView Ti25 infrared camera.
Therefore, an experimental calibration based on the change of the emissivity value was carried out
to find close results between both temperature measurements. It appears that = 0.95 can be used
for all the tested polymers. The apparent temperature of the surface is measured with an accuracy of
2% and a resolution of 0.1C. Several images were captured during each test. The analysis of
these thermographs and the maximal Temperature obtained was done with the SmartView 1.9FLUKE software.

Key Engineering Materials Vol. 446

65

Fig.1 Dimensions in mm of tensile specimens

Fig.2 Tension machine INSTRON 5567 and Thermal IR Imager Ti25 FLUKE
Scanning Electron Microscopy (SEM) analysis has been performed on the samples to have a better
understanding of the effects of strain rates on fracture morphology. The SEM utilized is a Leica
Stereoscan 430i model.

Results and discussion


The tensile Cauchy stress versus true strain of the investigated polymers is described by Fig. 3, Fig.
5 and Fig. 7. These tests are not conducted up to failure. The expression of the true strain is given
by the classical expression of the logarithmic strain while the assumption of the volume
conservation is adopted to calculate the stress even if it is well known that this assumption is
incorrect for polymers [1,10]:

d =


F 
= Ln and = .

S0  0
0

d

(3)

Where F is the force,  0 and  are the initial and current specimen length gage, S0 is the initial
specimen cross section. For polymers, the stress that corresponds to the beginning of inelastic strain
is often called yield stress ST. The experimental determination of yield stress is subjective since it
depends on the choice of the minimal inelastic strain value, on the measuring instrument precision
and on the adopted experimental methodology used for its detection. In this study it is assumed that
the yield stress is the intersection of the curve and 0.3% offset line which is parallel to the
elastic portion of the curve [1].

Structural Analysis of Advanced Materials

70

2.6 10-4 s-1

80

1.62 10-3 s-1

70

5.44 10-3 s-1


1.76 10-2 s-1

60

6.5 10-2 s-1


50

4.0

Yield stress
Yield stress ST(MPa)

True stress (MPa)

90

1.32 10-1 s-1

40
30
20

65

Youngs modulus

3.5

60

3.0

55

2.5

50

2.0

Youngs modulus E (GPa)

66

10
0
10

15

20

25

30

35

45
0.000

40

0.001

True strain (%)

0.100

1.5
1.000

Fig.4 Yield stress and Youngs Modulus of PC

100

60

2.6 10-4 s-1


1.62 10-3 s-1
5.44 10-3 s-1
6.5 10-2 s-1
1.32 10-1 s-1

80
70
60

4.0

Yield stress
55

Yield stress ST(MPa)

90

True stress (MPa)

-1

Strain rate (s )

Fig.3 Tensile tests of PC

50
40
30
20

Youngs modulus

3.5

50
45

3.0

40
2.5

35
30

2.0
25

10
0
0

10

15

20

25

30

35

20
0.000

40

0.001

True strain (%)

0.010

-1

0.100

1.5
1.000

Strain rate (s )

Fig.5 Tensile tests of PA66

Fig.6 Yield stress and Youngs Modulus of PA66

90

2.6 10-4 s-1

90

1.62 10-3 s-1

85

80

5.44 10-3 s-1

70

1.76 10-2 s-1

60

6.5 10-2 s-1

50

1.32 10-1 s-1

4.0

Yield stress
Yield stress ST(MPa)

100

True stress (MPa)

0.010

40
30
20

Youngs modulus

3.5

80
75

3.0

70
2.5

65
60

2.0
55

10
0
0

Youngs modulus E (GPa)

True strain (%)

Fig.7 Tensile tests of PMMA

50
0.000

0.001

0.010

-1

0.100

Youngs modulus E (GPa)

1.5
1.000

Strain rate (s )

Fig.8 Yield stress and Youngs Modulus of PMMA

The analysis of the tensile curves shows that:


The tensile behavior of these polymers is strongly nonlinear. The tensile curves of PA66 and of
PMMA do not show an intrinsic yield drop whereas for PC a peak followed by a drop is observed.
The Youngs Modulus of PC and PA66 is independent of the strain rate while a dependency is
observed for PMMA (Fig. 4, Fig. 6 and Fig. 8). However, this dependency remains relatively not
significant and can be neglected in the constitutive equation devoted to the prediction of the
mechanical behavior of polymers.
Fig. 4, Fig.6 and Fig.8 show that the yield stress ST increases by increasing the strain rate.
According to the above observations, it can be assumed that at the early stage of deformation the
mechanical behavior is independent of strain rate while as soon as an inelastic strain appears the
response of the polymers depends strongly on the strain rate suggesting that elastic-viscoplastic
constitutive laws can be used to reproduce the response of such materials as proposed by [1].

Strain rate dependence of the yield stress. It is well known that the dependence of yield stress
of glassy and semicrystalline polymers on temperature and strain rate is a thermally activated rate
process following the formalism developed by Eyring. Bauwens et al. [11] and Ghorbel et al. [12]
have shown that the yield behavior of PC tested in tension at low strain rate  and at temperatures

Key Engineering Materials Vol. 446

67

from -20 to 145C can be expressed by the exponential form of the Eyring equation:
ST
Q

= A Ln2C  + .
T
RT

(4)

Where T is the absolute temperature, R is the universal gas constant. The parameters A, C and the
activation energy, Q of the molecular movement of glassy transition , are constants for the
material. Thus, the Eyring equation describing the yield stress can be simplified in the form of Eq.5:

ST = A + B Ln( ) .

(5)

Fig.9 Yield stress as a function of strain rate for PC, PMMA and PA66.
Fig. 9 shows, for PC, PA66 and PMMA, the effect of strain rate on the tensile yield stress. The
Eyring model provides an excellent agreement with the experimental data for different strain rates
for PC and PMMA while it does not fit well with the experimental data for PA66. This last result is
in accordance with observations made by Sahn et al. [13]. A detailed investigation of the yielding
behavior of over a wide range of strain rate and temperatures establishes that the yield stress
increases more rapidly at higher strain rates or lower temperatures than it does at lower strain rates
or higher temperatures. Hence it appears that the assumption according to which one activated rate
process is required to describe yielding is not valid. Since several authors proposed the use of
models based on several activation processes for the modeling of the yield stress. Among the
previous models given an acceptable prediction for the yield stress, the Ree-Eyring theory involving
two activation processes, and is successfully used for PC [14]. It is given by:

Q
ST
Q

= A Ln2C  + + A sinh 1 C  exp


.
T
kT

kT

(6)

Where ST is the yield stress under uniaxiale loading, T is the absolute temperature, k is the
Boltzmann constant,  is the strain rate, Qi(i= , ) are activation energies for the two processes
and , and finally Ai and Ci are activation parameters. The process refers to low strain rates and
high temperatures, whereas the process refers to high strain rates and low temperatures. Within
the past 10 years, it has been shown, for amorphous polymers and for semi-crystalline polymers that
the two Eyring processes acting in parallel for the description of the yield stress on a wide range of
temperatures and strain rates can be substituted with the Cooperative model. It is assumed that an
internal stress i exists therefore that the effective stress * is given by:
* = ST i .

(7)

68

Structural Analysis of Advanced Materials

This internal stress is a structural parameter, which depicts the arrangement of defects inherited
from past thermal history. Where k is the Boltzmann constant and  * is the characteristic strain
rate, we obtain the following form for Cooperatives model [14]:
1/ n

ST i 2k

= + sinh 1 .
T
T T
 *

(8)

Richeton et al. [15] have used and compared both models to predict compressive yield stresses
obtained for PC and PMMA. They show that the cooperative model is more valid for a wide range
of strain rates and temperatures than the Ree-Eyring model. The results obtained in this study are in
accordance with the literature [16,17,18]. Moreover it appears that Ree-Eyring model provides good
prediction of tensile yield of PA66 even at moderate strain rates (Fig. 10). Hence, it can be
envisaged to introduce this model in the constitutive equations dedicated to model the thermomechanical behavior of glassy and semi-crystalline polymers.

Fig.10 Evolution of the yield stress as a function of logarithm of strain rate.

Strain rate dependence of the self heating. The effect of strain rate on the room temperature
response of PMMA, PC and PA66 has been investigated. Tensile tests are performed at different
strain rates for true strain less than 30%. The specimen surface temperature is monitored and the
increase of temperature versus time is recorded in Fig. 11, Fig. 12 and Fig. 13.
True Stress for 1.76 10-2 s-1

True Stress for 3.4 10-2 s-1

Temperature for 2.6 10-3 s-1

Temperature for 1.76 10-2 s-1

Temperature for 3.4 10-2 s-1

0.9

90

0.8

80

0.7

70

0.6

60

0.5

50

0.4

40

T (C)

True Stress (MPa)

100

True Stress for 2.6 10-3 s-1

0.3

30
20

0.2

10

0.1

0
0

True Strain (%)

Fig.11 Self-heating of PMMA induced by tension loading

Key Engineering Materials Vol. 446

True Stress for 2.6 10-3 s-1

True Stress for 1.76 10-2 s-1

True Stress for 3.4 10-2 s-1

Temperature for 2.6 10-3 s-1

Temperature for 1.76 10-2 s-1

Temperature for 3.4 10-2 s-1

100

25

90

50
40

10

30
20

T (C)

60
15

True Stress for 1.76 10-2 s-1

True Stress for 3.4 10-2 s-1

Temperature for 2.6 10-3 s-1

Temperature for 1.76 10-2 s-1

Temperature for 3.4 10-2 s-1

3
2.5

80

20

70

True Stress (MPa)

True Stress (MPa)

80

True Stress for 2.6 10-3 s-1

70

60
50

1.5

40
1

30
20

10

T (C)

90

69

0.5

10

0
0

10

15

20

25

30

35

0
0

10

True Strain (%)

15

20

25

30

35

True Strain (%)

Fig.12 Self-heating of PC during tensile tests

Fig.13 Self-heating of PA66 during tensile tests

It appears that isothermal deformation conditions are verified for PMMA loaded under tension until
failure which occurs for true strain <6% (Fig. 11). For PC and PA66, it can be observed that as
soon as the polymers are inelastically strained, the temperature rises. Increasing the strain rate
induces a greater temperature rise due to plastic work. The temperature rise depends on the material
and contributes to the polymer softening. Hence, at the same strain rate and true inelastic
deformation, the strain-rate response of PC is affected by the self heating resulting from plastic
work in a manner which is more relevant by comparison to PA66. At large deformations (Fig. 14
and Fig. 15) competing effects will occur: thermal softening and strain hardening. For example, it
can be noticed, for a strain rate of 5.44 10-3 s-1, a rise in the maximum temperature exceeds 22C
before the failure of the specimen, in the case of PA66 (Fig. 14). It is interesting to note that this
phenomenon is more significant when the strain rate increases. Assuming that such tests are
isothermal is then debated. Hence, it appears that at the range of large deformation a thermomechanically coupled behavior can be considered while in the range of small deformation no self
heating due to elastic deformation is recorded.
140

45

True stress
Temperature

120

40

30
80

25

60

20

T (C)

True stress (MPa)

35
100

15
40
10
20

0
0

15

30

45

60

75

True strain (%)

Fig.14 Mechanical and self heating of PA66 for a strain rate of 5.44 10-3 s-1 under tension

70

Structural Analysis of Advanced Materials

160

30

A
140

True stress
Temperature

C
25
20

100
80

15

60

T (C)

True stress (MPa)

120

10

40
5

20
0

0
0

15

30

45

60

75

True strain (%)

Fig.15 Mechanical and self heating of PC for a strain rate of 5.44 10-3 s-1 under tension
For PC, further extension beyond the knee point (Fig. 15) leads to localization regions which appear
in the form of micro shear bands causing the softening of stress-strain response: this is the material
strain softening. Fig. 15 shows that as the PC is inelastically strained (region B), the temperature
rise leads to the thermal softening of the polymer. The thermal softening is important and is severe
in region B for both polymers. In region C strain hardening occurs due to the alignment of the
molecules and to the chain orientation with plastic stretch. Thermal softening and strain hardening
have competing effects on the response of polymers. Thermal softening seems to be predominant in
region B while strain hardening is the main phenomenon in region C. For PA66, the thermographic
images illustrate that the self heating related to plastic strain is homogeneous and becomes
heterogeneous as soon as strain hardening begins (Fig. 14). For PC, beyond the knee, the self
heating is localized in regions where generally shear bands appears and then is extended as inelastic
strain increases which corresponds to the process of neck stabilization.

Key Engineering Materials Vol. 446

PMMA

PC

6.5 10-2 s-1

1.76 10-2 s-1

5.44 10-3 s-1

1.62 10-3 s-1

2.6 10-4 s-1

PA66

71

Fig.16 Fractography for different strain rate of PA66, PMMA and PC

SEM micrographs. The effect of strain rates on failure modes of polymers subjected to tension
is analyzed by scanning electron microscopy SEM (Fig. 16). A change on the fracture patterns is
typically observed for strain rates higher than 1.76 10-2s-1 for PA66. The surface and morphology of
the failed specimens reveal ductile zones. This brittle-ductile transition can be explained by the self
heating observed with a significant manner in PA66. By increasing the strain rate the temperature
rises resulting in important softening of the polymer and therefore a change in failure mode from
brittle to ductile. The strain rate does not affect significantly the failure behavior of PC and PMMA.
Fracture surfaces (Fig. 16) exhibit brittle fracture when these polymers are subjected to tension.
Conclusions
It has been established in this investigation that the Ree-Eyring theory can be successfully used to
describe the tensile yielding behavior of thermoplastics even at moderate strain rates. The Youngs
Modulus of PC and PA66 is independent of the strain rate while a dependency is observed for
PMMA. A self heating resulting from plastic deformation appears at tensile stress higher than the
yield stress. An important thermal softening is observed after yielding for PC and PA66 which is
magnified by increasing the strain rate. For PMMA the tensile test remains isothermal. Competing
effects of thermal softening and strain hardening are observed and analyzed. In the light of these
results, it appears that thermo-mechanical coupling effects must be introduced to model the
mechanical behavior of polymers at large deformations.

72

Structural Analysis of Advanced Materials

References
[1] E. Ghorbel: Int J of Plast Vol. 24 (2008), p. 2032
[2] Y. Tomita: Int J of Plast Vol. 42 (2000) , p. 1455
[3] I. Hadriche, N. Masmoudi, E. Ghorbel and F. El Halouani: Experimental characterisation of
the deformation behaviour of thermoplastic polymers under torsion, Proceedings of the 4th Int.
Conference on Advances in Mechanical Engineering and Mechanics, MATE21 (2008).
[4] E.M. Arruda, M.C. Moyce and R. Jayachandran: Mecha. of Mater. Vol.19 (1995), p. 193
[5] Z. Li, J. Lambros: Int. J. of Solids and Struct. Vol. 38 (2001), p. 3549
[6] D. Rittel: Mecha. of Mater. Vol. 31 (1999), p. 131
[7] M. Clark, D.Mc Cann and M. Forde: NDT&E Int. Vol. 36 No.4 (2003), p. 265
[8] L. Li, B. Li and F. Tang: Eur. Polyme. J. Vol. 43 (2007), p. 2604
[9] X. Zhang, Y. Li, G. Lv, Y. Zuo and Y. Mu: Polyme. Degrad. and Stab. Vol. 91 (2006) , p. 1202
[10] C. G'Sell, J.M. Hiver and A. Dahoun: Int. J. of Solids and Struct. Vol. 39 Issues 13-14 , JuneJuly (2002) , p. 3857
[11] C. Bauwens-Crowet and J. C. Bauwens: J. of Mater. Sci. Vol. 14 (1979), p. 1817
[12] I. Ghorbel, F. Thominette, P. Spiteri and J. Verdu: J. App. Polyme. Sci. Vol.55 (1995), p. 163
[13] G.F. Shan, W. Yang, M. Yang, B. Xie, J. Feng and Q. Fu: Polyme. Vol. 48 (2007), p. 2958
[14] F. Rietsch and B. Bouette: Eur. Polym. J. Vol. 26 No.10 (1990), p. 1071
[15] J. Richeton, S. Ahzi, L. Daridon and Y. Rmond: Polyme. Vol. 46 (2005), p. 6035
[16] S. Deschanel, B.P. Greviskes, K. Bertoldi, S.S. Sarva, W. Chen, S.L. Samuels, R.E. Cohen and
M.C. Boyce: Polyme. Vol. 50 (2009) , p. 227
[17] F. Povolo, G. Schwartz and . B. Hermida: J. of Polyme. Sci. Part A: Polymer Chemistry Vol.
34 Issue 7 (1998), p. 1257
[18] C.M. Chalat, G. Michel-Amadry, P.J. Halley and R.W. Truss: Carbohydr. Polyme. Vol. 74
(2008), p. 366
[19]

Key Engineering Materials Vol. 446 (2010) pp 73-82


(2010) Trans Tech Publications, Switzerland
doi:10.4028/www.scientific.net/KEM.446.73

Dynamic Response of Symmetric and Asymmetric E-glass /epoxy


Laminates at High Strain Rates
M. Tarfaoui1,a, S. Choukri2,b and A. Neme1,c
1

ENSIETA, LBMS, Mechanics of Naval and Offshore Structures Department, Brest, France
2

Ecole Mohamadia des Ingnieurs, Rabat, Morroco

tarfaomo@ensieta.fr, bchoukri@emi.ac.ma, cnemeal@ensieta.fr

Keywords: Dynamic compression, SHPB, glass/epoxy composites, dynamic response, fibre


orientation

Abstract. The mechanical properties of E-glass/epoxy composite at high strain rates are important
in evaluating this kind of composite under dynamic and impulsive loading. The in-plane and out-ofplane compressive properties at strain rates from 300 to 2500 s-1 were tested with split Hopkinson
pressure bar. Samples were tested in the thickness as well as in-plane direction for seven fibre
orientations: 0, 20, 30, 45, 60, 70 and 90. The kinetics of damage and the failure
modes were identified using a high-speed photography, infrared camera, optical techniques and a
scanning electron microscope. Results of the study were analyzed in terms of maximum stress,
Strain at maximum stress, failure modes, damage history and fibres orientation effects. From the
experimental data, the stress-strain curves, compressive stiffness, and compressive strain of the
composite are rate-sensitive in in-plane and out-of-plane compressive directions. The failure and
damage mechanisms are implicitly related to the rise in temperature during static and dynamic
compression.
Introduction
Recently applications have exposed polymer matrix composite materials to very high strain
rate loading conditions, requiring an ability to understand and predict the material behaviour under
these extreme conditions. Many composite aircraft structures such as fuselage and wing skins,
engine nacelles, and fan blades are situated such that impacts at high strain rates are a realistic
threat. The choice of composite materials as a substitute for metallic materials in high technological
applications is becoming more pronounced especially due to the great weight savings these
materials offer. In many of these practical situations, the structures are subjected to high impact
loads like impact, explosions or blast effect. Material and structural response vary significantly
under impact loading conditions as compared to static loading. The mechanical characteristics of
these materials are well known for static loading, they are likely to evolve with the strain rate [1-6].
The design of structures is based almost solely on tabulated material data, usually in the
form of stress-strain diagram. Optimal designs require accurate and complete tables for materials
tested under a variety of conditions. The majority of available tabulated material properties were
gathered by deforming the specimen very slowly. It has been known for years that materials are
stronger at higher rates of strain, such as caused by an impact and blast effect. The behaviour of
structures subject to impact has been of interest to many scientists for purposes of design as well as
developing constitutive models of the materials tested. The study of the composite materials
behaviour at high strain rates is still relatively new and reliable data on strain rate effects is very
scarce. Although the problem of obtaining reliable data is accentuated by difficulties encountered in
the design and conduct of impact tests on composites [7], the increasing importance and use of
composites in aeronautic applications ensures continued interest in overcoming such problems.
In this study, specimens of glass/epoxy composite were subjected to a static and dynamic
compression loading. Samples were tested in the thickness as well as in-plane direction. The fibre

74

Structural Analysis of Advanced Materials

orientations of the samples were 0, 20, 30, 45, 60, 70 and 90. Stress-strain curves at
increasing strain rates were obtained and discussed for different cases.
Static tests
Static tests were performed on a standard test machine connected to an acquisition bay with
3 measurement channels. Two are reserved for the biaxial gauge assembled, an entry measures the
longitudinal deflection and the other entry measures the transverse deformation with a scale of
500mV/V correspondent to 500mm/m. The third entry is reserved for the analogue measurement of
the efforts with a scale of 10V correspondent to 100kN with a crushing speed 0.5mm/mn. Two
types of static compression tests are used to obtain the elastic properties of the lamina. In-plane
loading (IP), parallel to the plane of the plies; plane (1,2), given by figure 1a and out-of-plane
loading (OP), according to the thickness; direction 3, illustrated by figure 1b. Table 1 shows the
lamina elastic values resulting from experimental work.

(a)

(b)

Fig. 1. Sample loadings


Characteristics
Experimental

E1 (MPa)

E2 (MPa)

G12 (MPa)

12

46217

16086

2224

0.28

Table 1. Elastic properties


Dynamic tests
The split Hopkinson bar test is the most commonly used method for determining material
properties at high strain rates, Figure 2. This technique of characterization, based on the response of
a material to wave propagation for high strain rate from 100s-1 to 5000s-1, was improved by Kolsky
[8].

(a)

(b)

Fig. 2. Compressive Hopkinson bar

Key Engineering Materials Vol. 446

75

The dynamic tests carried out on the test set-up require a phase of preparation that consists of
polishing the sample to guarantee a parallelism between the faces of supports with the bars. The
dynamic compression test consists of placing a sample between two bars with a high elastic limit,
called input bar and output bar. These bars are correctly aligned and are able to slide freely on their
support. The composite specimen is not attached to the bar in order to prevent the perturbations of
measurements due to additional interfaces.
Stress, Strain and strain rate in the specimen

Dynamic loading of the composite panels is provided by a SHPB for compression tests. The
longitudinal impact load F0 of the striker generates the uniaxial stress pulse transmitted to the input
bar as:
(t ) =

F0
= (c 0 )Vp (t )
Sb

(1)

is the velocity of the wave pulse, bars of Sb is the bars cross-section area, Vp (t ) the particle
velocity, and the density of the striker material. Accurate measurements of the wave pulse and
particle velocity in dynamic impact studies are important since the wave pulse is in fact the
propagation of the disturbance or vibration of the particles. The amplitude of the incident wave
pulse depends on the impact velocity (a function of the applied air pressure) and material properties
of the striker.
c0

2u
t 2

c 02

2u
z 2

=0

c 0 = (E 0 )

1
2,

(2)
E0 is the Young modulus and u (z, t ) the longitudinal displacement. The solution

yields:
u
u
=

z
c 0 t

(3)

for wave propagation in the positive direction, = 1 for the wave propagation in the negative
direction. It leads to solutions: u = u(z c 0 t ) for = 1 and u = u(z + c 0 t ) for = 1
The particle velocity is as follow:
= 1

V p ( z, t ) =

u( z, t )
t

Velocities with the interfaces specimen/bars are V1 =

(4)
dZ 1
and V2 = dZ2 .
dt
dt

where Z1 = z1 + u(z1, t ) and Z 2 = z 2 + u(z 2 , t ) .


With the deformation, the length of the specimen is: L = Z 2 Z1 = u(z 2 , t ) u(z1, t ) + L 0
where L 0 = z 2 z1 . Then:
L = L0 +

0 ( V 2 V1 )dt

(5)

Equations (6) and (7) give respectively, the strain, the stress and the strain rate in the sample.
i , r , t being respectively the incidental and reflected deformations pulse at the input

76

Structural Analysis of Advanced Materials

bar/specimen interface and that transmitted by the interface specimen/output bar interface, Figure
4b. Those are determined by transport formulas of the signals of the gauges towards the interfaces
specimen/bars. This transport takes into account the correction of the dispersive character of the
propagation, of the deformations pulse in the bars, by optimization of the decomposition
coefficients, in series of Fourier, of the signals. The particles in the input bar will propagate to the
right at a relative velocity of Vb in the longitudinal direction of the wave pulse. With a specimen of
cross-sectional area S 0 , sandwiched between the input and output bars, equilibrium at the interfaces
is satisfied by the continuities of forces and velocities at the interfaces such that the particle
velocities, sample stress, strain and strain rate for a thin sample are derived as:

Vs = Vt (Vi + Vr )

with

Vi = c 0 i

Vr = c 0 r
V = c
0 t
t

Vs = c 0 [ t + r i ] = 2c 0 r

(6)

Assuming equilibrium in short specimen, t = i + r ; the strain, strain rate and stress is:
s =

2c 0
L0

r (t )dt
0

d s
2c 0
=
r (t )
dt
L0
s (t ) =

(7)

Sb
E 0 t (t )
S0

In order to express the signals recorded at the input bar/sample and sample/output bar interfaces,
it is necessary to propagate the waves in the material. The case can be treated as one-dimension. In
this situation, all the waves are propagated at the same velocity C0 in the bar. The non-dispersive
property of the propagation in the bars is roughly verified in experiment. However it is more
realistic to consider the problem as bi-dimensional [9]. In this case the speed of each wave depends
on its frequency w and the radius R of the bar. This theory works well for predicting the dispersive
nature of waves. A typical set of incident, reflected, and transmitted signals resulting from gauges A
and B recorded by the digital oscilloscope from a pulse-shaped SHPB experiment on the composite
at the strain rate of 336 s-1 for 0 IP test is shown in figure 3.

Fig. 3. Typical incident, reflected and transmitted pulse from the SHPB
The strain rate evolution is sensitive to the entry pressure in the chamber of compressed air P
(impact pressure of the striker on the input bar), the loading direction and the sample lay-up (angle

Key Engineering Materials Vol. 446

77

), Figure 4. We will analyze that in detail in the following sections. For the results analysis and
exploitation, we will be interested in the evolution of:
- Initial dynamic modulus: E dynamic witch is evaluated for the linear part of = f ( ) curves
for different  ,
- Maximum stress: Max ,
- Strain at maximum stress: Max ,

- Total strain: Max .


In a concern of simplifying the graphic representations, the analyzed parameters: E dynamic , Max ,
Max and Max will be given according to the impact pressure P and the fibre orientation for IP

and OP tests.
In plane tests

It should be noted that the representative axis of the strain rate effect corresponds to the axis of
the impact pressure. The curves in blue indicate the experimental curves, Figure 4. Let us notice the
good physical coherence of the evolution of surface suggested since the strain rate is increasing as
the impact pressure increases for a given orientation.

Fig. 4. Strain rate evolution versus fibre orientation and impact pressure
Figure 5 shows the (a) undamaged and (b) damaged sample with the corresponding curves. The
damage can be located by the presence of a second peak on the  = f (t ) curve. To the formation of
this peak a fall of stress corresponds. This fall is primarily due to the formation of the letter V and
delamination on the external layers.
Figure 6 shows the dynamic behaviour of [0]40, [20]20, [30]20, [45]20, [60]20, [70]20, [90]40
laminates for the 9 impact pressure. One can note that [20]20 - [30]20 and [60]20 - [70]20 have
the similar tendencies. For [90]40 laminate, one note a brittle behaviour witch is controlled by matrix
failure. The nonlinear part of the = f ( ) curves represents different damaging modes. The
analysis of these curves, translating the dynamic behaviour for the various orientations, is
summarized in figure7.
The failure modes preserved in the specimens after mechanical loading thus correspond to the
well-defined loading history. According to the results stated previously, one can conclude that the
damaging mode depends on the fibres orientation and the loading direction.

Structural Analysis of Advanced Materials

300

100

200

50

100

0
0.1

0.2

0.3

0.4

0.5

0.6

(a) 0 IP test, P=0.5bar

540

450

440

350

340

250

240

150

140

50
-50 0

740
640

550

40

0.7
-100

Time (ms)

840

Strain rate

650

-1

400

Strain rate (s )

150

-50

Stress

750

Strain rate

200

Stress (MPa)

500

Stress

Stress (MPa)

250

Strain rate (s-1)

78

0.1

0.2

0.3

0.4

0.5

0.6 -60

Time (ms)

(b) 0 IP test, P=1.4bar

Fig. 5. Maximum Stress - Strain curves evolution versus strain rate

20

45

90

Fig. 6. Stress - Strain curves evolution versus strain rate


Figure 7 shows the detailed trends of dependencies of E dynamic , Max , Max , and Max on both
fibre orientation and impact pressure. These trends agree well with the experimental results of fibre
orientation effects and strain rate effects. For in plane tests, there exist a transitional impact pressure
(1 1.2 bar) and a transitional fibre orientation in the trends, reflecting the dependencies on strain
rate and fibre orientation observed in experiments. Furthermore, Figure 7c shows that the maximum
failure strain is dominated by the matrix plasticity for low strain rate and by the damage mechanism

Key Engineering Materials Vol. 446

79

for high strain rate while the initial modulus of elasticity Figure 7b is independent on deformationinduced damage in specimen. However, the maximum failure strength in specimen is dependent on
all the material constants. For an angle ranging between 0 and 45 one had an increase in the
maximum stress and effective rigidity with the increase in the impact pressure (the strain rate). And
for an angle ranging between 45 and 90, the evolution law is reversed.
For the various orientations, the laminate stiffness ( E dynamic ), the maximum stress ( Max ) and
the corresponding strain ( Max ) increase with the increase of P until a pressure threshold from
which the tendency is reversed; i.e. they decrease with the increase of  . Concerning Max , it
increase with the increase of  and fibre orientation .

(a) Maximum stress

(b) Edynamic

(c) Maximum strain

(d) Strain at Max. stress

Fig. 7. Stress - E dynamic , Max , Max and Max versus P and


Out of plane dynamic tests

Figure 8 gives the experimental curves in blue and the envelope surface. The evolution of 
follows two phases: a first phase ( 0.5 P 1.2 bar ) where the increase is fast and the second phase
where the variation is less marked. The stacking lay up has an effect on the strain rate evolution.
Indeed, one notices on blue curves that the evolution of  is relatively similar for samples 20/70,
30/60 and 45/45 whereas for samples 0/90 it is different.
Figure 9 gives the evolution of the strain rate of [0/90], [20/70], [30/60] and [45/45] samples for
9 impact pressures. The appearance of a second peak is around 0.9bar for [0/90] and 1.4bar for the
other orientation. The dynamic behaviour for OP tests is dominated by compressive properties of
the polymeric matrix and the damage is created for high impact pressure. The fibres orientation
affects the initiation and propagation of damage.

Structural Analysis of Advanced Materials

Strain rate (s-1)

80

Orien

tation

()

P
act
Imp

ress

u re

)
(bar

Fig. 8. Strain rate evolution versus fibre orientation and impact pressure
For undamaging tests, the fall of strain rate passes by negative values, which correspond to the
sample spring back. The glass/epoxy composite materials present a high strength on compressive
dynamic load for OP tests.

0/90

30/60

20/70

45/45

Fig. 9. Strain rate evolution versus impact pressure


In opposite to the IP tests, there is always an increase in E dynamic , Max and Max with the
increase in the impact pressure for OP tests. There is not threshold effect for OP loading, Figure 10.

81

Edynamic (GPa)

Max. Stress (MPa)

Key Engineering Materials Vol. 446

O ri

e nt

a t io

n (

Im

pa

tP

s
r es

e(
ur

ba

r)

O rie

n ta

( )

Im p

Pr
act

ess

ur e

(b a

r)

(b) Edynamic

Max. Strain (%)

Strain for Max Stress (%)

(a) Maximum Stress

tion

Or

ie n

tat
i

on

()

Im

pa

ct

Pr

( c ) Maximum strain

u
ss

re

(b a

r)

Imp

act

Pr e
ssu

re
(ba
r)

(
ion
tat
ien
r
O

(d) Strain at Max. stress

Fig. 10. E dynamic , Max , Max and Max versus P and , OP tests
Conclusions
In this study, the effects of fibre orientation and strain rates in a glass/epoxy composite material
were examined. Moreover, a family of compressive stressstrain curves at dynamic strain rates at a
series of fibre orientation was determined. All of the stressstrain curves had similar shapes: an
initial linear elastic portion followed by a nonlinear behaviour until failure. The stressstrain curves
of the composite materials show that the material is strongly sensitive to fibre orientation at the
same impact pressure: the initial modulus of elasticity, maximum failure stress, strain at maximum
stress and the maximum strain are all dependent on fibre orientation and strain rates. Within the
range of fibre orientation in this research, an approximate threshold impact pressure separated the
trends of dynamic parameters ( E dynamic , Max , Max ) dependencies into two portions. When
impact pressure was below the threshold value, the stressstrain curves exhibited increasing initial
modulus of elasticity, maximum stress and strain at maximum stress. Conversely, when impact
pressure was above the threshold value, the stressstrain curves did not exhibit hardening, but
softening: both the modulus of elasticity and maximum stress decreased, but maximum strain
increased at the impact pressure ranging of 0.5bar to 1.6bar for all orientations.

82

Structural Analysis of Advanced Materials

References
[1] Z. Rong, W. Sun, Y. Zhang. Dynamic compression behavior of ultra-high performance cement
based composites. International Journal of Impact Engineering, Volume 37, Issue 5, 515-520,
(2010).
[2] M. Shokrieh, M Omidi. Compressive response of glassfiber reinforced polymeric composites
to increasing compressive strain rates. Composite Structures, Volume 89, Issue 4, 517-523
(2009).
[3] Y. Luo, L. Lv, B. Sun, Y. Qiu, B. Gu. Transverse impact behavior and energy absorption of
three-dimensional orthogonal hybrid woven composites. Composite Structures, Volume 81,
Issue 2, 202-209 (2007).
[4] M. V. Hosur, J. Alexander, U. K. Vaidya, S. Jeelani. High strain rate compression response of
carbon/epoxy laminate composites. Composite Structures, Volume 52, Issues 3-4, 405-417
(2001).
[5] Qinlu Yuan, Yulong Li, Hejun Li, Shuping Li, Lingjun Guo. Strain rate-dependent compressive
properties of C/C composites. Materials Science and Engineering: A, Volume 485, Issues 1-2,
25 June 2008, Pages 632-637.
[6] H. Zhao. Material behaviour characterization using SHPB techniques, tests and simulations.
Computers and Structures, 81, 1301-1310 (2003).
[7] J. Harding. The Effect of High Strain Rate on Material Properties. Materials at High Strain
Rate, ed. T. Z. Blazynski. Elsevier Applied Science, London & New York, 1987.
[8] H. Kolsky. An investigation of the mechanical properties of materials at very high rates of
loading. Proc. Phys. Soc. , Vol. 62B, 676-700 (1949).
[9] D. Bancroft. The velocity of longitudinal waves in cylindrical bars. Physical Review, Volume
59, N 59, 1941, 588-593.

Key Engineering Materials Vol. 446 (2010) pp 83-90


(2010) Trans Tech Publications, Switzerland
doi:10.4028/www.scientific.net/KEM.446.83

Influence of process and material parameters on impact response in


composite structure: methodology using Design of Experiments
C.Garnier1,2,a, S.Mistou1,b and O. Pantal1,c
1

Universit de Toulouse, INP/ENIT, LGP, 47 avenue d'Azereix, BP 1629,


F-65016 TARBES cedex, France

Daher Aerospace, Research & Development Department, 23 route de Tours, BP 17,


Saint Julien de Chedon, F-41401 Montrichard Cedex, France
a

christian.garnier@enit.fr, bsebastien.mistou@enit.fr, cpantale@enit.fr

Keywords: impact response, RTM process, LRI process, composite structure, design of
experiments.

Abstract. Even if the mechanical performances of composite materials give new perspectives for
the aircraft and space design, the variability of their behavior, linked to the presence of initial
microscopic defects or led in service, constitute however a still important brake in their
development. As regards particularly the response to fatigue loads or ageing, the behavior of these
materials is affected by several sources of uncertainties, notably on the nature of the physical
mechanisms of degradation, which are translated by a strong dispersion in life time.
In aerospace industry, low energy impact phenomenon is not well known concerning
composite materials and composite structures. Many manufacturers use important safety factors to
design structures. The aim of this work is to define the most predominant parameters which permit
a good response of damage using experiences plans. The differences of these parameters by using
Resin Transfer Molding (RTM) or Liquid Resin Infusion (LRI) process than prepreg one is also
studied in this work.
Introduction
Nowadays, in aerospace industry, composite materials are used but their behavior is not well
known. During their life in aerostructure, these materials are subjected to many different loads as
maintainability impacts, hail, birds, and others. These loads can be classified according to their
energies or strain speed, as shown on figure 1 and 2.

Figure 1: Classification of dynamics phenomena according to the sollicitation time

84

Structural Analysis of Advanced Materials

Figure 2: Classification of dynamics phenomena according to the strain speed

Under high-velocity dynamics loads or medium-velocity dynamics loads, the structure is affected
and many damages are visible. In this case, this structure is automatically replaced. In case of quasistatic loads, the surface of the structure is few affected but damage can occur in it. The detailed
inspection (DET) and General Visual Inspection (GVI) inspection do not reveal systematically
these defects.
One of the main problems is to know how the structure responds on impact in terms of damage
produced inside and residual strength after impact. The knowledge of the entire phenomenon which
acts in this response is a good way to control it. First, we have to define these phenomena and all
the parameters linked to them. Then, we have to understand how they act to choose the best
configuration. To understand all the effects, we have to make experiments using a special test
apparatus. For the low energy impacts, a drop weight machine is used, and for upper energies, a gas
gun is used. Furthermore, we decided to know how the sample, impacted or not impacted, behave
under fatigue loads. The aim is to understand how the damage will initiate for no impacted samples
and how it will propagate for others.

State of the art.


Parameters acting on impact response.
In its final report in 1998, the FAA [1] does a state of the art about parameters which can modify
the impact response of composite structures. In its report they classify these parameters in three
parts. One part of these reported to the impactor, another one of these reported to the target and a
third one for parameters reported to the material and its manufacturing process.
In 1992, Demuts and al. [2] compare 2 resin/epoxy systems: graphite / epoxy and graphite /
BMI. They note that using the resin BMI doesn't give any advantage concerning residual strength in
compression after impact.
In 1994, Hitchen and al. [3] study the influence of the lay up on carbon/epoxy samples impacted
by 7 J. They denote that the layup modifies the shape and the size of the delamination. It also
modifies the energy level to initiate delamination. Besides, the layup acts on the structure strength
before and after impact.
In 1995, Hinrich and al. [4] work on the impact in composite panels, made by RFI (Resin Film
Infused) process, under compression load. They denote that thinner is the sample, higher is the
damage area. They also note that the contact force is affected by the shape and the speed of the
impactor. The prediction of the contact force shall be done separating the kinetic energy in two
energies : elastic energy and Hertz contact energy.
In 1998, Fuoss and al. [5] decide to study the influence of the stacking sequence on the impact
resistance of composites laminates varying three parameters: interface angle, ply orientation relative
to a fixed axis and ply grouping. Concerning angle interface, they denote that increasing the number
of angle interface reduces the damage area. For the ply orientation to a fixed axis, they note that
changing the angle influences the impact response. Changing the stacking sequence changes the

Key Engineering Materials Vol. 446

85

impact behaviour but each configuration is specific. Modifying specimen geometry or boundary
conditions will change the impact response. It is noted that the ply grouping reduces the damage
resistance and so, increases the damage area.
In 1998, Ambur and Starnes [6] work on plane and curvative samples. They show that the radius
of curvature doesn't act on the contact force but acts on the compression after impact residual
strength.
In 1999, Tai and al. [7] work on the influence of the thickness of composite on the response of
impact and on its fatigue behaviour. They predict that a linear relation exists between the stress
level and the life duration under fatigue load.
In 2002, Carti et al. [8] study the influence of resin and fiber properties. They use six carbon /
epoxy composites: four different resins and two different fibers. Impact tests were done using a
drop weight machine with a speed which varies between 1 and 3 m.s-1. They denote that, for CAI
tests, resin toughness is the most significant parameter. Changing type of fiber does not alter or few
alter compression after impact residual strength. Moreover, the type of resin acts on the level of
damage initiation load after impact.
In 2004, Mitrevski and al. [9] want to determine the influence of the impactor shape using three
different shapes: flat, hemispheric and conic and two impact energies. They also use two different
lay up for carbon/epoxy prepreg samples. They conclude that coarser is the indentation, higher is
the absorbed energy by the coupon. The hemispheric impactor increases the peak contact force,
decreases the contact duration and creates a smaller indentation.
Currently, many approaches around materials are made. Instead of using only thermosetting
resin, two components resins are used: thermosetting one with thermoplastic interleaving or
particulate thermoplastic polymer are locally located in the resin [10, 11, 12, 13].
Many parameters act on the response of impact of composite structures. In most of cases, authors
use prepreg to make their samples. To confirm that the same parameters act on RTM, as shown on
figure 3, or LRI processes, as shown on figure 4, a decision to make a design of experiments with
many parameters was done.

Figure 3: Principle of Resin Transfer Molding

Figure 4: Sample during Liquid Resin Infusion (LRI) process

86

Structural Analysis of Advanced Materials

Fatigue
Many investigations were carried out from fatigue loading on composite structure and their
behaviour [14-20]. Most of these investigations show that fatigue behaviour of composite structure
is a complex phenomenon. Many factors act on this behaviour, such as the loading level, the
frequency, etc. However, few studies take into account the worsening character of compression
during fatigue loading. Most of the pre-cited works deal with tension-tension fatigue loading. But,
when a fatigue cycle takes into account compression, local or global buckling shall appear [21] and
initiate in a premature way damages.
Concerning the damage development, many properties of the materials act on it. The
inhomogeneous microstructure and the variable quality of many commercial composite materials
[22], where there are small disparate porosities or cavities on it, the stiffness or strength of the
composite lead to complex damage development and to different modes of failure like
delaminations, disbondings, fibre breakage, etc [23].
About the experimental procedure, Bathias, Behesty and Tai [24, 25, 7] consider that the most
penalizing cyclic damage is fatigue under traction/compression. Moreover, they found that the
component in compression of a loading in traction/compression must be important to show its effect
on the life reduction after impact. Consequently they use a loading ratio of -1 between the minimum
and maximum stress.
In their works, Minnestyan, Philippidis and Post [26- 28] make a non exhaustive list of models
which predict residual strength. Existing models as Sarkani et al., Broutman and Sahu, Schaff and
Davidson, Hahn and Kim, Rotem, Yang et al.,interaction model, Sendeckyj, modified Broutman
and Sahu model, modified Hahn and Kim, modified interaction model and non linear model were
compared. Different types of coupons were tested and compared with numerical models. They note
that most of them are not completely reliable concerning residual strength after fatigue. Some
models correlate well with the experiment curves of life at the beginning while others correlate near
fracture. The experimental results show that the dispersion of residual static strength increases with
the accumulation of the damage during the increase in the number of endured cycles. Most of the
cited models follow this trend.
All the precited models and comparisons of them with experiments were carried out from
prepreg Fiber Reinforced Plactics. None of them were compared with RTM-FRP or LRI-FRP.

Testing methodology
In order to do design of experiments, it has been decided to classify the parameters in 3 classes:
process, materials and testing parameters. The first class includes the entire parameters which act
during the manufacturing of the structure. The temperature of the resin, the temperature of the
mold, the room temperature, the pressure, the time of polymerization, the time to reach the
temperature of polymerization can be cited. All the data which can change or variable during the
manufacturing can be integrated on the design of experiments. The second class involves the nature
of the fibre, the tow, the density and the thickness of a ply, the basis weight, the different
mechanical and chemical treatment of the fiber, the stacking sequence, the nature of the resin, the
number of components of the resin. And for the third class, the weight and the shape of the
impactor, the boundary conditions of the sample, the impact energy shall be parameters. The aim of
the design of experiments (DoE) is to know the most important parameters acting on the impact
response for two different processes: LRI (Liquid Resin Infusion) and RTM (Resin Transfer
Molding). Two considerations in the building of the DoE are the fact that some parameters of these
processes are different and that the direction of the resin injection is different. So it has been
decided to build two DoE: one for LRI process and other one for the RTM one, keeping the same

Key Engineering Materials Vol. 446

87

parameters for the materials (tow, basis weight, ...) and for the impact testing ( impactor weight,
radius of impactor tap, ...) .
As seen before, many parameters should be used in our Design of Experiments. In order to
predict the influence of each parameter, the coupling between them and the curvature of the DoE, it
has been decided to reduce the number of factors and so, the number of specimen to produce. Here
is the general response of the DoE:
Y B0 Bi X i Bij X i X j Bii X i2

(Eq.1)

where Xi is the parameter i, Bi the influence of the parameter Xi, Bij the interaction between Xi
and Xj, Bii the curvature of the plan and Y is the response.
First, the way to reduce the high number of parameters for the material class has to be chosen.
Nowadays, most of composite aerostructures are made with epoxy resin and carbon fiber.
Moreover, only few resins are qualified for aircraft. Deciding to use only epoxy resin seems to be a
good choice to reduce the number of parameters. Concerning fibers, we decided to choose carbon.
Secondly, concerning intrinsic fiber parameter, the study shall be done varying parameters which
are easy to change. Some parameters shall be changed in concert with suppliers on basis of the
existing fibers. Moreover these parameters must be, as far as possible, independent from one to
other. So we choose these four parameters: tow, basis weight, type of fiber, type of fabric. One
more chosen parameter for the experiments plan is the powdering of the fabric. The aim of this
parameter is to predict the level of joining between the powder and the resin for LRI and RTM
process. As we note in [Fuos98], the stacking sequence works on the impact response of composite
structure. So it will be added in the design of experiments.
On the implementation of composite materials, we will take as the basic the polimerization cycle
of epoxy resin. This cycle is composed of 3 stages: temperature rise, polymerization step and
decreasing temperature. For the first one and the third one, the speed will be introduced as
parameter. But for the polymerization step, we add time and temperature.
The main difference between the two processes is the pressure. LRI used a vacuum pressure
while RTM use pressure. Each one will be a parameter for the concerned process.
For each parameter, a field variation will be chosen using for each one 2 or 3 levels.
To know how the structure reacts on impact, we have to make experiments. Low energy impact
testing will be made on samples. We choose energy which will approximate the Barely Visible
Impact Damage (BVID). So we will impact the samples using a drop weight impact tester (fig. 1)
able to keep the drop weight after the first impact (brake). According to the AITM 1.0010, the
measurements of samples are 100 mm * 150 mm with a thickness of 4 mm. Concerning the
impactor, we will use one, made in steel, with hemispheric tap which weight 1 to 3 kg. This testing
will take place on new samples and also on old samples, ageing under fatigue loading, impacted or
not impacted.
To define the influence of each parameter, we have to measure responses of the experiments plans.
We decided to measure the health of the samples before impact, three-dimension geometry of the
damage produced by impact, and residual mechanical properties.

88

Structural Analysis of Advanced Materials

Figure 5: Drop weight test machine

Conclusions and perspectives


The impact phenomenon is unknown on composite structure and many parameters act on it. This
paper has 3 interests.
Firstly, it establishes a link between impact response phenomenon to the manufacturing of the
process. It develops all the parameters acting on the manufacturing that act on impact response on
composite structures.
Secondly, a new approach for impact response is carried out. A design of experiments is realized in
order to analyze in a statistical way the influence of all the precited parameters.
Thirdly, it shows that no works on impacted composite were carried out during fatigue loading.
The high degree of generality is presented in order to prepare the experiments. This work, linked to
the experimental results, will clearly predict the impact response due to manufacturing parameters.
A mathematical and statistical predicting model is carried out of this method.

Key Engineering Materials Vol. 446

89

References
[1] H.P. Kan, Final report DOT/FAA/AR-97/79 (1998).
[2] E. Demuts, R. S. Sandhu, J.A. Daniels, report N DOT/FAA/CT-92-25 (1992) 1097 - 1104.
[3] S.A. Hitchen and R.M. Kemp, Comp. 26 (1994) 207-214.
[4] GS. Hinrich, V. Chen, D. Jegley, L.C. Dickinson and K. Edward, NASA conference
publication 3294 (1995).
[5] E. Fuoss, P.V.Straznicky, C. Poon, Comp. Struc. 41 ( 1998 ) 67 -77.
[6] D.R. Ambur and J.H. Starnes Jr,, In structures, structural dynamics, and material conference
(1998).
[7] N.H. Tai, C.C.M. Ma, J.M. Lin and G.Y. Wu, Comp. Sci. and Tech.59 (1999) 1753-1762. [4]
GS. Hinrich, V. Chen, D. Jegley, L.C. Dickinson and K. Edward, NASA conference
publication 3294 (1995).
[8] D.D.R. Carti, P.E. Irving, Comp. part A 33 (2002) 483 - 493.
[9] T. Mitrevski, I.H. Marshal, R. Thomson, R. Jones and B. Whittingham,, comp. struct. (2004).
[10] G. Caprino, P. Iaccarino and A. Lamboglia, Composite structures 88 (2009) 360-366.
[11] A. Duarte, I. Herszberg and R. Patton, Composite structures 47 (1999) 753-758.
[12] J. Masters, Key Eng. Mater. 37 (1989) 317-348.
[13] H. Kishi, M. Kuwata, S. Matsuda, T. Asami and A. Murakami, Composites Science and
technology 64 (2004) 2517-2523.
[14] Reifsnider, KL, Lesko J, Case S. IUTAM - Symposium on Mechanics of Composite Material
(1983); 399-420.
[15] Talreja R, Yalvac S, Yats LD, Wetters DG. Transverse cracking and stiffness reduction in
cross-ply laminates of different matrix toughness. J of comp mat 1992; - 26(11): 1644-1663.
[16] Gathercole N, Reiter H, Adam T, Harris B. Life prediction for fatigue of T800/5245 carbonfibre composite : I. Constant-amplitude loading. Int J of Fatigue 1994; 16(8):523-532.
[17] Mao H, Mahadevan S. Fatigue damage modelling of composite materials. Comp Struct 2002;
58:405410.
[18] Dzenis YA. Cycle-based analysis of damage and failure in advanced composites under fatigue:
1. Experimental observation of damage development within loading cycles. Int J of Fatigue
2003;25(6):499-510.
[19] Gros XE, Bousigue J, Takahashi K. NDT data fusion at pixel level. NDT & E Int
1999;32(5):283-292.
[20] Vary A. Acousto-Ultrasonics, In: Nondestructive testing of fiber reinforced plastics.
Composites 1990;2:1-54.
[21] Lachaud F., thse de doctorat, Universit Paul Sabatier Toulouse (1997)
[22] Smith T.R. and Owen M.J, proceedings of the sixth International reinforced plastics Congress
(1968).
[23] B. Harris, Fatigue in Composites (2003).
[24] C. Bathias, Eng. Fract. Mech. 40 (1991) 757 -783.
[25] M.H. Behesty and B. Harris, In : ICFC 1 (1997) 365-372.

90

Structural Analysis of Advanced Materials

[26] L. Minnetyan, Computational


NASA/CR_2005-213573 ( 2005).

simulation

of

composite

structural

[27] T.P. Philippidis and V.A. Passipoularidis, Int. J. Fat. 29 (2007) 2104-2116.
[28] N.L. Post, S.W. Case, J.J. Lesko, Int. J. Fatigue ( 2008 ) 2064 - 2086.

fatigue,

report

Key Engineering Materials Vol. 446 (2010) pp 91-99


(2010) Trans Tech Publications, Switzerland
doi:10.4028/www.scientific.net/KEM.446.91

Approximate solution of the structural problems using


probabilistic transformation
S. Ouhimmou1, a, A. El Hami2, b, R. Ellaia1, c and M. Tkiouat1, d
1

Mohammed V University - Engineering Mohammedia School, LERMA


BP: 765, Ibn Sina Avenue, Agdal, Rabat, Morocco.

Laboratory of Mechanics of Rouen, National Institute for Applied Sciences

BP 08, university avenue 76801, St Etienne du Rouvray Cedex, Rouen, France.


a

ouhimmousiham@yahoo.fr, b aelhami@insa-rouen.fr, c ellaia@emi.ac.ma, d tkiouat@emi.ac.ma

Keywords: Reliability Methods, Probabilistic Transformation Method, Finite Element Analysis,


Finite Element software, Matlab, structures.
Abstract. The aim of this paper is to present a new methodology for the evaluation of the statistical
proprieties of the response of structures, based on The Finite Element Analysis (FEA) coupled with
the Probabilistic Transformation Method (PTM). Uncertainty modelling with random variables
motivates the adoption of advanced PTM for reliability analysis to solve problems of mechanical
systems. The PTM is readily applicable in the case where the expression between input and output
of structures are available in explicit analytical form. However, the situation is much more involved
when it is necessary to perform the evaluation of implicit expression between input and output of
structures through numerical models. For this we propose technique that combines the FEA
software, and the PTM program to evaluate the Probability Density Function (PDF) of the response
where the expression between input and output of structures is implicit. This technique is based on
the numerical simulations of the FEA and the PTM by making an interface between Finite Element
software and Matlab. Some problems of structures are treated in order to demonstrate the
applicability of the proposed technique.
1- Introduction
The uncertainty and reliability analysis in structural engineering have been vibrant topics of
research for several decades ([1], [3], [7]).
Considering the properties of the structural model realistically it is necessary to take into account
some uncertainty. This uncertainty can be conveniently described in terms of probability measures,
such as distribution functions. It is a major goal of reliability methods to relate the uncertainties of
the input variables to the uncertainty of the structural performance.
A fundamental problem in structural reliability analysis is the computation of the probability
integral:
Pf = Pr ob[G ( X ) 0] = f ( X )dX
G ( X )0
(1)
T
X = [ X 1 ,.., X n ] in which T is the transpose, is a vector of random variables representing the
Where
uncertain parameters of considered structure, f ( X ) is the probability density function of X , G ( X )
is the Limit State Function defined such that G ( X ) 0 the domain of integration denoted the
P
failure set, and f is the probability of failure. The difficulty of computing this integration led to
development of various methods of reliability analysis such as Monte Carlo simulation that is the
reference method, FORM and SORM ([4], [10]), and the PTM [12].

92

Structural Analysis of Advanced Materials

The PTM is an efficient reliability method to solve problems of mechanical systems with uncertain
parameters. The advantage of this method is finding the ''exact'' expression of the Probability
Density Function PDF of the solution when the PDF of the input is known. In many cases, the
structural response cannot be expressed explicitly and some numerical calculations are necessary.
The coupling of Finite Element Analysis (FEA) software with The Probabilistic Transformation
Method (PTM) program is therefore necessary.
In this paper, a proposed technique: the Finite Element Analysis (FEA) software coupled with the
Probabilistic Transformation Method (PTM) program is applied in order to evaluate numerically the
probabilistic and statistical characteristics of the response of stochastic mechanical system. It
involves four main steps: (1) sampling on input random variables, (2) using Finite Element Analysis
(FEA) software to have the response variable of system (3) estimating the Probabilistic Density
Function (PDF) of this response variable using the Probabilistic Transformation Method PTM
program, (4) concluding the probability of failure and reliability of systems. To demonstrate the
efficiency of the proposed technique, we have carried out different applications to cover several
structural problems.
otation
PTM: Probabilistic Transformation Method.
FE: Finite Element.
PDF: Probability Density Function.
FEA: Finite Element Analysis.
FEACPTM: Finite Element Analysis Combined to Probabilistic Transformation Method.
FORM: First Order Reliability Method.
SORM: Second Order Reliability Method.
G (.) : Limit State Function.
Pf : Probability of failure.

2- Method of Analysis
2.1 Finite Element Analysis

The Finite Element Analysis is the standard tool for certain classes of partial differential equations
arising in various fields of engineering and in particular for those arising in solid mechanics. For
linear systems enforcing global static or dynamic equilibrium the FE Analysis leads to a system of
linear equations, respectively
..

KU = F
(2)
M U (t ) + C U (t ) + KU (t ) = F (t )
(3)
Where the matrices K and M are the global stiffness and mass matrices, respectively, obtained by
adding the contributions of all element matrices
K = Ke
(4)
M = M e
(5)
e

The latter matrices have the form,

Ke =

B e d e

(6)

Me =

H eT H e d e

(7)

Where B e is the matrix relating element displacements and strains, D e is the elasticity matrix
relating stresses and strains, e is the mass density, H e is the shape functions and e is the spatial
domain of the element. The global damping matrix C is typically formulated in terms of M and K .

Key Engineering Materials Vol. 446

93

2.2 Probabilistic Transformation Method PTM

The Probabilistic Transformation Method is based on the following theorem


Theorem: Suppose that X is a continuous random variable with PDF (Probability Density
Function) f ( x) and A is the one dimensional space where f ( X )  0 , is differentiable and
monotonic. Consider the random variable Y = u ( X ) , where y = u ( x) defines a one-to-one
transformation that maps the set A onto a set B so that the equation y = u ( x) can be uniquely
solved for x in terms of y , say x = u 1 ( y ) . Then, the PDF of Y is

f Y (Y ) = f X [u 1 ( y )] J

(8)

dx du 1 ( y )
=
is the transformation Jacobean, which must be continuous for all
dy
dy
points y B .
The PTM is based on one-to-one mapping between the random output(s) and input(s) where the
transformation Jacobean J can be computed. The PDF of the output(s) is then computed through the
known joint PDF of the inputs multiplied by the determinant of transformation Jacobean matrix.
The idea of PTM is based on the following formula [6]
f u (u ) = J f z ( z )
(9)
Where J =

dz
(10)
du
f u (u ) is the probability density function of the variable u.
f z (z ) is the probability density function of the variable z.
The general steps in the application of the PTM program: (1) the random variable input is
generating and the stochastic equation of equilibrium is solved first using FEA software. (2)This
solution is used to compute numerically the function between the input and the output by the Cubic
Spline Interpolation, which is then inverted for the calculation of the determinant of the
transformation Jacobean. Finally, the PDF of the response at any point in the domain can be
deduced by using the formula (9).This approach has the advantage of giving a closed-form of the
density function of the response, which is very helpful for reliability analysis of mechanical systems
([6],[11]).
The PTM is one of the most widely used methods in reliability analysis. However, this method has
drawbacks in the solution of reliability problems. It requires the evaluation of the explicit response
functions with respect to the random variables that is very difficult in analysis of complicated
structures. To overcome these drawbacks and for to generalize the application of PTM; we propose
in this paper an interface between the FEA software and the PTM program.
In order to get information about the uncertainty of the structural response evaluated either in the
form of uncertainty range, or in the form of a PDF. We used the PTM program that gives the PDF
of the response variable. However, the situation became more and more complicated when it is
necessary to perform the evaluation of implicit expression between input and output of structures
through numerical models. To overcome this difficulty we proposed the FEACPTM technique.
J =

3- Finite Element Analysis Coupled with Probabilistic Transformation method :


FEACPTM
By the direct coupling method, we mean any reliability procedure based on a Probability of failure
search algorithm using directly the FEA software each time the output of system has to be
evaluated. The FEA software is used to perform the structural analysis to obtain the structure
weight, maximal displacement, and maximal stress, corresponding to a set of given design
variables. These analysis results are sent to the PTM program to conduct the Probabilistic Density
Function PDF, and the probability of failure and generate new random variables. The newly
generated variables are then used to update the input file. The FEA software is then invoked again
to perform the structural analysis with the new input parameters. This process is repeated until

94

Structural Analysis of Advanced Materials

satisfactory results are obtained. For the application of PTM program we need to compute the
function between the input and output variables of system and the determinant of Jacobean of
variables, in the case where the function between input and output variables is implicit then we
applied the Spline Interpolation to approximate this function, and calculating numerically the
Jacobean of the input and output variables. These steps are discussing in previews sections.
3.1 Finite Element Software

The number of existing FE software codes is very large, but since they are merely different
implementations of the same numerical modelling and analysis methodology. In particular, most FE
software codes involve three main phases, namely (i) the pre-processing, (ii) the assembly and
solution, and (iii) the post processing. The entire set of definitions is usually gathered into one or
more input files for the finite element program, which can be used to execute the analysis in batch
mode. As far as the pre-processor concerns, the finite element model is usually saved in a database
file. In the solution phase, the FE solver uses the specifications of the pre-processing phase to
assemble the element matrices corresponding to the adopted formulation, e.g. the mass, damping
and stiffness matrices. The final step of the FE analysis, post-processing, consists in the recovery of
the derived quantities of interest from the solution vector, which in most FE codes corresponds to
the vector of the displacements at the nodal DOFs. Furthermore, post-processing involves the
visualization of results to facilitate their interpretation by the analyst.
3.2 Interface between the FEA Software and the PTM program

A fundamental characteristic of a matlab code (PTM program) for reliability analysis of structural
engineering applications consists in the way it interfaces with the software that gives the finite
element modelling and solution. In this type of implementation the FEA is viewed as a black box in
the analysis process and the FE code is communicated through a generic interface, through the input
files of the latter. The PTM program controls the FE code by automatically modifying the input
files, set identifiers, which govern the automatic generation of input file samples by the stochastic
solver, using pattern matching and replacement.
3.3 Algorithm of method FEACPTM

The outline of the proposed technique FEACPTM is as follows:


1. Generate the input random variables;
2. Calculate the value of output variables by FE software, for each value of input the correspondent
value of output is estimating using FEA and it is stocking in solution file;
3. Approximate the function between input and output variables using Cubic Spline Interpolation;
4. Calculate the determinant of Jacobean of input and output variables (equation 10);
5. Apply the basic relation of PTM (equation 9);
6. Evaluate the graphic of PDF of output variable in function of this output variable (in our case
the PDF of displacement in function of displacement);
P
7. Approximate the Probability of failure f .
The algorithm of proposed technique is illustrating by this Figure 1

Key Engineering Materials Vol. 446

95

Modeling of structure

Generate a set of
random variable X
File: variables. out

FEA

Read output variable and


interpolate f with f ( X ) = Y

Numerical calculus of Jacobean


Static Analysis
Application of equation 9
File: response. out
Get the PDF of response and

Get structure response

Pf
FE software

Matlab

Figure 1: Algorithm of FEACPTM


We start with the creation of a file Variables.out that serves to stock the random variables
generating by PTM program in Matlab in order to use them by the executable file under FEA
software. Once the FEA software receives these variables it introduces them in the modelling of the
considered structure, and by the FEA and Static analysis done with the FEA software we get the
results (nodal displacement, strains, and so on).
These results are stocked in the file response.out. This later file is communicated to the PTM
program in order to calculate the PDF of the output variable of the considered structure (in our case
the vertical displacement Uy) and the probability of failure of this structure.

4- Application
The examples to be presented in this section all have implicit function of output variable in terms of
the input random variable. The proposed technique use the FEA for modelling the structure in every
iteration and calculate the value of displacement (output variable) corresponding at each input
variable value, and in other hand it would be necessary to use the PTM program for estimating the
Pf
value of
of the considered structure. In order to validate the FEACPTM technique and to
compare the results, the Monte Carlo Simulation is also employed as reference method for
comparison.
4.1. 3bar truss structure
As an example of the application of our technique proposed, the three-bar structure (Figure 2) is
analyzed using the FEA software for structural modelling, and static analysis in which the FEA
software are used to approximate the structural response and the PTM program is performed for
computing the probability of failure. For that purpose, statistic models must be defined for each
random variable involved. Hence, uniform distribution is assumed for the Young's modulus E and
the other parameters are considered as deterministic variables. These parameters (Geometrical and
material proprieties) are:
The section of each beam A = 1m2.
The length l = 3m.
The load W = 1,2N.
The mass density =7800kg/m3.

96

Structural Analysis of Advanced Materials

Figure 2: 3bar structure


The Young's modulus E is uniformly distributed in the range [5; 10]. Using the proposed technique
FEACPTM, we obtain the following graph:

Figure 3: PDF (u) when E is uniformly distributed


From the Figure 3, the PDF's of the normalized vertical displacement u y of the 3bar structure
calculated by our proposed technique FEACPTM are independent and uniformly distributed in the
range [1.2505; 2.5008]. Also in this case the obtained results are accurate as shown by favourable
comparison with classical Monte Carlo simulations. For Reliability Analysis, let us suppose that the
limit displacement is u lim it =2mm, It is required to find the failure probability Pf = P(u u lim it ) .
The numerical values of probabilistic characteristics of the displacement of this 3bar structure are
listed in this table.
Table1: Numerical results for proposed FEACPTM and Monte Carlo simulation
FEACPTM
Monte Carlo simulation
2.5009
2.5008
U max

U min
U mean
Standard deviation
Pf

1.2505
1.7349

1.2505
1.7334

0.3556
0.2536

0.3496
0.2504

Key Engineering Materials Vol. 446

97

4.2. Spatial truss


This application treats structural analysis on one hand of the arrow of a crane of construction that
one assimilates to a spatial truss. It is constituted by beams that are identical. To the extremity of
this truss is applied a load M = 5t. The bars forming the structure are in steel of which Young's
modulus E is uniformly distributed in the range [100GPa;300GPa ] and the Poisson
coefficient = 0.29 . The bars have circular sections of which the big one is worth A1 = 22mm 2 and
the small one A2 = 11mm 2 .The weight of each of the bars of the truss is not negligible in front of the
load M. The goal of this analysis is to determine the maximum displacement generated by the
applied load to his extremity.
This structure is analyzed using the FEA software for structural modelling, and static analysis in
which the FEA software are used to approximate the structural response, this response is used by
the PTM program for computing the probability of failure. For that purpose, statistic models must
be defined for each random variable involved. In this case, the Youngs modulus E is uniformly
distributed in the range [100GPa;300GPa ] .

Figure 4: the spatial truss.


The Young's modulus E is uniformly distributed in the range [100GPa;300GPa ] . Using the
proposed technique FEACPTM, we obtain the following graph

Figure 5: PDF (u) when E is uniformly distributed


From the Figure 5, the PDF's of the normalized vertical displacement u y of the spatial truss
calculated by our proposed technique FEACPTM are independent and uniformly distributed in the
range [0.00704, 0.02027]. Also the obtained results are accurate as shown by favourable
comparison with classical Monte Carlo simulations. For Reliability Analysis, let us suppose that the

98

limit

Structural Analysis of Advanced Materials

displacement

is

u lim it

=0.015

mm,

It

is

required

to

find

the

failure

probability Pf = P(u u lim it ) .


The numerical values of probabilistic characteristics of the displacement of this spatial truss are
listed in this table.
Table2: Numerical results for proposed FEACPTM and Monte Carlo simulation.
FEACPTM
Monte Carlo simulation
2.027e-2
2.11e-2
U max
7.04e-3
7.039e-3
U min
1.120e-2
1. 156e-2
U mean
Standard deviation
Pf

3.6e-3
0.2031

3.75e-3
0. 203092

5- Conclusion
In this paper, an efficient, robust technique is proposed to solve the reliability problems that
required the evaluation of implicit function between the input variable of the structure and his
response variable. This technique has been developed with the motivation to facilitate the
application of structural reliability analysis methods to numerical models of mechanical structures.
It integrates the treatment by the FEA software and the PTM program. In fact, the FEA is used to
estimate the structural response function, once the implicit response function is found numerically
the PTM program can easily be applied to solve the structural reliability problem.
In order to validate our technique, we are using the comparison with 10000 Monte Carlo
Simulations through numerical examples of structures such as 3bar structure and spatial truss.
Acknowledgements
This work was supported by CMIFM, A.I. EGIDE, Numero: MA/07/173.
References
[1] M. Lemaire: Evaluation of reliability index associed to structural mechanical models, French
journal of mechanic.
[2] B. Sudreta, A. Der Kiureghian: Comparison of finite element reliability methods Probabilistic,
Engineering Mechanics. (17) (2002) 337-348.
[3] A. Der Kiureghian, J-B. Ke: The stochastic finite element method in structural reliability. Prob.
Engng Mech.; 3(2) (1988) 83-91.
[4] A. Hasofer, NC. Lind, Exact and invariant second-moment code format. ASCE, J Eng. Mech.
Div; (21) (1974) 100-111.
[5] H. Madsen, S. Kenk, NC. Lind: Methods of structural safety. Prentice-Hall Inc.; 1986.
[6] S. Kadry: A Proposed Technique to evaluate the Stochastic Mechanical Response based on
Transformation with Finite Element Method: International Journal of Applied Mathematics and
Mechanics, 2006.
[7] R. Rackwitz: Reliability analysis : a review and some perspectives. Structural Safety. (23)
(2001) 365.
[8] G. Muscolino, G. Ricciardi and N. Impollonia: Improved dynamic analysis of structures with
mechanical uncertainties under deterministic input. Structural Safety, (15) (2000) 199-212.
[9] Eurocode 3: design of steel structures. European Committee for Standardization, Brussels; 1992.

Key Engineering Materials Vol. 446

99

[10] M. Lemaire, A. Mohamed and O. Flores-Macias: The use of finite element codes for the
reliability of systems. In: Frangopol, 1997.
[11] A. Mohamed and M. Lemaire, Linearzed mechanical model to evaluate reliability of offshore

structures, Structural Safety , (17) (1995), pp. 167-193.


[12] S. Kadry, On the generalization of probabilistic transformation method, of Engineering
Mechanics, 2006.
[13] M.F. Pellissetti and G.I. Schuller: Scalable uncertainty and reliability analysis by integration
of advanced Monte Carlo simulation and generic finite element solvers, Computers and Structures.
87 (2009) 930947.
[14] CHENG Ying and TU Hong-mao, FAN Hong-li : Implementation of an optimum algorithm for
structural reliability analysis based on FEM. Mech. Eng. China (4) (2006) 468471.

Key Engineering Materials Vol. 446 (2010) pp 101-110


(2010) Trans Tech Publications, Switzerland
doi:10.4028/www.scientific.net/KEM.446.101

Structural Shape Optimization using an Adaptive Simulated Annealing


W. El Alem1,2,a, A. El Hami2,b and R. Ellaia1,c
1

Mohammed V university - Engineering Mohammedia School, LERMA


BP. 765, Ibn Sina avenue, Agdal, Rabat, Morocco.

Laboratory of Mechanics of Rouen, National Institute for Applied Sciences

BP 08, university avenue 76801, St Etienne du Rouvray Cedex, Rouen, France.


a

welalem@yahoo.fr, b aelhami@insa-rouen.fr, c ellaia@emi.ac.ma

Keywords: global optimization, simulated annealing, triplate, structural optimization, simultaneous


perturbation stochastic approximation.

Abstract. In structural design optimization, numerical techniques are increasingly used. In typical
structural optimization problems there may be many locally minimum configurations. For that
reason, the application of a global method, which may escape from the locally minimum points,
remain essential. In this paper, a new hybrid simulated annealing algorithm for global optimization
with constraints is proposed. We have developed a new algorithm called Adaptive Simulated
Annealing algorithm (ASA); ASA is a series of modifications done to the Basic Simulated
Annealing algorithm ( BSA) that gives the region containing the global solution of an objective
function. In addition, the stochastic method Simultaneous Perturbation Stochastic Approximation
(SPSA), for solving unconstrained optimization problems, is used to refine the solution. We also
propose Penalty SPSA (PSPSA) for solving constrained optimization problems. The constraints are
handled using exterior point penalty functions. The proposed method is applicable for any problem
where the topology of the structure is not fixed, it is simple and capable of handling problems
subject to any number of nonlinear constraints. Extensive tests on the ASA as a global optimization
method are presented, its performance as a viable optimization method is demonstrated by applying
it first to a series of benchmark functions with 2 - 30 dimensions and then it is used in structural
design to demonstrate its applicability and efficiency. It is found that the best results are obtained by
ASA compared to those provided by the commercial software ANSYS.
1- Introduction
The increasing need for optimum structural designs without violating constraints has given rise to
several researches going on in the field of structural optimization. A good deal of effort has been
centred toward the search of structural shape optimization ([2], [3], [11], [14]). In many of the
previous studies the authors have used local search algorithms, such algorithms can only be
successful if it is used to improve the current design or only a small segment of the boundary is
allowed to move or if the objective function is convex. In fact, such an assumption doesn't hold, a
design engineer working in the field of research and development has to often design completely
new structures. The essential goal of a designer in using an optimization algorithm is just to state
the boundary conditions and let the algorithm do some iterations without human intervention until
automatically produce the best design. In this respect, the previous studies had only a relative
success. Many of them relied too much on designer's intuition including the choice of initial design,
or imposed tight restrictions on the movements of boundary. For these reasons, global optimization
should take part in structural problems. In some of the previous studies global search algorithms
were used, but the accuracy remained questionable. During a shape optimization process, geometry
of the structure may undergo substantial changes, such that it can no longer be considered as a
viable structure, e.g. geometric model may become unfeasible [17,9], stresses may exceed the

102

Structural Analysis of Advanced Materials

admissible stress level [17,9,4,15], natural frequency may be lower than a certain limit [4],
displacements may be too large [4,15]. In these cases, a robust optimization method should be used.
In this paper, a new hybrid simulated annealing algorithm for global optimization with
constraints is proposed. Simulated Annealing (SA) has been developed rapidly since the past 20
years [12] as an effective and simple optimization technique, the superiority of SA lies in its good
robustness and convenience to realize global optimization, which conventional optimization
techniques can seldom attain. This algorithm, among few other heuristics, is suitable for
complicated problems where global optimum is hidden among many local optima. The idea of the
method is an analogy with the way molten metals cool and anneal. For slowly cooled process,
system is able to find the minimum energy state. So slow cooling is essential for ensuring that a low
energy state is achieved. Although, simulated annealing is a global optimization method but it
suffers from some disadvantages such as accuracy or slow convergence to the global minimum. In
order to overcome these concerns many modifications have been proposed viz, (a) a new stopping
rule and (b) the application of a local search method. The present article focuses on the
enhancement of simulated annealing algorithms by proposing the new hybrid method ASA. The
method was designed in order to find the absolute minimum of an objective function without being
sensitive to the starting point, it's capable of handling problems subject to any number of design
variables or equality/inequality constraints. The present method can help researchers and
practitioners devise optimal solutions to countless real-world problems. Numerical results
demonstrate the efficiency, accuracy and applicability of the suggested method for structural shape
optimization.
The rest of this article is organized as follows: in Section 2 the basic simulated annealing
algorithm is explained in detail, in Section 3 the proposed modifications are described, in Section 4
the test problems are listed accompanied with the experimental results and in Section 5 some
conclusions are derived.
2- The basic simulated annealing
The BSA used in this paper is a simulated annealing (Figure1) based on the work of Van Laarhoven
and Aarts [22], the algorithm is also well described by Sitarz [18]. A brief summary of the
algorithm is given below for completeness.

Figure1: The basic steps of simulated annealing


begin
Initialization (1)
while not Termination (2) do
while maximal random perturbations not reached do.
snew := RAD ((s)) (3)
if snew is better than s
s := snew
else
s := snew with probability P (T, s, snew) (4)
end if
end while
Update T (5)
end while
end

Key Engineering Materials Vol. 446

103

The following is a description of the periods of the algorithm in detail:


(1) Initialization: Generate initial solution s and initial parameters.
(2) Termination: The algorithm terminates when T reaches the value of the minimal
temperature allowed.
(3) RAD ((s)): Randomly choose a solution from N(s). Where N(s) is the neighborhood of s.
(4) P (T, s, snew): P (T, s, snew) is the acceptance probability, which is defined as:
P (T, s, snew)= exp( E / k b .T ) ,
Where kb is Boltzmann constant. E = E(snew) - E(s), is the amount of increase in the objective
value caused by the uphill move and T is a parameter referred to as "annealing temperature".

(5) Update T: Updating Temperature means cooling schedule. Tk temperature in the iteration k
fulfils the following conditions:
Tk 0 and lim Tk = 0
k

The

cooling

schedule

may

be

classic

([5],

[13])

Tk = T0 k

Logarithmic

[16]

Tk = T0 / log(k + 1) and so on. Where is some constant such that 0 < < 1 , usually in the range of
0.90 - 0.99. T0 = 0.01, 1, 100.
BSA Algorithm starts from an initial solution s at a high temperature T, and makes a series of moves
according to RAND (N(s)). The change in the objective function values E is computed at each move. If
the new solution results in decreased objective function value, it is accepted with probability 1. If the new
solution yields increased objective function value, it is accepted with probability P (T, s, snew). To avoid
accepting large uphill move in the later stage of the search, the parameter T will be decreased over time by a
schedule which is called "the cooling schedule". By accepting worse solutions with a certain probability, SA
can avoid being trapped on a local optimum. There are two drawbacks for applying BSA Algorithm directly:
one is its slow convergence; the other is its accuracy. To obtain faster convergence and to ensure that the
region containing the optimal solution was found, we first accelerate the BSA according to the new stopping
rule that we have called Fast Simulated Anneling (FSA) and in the final stage, we apply PSPSA on the
solution found by FSA to improve the final result.
In this paper, a new ASA method is proposed. From numerical experiments, ASA shows its powerful ability
of global optima searching and wide applicability.

3- The proposed modifications


Since the BSA method approaches the neighbourhood of the global minimum but has a difficulty in
obtaining some required accuracy, it seems judicious to modify the basic algorithm according to a new
stopping rule, then finish the algorithm with a faster convergent method. According to this idea, we modify
BSA method to obtain the ASA method as follows:

104

Structural Analysis of Advanced Materials

Figure 2: The basic steps of ASA


Step 1 : Initialization and coefficient selection.
Initialize the starting point at random and the initial temperature T
Step 2 : Search the region containing the global solution.
Use FSA to reach the region containing the global solution according to the stopping rule
Step 3 : search the global solution
Keep the solution founded by FSA and restart the search with PSPSA method
Step 4: Iteration or termination.
Terminate the algorithm if there is little change in several successive iterates or the maximum
number of iterations has been reached.

Commonly, structural problems are constrained. For the reason that SPSA method involves penalizing
constraints a penalty method is used, the constrained problem is then converted into an unconstrained
problem which allow as to design a new method for constrained optimization problems, called Penalty
Simultaneous Perturbation Stochastic Approximation (PSPSA) method.
3.1 Penalty Method
Penalty method is a procedure for approximating constrained optimization problems by
unconstrained ones. The approximation is accomplished in the case of penalty methods by adding to the
objective function a term that prescribes a high cost for violation of the constraints.
Consider the problem

Minimize f ( x) , x IR n

Subject to : x S

(1)

Where f is a continuous function on IR n and S is a constraint set in IR n . The idea of a penalty method is
to replace problem (1) by an unconstrained problem of the form, for further details about the equivalence

and the approximation used see [10].


Minimize

f ( x) + P ( x). (2)

Where is a positive constant and P is a function on IR n satisfying:


(i) P is continuous
(ii) P ( x ) 0 for all x IR n
(iii) P ( x ) = 0 if and only if x S
Suppose that S = {x : g ( x) = 0}. Problem (1) can be then, replaced by the following unconstrained problem:

Minimize

f ( x) + g 2 ( x)
is l arg e enough
n
x IR

Suppose now that S is defined by a number of inequality constraints

S = {x : g i ( x) 0,
A very useful penalty function in this case is

i = 1,2,............., p}

Key Engineering Materials Vol. 446

105

P( x) =

(max [0, g ( x)])

i =1

For large it is evident that the minimum point of problem (2) will be in a region where P is small. Thus,
for increasing it is expected that the corresponding solution points will approach the feasible region S
and, subject to being close, will minimize f. Ideally then, as the solution of the penalty problem (2)
will converge to the solution of the constrained problem (1). More generally if the subset S is defined as

S = {x : g i ( x) 0,

i = 1,2,........., m

and h j ( x) = 0, j = 1,2,........., p}.

Then the problem (1) is equivalent to that of (2) with a penalty function of the form:
m

P( x) =

(max [0, g i ( x)]) 2 +


i =1

h j ( x)

j =1

3.2 SPSA method

SPSA is based on a highly efficient and easily implemented simultaneous perturbation


approximation to the gradient: this gradient approximation uses only two loss-function
measurements, independent of the number of parameters being optimized. The following general
algorithm of SPSA is based on the work of J. Spall ([19], [20]).
Figure 3: The basic steps of SPSA
Step 1: Initialization and coefficient selection.
Set the SPSA gain sequences a k = a / ( A + k ) and c k = c / k .
Step 2: Generation of the simultaneous perturbation vector.
Generate an n-dimensional random perturbation vector k where each of the n components of k
is independently generated.
Step 3: Loss function evaluations.
Obtain two measurements of the loss function y (.) based on the simultaneous perturbation around
the current x k : y ( x k + c k k ) and y ( x k c k k ) with c k and k from Steps 1 and 2.
Step 4: Gradient approximation.
Generate the simultaneous perturbation approximation to the unknown gradient g ( x k ) :

k11

.
y ( x k + c k k ) y ( x k c k k )
.
g k ( x k ) =
2 ck

.
1
kn
where ki is the ith component of the k vector
Step 5: Updating x estimate.
Use the standard SA form:
x k +1 = x k a k g k ( x k ) to update x k to a new value x k +1 .
Step 6: Iteration or termination.
Return to Step 2 with k + 1 replacing k. Terminate the algorithm if there is little change in several
successive iterates or the maximum allowable number of iterations has been reached.

106

Structural Analysis of Advanced Materials

4- umerical results
4.1. Benchmark test functions
Several tests have been performed on some well-known benchmark test functions; see
Appendix, with known global optima, in order to demonstrate the efficiency as well as the
accuracy of the proposed method. Basic information about the benchmark functions are reported
in Table 1, global optimization methods used for performance analysis are provided by Table 2
and computational results are summarized in Table 3 for each problem.
Table 1: Basic information about 4 Benchmark functions
Function
(Name)

Search space

DJ (De Jong)
CA (Camel)
S5 (Shekel)
S7 (Shekel)
S10 (Shekel)
SHU (Shubert)

[-100,100]n=30
[-10, 10]n=2
[0,10]n=4
[0,10]n=4
[0,10]n=4
[-10, 10]n=2

Number
of local
minima
1
6
5
7
10
760

The global optimum

Reference

f min

(0, 0,..,0)T
(0.089842, 0.712656)
(4, 4, 4, 4)
(4, 4, 4, 4)
(4, 4, 4, 4)
n.3n (-1.42513,-0.80032)

0
-1.031628
-10.1532
-10.40294
-10.53641
-186.730909

[23]
[23], [7]
[8], [16]
[8], [16]
[8], [16]
[6], [7]

Table 2: Global optimization methods used for performance analysis


Method
ePSO
ODE
GEN_S_M_LS

Name
extrapolation Particle Swarm Optimization
Orthogonal based Differential Evolution
GENetic algorithm using Stopping rule, Mutation
mechanism and a Local Search procedure
New Filled Function
Adaptive Simulated Annealing

NFF
ASA

Reference
[1]
[5]
[21]
[24]
This paper

Table 3: Performance of ASA on 4 test functions


Test functions
DJ
CA
S5
S7

Number of variables
30
2
4
4

ASA
5.6516e-036
-1.03162841
-10.1531996
-10.40294

S10
SHU

4
2

-10.5364098
-186.730908831023

Other methods
ODE : 2.06e-23
ePSO : -1.031604
GEN_S_M_LS : -10.107749
ePSO : -9.999624
GEN_S_M_LS : -10.342378
GEN_S_M_LS : -10.536410
NFF : -186.730908830971

As shown in Table 3, the ASA can converge fairly close to the global optimum (verified by the
analytical solutions). We also notice that, when compared with other methods ASA demonstrate its
accuracy in achieving the global solution.

Key Engineering Materials Vol. 446

107

4.2. Optimal design of a triangular plate


4.2.1 Problem description
The problem considered is a hexagonal steel plate, using thickness T1 and fillet radius FIL as the
optimization parameters (see Figure 4). This problem uses a 2-D model and takes advantage of
symmetry. The loading is tensile pressure (traction) of 50 MPa at the three flat faces. The purpose
of this optimization problem is to minimize the volume of the triangular plate without exceeding the
allowable stress ad . The stress anywhere in the triangular plate shouldn't exceed ad =150 MPa.
The optimum set of design variables can be achieved by solving the following minimization
problem:
Minimize Volume

von ad
Subject to :

20 T1 40

5 FIL 15

(3)

4.2.2 umerical results


In order to solve problem (3), we have applied two schemes:

Scheme 1: both Structural Analysis and Optimization done using the commercial software FEA
code ANSYS
The Scheme 1 is a design optimization using the First Order (FO) optimization method integrated
in ANSYS. The optimization method eliminates, iteratively, elements which are not needed from
the design space to obtain the optimum shape for the given problem.

Scheme 2: Structural Analysis done in the commercial software FEA code ANSYS but an external
Optimization code is used
In Scheme 2 we use FEA software to solve the modal analysis and get relevant data to feed it to our
optimization method, ASA, written in MatLab. For this reason, we decided to do an interface
between Ansys and MatLab. This interface allows MATLAB to set the value of the design
variables, obtained by ASA program, in designvar.txt and let ANSYS to get these values and to
create the volume.out and the stress.out. These values are used together with the design variables
values to run the ASA optimization method and to provide the new design variables. Therefore, a
new shape will be created and analyzed. This procedure goes on until convergence.
Table 4: Scheme 1 vs. Scheme 2.
Initial point

Scheme 1

Scheme 2

Volume

22967.2

17979

17952

von

56.727
30
10

150.42
20.064
7.2210

149.8627
20.0001
7.3160

T1
FIL

108

Structural Analysis of Advanced Materials

Table 4 shows the triplate results when using FO method implemented in ANSYS and our method.
The optimal volume obtained by our method is less than that obtained by the FO method. We can
also notice that the von = 150.42 when using FO method which is superior to ad while all the
constraints are satisfied and with less volume when using our proposed method. Calculation shows
that the reduction from initial shape is 21.84% (see Figure 5 and 6), which leads to economic
structure.
Figure 5: The Graph of a
Figure 6: The Graph of the
Figure 4: Triplate
Hexagonal plate with
optimal hexagonal plate
initial shape
using ASA

Conclusion
In this paper a new global hybrid method ASA has been proposed. The new algorithm can be
widely applied to a class of global optimization problems for continuous functions with constraints.
The experimental results show that the present method has proved the robustness and high
performance of its algorithm. We can clearly see from numerical and graphical results that the
algorithm yield the global optimum with high accuracy. The present investigation forms the basis
for further study to look at more complex real structures.

Acknowledgements
This work was supported by CMIFM, A.I. EGIDE, Numero: MA/07/173.

Appendix

5
SHU = i cos
i=1

[ (i +1) x1 +i] i cos [ (i +1) x2 +i] ,


5

i=1

1
CA = 4 x 2.1 x + x16 + x1 x 2 4 x 22 + 4 x 24 ,
3
2
1

4
1

The Shekel function coefficients a ij , ci are:

DJ = xi2
i =1
m

Sm =
i =1

1
, x IR n
( x ai ) ( x a i ) + ci
T

Key Engineering Materials Vol. 446

109

Table 5: Shekel function-four-dimensional

a i ,1

a i,2

a i ,3

a i,4

ci

4.0

4.0

4.0

4.0

0.1

1.0

1.0

1.0

1.0

0.2

8.0

8.0

8.0

8.0

0.2

6.0

6.0

6.0

6.0

0.4

3.0

7.0

3.0

7.0

0.4

2.0

9.0

2.0

9.0

0.6

5.0

5.0

3.0

3.0

0.3

8.0

1.0

8.0

1.0

0.7

6.0

2.0

6.0

2.0

0.5

10

7.0

3.6

7.0

3.6

0.5

References
[1] M.S. Arumugam, M.V.C. Rao, A.W.C. Tan, A novel and effective particle swarm optimization
like algorithm with extrapolation technique, Applied Soft Computing 9 (2009) 308-320.
[2] J.A. Bennett, M.E. Botkin, Structural shape optimization with geometric description and
adaptive mesh refinement, AIAA J. 238 (3) (1984) 458-464.
[3] M. Bruyneel, P. Duysinx, C. Fleury, A family of MMA approximations for structural
optimization, Struct Multidisc Optim. 24 (2002) 263-276.
[4] K. Dems, W. Gutkowski, 2D shape optimization with static and dynamic constraints, Struct.
Optim. 15 (1998) 201207.
[5] W. Gong, Z. Cai, L. Jiang, Enhancing the performance of differential evolution using orthogonal
design method, Applied Mathematics and Computation 206 (2008) 56-69.
[6] R. Horst, P. M. Pardalos, H. E. Romeijn, Handbook of Global Optimization, Springer Publisher,
Kluwer; 1995.
[7] Z. Huang, X. Miao, P. Wang, A revised cut-peak function method for box constrained
continuous global optimization, Applied Mathematics and Computation 194 (2007) 224-233.
[8] Y. Kao, E. Zahara, A hybrid genetic algorithm and particle swarm optimization for multimodal
functions, Applied Soft Computing 8 (2008) 849-857.
[9] E. Kita, H. Tanie, Shape optimization of continuum structures by genetic algorithm and
boundary element method, Engrg. Anal. Bound. Elem. 19 (1997) 129136.
[10] D.G. Luenberger, Introduction to Linear and Nonlinear Programming, Addison Wesley,
1973.
[11] R. Meske, J. Sauter, E. Schnack, Nonparametric gradient-less shape optimization for realworld applications, Struct. Multidisc. Optim. 30 (2005) 201-218.
[12] N. Metropolis, A.W. Rosenbluth, M.N. Rosenbluth, A.H. Teller, Equation of State
Calculations by Fast Computing Machines, J. Chem. Phys. 21 (1953) 1087-1092.

110

Structural Analysis of Advanced Materials

[13] N. Mladenovic, M. Drazic, V. Kovacevic-Vujcic, M. Cangalovic, General variable


neighborhood search for the continuous optimization, European Journal of Operational Research
191 (2008) 753- 770.
[14] J.P. Queau, P.H. Trompette, Two-dimensional shape optimal design by the finite element
method, Int. J. Numer. Methods Engrg. 15 (1980) 1603-1612.
[15] A. Schleupen, K. Maute, E. Ramm, Adaptive FE-procedures in shape optimization, Struct.
Multidisc. Optim. 19 (2000) 282302.
[16] P.S. Shelokar, P. Siarry, V.K. Jayaraman, B.D. Kulkarni, Particle swarm and ant colony
algorithms hybridized for improved continuous optimization, Applied Mathematics and
Computation 188 (2007) 129-142.
[17] P.Y. Shim, S. Manoochehri, Generating optimal configurations in structural design using
simulated annealing, Int. J. Numer. Meth. Engrg. 40 (1997) 10531069.
[18] S. Sitarz, Ant algorithms and simulated annealing for multicriteria dynamic programming,
Computers and Operations Research 36 (2009) 433-441.
[19] J.C. Spall, Multivariate Stochastic Approximation using a Simultaneous Perturbation Gradient
Approximation, IEEE transactions on automatic control 37 (1992) 332-341.
[20] J.C. Spall, Adaptive Stochastic Approximation by the Simultaneous Perturbation Method,
IEEE transactions on automatic control 45 (2000) 1839-1853.
[21] I. G. Tsoulos, Modifications of real code genetic algorithm for global optimization, Applied
Mathematics and Computation 203 (2008) 598-607.
[22] PJM. Van Laarhoven , EHL. Aarts, Simulated annealing: theory and applications, Dordrecht:
D. Reidel Publishing Company, Kluwer; 1987.
[23] L.Wang, K. Chen, Y. S. Ong (Eds.), Advances in Natural Computation, Part III, Springer
Science & Business Publisher, Changsha, China, 2005.
[24] C. Wang, Y. Yang, J. Li, A new filled function method for unconstrained global optimization,
Journal of Computational and Applied Mathematics 225 (2009) 68-79.

Key Engineering Materials Vol. 446 (2010) pp 111-119


(2010) Trans Tech Publications, Switzerland
doi:10.4028/www.scientific.net/KEM.446.111

Graft Interpenetrating Continuous Epoxy-Polysiloxane polymeric


network
1,a
T.Djilali , V.Nassiet 1,b, B.Hassoune-Rhabbour 1,c
1, Laboratoire de Gnie Production 47 Av.dAzereix 65000 Tarbes. France.
a
email : toufik.djilali@enit.fr, b email : valerie.nassiet@enit.fr,
c
email : bouchra.hassoune@enit.fr,
Keywords: hybrid adhesives, Interpenetrated Continuous network, thermomecanical properties,
Multi-module Bond Line, finite element.

Abstract. The objective is to design a joint, suitable for use from low to high temperature by
combination of two adhesives along the overlap length in single lap joint. This mixed modulus
concept is called Multi-module Bond line (MMBL). At high temperatures, a brittle adhesive (high
modulus) in the middle of the joint retains the strength and transfers the entire load. At low
temperatures, a ductile adhesive at the ends of the joint is the load-bearing adhesive. The first part
of this work deals with the formulation of adhesives with differend stiffnesses to be used in the
MMBL concept. Starting from a DGEBA resin/DETDA hardener system, different contents of
amine terminated polysiloxane modifiers are added to the original mixture. A phase-separated
structure is observed via scanning electron microscopy. The thermal, mechanical and dynamic
viscoelastic properties of polysiloxane modified epoxy networks are studied. The second part of this
paper will present the infinite element study of the assembly with two formulated adhesives in order
to verify if they respect the MMBL concept.
Introduction
In order to guarantee suitable bonding when there is a wide range of temperature changes, the
Multi-Module Bond Line (MMBL) concept can be used. The concept consists in combining two
adhesives. The first one, called High Temperature Adhesive (HTA), is brittle and exhibits high
tensile and shear strength under both dynamic and static loading, but has a poor peel and cleavage
strength in the temperature range. The second one, called Low Temperature Adhesive (LTA), is
more ductile and flexible with low tensile and shear strength, but develops high peel and cleavage
strength. The HTA should be in the middle of the bondline, while a LTA is set on the edges so as to
lower stress concentration. The MMBL was optimized in previous studies [1, 2, 3] to improve stress
distribution and ensure strong structural strength when thermal cycles or thermal shocks are applied.
In recent years, Adams and Da Silva have explored the MMBL possibilities for supersonic
applications. The results are hopeful to ensure structural bonding (high modulus) but not enough
flexibility [4, 5].
The aim of this paper is to formulate hybrid adhesives to be used in MMBL configuration.
Numerical simulations will show the ability of this concept in extreme temperature conditions even
if the adhesives behaviours are different.

112

Structural Analysis of Advanced Materials

Fig.1 illustrates the MMBL behaviour according to temperatures. The HTA has the highest elastic
modulus whatever the temperatures. The stresses applied to the assembly are transferred by the
HTA at high temperature but low temperature exposure reduces its strength by increasing its
brittleness.
TvLTA

TvHTA

Phonomenon intensity

Brittleness

Elastic
modulus

Ductility

Stress
resistance
Temperature

LTA
Work zone
HTA

Fig. 1 : Variation of mechanical characteristics variation according to temperature.


TvLTA is the LTA glass transition temperature, TvHTA is the HTA transition temperature.
The LTA is more ductile than the HTA whatever the temperature is (Fig.2). When the temperature
decreases, its modulus and its mechanical resistance increase. So, at low temperature, a higher
proportion of the stresses applied to the assembly are transferred by the LTA and the HTA is
protected. Furthermore, at this temperature, the only LTA is subjected to peeling stresses.

LTA

HTA

LTA

Fig. 2: Diagram of a Single-lap assembly according to the MMBL concept.


The two adhesives to be formulated have to cure simultaneously. So, the same basic epoxy resin
with two different hardeners is chosen. The resin with the first hardener will give stiff high Tv
network .The second flexible hardener must be added to the stiff network to give ductile low Tv
network.
These epoxy resins have specific chemical and mechanical characteristics. Because of the absence
of volatile or by-products during curing reactions, low shrinkage occurs. Epoxy resins can be cured
over a wide range of temperatures and the degree of cross-linking can be controlled. However,
cured epoxy systems have one main drawback: their strong brittleness, which shows poor fracture
toughness [6], poor resistance to crack propagation and low impact strength [7]. Therefore, in the
last few decades, much attention has been given to improve the thermal and mechanical properties

Key Engineering Materials Vol. 446

113

of epoxy resins, particularly their toughness. In recent years, much attention has been given to
siloxane oligomers as a specific type of rubber modifier because of their unique characteristics.
They can be functionalized and they exhibit low glass transition temperature (-120C), moisture
resistance, excellent electrical properties, low stress, high flexibility, and good thermal and
oxidative stabilities [8]. Numerous studies have dealt with the synthesis of epoxy-siloxane structure
networks leading to as many different conclusions. For Yorkgitis [9] flexural modulus and Tg
decrease with the increasing modifier contents, which indicates a partial miscibility between the
siloxane and the epoxy phases. For others [10] this reaction is thermodynamically impossible to
perform. Yayun and Menghuo [11, 12] study the thermal and mechanical properties of methoxylterminated polymethylphenylsiloxane-modified epoxy resins. The thermal stability and the fracture
are improved without drastic lowering of Tg and flexural modulus.
The first part of this work deals with the formulation of appropriate adhesives to be used in the
MMBL concept. To do that, a HTA polymer made of epoxy resin and a hardener is formulated.
Then, different proportions of a flexible modifier are added. The thermomechanical characterization
will enable to choose the appropriate network to be used as a LTA. This type of network will be
named Graft Interpenetrated Continuous network (GIC). The second part of this paper introduces a
simulation study by finite elements of the MMBL concept with the two manufactured adhesives.
The purpose is to confirm the MMBL behaviour with these adhesives.
Formulation
The well-known DiGlycidyl Ether of BiphenolA DER332 (Dow plastics) is used as epoxy resin.
DiEthylTolueneDiAmine DETDA80 from Lonzacure is used as a curing agent. These two
components make up the neat polymer. The chosen Polysiloxane modifier is the
AminoPropylTerminated PolyDimethylSiloxane (APTPDMS). Molecular structures and weights
are summarized in Table 1.
Components

DER332

Molecular
weights

Structural formula

348

n =0.03
DiGlycidylEther of Biphenol A

DETDA

178
DiEthylTolueneDiAmine

APTPDMS

900
AminoPropylTerminated PolyDimethylSiloxane

Table 1 : Components, molecular weights and structural formulae of the products.


Adhesives Preparation.
The DER332 epoxy resin being in crystallized state at room temperature, a heating at 50C is
required to obtain an efficient mixing. Then, a stoichiometrical proportion of epoxy resin, DETDA
and Polysiloxane is poured into a dig and vigorously stirred at 125C. Several Polysiloxane fraction
blends (0%, 10%, 30%, 50%, 70% and 100%) are prepared and tested. More details will be given in
the Preliminary reaction results section.

114

Structural Analysis of Advanced Materials

After preliminary reaction, the mixture is stirred in a warm aluminium mould (125C) covered with
a fluorochemical film and degassed at around 100C, at a 4mbar pressure. Heating the mould helps
the polymer outflow and spreading. It also prevents skin layer forming. The duration of preliminary
reaction and degassing is proportionally inversed to Polysiloxane fraction. This blending is cured
for 2h at 100C, 2h at 120C to promote interphase formation [13] , 2h at 160C, 4h at 175C and
1h post curing at 180C.
Dynamic Thermomechanical Analysis. The rheological characterizations of epoxy/silicone
adhesive systems have been performed with a strain-imposed dynamic rheometer (Rheometrics
ARES). Thermomechanical characterizations of the crosslinked samples were performed with
rectangular torsion geometry. The sample dimensions average are 2x12x45mm3. The thermal range
goes from 25C to 200C. The angular frequency is kept constant to = 1 rad/s and the temperature
ramp is equal to 2C/min.
Differential Scanning Calorimetry. The calorimetric study of the DGEBA/DETDA system has
been performed using a Shimadzu DSC 50. 10 mg of the non-cured mixture are set in an aluminium
cup placed in a measurement heating cell. An empty cell is used as a reference. All experiments
have been carried out under nitrogen atmosphere, in dynamic mode at 10C/min heating. A
microcomputer automatically registers the different points of enthalpy reaction according to
temperature and the glass temperature Tg so as to define the best curing process.
Thermal Mechanical Analyser. The Coefficient of Thermal Expansion (CTE) is determined using
Thermal Mechanical Analyser (Perkin Elmer TMA). The specimens thickness is around 2mm. The
applied force through the crystal probe is 20mN under a Helium pressure of 0.3bar and a
temperature ramp of 2C/min.
Results and discussion
Preliminary Reaction. Thermodynamic incompatibility of Polysiloxane with epoxy resin favours
phase separation. It is then needful to improve compatibility to enhance mechanical properties. The
flexibilizing agent in GIC network is amine terminated Polysiloxane chain ensuring chemical
linking with the network. It is known that dispersed particles, like CTBN, ATBN or ETBA [6, 14,
15] under strain generate cavitations, preferential ways to moisture diffusion. Furthermore,
unsaturated structures like CTBN contain atomic voids which easily accept oxygen atoms, leading
to oxidation. In our case, resin epoxy reacts with DETDA (network1) and ATPDMS (network2). B
X% for Blend X% means GIC network with X% of APTPDMS in the hardener. Polymerization
is carried out by stirring and heating at 125C to make the bulkier network formation easier
(network2) and bring about phase inversion and then increase compatibility of the reactive
Polysiloxane with epoxy resin (Fig.3).

Fig.3 : DSC thermograms of the network 1 & 2.

Key Engineering Materials Vol. 446

115

The preliminary reaction is performed by vigorous stirring and heating during decreasing time with
an increasing Polysiloxane fraction (1h30 for the neat polymer). Many conclusions lead to consider
the strong interaction between temperature, time and Polysiloxane fraction. In fact two processes
are suitable for obtaining such a network:
- adjusting temperature with a constant heating and stirring time () or
- adapting time to the cure rate at a constant temperature ().
The () process is chosen because the chemical species reactivity and the reticulation kinetics
remain constant for a constant temperature. These conditions avoid, at high temperatures, skin
forming, network degradation caused by a too high exothermic reaction.
The maximum conversion rate of preliminary reaction in the melt state increases as the molecular
weight of the Polysiloxane decreases. The low molecular weight Polysiloxane chains form smaller
particles due to better compatibility, hence a larger specific area to increase future adhesion..
In the first stage, network2 begins to grow; Polysiloxane particles are tiny and white. The
DGEBA/DETDA mixture is the transparent phase. Stage 2 shows a strong development of
network2 testified by a loss of transparency. By maintaining stirring and heating, phase inversion
occurs. Stage 3 shows an apparently homogeneous network which is in fact an interpenetrated
mixture of two networks (Fig.4).

Fig.4 : SEM Observation of B50% during preliminary reaction : (a) stage 1, (b) stage 2, (c) stage 3
Dynamic Mechanical Analysis. Fig.5 shows the dynamic mechanical properties of cured samples
including the shear modulus (G) and the loss tangent (tan) versus temperature. The cure process is
defined in Adhesive preparation section. Due to the shrinkage induced by Polysiloxane high
contents, the manufacturing of specimens becomes complex. For thermomechanical experiments,
the rectangular samples of Blends 70% and 100% bend when manufactured. The dynamic
mechanical properties of modified cured epoxy, for B0%, B10%, B30%, B50% are summarized in
Table 2.
According to mechanical spectroscopy (Fig.5) as well as DSC (Table 2), incorporation of
Polysiloxane chains strongly decreases the glass transition temperature (Tg) which is unique. Fig.5
shows an important widening of the damping area according to the Polysiloxane content [16, 17]
due to the network polymolecularity. The resulting network behaviour is intermediate between the
properties of networks 1 & 2.

116

Structural Analysis of Advanced Materials

1,00E+09

0,9
0,7
0,5

1,00E+08

0,1

Tan(d)

G' (Pa)

0,3

1,00E+07
-0,1
-0,3

Neat polymer

10%

50%

30%

1,00E+06

-0,5
25

45

65

85

105

125

145

165

Neat
Blend
10%
Blend
30%
Blend
50%

G25C[Pa]
1 109

Tan()25C
0.017

T
182

Tg
174

9.7 108

0.018

157

151

5.6 108

0.02

127

126

4.5 108

0.03

93

92

185

Temperature (C)

Fig. 5 Dynamic mechanical properties of


various APTPDMS fraction

Table 2 Thermomechanical properties of


GIC with different ATPDMS fraction.

Dilatometry. Linear coefficients of thermal expansion between room temperature and the
relaxation of each blend are presented in Table 3.
Blend X%

CTEa
(10-6 K-1)

CTEb
(10-6 K-1)

100%

285

287

50%

128

266

30%

106

253

10%

92

239

0%

82

224

Table 3 Linear Coefficient of Thermal Expansion for several Blends, CTEa below relaxation and
CTEb above relaxation.
The coefficient of thermal expansion increases with Polysiloxane content below Tg as well as above
Tg. These results are in agreement with those reported in earlier studies [18]. It is known that the
addition of a soft phase in an epoxy matrix with a CTE superior to that of the matrix, leads to
increase the local free volume in the epoxy matrix. It also increases the CTE of the whole network.
The MMBL Concept Finite Element Study
HTA and LTA Determination. Knowing the formulated adhesives thermomechanical behaviours,
we have to choose a grade between B0% and B50% to be used as HTA and LTA in MMBL joint
working on a large temperature range (working zone). First, simple lap joints (SLJ) using
aluminium substrates are manufactured with B0% and B50% separately. Then, the assemblies are
tested at 160C and -40C to determine the shear strength of every assembly. Fig.6 shows the
results obtained.

Key Engineering Materials Vol. 446

B0% at 160C
B50% at 160C

30

Shear stress (MPa)

Shear stress (MPa)

35
25
20
15
10
5
0

18
16
14
12
10
8
6
4
2
0

B0% at -40C
B50% at -40C

2
3
4
Shear strain (%)

117

0,2

0,4

0,6

0,8

1,2

1,4

Shear strain (%)

Fig. 6 Curves of shear strain versus shear stress for B0% and B50% at 160C and -40C.
At high temperature, the B0% break stress is around 27MPa 1 with a strain of 3.6 % 0.2. For the
B50 %, the break stress is equal to 4MPa with a more important strain (8.5 %). At low temperature,
the curve is typical with elastic linear behaviour for both adhesives. Furthermore, their break
stresses are equivalent, 16.3 MPa 1.3 for B50 % and 14.2 MPa 1.6 for B0% with an elongation
at break higher for B50%.
Thanks to these observations, the B0% is the suitable product to be the HTA and the B50% to be
the LTA. A finite element modelling study will show the stress distribution in the three MMBL
joints. This study will be carried out with B0% and B50% adhesives at high and low temperatures.
Finite Element Modelling Conditions. The purpose in this part is to verify that the formulated
adhesives work according to the MMBL concept. Particularly we have to verify that at low
temperatures, more load will be carried by the LTA and less by the HTA.
A study of three kinds of single lap joint (SLJ) assemblies is made. The first SLJ contains one joint
HTA, the second contains one LTA joint and the third is an MMBL with both adhesives. These
assemblies are tested on DMTA by creep. to measure the parameters of the generalized Maxwell
function (Eq.3). Then, this geometry is used in finite element study to determine the stress
distribution inside the joints at high and low temperatures (Fig.7). The weak overlap length (6mm
by taking into account the three joints) is imposed by the industrial application of this study. Spaces
(Figure 7) are inserted between joints to avoid adhesives mixing before curing, likely to lead to non
stoichiometrical proportions, hence a low conversion rate.
25mm
2mm

2mm

2mm
2mm

0,35mm

LTA

HTA

LTA

2mm

10mm

Fig7. Geometry of the SLJ with mixed adhesive joints.


A plane stresses model was made with ANSYS.11 software, which considers a linear elastic
mechanical behaviour for the aluminium substrate and elastic or viscoelastic behavior for adhesives
depending on the temperature. We modelled a 2D section in the plan Oxy and a finer meshing was
made for the zones of high stress concentration (Fig.8). A four-node isoparametric finite element
model was used (Plane182). The elastic and viscoelastic materials behavior can be modelled by this
element. For the viscoelastic element, the material properties are expressed in integral form using
the kernel function of the generalized Maxwell elements as Eq.1:
n
t
G (t ) G Gi exp
i 1
i

(1)

118

Structural Analysis of Advanced Materials

Where G(t) is the shear modulus measured by relaxation experiment. G is the asymptotic value of
shear modulus obtained at long time as the polymer behaves like a viscoelastic solid. Gi (shear
modulus) and i (relaxation time) are the Maxwell elements and are obtained by fitting the
relaxation curve.
The boundary conditions were those encountered in the experimental test conditions : i.e. no
displacement was allowed in the three directions and no revolution movement for (a) and (b) nodes.
Nodes in (c) can move along Oy axis and a force (F) was applied on the moving substrate.

Fig.8 Model and boundary conditions used for the finite element study.
Numerical Results. The finite element investigation yielded shear stress distributions on the
median plane along the overlap length at high and low temperatures. The calculated tensile stresses
values are weak as the force applied in the DMTA rheometer during creep experiments. To draw
the stresses profile, the values are normalized by the maximal value obtained for HTA at high and
low temperatures.
The stresses profile obtained within joints are typical (Fig.9) [19]. At high temperatures, the stresses
in the LTA represent 6 % of those within the HTA, while at low temperature, the stresses in the
LTA represent 68 % of those within the HTA (Fig.9 a). The HTA ensures the transfer of the stresses
at high temperature, while the LTA mechanical structural function doesnt exist. At low
temperature, the LTA carries more load (68 %).)

(a)

(b)

Fig. 9 : Shear stresses within the three joints : (a) at high and low temperature with a thickness joint
equal to 0.3 mm whatever adhesive) ; (b) at low temperature with LTA (thickness : 0.2 mm) and
HTA (thickness : 0.3 mm).
In order to increase the LTA participation at low temperatures, joint thickness is reduced. The
Volkersen expression (2) shows that stresses concentration increases while the joint thickness
decreases [20].
r

max
Gl 2

average
2Ete

(2)

Key Engineering Materials Vol. 446

119

where r is the stress concentration factor, max and average the respective maximum and average
shear stresses. G is the shear modulus of the adhesive, E the Young modulus of the substrates, l the
lap joint overlap, t the substrate thickness and e the joint thickness.
So, when the LTA joint thickness is 0.2 mm with 0.3 mm for HTA joint, the shear stress
distribution is shown in Fig.9 b.
Decreasing the LTA joint thickness allows to balance and to homogenize the stresses between three
joints and to make the HTA load smaller because it becomes very brittle at low temperatures
(Fig.9 b).
Conclusion
Grafted interpenetrated mechanically continuous polymer network has been processed by addition
of small molecular weight functionalized Polysiloxane in an epoxy matrix. Then it is necessary to
adjust the pre-polymerization accurately to bring about phase inversion, hence increase the
compatibility of the reactive Polysiloxane with epoxy resin.
Two functionalized Polysiloxane adhesives are used in the MMBL concept. The finite element
study shows that B0% can be used as HTA and B50% as LTA. At high temperature, HTA ensures
the structural bonding. At low temperature, the stresses distribution is more homogeneous between
the three joints, decreasing the stress carried by the very brittle HTA. So, these two adhesives are
appropriate for the MMBL concept.
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
[20]

S. Semerdjiev,UK: business book limited, (1970).


S. Srinivas, NASA TN D-7855,(1975).
R.L. Patrick, Vol.4. New York : Marcel Dekker, Inc, (1976).
R.D. Adams, Int. J. Adhes Adhes., 59, pp.171-182, (1996).
L.F.M da Silva, R.D. Adams, Int. J. Adhes Adhes., 27, 5, pp.362-379, (2007).
R.A Pearson, and A.F. Yee,: J.Mater.Sci., 24, 2571, (1989).
Lee H, Neville, K, Handbook of epoxy resins. New York: McGraw-Hill, (1967).
E.L. Warrick, O.L. Warrick, K.E. Pierce, J.C. Saam Polmanteer,: Rubber chem.;Technol.,
52, 437, (1979).
E.M. Yorkgitis, N.S.Eiss, C.Tran, G.L.Wilkes, J.E. McGrath : Adh in Pol Sci,72,79,(1985).
E. Yilgr, I. Yilgr : Polymer, 39, pp.1691, (1998).
L. Yayun : Master thesis, Syracuse University, Syracuse, NY, (1999).
C. Menghuo : Master thesis, Syracuse University, Syracuse, NY,(2001).
S. Bentadjine, Master thesis, Universit Claude Bernard, Lyon I,(2000).
D. Vachere, H.Sautereau, J.P. Pascault : J.Apply.Polym.Sci, 41,467,(1990).
D.Vachere, J.P. Pascault, H.Sautereau, S.M.Mochiar, C.C.Riccardi, R.J.J.Williams:
J.Appl.Polym.Sci., 43, 293,(1991).
P. Sung, C-Y. Lin, European Polymer Journal, Vol 33, Issue 6, pp.903-906, (1997).
P. Sung, Shin-Yu W, Polymer, Volume 39, Issue 26, pp.7033-7039, (1998).
T-H. Ho, C-S.Wang, Eur.Polym.J, 37, pp. 267-274, (2001).
L.J. Hart-Smith : NASA CR-11 2235, (1974).
O.Volkersen : Luftfahrorschung, 15 :41-7,(1938).

Key Engineering Materials Vol. 446 (2010) pp 121-130


(2010) Trans Tech Publications, Switzerland
doi:10.4028/www.scientific.net/KEM.446.121

Experimental and numerical modelling of LRI process


Romain Brault1, a, Michel Niquet1, b, Sbastien Mistou2, c
1

CRC Composites, 1 Rue Aristide Berges, 65016 TARBES Cedex, France


ENIT-LGP, PRES Univ. de Toulouse, Av Azereix, BP1629, 65016 TARBES Cedex, France

romain.brault@enit.fr, bmichel.niquet@crccomposites.com, cmistou@enit.fr

Keywords: Composite, LRI, permeability, process modelling, filling simulation

Abstract. The aim of this study is to gain knowledge concerning the process and its physics, as well
as to become able to optimize the fabrication of large and complex composite parts in aeronautics
applications. Composite materials have many advantages and the use of this technology is
increasing in the aeronautic industry. In the L.R.I. process, dry textile preforms are impregnated by
a thermoset liquid resin. All the elements are enclosed in a vacuum bag of known pressure. Once
preforms are totally impregnated, the resin system begins the curing reaction to obtain the
composite part. This study contains two major sections. Firstly, numerical modeling was done with
the Pam- Rtm finite element code to determine the evolution of the flow front during the infusion.
Simulations were performed to analyze the infusion of sandwich composite parts with perforate
foam, which allows the inferior skin to be impregnated in the same operation. Secondly,
experimental work was conducted to confirm the numerical results.
Introduction
Composite materials have many advantages over classical materials used in aeronautic industry.
That's why, since few years, composite manufacturing processes have changed in many ways.
Different new processes were developed in relation with the aeronautic manufacturing constraints.
Among them, the composite liquid processes have been significantly developed: Resin Transfer
Molding (RTM), Liquid Resin Infusion (LRI), Resin Film Infusion (RFI), etc... [1] The _ability of
these processes is ever too weak to optimize the manufacturing of parts in composite materials by
the liquid way. Many authors have already worked this subject [1, 2, 3, 4]. The key issue to achieve
a realistic modeling of liquid composite processes is the determination of permeability values.
Many experimental setup and method have been developed to make accurate permeability
measurement [5, 6]. Some paper deals with also the permeability evaluation but by the numerical
way, to save experimental cost and improve numerical models. [7, 8] Authors have also treated LRI
simulation by the hydro-mechanical coupling between resin propagation through a compacted fiber
preform. In this paper, a numerical work associated to the developpement of an exeprimental
method allows to evaluate different issues in association with the LRI process understanding.

1 Numerical modelling
This chapter deals with the numerical modeling of LRI process on monolithic and sandwich
composite parts. PAM-RTM is based on the following Darcy's Law.

(1)

122

Structural Analysis of Advanced Materials

In the Darcy's law, V (m/s) is the fluid velocity, K (m) is the permeability tensor of the fiber
preform, (Pa.s) is the fluid viscosity, P is the gradient between injection point and event.
The method employed to simulate the manufacturing of these parts will be developed and so, the
main difficulties in relation with the LRI numerical modelling will be exposed.
1.1 Infusion of a monolithic specimen. The monolithic specimen is the first simulation. It is a
simple model, which allows understanding the simulation of LRI process. First simulation works
concern the development of a model to simulate the LRI process on a monolithic part. The program
used to perform these simulations is PAMRTM edited by Esi-Group. This monolithic part has a
very simple geometry, to easily make the correlation with the experimental results. The used
geometry is a plate of dimensions 500x200x2mm. The boundary conditions applied during this
modelling are two: pressure inlet and event. The mesh of this plate have been performed with IDEAS program, and it was build with 2D triangular elements, which are the only accepted by
PAM-RTM. In regard to the materials used, the properties are those of carbon fiber UDC300 and
SR1710 resin, manufactured by SICOMIN. The mesh used in this model is represented by the
following Figure 1. Table 1 presents numerical parameters of theses simulations.

Table 1: Numerical simulation parameters

Figure 1: Mesh of monolithic model


1.1.1 Boundary conditions. Two kinds of boundary conditions have been used to perform these
simulations. The first is the pressure of the fluid during the process. Its value is fixed to 0.6 bar, a
classical value for composite made by LRI process. The second boundary condition is the event of
the process, equal to the atmospheric pressure. The Figure 2 shows the geometry with the applied
boundary conditions. Concerning the parameters in PAM-RTM, the permeability value is the
most complicated to give. Indeed, the determination of the 3D permeability tensor [K] can be done
only by the experimental way. The measurement method is complicated and expensive. This
problem has been bypassed by a sensibility study in relation with the experimental results. Starting
with the experimental filling time, the value of 1.10-9m2 is kept in PAM-RTM.

Key Engineering Materials Vol. 446

123

Figure 2: Boundary conditions of monolithic model


1.1.2 Results. PAM-RTM gives the filling evolution of the considered part with a good precision.
This simulation allows plotting the evolution of the fluid speed against time, as shown with Figure
3. This plot must be related to the different parameters of the modeling: fiber rate, permeability,
charge loss. This graph shows that the resin speed decreases with time, in function of the charge
loss. This fact is amplified if the permeability is more important. That is why the determination of
the permeability tensor is critical for the filling time prediction. [10] This is one of the most
important problematic in the LRI process study.

Figure 3: Resin speed evolution during monolithic infusion simulation


1.2 Infusion of a sandwich specimen. The sandwich simulation is interesting because, there are
many differences in the interpretation of results. The infusion of sandwich generates specifically
problem to consider, like the delay between the impregnation of the upper layer and the bottom
layer.
1.2.1 The delay problem. The major difficulty in the modelling of the sandwich infusion is to
control the impregnation of the upper and bottom layer. If the sandwich is made with foam, it is
necessary to realize many holes through the thickness to let the resin flow to the bottom layer. [11]
Indeed, the LRI process principe is to distribute the resin on the upper layer with a draining
component, and then it is flowing to the bottom. Concerning the sandwich part, the thickness is
generally more important than the monolithic composite. This important thickness generates the
delay problem in sandwich materials infusion, because there is a significative time during which the
resin flows through the foam by the micro-holes. The Figure 4 shows this problem.

124

Structural Analysis of Advanced Materials

Figure 4: Problem of delay in sandwich infusion


1.2.2 Modelling. Sandwich infusion analysis was made with an elementary volume, which is
representative of the foam, with micro-holes, upper and bottom layer. The layers and the foam are
considered in two different areas, with different materials properties. Layers properties are the same
as these used for monolithic simulations. Concerning the foam, it is an AIREX R82. The
dimensions of this model are a rectangular volume of 100x100x17mm. The thickness of the foam is
15mm, so each layer is 1mm thick. The particularity of this model is only the wall of the holes has
been drawn. It is not necessary to model the foam, because there is no fluid flowing through. The
simulation is concentrated on the flow behavior. Figure 5 presents this numerical modelling.

Figure 5: Mesh of sandwich model

1.2.3 Boundary conditions. The boundary conditions are the same as the monolithic model, except
event is applied on upper and bottom layer, as shown by the Figure 6. The resin injection remains
on the upper layer because of the draining fabric.

Key Engineering Materials Vol. 446

125

Figure 6: Boundary conditions of sandwich model


1.2.4 Results. One of major aim of these sandwich infusion simulations is to identify the delay
between the upper layer and the bottom impregnation. This goal is reached, as we can see on the
Figure 7, the upper layer is full (red color) before the bottom layer. This one is impregnated by
growing points when the resin fills the holes.

Figure 7: Sandwich simulation filling result


This fact is completely coherent with the experimental behavior. Nevertheless, there are some
problems to analyze on the delay plot (Figure 8). Indeed, the delay is not constant in time. This is
identified by rebounds on the curve. After 40s of infusion, rebounds disappear, which means that
there is no delay after 40s. This is no coherent with real results, the delay is observed up to the end
of the process. In simulation, after 40s, there is delay and at the end the resin flow by the two layers
at the same time, which is impossible in real process. It means that the model do not transcribe the
good behavior of sandwich infusion, for many reasons. First, if this phenomenon can be observed, it
will be in very special permeability and draining conditions. Another reason is that draining product
is not a parameter of the simulation, so PAM-RTM considers in the same way the flow through
the two extreme layers. Finally, this software considers infusion more than an injection under
pressure with vacuum assistance (process VARTM) instead of the real LRI process.

126

Structural Analysis of Advanced Materials

Figure 8: Evolution of delay in sandwich simulation

2 Experimental study
The goal of the experimental study is to be capable of measure significative parameters during a
composite part infusion and make the correlation with numerical results to help understanding LRI
process behavior. This correlation will allow identifying the major issues of this study.
2.1 Experimental equipment. A test bench was made to realize measurements during monolithic
and sandwich infusions. The mold which has been used is glass plate. All tests were performed with
this very simple mold. To follow the resin front during impregnation, a camera and a ruler were
installed on top and under the mold to take pictures at fixed time steps. The temperature was
measured with thermocouples regularly distributed on the part. All sensors were connected with
computers to make a real time acquisition of data. The test bench was also equipped with all
injection equipment necessary to the LRI process: vacuum system, resin trap, etc... This test device
is in a gray room, with controlled environmental conditions. The Figure 9 presents this test
equipment.

Key Engineering Materials Vol. 446

127

Figure 9: Experimental equipment


2.2 Experimental results. During experimental campaign, monolithic and sandwich infusions were
performed, but only monolithic results are exploitable for a quantitative correlation, because not
enough sandwich results were available.
2.2.1 Monolithics. These tests allow plotting the evolution of filling and fluid speed against time
for monolithic infusion, represented by the Figure 10. It shows that the numerical and experimental
results are similar. These plot mean also that carbon fiber have a permeability lower than glass
fiber. This is important for the determination of LRI process parameters: resin viscosity,
temperature, draining strategy, etc... The test bench allows also analyzing the evolution of
temperature during the LRI process, to complete his understanding.

Figure 10: Evolution of filling and resin velocity evolution

128

Structural Analysis of Advanced Materials

2.2.2 Sandwiches. If there is not exploitable quantitative measure, some observations can be made
during sandwich infusion tests. First, concerning the delay problematic, the experimental results
confirm that resin fill the bottom layer by growing points, and then the impregnation front is
circular, around micro-holes. It is the same behavior in sandwich infusion simulation, at the
beginning of filling, which means this modeling is acceptable up to a limit of part length.
3 Correlation of results
This section deals with the comparison between numerical and experimental results. Also, some
issues link to the LRI modelling will be identified to share on the study of this liquid composite
process.
3.1 Simulation/Experience comparison. As it is shown on graph provided by previous chapters,
the correlation of results is good. Indeed, concerning the monolithic results, the prediction of filling
and the resin speed are close between the both evaluation techniques. For sandwich infusion, the
results are also good, by visual observations.
These results allow validating many things. First, the monolithic modelling is completely
exploitable, for 2D geometry. This tool can be used for the cycle prediction in the case of simple
parts. Mathematic law can be identified on graph, and a further study is necessary to affine
coefficients and identifies terms in physical LRI parameters.
3.2 Identification of new issues. This study allowed to identify many issues, which must be solved,
to make an accurate prediction of the LRI process behavior. These issues can be divided in two
categories, numerical and experimental issues.
3.2.1 Numerical issues. Concerning the numerical modelling, there are many ways to optimize the
simulation results. Softwares will take in account draining component to be efficient in the
prediction of LRI behavior, especially for sandwich modelling. Also, the modelling of flow through
the micro-holes must be developed to solve errors in the impregnation delay prediction. The major
work to be more efficient in LRI simulation is probably the permeability measure. It seems clear
that it is the more influent parameter. All these modifications should allow developing a reliable
numerical model.
3.2.2 Experimental issues. In the experimental study to, the major research axis is the permeability
determination. That is a complicate measure, which requires many precautions:
- Side effect
- Flow control
- Injection section
- Pressure measurements
It will be interesting to measure the effect of fibers compression under the resin action, this
parameter modify the preform permeability, and so the filling time [11]. An experimental technique
which measures in real time the compression stress on fiber section would allow deepening LRI
analysis.
Research axes are necessary to develop reliable techniques for LRI cycle predictions. This is a
process more and more used in aeronautic industry, and so, many companies will be interested by
the prediction of part made by LRI process in few years.

Key Engineering Materials Vol. 446

129

Conclusion
This six months study dealt with the experimental and the numerical modelling of LRI process, a
liquid composite molding technique. The simulation work allows evaluating performance of PAMRTM software used to optimize RTM and LRI composite manufacturing. Different modelling
have been realized, for monolithic and sandwich composites. The results of these models are the
prediction of filling time and resin speed evolution, with an acceptable accuracy. The experimental
campaign is the second research way to understand the LRI process. This one, by a test bench
developed for this study, allows making a correlation with the numerical results, to validate PAMRTM models. The correlation was good for monolithic infusion, but the sandwich simulations
gave errors at the end of process. The delay problem was not ideally modeled. The last part of this
paper gives several ways to be developed for the numerical and experimental work, with the aim to
be more efficient in the LRI process understanding and prediction.

References
[1] P.Celle. Couplage fluide/milieux poreux en grandes dformations pour la modlisation des
procds d'laboration par infusion. Thse de doctorat, Ecole Nationale Suprieure des Mines de
Saint- Etienne, 2006
[2] T.Ouahbi, A.Saouab, J.Breard, P.Ouagne, S.Chatel. Modelling of hydro-mechanical coupling in
infusion processes. Composites Part A, Vol.38, pp.1646-1654, 2007
[3] S. Comas-Cardona : Modlisation, simulation et contrle du couplage hydro-mcanique pour le
moulage de composites. Thse de doctorat, Universit des Sciences et Technologies de Lille, 2005
[4] J. Echaabi, M. Ouadi Bensalah : Simulation du front d'coulement dans les procds de moulage
des composites liquides. C.R. Mcanique 333, pp. 585-591, 2005
[5] H. Talvensaari, E. Ladstatter, W. Billinger, Permeability of stitched preform packages,
Composite Structures Vol. 71, pp. 371_377, 2005
[6] J.M. Lawrence, J. Barr, R. Karmakar, S.G. Advani, Characterization of preform permeability in
the presence of race tracking, Composites: Part A Vol. 35, pp. 1393_1405, 2004
[7] S. Song, K. Chung, T.J. Kang, J.R. Youn, Prediction of permeability tensor for three
dimensional circular braided preform by applying a finite volume method to a unit cell, Composites
Science and Technology Vol. 64, pp. 1629_1636, 2004
[8] M.A. Choi, M.H. Lee, J. Chang, S.J. Lee, Permeability modeling of fibrous media in composite
processing, J. Non-Newtonian Fluid Mech. Vol. 79, pp. 585-598, 1998
[9] S. Morel, C. Binetruy, P.Krawczak : Compression des renforts dans les procds LCM : 1.
Analyse mcanique et microstructurale. Revue des Composites et de Matriaux Avancs, Vol. 12,
n2, pp. 243-263, 2002

130

Structural Analysis of Advanced Materials

[10] J.C. CharpentieR : Elments de mcanique des fluides : Application aux milieux poreux. Les
Techniques de l'Ingnieurs, J1-065 :12, 2004
[11] R. Brault. Mmoire de fin d'tudes. Modlisation du procd d'infusion de composites, Ecole
Nationale d'Ingnieurs de Tarbes, 2008

Key Engineering Materials Vol. 446 (2010) pp 131-136


(2010) Trans Tech Publications, Switzerland
doi:10.4028/www.scientific.net/KEM.446.131

BEM Simulation Of 3D Updated Resin Front For LCM Processes


R. Gantois1, a, A. Cantarel2,b, J.-N. Flices2,c , N. Pirc1,d, F. Schmidt1,e
1

Universit de Toulouse ; INSA, UPS, Mines Albi, ISAE ; ICA (Institut


Clment Ader); Campus Jarlard, F-81013 Albi cedex 09, France
Ecole des Mines Albi, Campus Jarlard, F-81013 Albi, France

Universit de Toulouse; INSA, UPS, Mines Albi, ISAE; ICA (Institut


Clment Ader); 1, rue Lautramont, F-65016 Tarbes, France

renaud.gantois@mines-albi.fr, barthur.cantarel@iut-tarbes.fr, cjean.noel.felices@iut-tarbes.fr,


d
nicolas.pirc@mines-albi.fr, efabrice.schmidt@mines-albi.fr,

Keywords: Numerical simulation, Boundary Element Method, Moving Mesh, Liquid Composite
Molding, Darcys law

The use of composite materials in large structures has increased in recent years. Aircraft
industry has recently begun to investigate the field of Liquid Composite Molding (LCM) through
research programs because of its ability to produce large parts at low cost. The present paper
focuses on modeling a 3D radial impregnation through an anisotropic fibrous perform. As a
preliminary work, it is assumed an isothermal flow and no hydro-mechanical coupling. Governing
equations are Darcys law and mass conservation. Simulation is performed combining Boundary
Element Method (BEM) with a lagrangian moving mesh method. An analytical solution is
developed to assess the numerical model.
Abstract.

Introduction
Composite materials have been used in the aircraft industry over the past three decades for their
excellent specific mechanical properties. Large structural parts made of resin/fiber composites
require specific forming processes. The traditional prepreg autoclave process produces high quality
parts, but involves heavy manufacturing cost due to autoclave investment and use of high value
added raw materials. As an alternative, recent technologies in Liquid Composite Molding (LCM)
have become established recently under the designation of infusion processes. Liquid Resin
Infusion (LRI) is one of them. It consists in forcing the liquid resin to circulate transversally into the
dry preform using a distribution medium, consecutively to the vacuum application [1].
Prediction of resin flow during the infusion filling stage requires a full three-dimensional model
in order to take into account the lag due to the distribution medium. Previous works revealed the
adequacy of Boundary Element Method (BEM) solving the Darcys flow [2-4]. It was performed
assuming a two-dimensional flow, neglecting the impregnation through the thickness.
The present paper is focused on a three-dimensional numerical modeling of the Newtonian
isothermal flow using constant Boundary Elements. Combining Darcys law with incompressibility
equation leads to Laplace equation which is solved using BEM at each step time on an updated
lagrangian mesh. Preliminary results are presented, i.e. BEM simulations performed in the case of
central injection. Computed results are compared with an analytical solution.

Governing equations
Modeling Assumptions. Impregnation is modeled as a Darcys flow. Fibrous perform is
regarded as a porous medium. Hydro-mechanical couplings are neglected, so that the medium is
assumed to be not deformable and of a constant permeability. The flow is considered as isothermal

132

Structural Analysis of Advanced Materials

and resin viscosity is assumed to be constant, neglecting chemical reactions effects. Consequently,
governing equations reduce to Darcys law and incompressibity equation.

Darcys law. Darcys law [5] states that the average velocity v is proportional to the pressure
gradient according to the following equation:


[K ]  p
v =

(1)

where [K ] is the permeability tensor, p the resin pressure and the liquid resin viscosity.
Combining Eq. 1 and incompressibility equation gives:

 [K ] 
.
p = 0

(2)

For sake of simplicity, we assume in this paper that the principal orthogonal axis of the
reinforcement coincides with the reference coordinates system, such as [K ] reduces to diagonal
matrix. Eq. 2 can then be reduced to Laplace equation in the isotropic equivalent domain by
stretching the coordinates. In the Cartesian coordinate system, the equivalent coordinates


xe = ( xe1 , xe2 ) may be deduced from the physical coordinates x = ( x1 , x2 ) using the following
transformation:

Ke
xi
Ki

xei =

(3)

where K i is a principal value of [K ] and K e the equivalent permeability given as K e = 3 K1 K 2 K 3 .


In the isotropic equivalent domain, Eq. 1 and Eq. 2 reduce simply to:

K 

ve = e p

(4)

p = 0

(5)

Analytical solution. Following the approach introduced by Adams et al. [6], an analytical
solution is used to calculate the evolution of the flow front versus time. Computations are
performed in the isotropic equivalent system, using the equivalent permeability K e and the
spherical injection gate of radius r0e . A constant pressure injection p0e is applied, so that the
macroscopically resin front shape is spherical of radius r f e . In such configuration the resin pressure
varies only in the radial direction according to:

1 1

p pf
r0 e r
= 1
1
1
p0 p f

r0e r f e

(6)

Substituting the expression of the gradient pressure into Darcys law and introducing the system
porosity gives the radial component of the front velocity v f e :

v fe =

Ke

(p

pf

1
1
r fe

r0

e r fe
2

The evolution of the flow front is governed by the following differential equation:

(7)

Key Engineering Materials Vol. 446

K
dr
= e
dt fe

(p

pf

133

)
(8)

1
1
r fe
r0 r f
e
e
2

Finally, the time dependant flow front position can be deduced by integrating the previous equation
during the filling stage. This leads to the following solution:

rf e

r0e

K e ( p0 p f )
rf

.2 e 3 + 1 = 6
t
2
r

r
0
e

0e

(9)

Eq. 9 gives immediately the flow front which is spherical in shape for an isotropic reinforcement.
For the orthotropic case, the reciprocal transformation is applied to give the three semi-axis of the
flow front which is ellipsoidal in shape.
umerical Simulation
Boundary Element Method. The calculation domain is delimited by its boundary

= p q A know pressure p (resp. normal pressure gradient q ) is prescribed on the


*

boundary surfaces p (resp. q ). Eq. 4 is multiplied by the Green function p . Integrated twice
over the calculation domain leads to the well-known Somigianas equation [7,8]:

ci pi + pq * d = p * q d with ci =

(10)

Where pi is the value of the pressure at a point M i located on the boundary (which is supposed to
*
*
be smooth), q the pressure gradient associated with p and is the internal angle of the corner
*
*
in radians. For a three-dimensional domain, p and q are given as:

p* =

1 1
1 r.n
*
and q =
4 r
4 r 3

(11)

where r = M i P , M i the point of application of the Dirac delta function and P the point under
consideration. Meshing the boundary into  constant boundary elements and applying Eq. 10
gives:


ci pi + p j q * d = q j p * d
j =1

j =1

(12)

2
Introducing the  matrix H and G , Eq. 12 can be transformed into:

j =1

j =1

p j H ij = q j Gij
where H ij =

(13)

1
ij + q * d and Gij = p * d . Finally Eq. 13 is reordered to take the form of a
2
j
j

linear system as:

AX = F

(14)
2

Where X is the vector of  unknowns, A is a  matrix obtained by reordering H and G . F


is the known vector computed from the boundary conditions, H and G matrix.

134

Structural Analysis of Advanced Materials

Fluid domain updating. The fluid domain is continuously expanding as it flows, constraining to
update the resin flow front. The filling stage is regarded as a succession of quasi-steady states, so
that the geometry is redefined at each time step. Unknown pressures p and normal pressure
gradients q are calculated using Eq. 14. Nodes on the moving boundary are relocated using a
backward Euler integration scheme:


 
x (t + t ) = x (t ) + t (v .n ) n
(15)
Combining Eq. 15 and Eq. 4 yields to:



q 
x (t + t ) = x (t ) + t ( K e n )

(16)

The procedure updates the resin flow front, but induces confined mesh degeneration. To overcome
this difficulty, an additional adjustment is performed relocating the nodes in contact with the mold.
Algorithm. The implemented algorithm is summarized below:
Initial computation domain is meshed.
Stationary boundary conditions p or q are applied on each BEM element.
Nodes coordinates are transformed according to Eq. 3.
Normal pressure gradients q are calculated according to Eq. 14.
Normal pressure gradients are prescribed on each node averaging values from surrounding
elements.
Nodes on the moving boundary are relocated according to Eq. 16. Nodes in contact with the
upper mold wall are relocated according to a homogenization procedure.
New normal pressures p and normal pressure gradients q are calculated next steps. The
three last steps are repeated until the allotted mold-filling time is reached.
Nodes coordinates are transformed according to the reciprocal transformation.
Validation of numerical simulation
Boundary conditions. The simulation consists in propagating the resin front into an anisotropic
reinforcement characterized by its permeability tensor [K] and its porosity . A constant pressure
p 0 is prescribed at the spherical inlet of radius r0 . On the flow front, the pressure p f is constant
and set by the vacuum pump. Processing and rheological parameters assigned are referenced in
Table 1. They keep close to the LRI process conditions but the distribution medium is not taken into
account as a first approach.
Fig. 1 shows the boundary conditions prescribed on the boundary of the initial fluid domain. The
non penetration condition is assumed assigning q = 0 at the upper mold wall.

[K ]

(10 -11m)

0.1

4 0 0
0 2 0

0 0 1
p0

(Pa)

1.10-5
Figure 1: Boundary conditions

(Pa.s)

pf

(Pa)

-1.10-5

0.5

r0

(m)
3.10-2

Table 1: Processing parameters

Key Engineering Materials Vol. 446

135

Comparison with analytical solution. For symmetry reasons, the simulation is performed on a
quarter domain meshed using 540 elements as shown in Figure 2. The step time is set at
10-2 s. CPU time for the complete simulation is around one hour on Intel Core2 Duo CPU (2.26
GHz, 1.93 Go of RAM) using Matlab environment. The computed flow fronts are in fair
agreement with the analytical one as shown in Figure 3. At 10 s elapsed time, the deviations
between the numerical and analytical flow fronts positions do not exceed 1%. Prolongating the
simulation until 60 s gives deviations around 3%, due to the gradual mesh degeneration as the
iterations go along. Figure 4 illustrates the flow front evolution during the filling stage, comparing
the front shapes at t=0.5 s and t=10s.

x axis

y axis

z axis

Figure 2: Initial mesh domain

Figure 3: Comparison with analytical solution

The initial mesh generation is performed on the equivalent isotropic domain. In order to assess the
mesh effect on the numerical method, some simulations have been carried out meshing the physical
domain. Deviations are all the more important that the permeability values are different. This is due
to the transformation from Eq. 3 which stretches the mesh, degenerating elements before any
calculation.

Flow front at t=0.5 s

Flow front at t=10 s

Figure 4: Flow front computed at t=0.5 s and t=10 s

136

Structural Analysis of Advanced Materials

Conclusion
The 3D radial front propagation is accurately simulated. The proposed technique is able to give a
high accuracy of the front position, making it suitable for front tracking in permeability
measurement [9]. It is a great advantage comparing to Volume Of Fluid techniques.
However, the proposed method is not directly applicable to complex shapes, because there is no
contact algorithm implemented. In order to perform it, current works [10] focus on fixed grid
approach, more suitable to handle contact (in particular merging and dividing fronts).
Future works will include hydro-mechanical couplings occurring during Liquid Composite Molding
process, such as fibrous perform deformations.
Acknowledgments
This work was supported by DAHER-Socata and CRCC-Resource and Competencies Center for
Composite Materials. The authors are especially grateful to M. Niquet for its assistance during this
works.
References
[1] N.C. Correia, F. Robitaille, A.C. Long, C.D. Rudd, P. Simacek, S.G. Advani, Analysis of the
vacuum infusion moulding process: I. Analytical formulation, Composites: part A 36 (2005).
[2] M.-K. Um and W.I. Lee, A study on the Mold Filling Process in Resin Transfer Molding,
Polymer Engineering and Science, Vol. 31, No. 11 (Mid-June 1991).
[3] F.M. Schmidt, P. Lafleur, F. Berthet, P. Devos, Numerical Simulation of Resin Transfer
Molding Using Linear Boundary Element Method. Polymer Composites. Vol. 20, No. 6, 725732 (1999).
[4] S. Soukane, F. Trochu. Application of the level set method to the simulation of resin transfer
molding. Composites Science and Technology 66, 1067-1080 (2006).
[5] H. Darcy, Les Fontaines Publiques de la Ville de Dijon. V. Dalmont, Paris, 1856 (in French).
[6] K.L. Adams, W.B. Russel, L. Rebenfeld. Radial penetration of viscous liquid into a planar
anisotropic porous medium. International Journal Of Multiphase Flow, Vol. 14, No. 2 (1988).
[7] C.A. Brebbia, J. Dominguez, Boundary Elements : an introductory course. 2nd ed., McGrawHill Company: Computational Mechanics Publications (1992).
[8] N. Pirc, F. Bugarin, F. Schmidt, M. Mongeau. BEM-based cooling optimization for 3D
injection molding. International Journal for Simulation and Multidisciplinary Design
Optimization, Vol. 2, No. 3 (2008).
[9] P.B., Nedanov, S.G. Advani. A method to determine 3D permeability of fibrous
reinforcements. Journal of composite materials (2002).
[10] R. Gantois, A. Cantarel, G. Dusserre, J.-N. Flices, F. Schmidt. Numerical Simulation Of
Resin Transfer Molding Using BEM and Level Set Method, 13th ESAFORM Conference on
Material Forming, Proceedings (2010).

Key Engineering Materials Vol. 446 (2010) pp 137-145


(2010) Trans Tech Publications, Switzerland
doi:10.4028/www.scientific.net/KEM.446.137

FEA of dynamic behavior of Top hat bonded stiffened composite panel


G. Mohamad1, a, M. Tarfaoui1, b and V. Bertram2, c
1

ENSIETA, LBMS, Mechanics of Naval and Offshore Structures Department, Brest, France
2

Germanischer Lloyd, Vorsetzen 35, D-20459 Hamburg, Germany

mohamagh@ensieta.fr, btarfaomo@ensieta.fr, cvolker.bertram@gl-group.com

Keywords: Top-hat, dynamic loading, stiffened structure, naval construction, composite

Abstract. In this work the dynamic responses of bonded top hat stiffened panel have been studied
by using finite element analysis model (Abaqus). A symmetrical 2D model was performed and used
in the simulations. In the first part, dynamic behavior and impact speed effects at such composite
structure have been studied. In the second part, other simulations were carried out by using cohesive
elements proposed to predict the delamination might happen under such loading.
Introduction
Actually, composite materials have a great success in the advanced structures. All this structures
need the satisfaction of the above properties of materials to the maximum extent. In the naval
vessels and speed aircraft, explosion might be happened at any time, so it is very important to study
the materials and the structure under such loading.
To investigate the dynamic response, S. Goswami [1] has studied laminated shells under impact
loading. He has studied also transient dynamic response of laminated composite stiffened shell [2].
He proposed a formulation with which the stiffeners can be placed anywhere inside the shell
element giving more flexibility in mesh division. They found also that by increasing the number of
stiffeners, the structural stiffness can be increased with reduction of deflection values, and that the
positioning of the stiffeners is also of prime importance. Also, it is found that the ply orientation of
[45/-45]s present lower deflection values than the other laminate orientation of [90/0]s. Many other
researchers had been interested by studding the dynamic responses of composite sandwich panels
[3-6]. Dynamic responses of top hat bonded stiffened panel have been studied in this work. First of
all, one has tried to study the response of this stiffened structure under various impact speeds to
have information about impact speed effect at such composite structure. In the second and the third
parts, cohesive elements were used to predict delamination and interface failure under dynamic
loadings.
In the present study, the dynamic flexural behavior of top hat bonded stiffened panel, with
composite face-sheets and a balsa core, was analyzed by developing a 2D finite-element model.
Two cases are considered. The first, the model is based on a linear elastic behavior of materials and
without taking into account of the damage and the second is based on linear elasticity fracture
mechanics (LEFM) to describe the structural damaging modes. To model separation between
different elements of structure, a cohesive zone model (CZM) was used. The CZM were also used
to predict the delamination of the face-sheets. Dynamic bending simulations were performed for
various velocity impacts.
Comparison between different cases in terms of contact-force histories, peak-force values, and
maximum displacement was performed. A scenario of damage is proposed for different velocity
impact. The results of these analyses are then studied at various velocity impacts to establish the
significance of the dynamic consideration in predicting the structural response of such structures.
Material
Sandwich panels with balsa core are employed in this study. Woven glass impregnated in
polyester resin is used like composite material in the top-hat stiffener and the skins. To fill the space

138

Structural Analysis of Advanced Materials

between the stiffeners, the core and the adhesive joint one used resin material with mechanical
property presented in Table 1 and 2.
Table 1. Material property
Material

Location

Araldite 2015

Adhesive joint

Resin

Fillet

Property
E [MPa]

E [MPa]

Value
2000
0.277
1369

0.34

Table 2. Material property


Mechanical property
E1
E2
E3
12
13
23
G12
G13
G23

Composite
22000
22000
8900
0.27
0.38
0.38
5300
3170
3170

Balsa
50
3518
50
0.02
0.5
0.02
157
157
157

where:
E1 , E 2 and E 3 : Young Modulus in three directions
12 , 13 and 23 : Poisson's ratio
G12 , G13 and G23 : Shear Modulus
Numerical investigations
Some dynamic experimental tests on top-hat stiffener are expected to do in the future, Fig. 1. In
order to prepare for the experimental test, some numerical simulations were done. An impact with a
three point bending was used in this finite element analysis. The structure was impacted at the
middle of the lower surface of the panel, Fig. 1.

Fig. 1. Impact test apparatus


A symmetrical 2D explicit model with quadratic elements, standard CPS4R and elements CPS3
are used to mesh top-hat stiffener, core, base panel and fillet. On the other hand rigid elements types
R2D2 are used to model the two supports and the load application, Fig. 2. On figure 2, examples of
mesh are given. For model without damage, 1245 quadratic elements for meshing the panel, 450
elements for the stiffener, 23 elements for adhesive, 607 elements for stiffener core and 3 linear
triangular elements with 2 linear quadrilateral elements in the fillet were used to mesh the structure,
Fig. 2a. In the other hand, 7875 quadratic elements for meshing the panel, 2337 elements for the

Key Engineering Materials Vol. 446

139

stiffener, 178 elements for adhesive, 708 elements for stiffener core and 63 elements in the fillet
were used to mesh the model with damage, Fig.2b.

(a) FEM without damage

(b) FEM with damage

Fig. 2. Stiffened panel under dynamic load


Model without damage of dynamic responses under different impact speeds
In this part, a linear elastic model is developed. One will be interested in the localization of the
zones of stress concentration. The stiffened panel was impacted under several speeds: 1, 27,
10, and 15 m/s. The principal aim is to have general information about structural response without
damage. In first time, a comparison between Von-Mises stress values at the contact point was done.
Figure 3 shows that the stress level increases when the impact speed increases. The maximum VonMises level was found near the impacted zone.
450

3m/s

400

5m/s

Von mises stress (MPa)

10m/s

350

15 m/s

300
250
200

150
100
50
0
0

0,1

0,2

0,3

0,4

0,5

Time (ms)

(a) VonMises stress

(b)Von-Mises stress compraison (first contact)

Fig. 3. Von-Mises stress evolution at impact point


200

70
60

CPRESS (MPa)

50
CPRESS (MPa)

3m/s

180

5 m/s

40
30
20

160

5m/s

140

10m/s

120

15m/s

100
80

60
40

10

20
0

0
0

0,5

1,5

0,1

Time (ms)

0,2

0,3

0,4

Time (ms)

(a) CPRESS at the impact point

(b) CPRESS under differents impact speed

Fig. 4. Contact pressure (CPRESS)

140

Structural Analysis of Advanced Materials

The structural behavior under dynamic loading follow three steps: first contact, separation and
the last contact, Fig 3-4.
In the adhesive joint, the maximum Von-Mises stress has been registered under different impact
speed. Stress level increase when the impact speed increase, Fig 5. The evolution of Von-Mises
stress was analyzed at a critical zone in the adhesive joint which is localized under the fillet.
7m/s

Impact speed (m/s)

6m/s
5m/s

4m/s
3m/s
2m/s
1m/s
0

10

15

20

25

30

Maximum Von Mises stress (MPa)

Fig. 5. Maximum Von-Mises stress in the adhesive joint


10m/s

Impact speed (m/s)

7m/s
6m/s
5m/s
4m/s
3m/s

2m/s
1m/s
0

100

200

300

400

500

Von Mises stress (MPa)

(a)

(b)

Fig. 6. Max Von- Mises stress in the structure


The maximum Von-Mises stress in the whole structure has been also analyzed and has been
compared under several impact speeds. Figure 6 shows that the maximum stress level increases
when impact speed increase. This maximum stress occurs near the impact point, Fig 6b.
This information will help us to detect stress concentration structures zones where the structure
risks having damage. Cohesive elements have been created in these zones to reproduce possible
damage.
Damage modeling with traction separation law
Interface zone laws are used to describe the relationships between tractions and displacement
jumps through the interface during the debonding process. This section intends to present the
bilinear cohesive law based on Geubelle and Baylor [7] and De Moura et al. works [8]; it describes
the multilinear traction, which links the normal (Mode I) or shear (Mode II) stresses, acting in the
crack plane, and the respective normal or shear strains. The interface law was applied to model the
ply delamination in laminated fiber-reinforced composites [9-11] and the behavior of adhesively
bonded assemblies [12]. The cohesive model formulation is based on the damage mechanics of
continuum media and Linear Elasticity Fracture Mechanics (LEFM).
When the interface is subjected to pure I, II, and III loading Modes, the delamination occurs if
the interlaminar associated stress reaches its maximum interfacial value. Obviously, under mixed-

Key Engineering Materials Vol. 446

141

Mode loading, delamination onset may occur before those maximum values are obtained. The
interaction between the stress components under mixed-Mode loading has to be taken into account
by using a multi-axial stress criterion given as:
2

t1 t 2 t 3
0 0 0 1 .
(1)

t1 t 2 t 3
where t 10 , t 20 and t 30 are maximum interlaminar traction stress vector components associated with
the three directions. These parameters can be put linearly in equation with their respective
displacement 03 , 02 and 01 ( 0i 0 ; i 1,2,3 ). Symbols . stand for the positive and negative part
2

of the argument and defined as : x

x x 2 .

Delamination propagates when the energy release rate equals its critical value under pure Mode I
or Mode II fracture or Mode III fracture. Generally, delamination growth occurs under mixed-mode
loading. In that case, the delamination process may take place before any of the energy release rate
components attains its individual critical value, Fig. 7. The most widely used criterion to predict
propagation of delamination under mixed-Mode loading is the power law criterion based on the
proposal of Reeder [13]:
GI

G
Ic

G II

G
II c

G III

G
III c

1.

(2)

where n is an empirical parameter, which is considered as characteristic of the existing coupling


phenomenon between the modes. GI c , G II c and G III c are the critical energy release rate for mode I,
II and III.

Fig. 7. Cohesive element behavior in mixed mode


Model with damage
After that stiffened panel has been numerically tested with a non damage model, thin cohesive
zones were created between the different composite plies and between the different structure parts.
The objective of this operation is to predict delamination and separation between different structure
elements under such loads.
Quadratic linear cohesive elements types COH2D4 are used. These elements can reproduce the
delamination only under traction loading, with a failure energy criterion type Benzeggagh-Kenane
(BK). 7875 elements have been used for meshing the panel, 2337 elements for the stiffener, 176
elements in the adhesive, 708 elements in the core and 63 elements in the fillet within 3 linear
triangular elements and 60 linear quadrilateral elements.

142

Structural Analysis of Advanced Materials

The structure has been impacted by several impact speeds like in the first part of this work. First
of all the contact force evolution was analyzed, at the contact impact point, with displacements
evolution, Fig. 8a. In figure 8b, the evolution of max-mid-deflection is plotted as a function of
impact speed. Table 3 present maximals contact forces and maximals displacements under
different impact speeds.
25

22

4m/s
5m/s
10m/s
15m/s

Contact force (kN)

18
16

20
Deflection (mm)

20

14
12
10
8

15

10

4
2

0
0

10

12

14

16

Impact speed (m/s)

Displacement (mm)

(a) Contact Force vs. Displacement

(b) Max-mid-deflection

Fig. 8. Contact force and Max-mid-deflection under different impact speeds


Table 3. Load and displacement
Impact speed (m/s)
Maximal load (kN)
Maximal displacement (mm)

4
12.58
1.62

(a) Stress concentration, impact speed 4m/s

5
14.99
1.90

7
19.82
2.69

10
27.24
3.89

15
41.12
5.80

(b) First damage observed under 5m/s

(c) A separation under a speed of 20m/s

Fig. 9.Damage examples under different impact speeds

Key Engineering Materials Vol. 446

143

The first damage onset was observed under impact speed of 5 m/s. For lower velocity impact,
there was stress concentration in the same region but there was no damage in the structure, Fig. 9ab.The first damage was observed as a separation between top-hat and fillet, Fig. 9b. After that
composites plies delamination in the composite top-hat in the curved zone may take place. When
this bonded stiffened panel was impacted with high impact velocity as 15 or 20 m/s, a total
separation between the top hat stiffener and the base panel happed Fig. 9c.
Figure 10 shows that damage initiation increase by increasing impact speed. The stiffness
increase to have a maximal under certain impact speed and after that it become less than before.

45
4m/s
5m/s
10m/s
15m/s

40
35

Load (kN)

30
25
20
15
10
5
0
0

Displacement (mm)

Fig. 10. Load vs. displacement under four impact speeds

Damage history
To track the damages scenario of such composite structure under this impact loads, first of all,
stress concentration zones were tracked under different impact speeds. Then we have localized the
first structural damage and the correspondent impact speed. Afterward, we have increased impact
speed to study the damages evolution and to localize final damage (structure failure), Table 4.

144

Structural Analysis of Advanced Materials

Table 4. Damage history


Impact speed
(m/s)

Stress concentration
before damage onset

Initial damage

Final damage

No damage

No damage

Interface hat stiffener /


adhesive joint

Interface hat stiffener / Fillet

Interface hat stiffener / Fillet

Delamination in the first ply


of the stiffener

In the adhesive joint

Interface hat stiffener /


adhesive joint

Delamination in the plies of


the stiffener

In the curved zone

Interface stiffener / Fillet

Delamination in the first ply


of the stiffener

In the curved zone

In the curved zone

In the curved zone

10

15

Summary
A dynamic numerical work on a composite bonded stiffened structure has been presented in this
paper. In these work certain dynamic responses of such structure under different impact speed were
studied. Cohesive elements have been used to simulate the separation between the different parts of
structure and damage scenario has been analyzed. The general conclusions are:
The maximal Von-Mises stress increase with the rise of impacts speed both in the
structure and in the adhesive joint,
Max-mid deflection rise when the impact speeds increase.
The first damage was observed under 5m/s.
The initial damage was observed in the curved zone of the composite top-hat.
The maximum force increase with the increase of impact speed, but one have
observed stiffness decrease after certain impact speed value.
In order to evaluate the reliability of the numerical model, dynamic three-point bending tests
were carried out.

Key Engineering Materials Vol. 446

145

References
[1] Goswami S. Finite Element Modeling for the Impact Dynamic Response of Laminated
Composite Stiffened Shells, Mechanical Engineering Department, National Chiao Tung
University, Hisin Chu 30050, Taiwan, China.
[2] Goswami S. Geometrically Non- Linear Transient Dynamic Response of Laminated Composite
Stiffened Shells, Department of Ocean Engineering and naval Architecture, Indians Institute of
Technology, Kharagpur 721302, Onest Bengale INDIA.
[3] Wahab M. Study of the dynamic behaviour of sandwich plates under impact, Bordeaux, 1997.
[4] Song J., Wong H. and Belytschko T. A comparative study on finite element methods for
dynamic fracture, Springer- Verlag 2007.
[5] Kevin M, Tuner A., Vizzini J. Response of Impacted Sandwich Panels With Integral Stiffeners,
Journal of Sandwich Structures and Materials, 6,313, 2004.
[6] Vaidya A.S., Vaidya U.K., Uddin N. Impact response of three-dimensional multifunctional
sandwich composite, Journal of Sandwich Structures and Materials, Materials Science and
Engineering A, Volume 472, 52-58, 2008.
[7] Geubelle P. H. and Baylor J. S. Impact-induced delamination of composites: a 2d simulation.
Composites Part B, 29B:589-602, 1998.
[8] De Moura M. F. S. F., Conalves J. P., Marques A. T. and de Castro P. T. Prediction of
compressive strength of carbon-epoxy laminates containing delaminations by using a mixedmode damage model. Composite Structures, 50:151-157, 2000.
[9] Camanho P. P. and Davila C. G. Mixed-mode decohesion finite elements for the simulation of
delamination in composite materials. Technical Memorandum-2002-211737, NASA Langley
Research Center, 2002.
[10] Turon A., Davila C. G., Camanho P. P., and Costa J. An engineering solution for using coarse
meshes in the simultaion of delamination with cohesive zone model. Technical Memorandum2005-213547, NASA Langley Research Center, 2005.
[11] Hamitouche L., Tarfaoui M., Vautrin A. An interface debonding law subject to viscous
regularization for avoiding instability: Application to the delamination problems, Engineering
Fracture Mechanics, Volume 75, Issue 10, Pages 3084-3100, 2008.
[12] Campilho R. D. S. G., de Moura M. F. S. F. and Domingues J. J. M. S. Modelling single and
double-lap repairs on composite materials. Composites Science and Technology, 65:19481958, 2005.
[13] Reeder JR. An evaluation of mixed-mode delamination failure criteria. Technical
Memorandum 1992-104210, NASA Langley Research Center, 1992.

Key Engineering Materials Vol. 446 (2010) pp 147-155


(2010) Trans Tech Publications, Switzerland
doi:10.4028/www.scientific.net/KEM.446.147

Experimental characterization of behavior laws for titanium alloys:


Application to Ti5553
Vincent WAGNER1,2,a, Maher BAILI1,b, Gilles DESSEIN1,c,
Daniel LALLEMENT2,d
1

Universit de Toulouse ; INP/ENIT ; LGP ; 47 avenue d'Azereix, BP 1629, F-65013 Tarbes


Cedex, France
2

Messier Dowty Bidos ; rue Guynemer, BP 39, F-64401 Oloron Sainte Marie
a

Vincent.Wagner@enit.fr, bMaher.Baili@enit.fr, cGilles.Dessein@enit.fr,


d
Daniel.Lallement@messier-dowty.com

Keywords: Ti5553; Cutting force; cutting temperature; cutting conditions; tool geometry; wear;
thermal effect; stress; tensile yield stress (TYS); ultimate tensile stress (UTS).

Abstract. The aim of this paper is to study the machinability of a new titanium alloy: Ti-5AL-5Mo5V-3CR used for the production of new landing gear. First, the physical and mechanical properties
of this material will be presented. Second, we show the relationship between material properties and
machinability. Third, the Ti5553 will be compared to Ti64. Unless Ti64 is + alloy group and
Ti5553 is a metastable, we have chosen to compare these two materials. Ti64 is the most popular of
titanium alloys and many works were been made on its machining. After, we have cited the Ti5553
properties and detailed the behavior laws. They are used in different ways: with or without thermal
softening effect or without dynamic terms. The goal of the paper is to define the best cutting force
model. So, different models are compared for two materials (steel and titanium alloy). To define the
model, two methods exist that we have compared. The first is based on machining test; however the
second is based on Hopkinson bar test. These methods allow us to obtain different ranges of strain
rate, strain and temperature. This comparison will show the importance of a good range of strain
rate, strain and temperature for behavior law, especially in titanium machining.
Introduction
The aircraft industry uses materials more and more efficient. This trend affects the majority of parts
such as structure parts. Titanium alloy (Ti-5AL-5Mo-5V-3CR) is now used for the production of
landing gear. The poor titanium alloys machinability is explained by Ezugwu in [1]. The Ti5553 can
be classified in hard to cut material and there is a low knowledge in Ti5553. Arrazola ([2])
compares the Ti5553 and the Ti64 alloys: the Ti5553 machinability is very poor compared to Ti64
one. No cutting force model has been defined in his work. However, some works have been made
on the cutting model and especially on Ti64 machining ([3]). The machining is a process where
stress appears according to strain, strain rate and temperature. The temperature is the result of the
high level of strain, strain rate and stress. The cutting model must have a law that defines this
phenomenon. There are a lot of behavior laws and they are never used in the same ways. Moreover,
few researchers study the influence of the behavior law in model of cutting force and cutting
temperature, especially, in hard machining material where the cutting forces are very sensitive to
cutting conditions. Changeux ([4]) shows the influence of behavior law in cutting force modeling
and suggest a johnson-Cook model as the well adapted one for machining. However, his work has
been made for austenitic steel. In his work, Poulachon ([5]) explains that the identification methods
of behavior law can be classified by the strain rate range. He shows the difference between
machining strain rate and identification methods strain rates. To allow a higher range of strain rate,

148

Structural Analysis of Advanced Materials

[3,6] detail a method to identify the constants of behavior law from machining. Fanning ([7])
depicts the Ti5553 material properties but for a not treated Ti5553.
There are many goals in this paper. First, the physical and mechanical properties will be presented.
We show the relationship between material properties and machinability. The Ti5553 will be
compared to Ti64. The Ti64 is + alloy group and Ti5553 is a metastable beta but we have chosen
to compare it with this material. Ti64 is the most popular of titanium alloys and many works were
made on its machining. After, we have shown the Ti5553 properties and we have detailed the
behavior laws. They are used in different ways, with or without thermal softening effect or without
dynamics terms. The target is to define the best model to use in cutting force model. Different
models are compared for two materials (steel and titanium alloy). Another important point is the
model definition. There are different ways to define the constant model. We compared two methods.
The first is defined from machining test and the second from Hopkinson bar test. This comparison
will show the importance of a good range of strain rate, strain and temperature for behavior law.
Physical properties
Phase diagram.
The Ti5553 is a metastable beta with a low transus beta (occurs at 845C). The titanium alloy
Ti64 belongs to + alloys group, its transus beta occurs at about 990 C. In machining, this
difference between the two alloys can be important. In his paper, Ozel ([3]) uses the Oxley model
([8]) to estimate the cutting temperatures. He shows in [3] that the cutting temperatures of Ti64
machining are range from 750C to 940C. Consequently, they are always less than the transus beta
temperature. So, the titanium structure + and its concentration do not change.
The mechanical properties of Ti5553 compared to the Ti64 alloy are higher and the physical
properties are almost the same (Table 1). The cutting temperatures will be also higher when
machining Ti5553 alloy. In addition to that, they are superior to the transus beta temperature; as a
result, the alloy structure will be changed. The beta phase concentration becomes more important.
For Arrazola ([2]), the differences of structures with a variable quantity of the alpha phase and the
morphology of the transformed + phase can explain the poor Ti5553 machinability. This
modification of phase will be important, especially in machining of structure part.
Thermal conductivity.
The Ti5553 alloy has a poor thermal conductivity and can be classified like heat resistant material.
The Figure 1(a) shows the thermal conductivities of Ti5553, Ti64 alloys and steel. Indeed,
according to temperature, we can see whatever the steel thermal conductivity is the highest. It
decreases when temperature increases whereas the titanium thermal conductivity increases. If we
compare the titanium values, the Ti5553 thermal conductivity is always higher than the Ti64 one.
This poor thermal conductivity can generate problem in machining. In fact, it limits the heat
evacuation in chip. In his paper [1], Ezugwu depicts the distribution of thermal load when
machining titanium and steel. In titanium machining, a large proportion (80%) of the heat generated
is conducted in tool. However, in steel machining only 50% of heat generated is conducted in tool.
This difference can be explained by the higher thermal conductivity of tool ( 60W / m / C )
compared to titanium alloy thermal conductivity ( 8W / m / C ). Whereas, for the used machining
tool, the thermal conductivity compared to steel one are quite the same. The heat is being quickly
conducted in tool, so the thermal softening can be easily obtained.
In their models, [8] and [10] demonstrate that temperatures in shear plan and at the tool-chip
interface are function of thermal conductivity (see Eq. 1 and Eq. 2).
TAB =

(1 ) FS cos
St u w cos( )

(1)

Key Engineering Materials Vol. 446

149

SVt u

(2)
tan
K

Where: TAB = Average temperature in shear plan; = Density; Heat capacity; S = Cutting velocity;
t u = Undeformed chip thickness; K = Thermal conductivity; = Shear angle; = Rake angle; w =

= 0.5 0.35 log

width of cut; FS = Shear force; V = Cutting speed.

(b)
(a)
Figure 1 : Thermal conductivity and heat capacity comparison [9]
Eq. 2 defines the heat rate conducted into the part. However, Eq. 1 represents the average
temperature in primary shear plan. When K increases, the proportion of heat conducted into the part
() is higher and leads to a reduction of temperature in primary shear plan.

Heat capacity.
The Figure 1(b) shows a comparison between heat capacity of steel and titanium alloys. To the
contrary of thermal conductivity, the heat capacity of the three materials increases according to
temperature. The steel heat capacity is higher for low temperatures ( 575 J / kg / C ) whereas the
Ti64 is the lowest ( 460 J / kg / C ). When the temperature increases, the Ti5553 heat capacity
becomes highest ( 775 J / kg / C at 600C ). This value is 20% upper than Ti64 one. For Ti5553, the
increasing heat capacity between low and high temperatures is more than 35%, whereas it is only
about 15% for Ti64.
In [8,10] models, the cutting temperature is inversely proportional to the heat capacity (see
Eq. 1). We can see that increasing heat capacity leads to reduction of the cutting temperature. Like
thermal capacity, with a high heat capacity it is difficult to obtain the thermal softening. The cutting
temperature can be limited by these high values and can be less than beta transus.

Mechanical properties
Tensile test.
Figure 2 shows the true stress according to true strain. This value was obtained with ambient
temperature and with constant deformation speed. The Young modulus, strain, tensile yield stress
(TYS) and ultimate tensile stress (UTS) were identified from the tensile test. The values are noted in
Table 1. First, we can say that the Ti5553 material properties values are higher compared to Ti64.
The ultimate tensile stress and tensile yield stress are 20% higher. The second point is the low
ductility: 1.5%. In comparison with Ti64, these values are very low and show that the Ti5553 is a
brittle material.

150

Structural Analysis of Advanced Materials

Figure 2 : Tensile tests Ti64 and Ti5553


The material properties influence the machining responses and can explain the high cutting force
values. The Ti5553 ultimate tensile stress is also higher even at 400C in machining and the cutting
forces are also higher. Another consequence of high mechanical properties and poor ductility, it will
be difficult to get a segmented chip.
Material

TYS (MPa)

Ti5553
Ti64

1250
1050

Table 1 : Material properties of Ti5553 and Ti64


UTS (Mpa)
Modulus
Elongation
(GPa)
(%)
1320
1200

110
105

1.5
18

Hardness
(HB5)

TYS at 400C
(Mpa)

370
241

860
550

Behavior law
There are two goals in this section. First, we will determine the best model defining the cutting
forces and temperature in machining. Secondly, we will propose the best solution to define the
behavior law.
The behavior laws are used to determine the stress according to strain, strain rate and temperature.
Some different laws are used in modeling of cutting forces and cutting temperatures. [8] uses the
Hollomon model and neglects the temperature and strain rate effects. [3], [11] and [12] use the
Johnson-Cook law. Other models can be considered. Lindholm model defines stress according to
strain and strain rate. However, it does not take into account the temperature. The Norton-Hoff law
can also be considered. Johnson-Cook model seems to be the best to model the cutting forces ([4]).
m

 T To
= A + B( )n .1 + C ln .1
(3)
0 Tm To

However the Johnson-Cook model (Eq. 3) is used in different ways. In their works, [11] and [12]
consider an adiabatic process and neglect the heat influence in primary shear plane. They determine
the shear flow stress along primary plan from the restricted Johnson-Cook law. The thermal
softening effect is not inserted in restricted Johnson-Cook model. In [3], Ozel used also this model
but in its complete definition with thermal softening effect. The target is to verify these assumptions
and to see if it is always true in high mechanical and poor thermal properties materials. The titanium
alloys have a poor thermal conductivity and a high heat capacity. The heat is rapidly conducted in
tool ([1]). The Johnson-Cook model must have the thermal softening effect to define the cutting
shear flow stress. The tensile test depicts that the titanium alloy are brittle. The strain rate constant
(C) must be integrated in the model.
In this first part, we set the parameters to insert into the model. To define the best model, we have
compared two materials: AISI1045 steel and Ti64 titanium alloy. For each material, the Johnson-

Key Engineering Materials Vol. 446

151

Cook model have been used in three different ways and we have compared the theoretical results
( JC ; JC1 ; JC 2 ) and the experimental results AB , where AB is effective flow stress in primary
shear plan defined in Oxley model from machining ([8]) represents the stress obtained with the
Johnson-Cook model. JC1 and JC 2 are the stress obtained without the thermal softening effect
and without the strain effect (Eq. 4 and Eq. 5). In [3], the Johnson Cook model used is defined by
Hopkinson bar test.
JC 2 = A + B( )n
(3)

n
JC1 = A + B( ) .1 + C ln
(4)
0

m
 T To
n
JC = A + B( ) .1 + C ln .1
(5)

0 Tm To

The results for the two materials and the three models are presented in Figure 3. Figure 3(a)
shows the comparison between theoretical results and the experimental results for Ti64. However,
Figure 3 (b) depicts the same comparison for AISI1045. For steel, the best model seems to be the
complete Johnson-Cook model. The errors between the values defined in machining and with the
Johnson-Cook models are JC = 7% , JC 2 = 9% and JC = 7% . For steel, the second model is the
worst one. The best values obtained for Ti64 are also obtained with the complete model (Eq. 6). The
errors between the values defined in machining and with the Johnson-Cook models are JC = 5% ,

[
[

]
]

JC 2 = 30% and JC = 5% . Like for steel, the worst model is the model without thermal effect.
First, this analysis shows the importance of thermal softening effects in shear plan model.
Without thermal effect, the effective flow stress values are higher. It seems to be important to use
the complete Johnson-Cook model in steel and in titanium alloy. Moreover, the mechanical
properties of titanium remains high even at high temperature.
Figure 3(b) depicts that the titanium model seems to be more sensitive to the behavior law. The
good fit for steel is due to the good range of strain rate, strain and temperature of Johnson-Cook
model. The used model is defined at test strain up to 10 4 s 1 and a temperature up to 600C . The
strain rate and the temperature defined in cutting modeling are often in this range
( 3399 <  AB < 12563 and 428 < TAB < 494 ). For Ti64, the material law is defined for a strain rate of
and a temperature up to 1100C . Our calculations show some strain rate values up to the strain rate
model 1780 <  AB < 84993 and a temperature between 531C and 671C . The difference between
model strain rate range and machining strain rate range is the error explanation. It seems to be very
important to define correctly the range of strain rate and temperature, especially, when machining
generates high strain rate.

(a : Ti64)

(b : Steel)
Figure 3 : Behavior law comparison

152

Structural Analysis of Advanced Materials

The constant of Johnson-Cook model can be defined by different methods. Poulachon ([5]) explains
that the method must be chosen according to the strain rate (as illustrated in Figure 6). The split
Hopkinson bar test is the most commonly used method for determining stress according to high
rates of strain. The Taylor test may be also considered. It allows obtaining higher strain rate
( 5 10 3 s 1 <  < 5 10 6 s 1 ) ([13]). Another method can be considered. Indeed, Ozel ([3]) and
Tounsi ([6]) have developed a method to define the Johnson-Cook material model from machining
test. There are different advantages with these methods. First, the strain rate and temperature are
always those obtained in machining. Secondly, the impact test is not always feasible for brittle
material. The increasing of temperature can raise the ductility of the material. Thus, impact test
becomes possible.

(a : Ti64)
(b : Steel)
Figure 4: Comparison between two identification methods
We have compared two material models. The first model has been defined with Hopkinson bar test
and the second from machining tests. We have chosen to use the model defined in [3]. This test has
been made on titanium alloy (Ti64). AB is the effective flow stress at primary shear zone and has
been determined from Oxley model. HB is the stress defined from Johnson-Cook model and ,
machining the stress obtained with behavior law identified from machining.
Figure 4 shows the stress obtained from the two models. The tests have been made in different
cutting speed (Vc) and different undeformed chip thickness (tu). For steel, machining and HB are
quite equal. The identification method does not seem to influence the results. For, Ti64, it can be
observed that sometimes is better and sometimes is the best. An analysis shows that the results are
function of strain rate values. For low strain rate values (Vc15-tu0.1, Vc15-tu0.08), the best model
is the model obtained with Hopkinson bar test. Whereas, for high strain rate (Vc15-tu0.02; Vc60tu0.02) the best model is one that is defined from machining test. This analysis shows the
importance of the good definition of model range. The Hopkinson bar test allows a maximum strain
rate of while strain rate machining is 8500 s 1 . Moreover in analyzing, the behavior law obtained
from machining test limits the influence of strain rate constant ( C HB = 0.028 and
C machining = 0.000002125 ). In high strain rate, the stress is also reduced with increasing of strain rate.

This observation seems to be coherent with machining tests and explain the stress decreasing.
The other points are the temperature effect. To explain the difference between the two materials
some response surfaces have been plotted. Figure 5(a) and Figure 5 (b) show the stress plotted
according to strain rate and temperature. In machining, machining is always lower at HB . The two
surfaces show the difference between two identifications methods. In studied range temperature and
strain rate, the difference between the stresses is different for the material. The Figure 5(a) shows
that the two surfaces obtained are close and almost parallel. They are closer to high temperatures. In
machining test, machining and HB are the same, the results seems to be coherent. For Ti64, the two
surfaces are more remote and the slope is different. Increasing of temperature reduces the gap
between the two surfaces. However, in comparison with steel, the gap is higher. This analysis shows

Key Engineering Materials Vol. 446

153

the importance of thermal softening above all in titanium alloy machining and the importance of a
good behavior law.

(a)

(b)
Figure 5 : Response surfaces

The modeling of cutting force in titanium alloy machining is more sensitive to behavior law. The
complete model must be used to define the shear flow stress along primary zone. The behavior law
must be defined with Hopkinson bar if strain rate values are good (Figure 6). However, another
point to consider is the temperature. It is very important above all in machining of titanium alloy
where the strain rate seems to be higher.

Figure 6 : Strain-rate range reached in metalworking and limitations of the standard mechanical
tests
For Ti5553, the first term of Johnson-Cook has been made. A is the yield stress when B and n
represent the effects of strain hardening. These values show the high properties of Ti5553. If
compared with Ti64 yield stress, the Ti5553 one is higher. The table 2 illustrates that the effect of
strain hardening in Ti5553 is higher compared to Ti64. These values show the poor ductility of
Ti5553. Moreover, the effect of strain hardening is higher for Ti5553.
Table 2 : Johnson Cook law
Material

Ti5553
Ti64

1250
783

403
498

0.47
0.28

In using those results and works established in [14], we have defined the strain obtained in Ti5553
machining. The test has been made with constant undeformed chip thickness ( t u = 0.2mm ), constant
rake angle ( = 13 ) and a low edge preparation. We can denote that Ti5553 machining creates a
very high strain rate (table 3). Those results show the importance of the behavior law definition. In
this case, Hopkinson bar test do not allows the good strain rate. To get the good strain rate, two
methods are possible, the Taylor test which permits us to get very high strain rate
( 5 10 3 s 1 <  < 5 10 6 s 1 ) [13] or to define model with machining tests. We have chosen to use

154

Structural Analysis of Advanced Materials

test machining to define behavior law. These methods allow us to define the behavior law in good
range (strain rate and temperature) but especially to increase material ductility with the increasing of
temperature.

Tests
Vc=25m/min-Vf=0.05mm/tr
Vc=35m/min-Vf=0.05mm/tr
Vc=45m/min-Vf=0.05mm/tr
Vc=55m/min-Vf=0.05mm/tr
Vc=65m/min-Vf=0.05mm/tr
Vc=65m/min-Vf=0.1mm/tr
Vc=65m/min-Vf=0.15mm/tr
Vc=65m/min-Vf=0.2mm/tr

Table 3 : Strain rates for Ti5553


Co
 AB ( s 1 )
2.01
3.53
2.57
3.55
3.53
1.92
2.01
0.81

5172
11173
10923
17643
20722
7070
4566
1488

Fc (N)
543
575
618
583
578
941
1384
1747

Ft (N)
129
140
177
141
141
125
304
384

Conclusion
This paper shows the influence of behavior law in cutting force model. We have compared different
models of shear stress along primary plan. These tests have shown that the model must have the
thermal effect especially in titanium alloys machining. Moreover, the model strain rate range must
be the same as machining. To limit error, we have shown that the best method to define the behavior
law is the definition from machining. This method allows the good stain rate and the good
temperatures. We have shown that observations are very important in Ti5553 machining. Indeed,
the strain rate seems to be very high compared to Ti64. So, we have chosen to define a complete
Johnson-Cook model based on machining tests. Other tests must be made to define the JohnsonCook law constants. They let us define the influence of cutting conditions on cutting forces ant
cutting temperature. The Johnson-Cook law will be defined from these tests where cutting
temperature has been recorded with thermocouples. The strain rate and strain will be defined from
Oxley model ([8]).

References
[1] Ezugwu, E., & Wang, Z. (1997). Titanium alloys and their machinability - a review. Journal of
Material Processinf Technology &o. 68 , 262-274.
[2] Arrazola, P., Garay, A., Iriarte, L., Armendia, M., Marya, S., & Le Matre, F. (2008).
Machinability of titanium alloys (Ti6Al4V and Ti5553). Journal of Material Processing
Technology .
[3] Ozel, T., & Zeren, E. (2006). A metothodology to determine wprk material flow stress and toolchip interface friction properties by using analysis of machining. Journal of manufacturing science
and Engineering, Vol. 128 , 119.
[4] Changeux, B. (2001). Loi de comportement pour l'usinage. Localisation de la dformation et
aspects microstructuraux. ENSAM Paris: PhD Thesis.
[5] Poulachon, G., Moisan, A., & Dessoly, M. (2002). Contribution l'tude des mcanismes de
coupe en tournage dur. Mcanique et Industrie Vol.3 , 291-299.
[6] Tounsi, N., Vincenti, J., Otho, A., & Elbestawi, M. (2002). From the basic mechanic of
orthogonal metal cutting toward the identification of the constitutive equation. International
Journal of Machine Tools & Manufacture, Vol.42 , 1373-1383.

Key Engineering Materials Vol. 446

155

[7] Fanning, J. (n.d.). Properties of TIMETAL 5553 (Ti-5Al-5Mo-5V-3Cr). ASM International, Vol.
14 , 788-791.
[8] Oxley, P. (1989). Mechanics of machining an analyticale approach to assessing machinability.
Ellis Horwood Limited.
[9] Tricot, R. (1988). Thermo-mechanical treatments of titanium alloys. Proc. 6th World Conf. on
Titanium.
[10] Boothroyd, G. (1963). Temperatures in orthogonal metal cutting. Pro. Inst. Mech. Eng. , 189802.
[11] Joyot, P. (1994). Modlisation numrique et exprimentale de l'enlvement de matire.
Universit de Bordeaux : PhD Thesis
[12] Puigsegur, L. (2002). Caractrisation thermique d'un procd d'usinage par tournage
approache analytique et par identification de systme non entiers. ENSAM Bordeaux: PhD Thesis.
[13] Nistor, I. (2005). Experimentale et simulation numrique de l'endommagement en dynamique
rapide : application aux structures aronautiques. Universit de Toulouse/ENP/ENIT : PhD Thesis.
[14] Johnson, G., & Cook, W. (n.d.). A constitutive model and data for metals subjected to large
strains, high strin rates and high temperatures.

Key Engineering Materials Vol. 446 (2010) pp 157-169


(2010) Trans Tech Publications, Switzerland
doi:10.4028/www.scientific.net/KEM.446.157

Effect of ductile damage evolution in sheet metal forming:


experimental and numerical investigations
Fethi Abbassi1, Olivier Pantal2, Sbastien Mistou2, Ali Zghal1, Roger
Rakotomalala2,
1
2

URMSSDT- ESST Tunis, 5 Avenue Taha Hussein, BP, 56, Bb Manara, 1008 Tunisia

Universit de Toulouse ; INP/ENIT ; M2SP-LGP, 47 Avenue dAzereix ; F-65016 Tarbes, France

Fethi.abbassi@ipeib.rnu.tn, pantale@enit.fr, mistou@enit.fr, ali.zghal@esstt.rnu.tn, roger@enit.fr

Keywords: Ductile damage, Fracture, GTN model, Necking, Sheet metal forming, DIC, Numerical
simulation

Abstract. The numerical simulation based on the Finite Element Method (FEM) is widely used in academic
institutes and in the industry. It is a useful tool to predict many phenomena present in the classical
manufacturing forming processes such as necking, fracture, springback, buckling and wrinkling. But, the
results of such numerical model depend strongly on the parameters of the constitutive behavior model. In the
first part of this work, we focus on the traditional identification of the constitutive law using oriented tensile
tests (0, 45, and 90 with respect to the rolling direction). A Digital Image Correlation (DIC) method is
used in order to measure the displacements on the surface of the specimen and to analyze the necking
evolution and the instability along the shear band. Therefore, bulge tests involving a number of die shapes
(circular and elliptic) were developed. In a second step, a mixed numericalexperimental method is used for
the identification of the plastic behavior of the stainless steel metal sheet. The initial parameters of the
inverse identification were extracted from a uniaxial tensile test. The optimization procedure uses a
combination of a Monte-Carlo and a Levenberg-Marquardt algorithm.
In the second part of this work, according to some results obtained by SEM (Scaning Electron Microscopy)
of the crack zones on the tensile specimens, a Gurson Tvergaard Needleman (GTN) ductile model of damage
has been selected for the numerical simulations. This model was introduced in order to give informations
concerning crack initiations during hydroforming. At the end of the paper, experimental and numerical
comparisons of sheet metal forming applications are presented and validate the proposed approach.

1 Introduction
The numerical simulation is a widely used tool to optimize the sheet metal forming process and to reduce the
development time and therefore the final cost of products in manufacturing such as, automotive and
aerospace. An accurate estimation of the material parameters for constitutive models is often a common need
for the engineering community [1]. With the growth of computing facilities, material models are becoming
more and more complex, leading to the need of more complete identification procedures. The identification
step precedes the design of the manufacturing process. The results of the identification procedure are used as
input data for the numerical simulation of the forming process. Finite Element Modeling (FEM) has been
widely introduced into the design of manufacturing products because of its high efficiency in predicting
several problems and major defects occurring in sheet metal forming manufacturing process like necking,
wrinkling, buckling and surface deflections. In order to predict those defects within a virtual manufacturing
system (numerical simulation), an accurate description of the sheet metal mechanical behavior is an essential
requirement.
In order to increase the predictive capabilities of the virtual forming tools, an accurate modeling of microdefects (i.e. damage) initiation and growth under large transformations have to be used. Concerning, the
ductile failure of materials, it is controlled by three stages: micro-void nucleation, growth and coalescence
mechanisms (Fig. 1). The damage evolution depends directly on the state of deformation. An explanation of
the internal degradation and a review of the mathematical modeling, and the recent development in
numerical tool can be found in many publications [2, 3, 4, 5]. In this investigation, according to the
metallographic observations and the spherical inclusion in the used material, a ductile micromechanical
damage model proposed by Gurson [6, 7] was selected.

158

Structural Analysis of Advanced Materials

Nomenclature

Objective function

Total number of responses

wr
hFEM

Vector of the responses weights

hexp

Vector of the experimental responses

Pressure

Dome height

Youngs modulus [MPa]

Poissons ratio

Hardening coefficient

Strength coefficient,

Vector of the simulated responses

Equivalent true strain


0

Pre-strain

Plastic potential

Void volume fraction


*

f
fc
f u*

Modified void volume fraction

ff

Void volume fraction

Critical volume fraction


Ultimate value of
Nucleation micro-void volume fraction
Plastic hydrostatic strain
Mean effective plastic strain
Standard deviation

(q1, q2, q3)

Fitting parameters
Volume of the material without defects
Volume of material

Hydrostatic stress [MPa]

Yield stress of matrix material [MPa]


von Mises equivalent stress [MPa]

rd

Die profile Radius [mm]

rb

Blank radius [mm]

e0

Sheet thickness [mm]

a
v (x, y ) u (x, y)

Die circular radius [mm]

a1, b1, a2, b2

Elongation terms
Shearing terms.

a3 and b3

Homogenous displacement field for one pattern

Key Engineering Materials Vol. 446

159

Fig. 1: (a) tensile test of bi-entailed plate specimen, (b) evolution of the damage process in a ductile metal
related to the macroscopic loading evolution.

2 Experimental technique and results


2.1.

Digital Image Correlation in 2D: Full-field optical measuring methods

The image correlation allows the measurement of the displacements fields on a planar surface: a single
camera acquires a sequence of images of a planar object under plane strain or stress during the deformation
process. The displacements of many points, distributed on the surface of the object are calculated from the
grey level analysis of the images. DIC is an application based on the comparison of two images acquired at
different states during the deformation. Two subsets are chosen respectively from the reference and the next
image used for the computation. The algorithm of correlation used to detect the local displacements of the
pixels by comparing the two subsets, is as follows. In practice, a single value is not a unique signature of a
point, hence neighboring pixels are used. The matching of images acquired by only one camera, at different
time, on an object which becomes deformed is called temporal matching, or tracking. From its principle, the
correlation technique can be used correctly only with objects having a surface with a sufficiently random
texture.
One CCD (Coupled Charge Device) camera acquires digital images with a 1280 by 1024 pixels definition.
These images are then analyzed by the Aramis software developed by the GOM society. This one gives a
measure of the displacements and strain fields on the surface of an object within a precision of 200 m/m.
Aramis software is able to match correspondent points of an image by grey level analysis, if the surface of
the specimen is covered with a black and white mapping, which forms a random grey-level (see Fig. 2). A
gray level coded on 8 or 12 bits corresponds to each pixel of the CCD sensor. A succession of several pixels
lying on the same line forms a grey level sequence. The fundamental principle lies in the fact that the
distribution does not vary during the deformation of the object. It is therefore sufficient to follow this
distribution of grey levels during their displacement to obtain the displacements of the corresponding point.
This is made possible by the use of correlation domains which are nn pixel gathering zones.

160

Structural Analysis of Advanced Materials

Fig.2 The variable grey levels on the surface of a tensile specimen


In the DIC method the relation between the deformed image and the undeformed one is illustrated in Fig. 3.
If we call G the central point of the subset in the initial configuration and G* the corresponding central point
of the subset in the deformed configuration, the relationship between the coordinates of these two points can
be expressed by:

x* x u (x, y)

(1)

y* y v (x, y )

where u(x,y) and v(x,y) represent the homogenous displacement field for one pattern (whole set of pixels).
O, O*

f(x, y)

Initial pattern

x, x*

B
u

G
D

f*(x*, y*)

C
v

B
*

A*
G*
P
*

Deformed pattern

C*

D*
Pixel

y, y*

Fig. 3 Relative location of sub-images of the deformed and the undeformed patterns on the surface.

In the Aramis software the displacement field is considered homogeneous and bilinear with respect to x and
y:

u ( x, y ) a1 x a2 y a3 x y a4

v( x, y ) b1 x b2 y b3 x y b4

(2)

Where a4 and b4 represent the terms of the rigid body motion, a1, b1, a2 and b2 represent the elongation terms
and a3 and b3 are the shearing terms.

2.2.

Tensile test

In the first part of this work, a large experimental work has been done. Several low rate tensile tests were
carried out in the laboratory on an INSTRON tensile test machine at a strain rate of 10 -3/s. 12 tensile
specimens were obtained using a laser cutting process from the sheet at 0, 45 and 90 with respect to the
rolling direction with standard ASTM-E8. A digital image correlation method was used to measure the
displacement fields at the surface of the specimens during the tensile tests. Figure 4 illustrates the specimens
geometry and the equipments used. During the tensile test, many data are collected such as the hardening
law, the anisotropic yield criterion and the necking of the material.

Key Engineering Materials Vol. 446

161

Fig. 4 Experimental setup of uni-axial tensile test

2.3.

Bulge test

The standard tensile test is one of the most important engineering procedures used to characterize the
mechanical behavior of materials [8]. As the range of stable uniform strain is restricted to less than half the
one sustainable under biaxial stress [9], it is obviously desirable to generate the required data directly from
biaxial tests. The circular hydraulic bulge test is widely used to determine the biaxial stress-strain curve for
sheet metal forming. This test has been used by many authors in order to study the material formability and
to illustrate several defects of sheet metal forming process predicted by numerical simulation softwares [10].
During this test (see Fig. 5 for geometry and parameters), a circular blank is clamped at its external boundary
between a die and blank holder by a drawbeads and a linearly increasing hydraulic pressure is applied on the
bottom surface of the blank. Both circular and elliptic dies have been used for those tests. The experimental
tests have been performed on a 304L stainless steel. Many experiments have been done in order to study the
influence of the rolling direction vs. the orientation of the major axis of the elliptic die (again, at 0, 45 and
90).

(a) The Principe of bulge test

Die profile Radius

rd =6mm

Blank radius

rb = 66.5mm

Sheet thickness (e0)

e0= 1mm

Die circular radius

a=55mm

Die elliptical major axis

110mm

Die elliptical minor axis

74mm

(a) bulge test geometry conditions

Fig. 5 Bulge test geometry and parameters

3 Parametric identification procedure


The procedure of a parametric identification problem consists in computing many numerical simulations of a
physical process, knowing a set of experimental results, and then to minimize the difference between the set
of numerical and experimental responses. The solution of the problem is then composed by the set of
parameter for which the minimal difference has been obtained during the process. Concerning the
identification procedure, many optimization methods have been developed in order to identify non linear

162

Structural Analysis of Advanced Materials

physical laws (behavior of material, damage, necking). The optimization methods can be classified in
three main categories [11]: classical structural optimization methods, linearization methods (GaussNewton,
LevenbergMarquardt) and convex approximations methods (sub-iterative Globally Convergent Method of
Moving Asymptotes (GCMMA)). In this contribution, we used a homemade tool developed in the LGP
laboratory called PILOTE. This one uses a combination of the Monte-Carlo and the Levenberg Marquardt
algorithms [12].
The Monte-Carlo method is a widely used tool based on a random number generation and a set of statistical
tools. It has been successfully applied in many fields such as mechanics, electrics or nuclear physics. In this
identification procedure, the idea is to use such a factor of chance to solve a physical problem using a set of
random parameters as input data for the numerical model. The application of the Monte-Carlo method
requires defining a set of initial parameters and the allowed range for each parameter. This method is a good
tool for a coarse research. Once the convergence criterion of the Monte-Carlo is satisfied the obtained set is
used as input data for the refinement of the solution by the Levenberg-Maraquardt algorithm. The
Levenberg-Marquardt optimization procedure is also widely used for nonlinear optimization. Its a stabilized
version of the Gauss-Newton method, which works very well in practice and has become the standard for
solving nonlinear least squares system of equation.
For the identification, the general procedure is schematically illustrated in Fig. 12. The experimental results
obtained by tensile test are injected into the identification algorithm as starting parameters for the
identification procedure.

Fig.6 Schema of parameter optimization by Monte-Carlo and Levenberg-Marquardt algorithm


The objective function can be defined as in equation 3 to minimize the difference between the numerically
computed and experimentally measured displacement of the centre point of the blank. This dome position h
can be recorded during the whole bulge test.

Key Engineering Materials Vol. 446

1
F
m

hFEM [ j ] hexp [ j ]

wr [ j ]

h
[
j
]
j 1
exp

163

(3)

where m is the total number of response, hFEM is the vector of the simulated responses, hexp is the vector of
the experimental response and wr is the vector of the responses weights.
In this investigation, an isotropic hardening is considered in the numerical model to describe the stress-strain
relation of AISI 304 metal. A subroutine (Uhard) has been implemented in the Abaqus. This relationship, a
Swift law, is described by the equation:
(4)
Table 1 reports the results of the parametric identification. Those parameters will be used in the numerical
model of the forming process.
Swift Law coefficients
0

K (Mpa)

Initial Parameters

0.08

1538

0.6

Identified Parameters

0.0368

1580.53

0.53

Table 1: Results of parametric identification process

3.1.

Continuum Damage Mechanics (CDM)

3.2.

Experimental observations

Onset of ductile fracture is initiated by void formation around non-metallic inclusions and second-phase
particles in metal matrix that is subjected to plastic strain under influence of external loading. Observations
of the damage nucleation mechanisms have been carried out through in situ tensile tests at room temperature.
The spherical shape of the second phase particles and non-metallic inclusions present in the stainless steel
can be identified as possible spots for the initiation of the ductile fracture (Fig. 7). Several damage models
were developed to describe this complex micro-mechanism in sheet metal used in forming process.

164

Structural Analysis of Advanced Materials

Spherical inclusions

Fig. 7 Scanning electron microscope (SEM) views of fracture in tensile specimen (spherical inclusion
observations)

3.3.

GursonTvergaardNeedleman (GTN) Model

The main purpose of this work is to use Continuum Damage Mechanics (CDM) to predict a sheet metal
failure. In 1977, Gurson [6] developed a constitutive model to describe the micro-growth nucleation
mechanical effects of damage in ductile metals, which was extended in 1984 by Tvergaard and Needleman
[7] to incorporate some additional parameters (q1, q2, q3). The yield function describing the plastic
constitutive model is represented as follows:

3
2
2 f * q1 cosh q2 m ( 1 q3(f * )2 ) 0
2
2
y
y

(5)

Where q1, q2 and q3 is a set of fitting parameters used to calibrate the model predicting the periodic array of
spherical and cylindrical voids. In this equation, q3= (q1)2, m is the hydrostatic stress, y is the yield stress of
the material, is the von Mises equivalent stress and f is the void volume fraction which is equal to 1
.
The term represent the modified void volume fraction given as follows:

if f f c
f

*
(6)
f
fu fc
f

(f

f
)
if
f

f
c
c
c

ff fc

Where
is the ultimate value of , is the void volume fraction corresponding to failure, and is the
critical volume fraction when rapid coalescence occurs.
The instantaneous rate of growth of the void fraction depends both on nucleation of new voids and on the
growth of pre-existing voids. It is given by:
*

df (1 f ) (d kkp ) a d

growth

Where

nucleation

is the plastic hydrostatic strain and a is defined as follows:

(7)

Key Engineering Materials Vol. 446

1 p
N
a
exp
2 SN
S N 2

fN

165

Where
is the standard deviation,
nucleation micro-void volume fraction.

(8)
is the mean effective plastic strain of nucleation and

is the

4 Numerical investigations
The objective of the numerical simulations is to predict the damaged zones of the forming parts. The
parameters of the GTN model have been taken from literature data [13, 14] and the comparison of finite
element prediction and experimental results was used to determine the critical damage parameters. Table 2
reports the material parameters used in the all numerical models used hereafter.
Parameters

Element number

Youngs modulus, E
Poissons ratio,

193 GPa
0.3
0.001

Initial void volume fraction, f0


Void volume fraction for coalescence, fc

0.1

Void volume fraction for element deletion, ff

0.12

Volume fraction of nucleating particules, fn

0.04

Mean nucleation strain, n


Standard deviation of nucleation strain, Sn
Tevergaard q parameters: q1, q2, q3

0.6
0.1
1.5, 1, 2.25

Table 2: Material parameters

4.1.

Tensile test

Fig. 8a presents the distribution of the equivalent plastic strain (PEEQ). These numerical results are obtained
with the GTN damage model in a tensile test simulation using the ABAQUS code.
A non uniform distribution is clearly obtained due to the localization of the deformation. The load drop due
to the presence of a diffuse necking and the localization is clearly captured by the use of the ductile damage
model. Fig. 8b shows the experimental plastic deformation field just before the rupture of the specimen
obtained from the optical measurement system. The good correlation between both results shows the
accuracy of the numerical prediction of the shear band.

(a)

(b)

Fig. 8: Tensile test (a) numerical results, (b) Measured distribution of the strain from the ARAMIS system

166

Structural Analysis of Advanced Materials

4.2.

Bulge test

The GTN damage model was used in the simulation of the elliptical bulge test and compared with the
experimental results. Fig. 9 (a) and Fig. 9 (b) present respectively the numerical results and the experimental
deformed part after fracture. We observe a good correlation between both results. As the numerical model
doesnt include crack propagation, only the initiation of the crack is reported by the model.

(a)

(b)

Fig. 9 (a) Numerical prediction and (b) experimental specimen after cracking for elliptical bulge test case
Fig. 10 reports the numerical and the experimental evolution of the dome position during the bulge test
according to the pressure load. The 3 experimental curves Exp_Ellip_90, Exp_Ellip_45 and Exp_Ellip_0
represent respectively the evolution of the dome position according to the internal pressure of the elliptical
bulge test with the major axes oriented 90,45 and 0 with the rolling direction of sheet. In table 3, a
quantitative comparison those results for three different levels of pressure in the bulge test (P=10 MPa,
P=20MPa and at fracture) is reported. A quite small difference between those results is observed, mainly
linked to the fact that the GTN model is isotropic.

Fig. 10 Numerical and experimental comparison of dome position evolution according to the pressure
load

Key Engineering Materials Vol. 446

Dome position (mm)

167

Pressure value at
fracture (MPa)

P=10 MPa

P=20 MPa

fracture

Experience

13.5

23.5

40

30.5

Numerical

13.89

21.99

40.67

29.79

Variation

2.88%

6.38%

1.67%

2.32%

Table 3. Comparison between numerical an experimental results in elliptical bulge test (direction 0)

The thickness (STH) variation is commonly used as an indicator for the appearance of defects during the
forming process. For these reasons, its recommended to evaluate the thickness variation during the
development of a manufacturing process. Gutscher founds that the behavior law parameters has significant
influence on the height and thickness in the bulge test [15]. In order to evaluate the identified parameters
obtained with the parametric method, we have measured the thickness of the specimen using an optical
system based on the use of a fringe pattern projection. The measures have been compared with the numerical
results obtained by the numerical modeling of the elliptical bulge test based including the ductile damage.
Results are reported in Fig. 11, and, again, a good correlation was observed between those results.

(a)

(b)

Fig. 11. Thickness variation in deformed parts (a) numerical results, (b) experimental results

4.3.

Erichsen test

Erichsen test, consisting in the deep drawing of a circular sheet using a spherical punch (Fig. 12), provides
the so-called Erichsen index (IE) obtained as the punch penetration before rupture [16]. This instant can be
associated to the maximal punch force.

Fig. 12: Erichsen test (a) Geometrical condition [16], (b) numerical model
The highly localized thickness reduction that takes place beyond this point causes, for slightly higher levels
of punch penetration, a sudden circumferential rupture of the specimen. Fig. 13 schematically depicts the
problem layout. In the numerical investigation, It is noted that GTN model it is an effective tool to predict
the rupture of the specimen.

168

Structural Analysis of Advanced Materials

Fracture zone

(a)

(b)

Fig. 13: Numerical results of Erichsen test (a) Evolution of volume fraction, (b) von Mises stress.

5 Conclusions
This paper presents an experimental and a numerical investigation of porous metal plasticity. Physical
mechanisms of ductile fracture of stainless steel sheet, nucleation, growth and coalescence of voids have
been presented. In this work, the GursonTvergaardNeedleman (GTN) model is used to study the damage
evolution and necking in sheet metal forming process. In the experimental section, several tensile tests were
presented in order to evaluate the heterogeneity of deformation and the diffuse and local necking processes.
Experimental results have been obtained with optical measurement techniques based on Digital Image
Correlation (DIC) and 3D digitalization.
A numerical approach has been proposed based on the use of the GTN model. The main result presented here
concerns the experimental measurement of the thickness variation of the whole deformed part during an
elliptic bulge test and the comparison with the numerical results. A good correlation between those results
was observed, therefore, we can conclude from this that the proposed numerical approach is quite reliable.

References
[1] A. Andrade-Campos, S. Thuillier, P. Pilvin and F. Teixeira-Dias, On the determination of

material parameters for internal variable thermoelastic-viscoplastic constitutive models,


International Journal of Plasticity, vol. 23 (2007) 13491379.
[2] J.L Chaboche.,. Continuous Damage Mechanics: a tool to describe phenomena before crack
initiation. Nuclear Engineering and Design 64, (1981) 233247.
[3] M. Brnig, Ricci S., Nonlocal continuum theory of anisotropically damaged metals.
International Journal of Plasticity 21, (2005)13461382.
[4] G. Rousselier, S. A. Leclercq, simplied polycrystalline model for viscoplastic and damage
nite element analyses. International Journal of Plasticity vol.22, (2006) 685712.
[5] C. Zhiying, D. Xianghuai, The GTN damage model based on Hill48 anisotropic yield criterion
and its application in sheet metal forming, Computational Materials Science 44 (2009) 1013
1021.
[6] A.L. Gurson, Continuum theory of ductile rupture by void nucleation and growth. I. Yield
criteria and ow rules for porous ductile media, J. Eng. Mater. Technol. Vol. 99 (1977) 215.
[7] V. Tvergaard, A. Needleman, Analysis of the cup-cone fracture in a round tensile bar, Acta
Metall. 32 (1) (1984) 157169.
[8] D. Celentano, E Cabezas., C. Garcia and A. Monsalve, Characterization of the mechanical
behaviour of materials in the tensile test: experiments and simulation. Modeling Simulation,
Mater. Sci. Eng. Vol. 12 (2004) 425-444.

Key Engineering Materials Vol. 446

169

[9] J.F. Michel, P. Picart, Size effects on the constitutive behaviour for brass in sheet metal
forming, Journal of Materials Processing Technology 141 (2003) 439446.
[10] P. Rui R. Cardoso, One Point quadrature shell elements for sheet metal forming analysis, Arch. Comput.
Meth. Engng. Vol 12, (2005) 3-66.

[11] J.P. Kleinermann, J.P. Ponthot, Parameter identification and shape/process optimization in
metal forming simulation, Journal of Materials Processing Technology 139 (2003) 521526.
[12] F. Abbassi, O. Pantal, O. Dalverny, A. Zghal, and R. Rakotomalala Parametric sheet metal
characterization by using Monte-Carlo and Levenberg-Marquardt: bulge test application
APCOM07 in conjunction with EPMESC XI, December 3-6, , Kyoto, JAPAN (2007).
[13] F. ABBASSI, Fdration des connaissances de mise en forme dans une plateforme de
prototypage virtuel. Thse de Doctorat, Institut National Polytechnique de Toulouse, 12
Novembre 2008.
[14] N. Benseddiq, A . Imad, A ductile fracture analysis using a local damage model, International
Journal of Pressure Vessels and Piping vol 85, 219-227, 2008.
[15] G. Gutscher, H. C. Wu, G. Ngaile and T. Altan, Determination of flow stress for sheet metal
forming using the viscous pressure bulge (VPB) test, Journal of Materials Processing
Technology, Volume 146, Issue 1, (2004) 1-7.
[16] D. Franois., Techniques de l'Ingnieur - Essais mcaniques des mtaux. "Essais d'aptitude
la mise en forme". M 125 07- (1984).

Keywords Index
A
Absorption Coefficient
Adhesive Failure
Adhesive Joint
Adhesives Formulations

1
43
43
51

B
Boundary Element Method (BEM)

131

C
Carbon Fiber
Carbonaceous Materials
CMS
Cohesive Failure
Composite Structure
Composites
Cutting Condition
Cutting Force
Cutting Temperature
Cyclic Voltammetry

23
23
33
43
83
137
147
147
147
23

Fracture
Fracture Toughness

157
11

G
Glass Composites
Global Optimization
GTN Model

73
101
157

H
Hybrid Adhesives
Hydrothermal Ageing

111
43, 51

I
Impact Response
Infrared Thermographs
Intercalation
Interpenetrated Continuous
Network
Irradiation

83
63
33
111
1

K
D
Darcys Law
Design of Experiments (DoE)
Detector
Differential Scanning Calorimetry
(DSC)
Digital Image Correlation (DIC)
DMSO
Ductile Damage
Dynamic Compression
Dynamic Load
Dynamic Response

Kaolinite
131
83
1
43
157
33
157
73
137
73

33

L
Liquid Composite Molding
LRI Process

131
83

M
MATLAB
Moving Mesh
Multi-Module Bond Line

91
131
111

N
E
Electrorheological Fluid
Epoxy Composite

33
73

137
157
131, 157

F
Fibre Orientation
Finite Element (FE)
Finite Element Analysis (FEA)
Finite Element Software

Naval Construction
Necking
Numerical Simulation

73
111
11, 91
91

Optimization of Polymerization

43

P
Photoconductivity

172

Structural Analysis of Advanced Materials

Probabilistic Transformation
Method
Pt Electrodeposition

91
23

Y
Yielding Behavior

R
Reliability Method
RTM Process

91
83

S
Self Heating
Semiconductor
Sheet Metal Forming
Shottki Barrier
Simulated Annealing (SA)
Simultaneous Perturbation
Stochastic Approximation
Sorption Models
Split Hopkinson Pressure Bar
(SHPB)
Stiffened Structure
Strain Rate Effect
Stress
Structure
Structure Optimisation
Synthesis

63
1
157
1
101
101
51
73
137
63
147
91
101
33

T
Technological Parameter
Tensile Yield Stress (TYS)
Thermal Effect
Thermomecanical Properties
Thermomechanical Behaviour
Thermoplastics
Thermosetting Resins
Ti5553
Tool Geometry
Top-Hat
Triplate

11
147
147
111
43
63
51
147
147
137
101

U
Ultimate Tensile Stress (UTS)

147

V
Vickers Micro-Indentation
Voltage

11
1

W
Wear

147

63

Authors Index
A
Abbassi, F.
Ali, M.

157
43

B
Baili, M.
Bertram, V.
Brault, R.

147
137
121

C
Cantarel, A.
Choukri, S.

131
73

147
111

E
El Alem, W.
El Halouani, F.
El Hami, A.
Ellaia, R.

131
33

131
83
51
23
63
11

Mares, M.
Masmoudi, N.
Mehrabova, M.
Mistou, S.
Mohamad, G.

11
63
1
83, 121, 157
137

Nassiet, V.
Neme, A.
Niquet, M.

51, 111
73
121

Ouhimmou, S.

91

P
Pantal, O.
Pikasi, A.
Pirc, N.
PUEL, B.

83, 157
23
131
43

Rakotomalala, R.
Reiss, T.

157
33

S
Schmidt, F.
Simitzis, J.

131
23

H
Hadriche, I.
Hassoune-Rhabbour, B.

G
Gantois, R.
Garnier, C.
Gaussens, C.
Georgiou, P.
Ghorbel, E.
Goanta, V.

147
33
43

O
101
63
91, 101
91, 101

F
Flices, J.N.
Fleureau, J.

Lallement, D.
Laribi, S.
Lorrain, B.

D
Dessein, G.
Djilali, T.

63
51, 111

Tarfaoui, M.
Tassin, J.F.
Tkiouat, M.

73, 137
33
91

K
Karama, M.

43

W
Wagner, V.

147

174

Structural Analysis of Advanced Materials

Z
Zghal, A.

157