Predicting the strength

of composite laminates
One of the vital tasks facing designers of composite components involves predicting
the strength of the laminate under two or three-dimensional loads from basic
information on the properties of the constituent layers. Peter Soden reports on the
results of a project known the ‘Failure Olympics,’ in which contributors around the
world tested the accuracy of various theories for predicting deformation and failure
strength of laminated composite structures.

O
ne major selling point for fibre Laminate theory can then be employed engineers have devoted much of their
reinforced polymer (FRP) com- to predict the stresses in the individual working lives to developing them.
posites is their high strength layers. From that point, some form of Practising designers will be familiar with
coupled with low weight. It is well failure criterion can then be used to com- many of the more famous theories and
known that the full benefit of the excep- pare the applied stresses with the people involved with them. The article
tionally high stiffness and strength of a strength of the layer in order to establish Composites Design in the Real World in the
unidirectional fibre reinforced composite the level of safety in the structure. September 2003 issue of Reinforced
is achieved only when loads are applied Since the introduction of composites, Plastics referred to one of these theories.
parallel to the fibre direction. For appli- numerous different failure criteria have Computer programs for analysing the
cations where the components are been proposed and some researchers and strength of composite structures some-
required to support high multiaxial loads
(which is the norm in real engineering
structures), laminate construction is
commonly used. Such laminates contain
fibres arranged in several different direc-
tions. Figure 1 is an example of a lami-
nated composite component that is sub-
ject to multiaxial stresses.
One of the vital tasks facing design-
ers of composite components is that of
predicting the strength of the laminate
under two or three-dimensional (3D)
loads from basic information on the
properties of the constituent layers. The
accuracy with which the designer can
predict the strength of the laminate
under these circumstances has a major
influence on the weight and cost of the
product and on the safety margins that
apply during its service life.
Modern computer aids such as finite
element analysis (FEA) allow the design-
er to predict the distribution of stresses Figure 1: A high-pressure composite vessel for a breathing apparatus. The low profile toroidal shaped
vessel, which fits snugly on a fireman’s back, is subjected to high biaxial stresses that vary with
in load bearing components and position around the minor circumference.
structures (see for example Figure 2).

42 REINFORCEDplastics February 2005 0034-3617/05 ©2005 Elsevier Ltd. All rights reserved.
 Predicting the strength of composite laminates

times offer the designer a selection of the

more popular failure criteria from which Box 1: Contributors to the World Wide Failure
to choose. However, within the comfort-
ing familiarity of these apparently well-
Exercise
established procedures lie traps for the Authors Organisation
unwary or inexperienced practitioner, C.C. Chamis, P.K. Gotsis and L. Minnetyan NASA Lewis, Cleveland, USA
and the predictions are not always as L.J. Hart-Smith Boeing, USA
accurate as might be expected. G.C. Eckold AEA Technology, UK
E.C. Edge British Aerospace, Military Aircraft
The WWFE Division, Warton, UK
Over the last 12 years. three engineers – L.N. McCartney National Physical Laboratory, London,
Mike Hinton, Peter Soden and Sam UK
Kaddour from the UK defence research A. Puck and H. Schürmann Technische Hochschule, Darmstadt,
organisation QinetiQ and the University Germany
of Manchester Institute of Science and W.E. Wolfe and T.S. Butalia Department of Civil Engineering,
Ohio State University, USA
Technology (UMIST) – have organised
C.T. Sun and J.X. Tao School of Aeronautics & Astronautics,
and coordinated a very extensive inter-
Purdue University, Indiana, USA
national collaborative activity. This
P. Zinoviev, S.V. Grigoriev, O.V. Labedeva, Institute of Composite Technologies
activity has become known within the L.R. Tairova Orevo, Moskovkaya, Russia
composites community as the ‘World
S.W. Tsai and K-S. Liu Aeronautics and Astronautics
Wide Failure Exercise’ (WWFE), or in Department, Stanford University,
some circles as the ‘Failure Olympics.’ California, USA
The objective of the WWFE was to A. Rotem Faculty of Mechanical Engineering,
establish the accuracy and capabilities of Technion-Israel, Institute of
the currently available theories for pre- Technology, Haifa, Israel
dicting deformation and failure strength R. Cuntze and A. Freund MAN Technologies, Germany
of laminated composite structures. T. Bogetti, C. Hoppel, V. Harik, J. Newill, B. Burns US Army Research Laboratory,
Within the exercise the leading failure AMSRL-WM-MB, Aberdeen Proving
Ground, MD, USA
theories for composite laminates were
S.J. Mayes and A.C. Hansen US Naval Surface Warfare Center,
described by their originators and used
West Bethesda, and Alfred University,
by them to solve a prescribed set of prob- USA
lems. The predictions of the various Z-M. Huang Department of Engineering
theories were subsequently compared Mechanics, Tongji University,
with one another and, at a later stage in Shanghai, China
the exercise, with experimental data. M.J. Hinton, A.S. Kaddour and P.D. Soden QinetiQ, Fort Halstead and
The contributors of theoretical papers (the organisers) Farnborough, UK, and UMIST,
Manchester, UK
included many internationally renown-
ed scientists and engineers from six
countries (see Box 1). In all, 15 groups The experimental results used in the exercise were obtained mainly from tests conduct-
took part and presented a total of 19 ed by four organisations: DFVLR, Stuttgart, Germany; QinetiQ, Fort Halstead and
different approaches. Farnborough, UK; University of Utah, Salt Lake City, USA; and UMIST, Manchester, UK.
Most of the leading approaches to
predicting failure in composites were
represented, including theories employ- approach. Some theorists concentrated (Chamis) started from constituent
ing the well-known Maximum Stress, on mathematical elegance (Tsai), some matrix and fibre properties and used
Maximum Strain, interactive and mech- on simplicity (Hart-Smith 10% rule), micro-mechanics to predict the layer
anistic failure criteria. Some less widely and some (Puck) on reflecting the range properties.
known methods were also represented. of physical failure mechanisms The performance of the theories
McCartney, for example, employed a observed in composites. Most have the depends upon much more than just the
damage mechanics/fracture toughness practical application of their theories as type of failure criterion employed. Some
approach that was in the early stages of a central concern. The majority of the aspects that have a significant influence
development, and Eckold agreed to theories took the laminate layers as the on the predictions are whether the theo-
present a simple standard design code basic building blocks, but a few ry assumes linear or non-linear material

February 2005 REINFORCEDplastics 43

Predicting the strength of composite laminates

laminates as employed in aircraft struc-

tures. A wide range of realistic uniaxial
and biaxial loading conditions was also
specified including tension, compres-
sion, shear and combined loading. The
participants were asked to predict the
deformation of the laminates, the loads
at which failure was expected to occur,
and the modes of failure expected under
the different types of loading. The for-
mat of the required output, i.e. as two-
dimensional failure envelopes or stress-
strain curves, was defined in detail. At
this stage, the participants were not pro-
vided with any experimental results.

Experimental data
In addition to theoretical work and
service experience with a variety of
Figure 2: Combined finite element and laminate analysis predictions for the distribution of shear composite products, a large body of
stresses parallel to the fibres in the innermost layer of a filament wound tube supported on a flat plate
and subjected to a local load by a spherical indenter. experimental data is available on the
behaviour and strength of composites.
The basic properties of the materials are
usually derived from tests on small lab-
oratory samples subjected to simple
loading, for example uniaxial tensile
tests on specimens cut from flat sheets
of composite material. In most practical
applications, the loading conditions
will be more complex than those that
pertain in the basic laboratory tests. In
practice, components are usually sub-
jected to biaxial or triaxial stresses and
loads may vary and may be applied for
long periods of time under severe envi-
ronmental conditions. Tests that simu-
late realistic multiaxial loading situa-
tions are difficult and expensive to carry
Figure 3: Test cases used in the World Wide Failure Exercise. These include 14 laminates and types
out but some facilities do exist for this
of loading. purpose. The most successful tests to
date have been on tubular and cruci-
form specimens (see Figures 4 and 5 on
behaviour and whether or to what extent unidirectional carbon fibre and two glass page 46).
the effects of thermal stresses and fibre fibre reinforced epoxy composites. They Experimental data for use in the exer-
realignment are included. Also of impor- were asked to solve a series of 14 chal- cise was gathered from four different
tance are the way in which laminate lenging problems specified by the organ- groups in the UK, the USA and Germany.
behaviour is treated after initial failure isers (see Figure 3). The data included properties of fibres,
has taken place and the type of numeri- The problems were intended to test resins and unidirectional fibre reinforced
cal algorithm employed. the theories to the full and were selected laminae that were used as input data for
to represent a variety of practical the theories. At a later stage in the
The test problems laminate configurations including uni- programme, when the initial round of
The participants were all given the same directional, cross-ply, angle-ply (as used predictions had been published, results
input data, including properties for two in pipework) and quasi-isotropic from the tests on laminates were

44 REINFORCEDplastics February 2005

Predicting the strength of composite laminates

provided to the participants for compari- laminate at much lower loads than those
son with their theoretical predictions. that produced final fracture, although
The majority of these test results were
A wide range of failure those predictions are often ignored in
obtained from experiments on tubes sub- stresses was predicted by practice. This highlights the need for
jected to internal or external pressure designers to pay careful attention to the
combined with axial tension or compres-
the different theories. nature of the predicted failures and
sion or torsion. Graphs of experimental whether they are significant for the par-
results were provided for all the test ticular application in hand.
problems including stress-strain curves There was often a large difference
and failure envelopes, which showed the between the loads at which initial failure
strengths of the laminates when they and final fracture occurred, up to a factor
were subjected to different types and of 50 between them in extreme cases.
combinations of loading. Information Use of the simple and familiar procedure
was also given on the types of failures for assessing laminate strength described
observed in the experiments on lami- in the introduction to this article, with
nates, including cracking, leakage and little or no allowance for progressive fail-
fracture. ure would, in many cases, predict initial
failure strengths which were more than a
Results factor of ten lower than the actual meas-
Superimposing the theoretical predict- ured final fracture strength.
ions upon the experimental results for a Figure 6 shows an example of some of
wide range of practical test cases gave a the predicted final failure strengths com-
good indication of the predictive abilities pared with the experimental results for
of the various theories, and also revealed Test Case 6, a (0°/±45°/90°) quasi-isotrop-
shortcomings in some of the experimen- ic carbon fibre/epoxy laminate that is
tal results. commonly used in aircraft construction.
A wide range of failure stresses was This illustrates the level of agreement Figure 4: A biaxial loading test using a planar
cruciform specimen. The figure shows the
predicted by the different theories and in with experimental results, which was larger of two biaxial testing machines at
most cases a sequence of different reasonably good in some cases and poor QinetiQ, Farnborough. This machine has a load
capacity of 1500 kN and is capable of applying
failures was predicted before the lami- in others. Nearly all of the theories either tensile or compressive loads on each of
the two orthogonal axes.
nates finally fractured. predicted resin cracking in this particular

Figure 5: Biaxial loading tests on a tubular test specimen. The figure shows how the basic test rig can be configured to apply any desired ratio of biaxial
stresses (SR). It can be used to explore the complete biaxial failure envelope.

46 REINFORCEDplastics February 2005

 Predicting the strength of composite laminates

shortcomings in many theories, and a

Box 2: The MMS5 project number of contributors were able to
improve their models and predictions
Some of the companies currently involved in the MMS5 projects and some of the
applications in which they are interested: during the course of the exercise (as
indicted by the Part B results in Figure 7).
Ford Motor Co: components for cars and commercial vehicles, panels, loadfloors and
bulkheads.
Vosper Thornycroft (VT) Holdings: offshore oil platforms, railway rolling stock, bridges Availability of results
and ships; blow-out blast and structural panels. All of the results have recently been
Ameron: GRP epoxy and phenolic pipe systems; pipes for offshore petroleum, chemical made available in a book, Failure Criteria
processing, fuel handling, marine and water industries. in Fibre Reinforced Polymer Composites: the
Dowty Aerospace Propellers: aerospace and marine applications; braiding hybrid World Wide Failure Exercise (ISBN: 0-08-
glass/carbon preforms; Formax fabric reinforcements, in particular non-crimp multiaxial 044475 75-X), edited by the organisers
warp-knit glass fabrics (for resin transfer moulding and prepreg). M.J. Hinton, A.S. Kaddour and P.D.
St Gobain-Vetrotex International: fabric reinforcements and thermoplastic prepreg Soden and published by Elsevier in 2004.
tows, specifically the Twintex™ range of commingled yarns, used in bumper systems,
Details are provided of the test cases and
load floors and sporting goods (eg. snowboards).
experimental results. All of the theories
are described in detail by their origina-
The theories usually gave good predic- number of different test cases. Some tors, and all of the theoretical predic-
tions for the shape of stress-strain curves theories performed very well considering tions are compared with one another
in the test cases where the behaviour was the wide range of problems considered and with experimental results for each
dominated by the fibre properties. (see Figure 7 on page 48). The highest test case. The accuracy of the predictions
However, in other cases, linear elastic ranked theories were able to predict the and the performance of the theories are
theories greatly underestimated the meas- experimental results to an accuracy of assessed and recommendations are made
ured deformation, and none of the theo- better than ±50% in more than 75% of on the uses of the theories in engineer-
ries gave satisfactory predictions of the the test cases. Some theories performed ing design.
large non-linear deformations that were poorly and some were awarded low over-
observed in some of the experiments. all scores because they were only able to Current projects
The overall performance of the vari- predict results for a few of the test cases. The WWFE considered a much wider
ous theories was ranked by comparing Comparing the theoretical predict- range of laminate types and loading con-
the ratio of predicted to measured ions with a wider variety and range of ditions than is usual in such studies.
strengths and deformations for a large experimental results than usual revealed Even so, it still only covered a small cross
section of the polymer composite mate-
rials currently employed in commercial
applications. For example, it did not
include thermoplastics or polyester
resins, chopped fibres or cloth reinforce-
ments.
A project on the design and failure of
material systems under multiaxial loads
is currently being carried out in the UK
under the Department of Trade and
Industry (DTI) Measurements for
Materials Systems (MSS) programme.
This particular project, designated
MMS5, is being run by a team from
QinetiQ, The University of Manchester,
Nottingham University, AEA Technol-
ogy, National Physical Laboratory (NPL)
and NetComposites. This project aims to
take up to date knowledge on compos-
Figure 6: Final failure envelopes for a quasi-isotropic (0°/±45°/90°) carbon fibre/epoxy laminate ites, drawn from the WWFE and experi-
subjected to range of biaxial loads (WWFE Test Case 6). The predictions of several different theories
are compared with experimental results. ence gained by the partner organisations
from previous work, and adapt this

February 2005 REINFORCEDplastics 47

Predicting the strength of composite laminates

Not attempted Grade B

Grade C Grade A
100%

90%

80%
Percentage score

70%
60%

50%

40%

30%

20%

10%

0%
Zinoviev
Bogetti
Puck
Cuntze
Tsai
Sun(L)
Edge
Huang
Mayes
Wolfe
Hart-Smith3
Chamis2
Rotem
Hart-Smith1
Hart-Smith2
Sun(NL)
Eckold
McCartney

Cuntze-B
Tsai-B
Mayes-B
Wolfe-B
Edge-B
Huang-B
McCartney-B
Theory descriptor

Figure 7: Quantitative assessment of overall performance of various failure theories.

knowledge for composite material sys- confidence to other similar systems and MMS5 IAG and some of their interests.
tems that are applicable to wider com- to other complex loading conditions. Industrial involvement is considered
mercial markets including marine, auto- Steps are being taken to make infor- vital in projects of this kind. The pro-
motive and civil engineering mation available in forms that are read- gramme aims to address the widest possi-
The programme involves consulting ily accessible and applicable by design- ble range of UK industry needs in this
material suppliers and users and survey- ers in a range of industries. The routes area. It is intended that substantial bene-
ing the wider industrial needs for multi- through which this is done include the fits will accrue to the industrial suppliers
axial data, and examining the findings of project workshops, the UK National and users who participate.
the WWFE, to determine where multi- Composite Design Tools, Design Codes, Improved understanding of the
axial data is most lacking and where it Standards and an Interactive Knowledge behaviour of industrial composites
would be of most benefit. Base (IKB) with internet access. The under multiaxial loads and the informa-
Facilities are available within the more conventional dissemination tion gained through the projects
partner organisations for multiaxial test- routes through conferences and journal described in this article should lead to
ing of planar cruciform and tubular test articles are also being used. One object- greater safety and more efficient, eco-
specimens and other forms of test are ive is to present the information in a nomic and widespread use of composite
under consideration. Test specimens are manner which will assist participating materials in an increasing variety of load
being manufactured and selected tests companies to assess which of the failure bearing applications. ■
carried out to fill gaps in existing know- prediction methods, described in the
ledge and to extend the database of WWFE, standards and elsewhere, will be
multiaxial test data to the alternative most appropriate for use in their own
If you have interests in the area of load
composite material forms requested by applications.
bearing composite components and
industrial users and suppliers.
would like these to be addressed, or if you
This test work is being carried out in Industrial involvement wish to contribute to the project or keep
conjunction with modelling activities, An Industrial Advisory Group (IAG)
up to date with its progress, please visit
the intent being to use the analysis to exists to provide industrial focus and
the website www.mms5.co.uk and com-
assist in the understanding of the experi- guidance to the projects and regular IAG
plete the survey at www/arpconsortium.
mental results and the extracting of meetings are held at different centres
org/mms5.asp.
maximum value from them. This will within the UK. Box 2 (page 47) shows
allow the data to be extrapolated with the current industrial membership of the

48 REINFORCEDplastics February 2005