See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/258660553

The predictive capability of failure mode concept-based strength conditions

for laminates composed of unidirectional laminae under static triaxial stress
states

Article in Journal of Composite Materials · September 2012

DOI: 10.1177/0021998312449894

CITATIONS READS

41 1,212

1 author:

Ralf Cuntze
Carbon Composites e.V., Augsburg
73 PUBLICATIONS 633 CITATIONS

SEE PROFILE

All content following this page was uploaded by Ralf Cuntze on 09 October 2018.

The user has requested enhancement of the downloaded file.

 THE PREDICTIVE CAPABILITY OF
FAILURE MODE CONCEPT - BASED STRENGTH CONDITIONS FOR
LAMINATES COMPOSED OF UD LAMINAE
UNDER STATIC TRI-AXIAL STRESS STATES

R. Cuntze
formerly MAN Technologie AG, Augsburg, Germany
D-85229 Markt Indersdorf, Germany. Tel.: 0049 8136 7754, Ralf [email protected]

Abstract
This paper represents the author’s contribution to the Second World-Wide Failure Exercise (WWFE-
II) using his Failure Mode Concept (FMC) modelling capability. The WWFE-II deals with the
behaviour of isotropic material and unidirectional (UD) as well as multi-directional UD laminae-
composed laminates subjected to three dimensional (tri-axial) states of stress. 12 challenging Test
Cases were provided by the organisers (Kaddour and Hinton) and those cover stress-strain curves
and failure envelopes.
The application of the FMC model involves the following: (a) proposing appropriate nonlinear stress-
strain curves for the UD lamina before and after initial failure occurrence, (b) taking into account the
effect of pressure on mechanical properties and (c) incorporating the five FMC modes of failure (two
Fibre Failures (FF) and three Inter-Fibre Failures (IFF)) into suitable form. The UD-lamina is
treated as a transversely-isotropic material which is consistent with the WWFE-I.
The FMC theory was capable of predicting the behaviour of the Test Cases. TC 10 was not tackled in
order to reduce the amount of work for the single author.

Keywords: UD- lamina failure conditions, multi-axial stress states, hydrostatic compression,
laminate behaviour, nonlinear behaviour

Nomenclature
as, bs : Parameters in softening regime
nd
a1, a2, a3 : Fitting parameters for 2 Tg effect

b ,b|| : UD material’s internal ‘friction parameters’ of the FMC-based model

E1 = E||, E2 = E3 = E : Elastic moduli of a UD lamina in the directions x1  1  || , x2  , x3  ,
E0 : Initial Young’s modulus
Esec : Secant elasticity modulus

Eff : Resultant (global) stress effort of all interacting failure modes

Effmode : Stress effort of a UD-lamina in a distinct failure mode
E||f : Elastic filament (fibre) modulus in fibre (x1) direction

Ef : Elastic filament (fibre) modulus transversal to fibre direction

e||t ,e||c : Tensile and compressive failure strain of a UD-lamina in x1 direction

F|| , F|| , F , F , F|| : Failure functions for FF and IFF modes

f2ndTg: reduction factor or function to consider the 2ndTg effect

G12 ; G23 : Shear modulus of a UD lamina in the x1-x2 plane (  G|| ); in the x3-x2 plane (  G )
WWFE-II December 2008 Final 1 of 77 29.12.2009
8:07
G12,sec : secant shear modulus in the x1-x2 plane

I1, I2, I3, I4, I5 : Invariants of the transversely-isotropic UD-material

I1, J2 : Invariants of the isotropic matrix material
M : Matrix weakening exponent
m : Mode interaction or rounding-off exponent
n : Repetitions of sub-laminates, Ramberg-Osgood exponent
p : Pressure on the faces of the specimen (absolute value)
phyd : Hydrostatic pressure (absolute value)
R , R : Average or typical strength (optimally, the mean of the basic population of the stochastic
quantity strength) for model validation by test data, strength design allowable for Design Verification
(an A- or a B-value)
Rp0.2 , Rc0.2 : Tensile, compressive stress values at 0.2 % plastic strain  yield strength ( R0.2 denotes
both properties)
R : termed here Cohesive strength of a ductile behaving material

R|| f : Strength of the filament (fibre)

R||t  X t , R||c   c : UD tensile and compressive strength parallel to the fibre direction

Rt  Y t , Rc  Y c : UD tensile and compressive strength transverse to the fibre direction
R||  S : UD in-plane shear strength, transverse/parallel to the fibre direction
[S] : denotation of compliance matrix
Tg : Glass transition temperature

Tr : Definition for the Triaxiality value Tr  ( I 1 / 3) / 3 J 2   m /  eqMises ,  eqMises  3J 2 or in Lode

coordinates  
Tr  2 / 3  (I 1 / 3) / 2J 2 uniaxial: Tr  1 / 3
t : Thickness of the laminate
Vf : Fibre volume fraction
x1, x2, x3 : Coordinates of the UD lamina (x1 = fibre direction ||, x2 = direction transverse to the fibre
 , x3 = thickness direction)

1, 2, 3: Normal strains of a UD lamina

 : degradation function in softening regime
12 = 21; 13 = 31; 23 = 32 : Shear strains of a UD lamina

12: Major Poisson's ratio in the WWFE-II (  || in the German VDI 2014 guideline)

  : Poisson’s ratio in the transversal plane (  23)

 : Parameter which represents the 120°-symmetry of brittle isotropic materials
 fp : designation of the fracture plane angle
 : internal friction coefficient of the isotropic material
 || ,   : internal friction coefficients of the transversely-isotropic material model

WWFE-II-Part A_dispatched Dec09 2 of 77 29/12/2009, 8:07

 ef : effective hydrostatic stress in the constrained matrix material (absolute value)

 hyd : hydrostatic stress (sign dependent)

1, 2, 3 : Normal stresses in a UD lamina
I, II, III : Principal stresses in an isotropic material

 2c , 1t : Compressive stress across and tensile stress along the fibre direction
||, : Stresses parallel and transverse to the fibre direction (symbolic denotation)

ˆ x ,ˆ y : Average in-plane stresses of the laminate

|| ||
 eqmod e : Equivalent stresses of a mode (  eq , eq , eq , eq , eq|| ) which include load-induced
mechanical stresses and residual stresses

1f , 2f : Filament stress in x1 direction; in x2 direction

  : Stress vector of the lamina

ˆ : Average stress loading a laminate’s face or cross section
12 = 21, 13 = 31, 23 = 32 : Associated shear stresses of a UD lamina in the elastic symmetry
directions. The first subscript locates the direction normal to the plane on which the shear stress is
acting; the second subscript indicates the direction of the shear force
 || ,  Shear stresses transverse/parallel and transverse/transverse to the fibre direction

Abbreviations
CLT : Classical Laminate Theory
CDS : Characteristic Damage State
CDM : Continuum Damage Mechanics
CrF : Crushing Fracture
CTE : Coefficient of thermal expansion
CME : Coefficient of moisture expansion
F : Failure function
FEA : Finite Element Analysis
FF, IFF : Fibre Failure, Inter-Fibre Failure
FMC : Failure Mode Concept
FRP : Fibre-Reinforced Plastic
MfFD : Multifold Failure Domain
MiFD : Mixed Failure Domain (interaction of several failure modes)
NF, SF : Normal fracture, shear fracture
UD : Uni-directional
wrt : with respect to

Subscripts, superscripts, and signs

c, t
: compressive, tensile (German Guideline VDI 2014)
cr : critical
WWFE-II-Part A_dispatched Dec09 3 of 77 29/12/2009, 8:07
eq : equivalent
ef : effective hydrostatic stress and effective temperature difference at curing (stress free
temperature – room temperature)
f, m : fibre (the term filament sometimes more accurate), matrix
fr
: fracture
fp : fibre parallel fracture plane
hard : hardening
hyd : hydrostatic
th
k : k lamina
L, R : denotes stress from external applied load, residual stress
n, t : normal, tangential to the fibre plane of the UD lamina
ns : n-fold symmetric lay-up
p, ps
: principal (in §2.2.1), pseudo (in §2.3.1)
p0.2 : tensile yield strength
Res : reserve
soft : softening
s : symmetric lay-up
trf : trigger factor
,
: indicates failure, either induced by Mohr’s shear or by Mohr’s normal stress acting at the
physical fracture plane
┴, || : transverse, parallel to fibre direction of a UD lamina
0.2 : 0.2 remaining plastic strain (technical onset of yielding)
1, 2, 3: indicates the directions in the UD lamina
I, II, III : indicates the principal stresses in an isotropic material
x, y, z : indicates the directions in the laminate
─
: typical or average (sometimes the statistical mean) needed when modelling
^ : over the thickness smeared (averaged)
 : designation of transverse plane (as subscript and superscript).

1. Introduction
1.1. Background to the WWFE-II
As an engineer in industry the author had to design to static and cyclic loadings. On the basis
of Structural Design and Analysis the structural part’s stiffness, stability and strength as well
as damage tolerance & fatigue life had to be demonstrated. This task includes the assessment
of stress states, stress concentrations (notches) and stress intensities (flaws, cracks). In order to
fully prove Structural Integrity by the Structural Design Verifications the demonstration of
the fulfilment of all static strength requirements, and of fatigue and damage tolerance
requirements is mandatory. The primary challenge of the WWFEs is the assessment of static
strength by realistic failure conditions for fibre-polymer composites.
For design engineers the choice of the macro-mechanical level in modelling is mandatory
because - in general - they obtain as input macro-mechanical stresses from stress analysis.
Hence strength analysis, basis of the design verification, can be and should be carried out on
the macro-mechanical level, too. But, one has to keep in mind two aspects: Describing failure
by macro-mechanical stresses is different to describing failure by the really failure
mechanism-causing micro-mechanical stresses in composite laminae. A macro-mechanical
failure description is not always accurate such as it is later delineated for the fibre failure

WWFE-II-Part A_dispatched Dec09 4 of 77 29/12/2009, 8:07

modes. In design, the knowledge about an identified failure mode and its importance is helpful
because the designer can better counteract. This importance is later described by a quantity
which the author termed ‘stress effort’ of the material.
Failure conditions for a lamina, utilized as a piecewise-homogeneous material, do not fully
cover the so-called ‘in-situ’ failure behaviour of the embedded laminae (perfect bonding of the
layers is assumed here). The different failure behaviour between a so-called isolated and an
embedded lamina comes from the different influence of the occurring micro-cracks. In the
isolated case, the critical one of them grows to macro-size and causes fracture, whereas in the
embedded case ‘just’ multi-site micro-cracking is generated and is acting at least up to the
level of the so-called critical damage state (CDS). Then, the number of micro-cracks
practically remains about constant, just their width increases sometimes followed by
delamination.
In this context, the properties used as input for the analysis are test results from isolated UD
lamina specimens such as a tensile coupon. They are load-controlled as they are results of
weakest link type (here, ’sudden death’ modelling is adequate) whereas the in-situ behaviour
of an embedded UD lamina is deformation-controlled and therefore of redundant type. This
fact shows up that a good mapping of the course of ‘isolated UD test data’ does not involve
the full information necessary for a qualified laminate analysis. The non-linear analysis in the
strain-softening domain of an embedded lamina has to consider the positive effect from
obtaining redundancy by the embedding.
For a better understanding it is to be re-called: From 1992 till 2003 the WWFE-I,
([Cun04a,b; Hin02; Hin04; Puc02]), was under way to monitor and check the current
capability of methods for predicting the strength of fibre composite laminates. WWFE-I was
organised at UMIST and QinetiQ, UK, by Prof. Dr. M. J. Hinton, Mr. P. D. Soden and Dr. A.
S. Kaddour. Now they organize WWFE-II, see Ref. [Kad07], in order to get an effortful
validation process to an end.
Main result of the WWFE-I was “The participating failure theories could not be tested to the
full” due to the fact that in the negative stress state domains the needed Test Cases were
missing. Now, well tested as well as well evaluated experimental data in the multi-axial
compressive domain shall be provided in the WWFE-II, for a real 3D validation of the
actually available and competing UD theories.
For the predictions, test cases were selected which include carbon and glass fibres, different
epoxy matrices, stacking sequences, and loading conditions involving uni-axial and bi-axial
tension and compression, torsion shear, hydrostatic loading as well as combinations of them.
In the WWFE-I the contributors could only check predictions against test cases in the
tensile/shear domain.

1.2. FMC theory and its success

The failure theory of this contribution consists of three parts: 1) The UD failure conditions
with input data, 2) The non-linear material modelling of lamina and matrix, and 3) The coding
which is responsible for an accurate solution of the analytical/numerical model. Therefore, a
WWFE is not just a competition of the predictive capabilities of the failure conditions because
‘judging of failure theories’ without viewing the nonlinear stress analysis is the half story,
only. A well established theory must predict accurate deformations and multi-axial failure
states.
Here, in addition to WWFE-I, knowledge about the pressure-induced effect of the shift of the
so-called 2ndTg (glass temperature) point of the matrix material below zero C°) of the matrix
WWFE-II-Part A_dispatched Dec09 5 of 77 29/12/2009, 8:07
on stiffness and strength of the UD composite is mandatory for the predictions if a high multi-
axial compressive loading is acting. The established models shall also describe the non-linear
post-failure behaviour or more accurate, the post-initial failure beyond IFF.
For a better understanding of the paper some definitions shall be given here: A lamina is the
building block when calculating the laminate. It may consist of several layers (plies) or also
describe just one part of a layer. Micro-mechanics is understood here as dealing with the
composite constituents fibre, matrix, and the interface layer or the interphase material. Meso-
mechanics is the treatment of the lamina’s internal fibre (textile) architecture, and is not
addressed here. Macro-mechanics is the treatment of the lamina - as a homogenized material -
in laminate analysis. With respect to the nomenclature the reader is also referred to the
WWFE-I paper [Cun04a]. In the text it will be distinguished between stresses (positive in
positively defined direction) and loadings (direction of action is of interest). Strengths are
always positive quantities.
According to the WWFE-II instructions the paper is intended ‘to describe failure theory and
method of application to laminae and laminates in sufficient detail to allow predictions to be
reproduced by others’. It will comment on the nature and the effects of the failures predicted
and, if appropriate, how the predictions could be used for design. The formulations shall be
given so that both industrial design and research should benefit.
The paper presents a set of three-dimensional (3D) failure conditions for UD laminae made
from fibre-reinforced plastics (FRP). The conditions are based on a concept termed the
Failure Mode Concept (FMC) which provides failure conditions formulated on UD lamina
level that allows for a prediction of critical lamina failure modes, and finally of laminate
failure. Following the FMC there are five independent conditions: For Inter Fibre Failure
(IFF) three, and for Fibre Failure (FF) two conditions. These are based on averaged lamina
stresses. The IFF conditions were basically developed in 1994 and than contributed as a
postrunner to the WWFE-I, [Cun04a, b]. Later, some simplifying modifications have been
made for a more engineering-like application, [Cun06]. In this context, for clarifying some
close links to A. Puck’s research work it is to be noted that - from 1992 until 1997 - the author
was the industrial leader of a research project on the Puck/Hashin IFF Action Plane Strength
Criteria [VDI97].
Strength failure conditions allow for an assessment of a multi-axial stress state on basis of
the uni-axially measured strengths, primarily. Such strengths belong to a distinct strength
failure mode of the types Shear Fracture (SF), Normal Fracture (NF) or Crushing Fracture
(CrF), if the designer is faced with a porous matrix in the test specimen which involves
cavities, voids. This has to be considered when mapping FRP behaviour in the FMC by the
relevant invariants of the UD material.
A failure condition is the mathematical formulation of a failure curve or a failure surface,
respectively. Existing failure conditions often map a course of multi-axial test data by one
global equation, such as Eq.(1). These are often termed ‘fully interactive failure conditions’,
and do not take care whether the data belongs to one or more failure mechanisms or failure
modes. Extrapolations out of the mapping domain, therefore, may lead to erroneous results.
Further, if a correction or change in the domain of one failure mode has to be made this
usually affects the domain of another independent mode. This is a mathematical consequence
but it is physically not correct.
Driven from the shortfalls of such a ‘global fitting’ the author looked for a ‘failure mode-
related fitting’ which then, however, will require a failure condition for each mode, Eq.(2),

WWFE-II-Part A_dispatched Dec09 6 of 77 29/12/2009, 8:07

 1 global failure condition : F ( {σ}, {R} ) = 1 (usual formulation), (1)
a set of mode failure conditions : F ( {σ}, Rmode) = 1 (FMC principle) . (2)
Such a set of mode failure conditions will piecewise build up the multi-dimensional failure
surface and also the failure curve. In the case of brittle materials - such as with the usually
utilized epoxy-based FRP composites - failure coincides with fracture. For the more
plastically behaving thermosets and rubber-modified epoxy matrices this has to be checked.
Fracture is understood in this paper as a separation of material. The material is assumed to be
initially free of damage such as technical cracks (size in the order of mm) and delaminations
but not free of small flaws (size in the order of microns) prior to loading. Types of fracture
which are recognised by fractography in case of pore-free (‘dense’) transversely-isotropic
ideal UD materials are brittle fracture failure (Normal Fracture (NF)  cleavage) and shear
fracture (SF) failure under compression.
To achieve a closed understanding some conclusions are recalled from WWFE-I:
• The IFF modes incorporate cohesive fracture of the matrix and adhesive fracture of the
fibre-matrix interface. Both together are often termed ‘matrix failure’, only.
• An IFF mode normally indicates the initial failure or onset of failure in a laminate. Two of
the 3 IFF modes, IFF1 and sometimes IFF3, can be tolerated if the effect on the design
requirements allows for it. The appearance of a FF mode in a well-(strength) designed
laminate usually marks the final failure of the laminate.
• Like the FFs, the 'explosive' effect of a so-called wedge fracture failure (the 2c-caused
IFF2) of an embedded lamina of the laminate may also directly lead [Puc96, VDI97] to
final failure, e.g. in the case of a torsion spring or, via local delaminations, to buckling of
the adjacent laminae and therefore to final failure of the laminate, too.

The FMC generates a phenomenological 3D lamina stress-based engineering approach for

the derivation of failure conditions. The FMC regards the mechanics (physics) in the pure
failure mode domains and applies simple probabilistics in the interaction zones of the failure
modes, thus finally leading back to a formulation that looks again like a formulation of a so-
called ‘global failure surface’. The system of strength failure domains is denoted: FF1→ F|| ,
σ 
FF2→ F|| , IFF1→ F , IFF2→ F , IFF3→ F|| . Here, the superscripts and indicate that
a normal stress or a shear stress-governed failure is given. In order to avoid a misuse of the
strength and elasticity properties these are written symbolically whenever necessary.
Main topics of the paper will be:
 a presentation of the fundamentals addressing lamina stresses, invariants, necessary
properties, and observed failure modes
 a comprehensive description of the invariant-based FMC (generally applicable to
materials which can be homogenized, for isotropic and anisotropic ones)
 a preliminary investigation for the author’s understanding of the hydrostatic pressure-
caused effects on matrix stiffness and strength
 the prediction of the Test Cases
 as Annex: determination of the two FMC-based friction parameters with representation
of the Mohr-Coulomb relationships.
In the following, the main aspects of the FMC shall be shortly recalled, its application and
success:

WWFE-II-Part A_dispatched Dec09 7 of 77 29/12/2009, 8:07

- The FMC, see [Cun04a], is a general concept (applicable not for composites only) for the
establishment of failure conditions (F = 1) for initial failure (corresponding for composites to
the IFF of the lamina) and final failure. The complete failure surface consists of piecewise
smooth regimes or partial failure surfaces. Each regime represents 1 failure mode and is
governed by 1 basic strength. Each failure condition describes the interaction of stresses
affecting the same failure mode and enables to determine the stress effort in a 'material point'.
It is an effective concept that utilizes a very strict failure mode thinking as well as the
application of material symmetry-related invariants which are dedicated to a volume change or
a shape change, or to material internal friction the homogenized material element may
experience. It finally captures all UD failure modes in one equation avoiding the shortfalls of
the usual ‘global fitting’ strength conditions. Failure mode identification is inherent to the
FMC.
The FMC uses a smeared macro-mechanical modelling of a UD material which sometimes
comes to its limit, e.g. in the mode F|| . In this case, failure occurs even for  1   ||  0 in
the case of bi-axial compression when the filament’s tensile strength R|| f is reached due to the
Poisson effect. This is of high importance in case of confining pressure.
- The FMC-based failure conditions are capable of predicting (curve fitting) the mechanical
behaviour under bi-axial and tri-axial stress states, if all physically necessary properties are
provided. Its excellent mapping of in-plane multi-axial strengths has been proven in the
WWFE-I (sometimes first, after being capable to re-evaluate the originally provided test data
due to finally provided additional information as for TC 2 and TC 7 in WWFE-I, Part B).
Sufficient for pre-dimensioning is the knowledge of the basic strengths. A remaining
unknown curve parameter b which concerns the material friction can be estimated.

1.3. Need to move to 3D failure analysis

Tri-axial failure states are encountered in submarines, bolted and screwed joints, bearings
(such as sealed polymer bearing cartridges pressurized up to 600 MPa), impact and ballistics,
and other applications like composite high pressure vessels. However in the past, mainly
experimental investigations have been performed including bi-axial and tri-axial compression
(such as hydrostatic pressure loading p hyd ) and not so much effort was put on analytical
investigations. In consequence, there is a strong need to validate failure conditions in the
compression domain.
Basically, high pressure and especially high hydrostatic pressure has several effects: firstly it
‘heals flaws’ and increases stiffness and strength, however, final failure behaviour becomes
more brittle (sudden fracture, damage tolerance capability of the structure is reduced);
secondly it elasto-mechanically ‘strengthens’ the compressed UD solid; thirdly it causes at
high pressures a weakening of the matrix with a following reduction of the UD material’s
stiffness and strength. Therefore, the effective ‘local’ hydrostatic pressure pef in the matrix has
to be determined for the consideration of the latter effect. As some actually available literature
on compressive hydrostatic loading may be cited [deTer99, Hop95, Pae77, Pae96, Par85,
Rhe01, Rhe04, Shi92a,b]. They fully prove that the behaviour of the UD composite in the
higher phyd regime cannot be accurately predicted without any knowledge on the effect of phyd
on a weakening of the matrix material.
Recently, in [Hine05] a survey was given on the effect of hydrostatic pressure on the
mechanical properties of UD-GFRP. It ended with the conclusion: “For a better utilization
WWFE-II-Part A_dispatched Dec09 8 of 77 29/12/2009, 8:07
of composites it is essential that the mechanical behaviour under tri-axial loading is
adequately understood”. Thus, the Hine paper may be interpreted to be the starter for the
WWFE-II organizers. Unfortunately various tri-axial test results, still compared by the
author, were not consistent. This makes the WWFE-II the more necessary.
Mind: Moving now from 2D-laminates in the WWFE-I to 3D-laminates which do not have
a ‘z-thread reinforcement’ requires an adequate consideration of the thickness direction
because its properties - in general –are lower than the in-plane ones. This would cause the
single lamina of the laminate to be modelled as an orthotropic material which is more
complex than the transversely-isotropic UD-lamina material. However, since the specific
thickness properties were practically not provided for Part A, the author necessarily sticks
to the transversely-isotropic UD lamina. So, WWFE-II is in line with WWFE-I.

2. Formulation of the FMC theory

2.1. Fundamentals on stresses, strengths, invariants and micro-mechanics
A UD-lamina consists of the constituents fibre, matrix and interphase (≠ interface). The
interphase ‘material’ is usually considered by the matrix material properties. After
homogenization the UD-lamina is called here UD material. For the UD material element, Fig.
1 depicts the 3D state of stress    ( 1 ,  2 ,  3 , 23 , 31 , 21 ) T . In the coming text for varying
quantities such as stresses the subscripts 1, 2, 3 are utilized, but for properties symbolic signs.
Because of the symmetries of the transversely-isotropic UD material, modelled an ideal
crystal, there are 5 basic strengths and 5 elasticity properties, only. Therefore, the
characterisation of the strength requires the measurement of five independent basic lamina
strengths (in brackets the US denotation): R||t (= X t ) and R||c (= X c ) as tensile and compression
strength parallel to the fibres; Rt (= Y t ) and Rc (= Y c ) as tensile strength and compressive
strength transversal to the fibre direction); and R|| (= S ) as in-plane shear strength. The five
strengths are given in the old symbolic denotation, used in the German guideline [VDI2014],
in order to avoid misunderstanding in the application of material properties as FEA input. The
measurement of above these five strengths in R  ( R||t , R||c , Rt , Rc , R|| ) T is standard.
R||t is governed by the strength of the constituent filament (fibre stands sometimes for a
bundle or a tow of filaments and, therefore, cannot be always used as a synonym). R||c is
dependent on the shear failure limit of the filaments in a stiff matrix, on the micro-buckling of
the filaments, and on the more structural failure of a bundle of filaments termed 'shear
stability' or ‘kinking’. This includes different micro-failure mechanisms: The matrix may
shear under loading and does not stabilise the generally somewhat misaligned filaments
embedded within. Hence it comes to splitting of the filaments under the pressure head. Also,
the load "grasping" filament as brittle constituent may shear under  ||c . This is filament (fibre)
material behaviour. Rt is determined by the relatively low strength properties of the matrix
(cohesive failure) and the interphase material (adhesive failure of interface filament-matrix
caused by a weak filament-matrix bond), as well as by the filaments acting as embedded stress
raisers.
Strength criteria (F < = >1) or failure conditions (F = 1) should be formulated by invariants
based on the macro-mechanical UD-stresses, see Fig. 1. Invariants have the advantage that the
failure condition does not depend on coordinate-system transformations, because invariants
are a combination of stresses – powered or not powered - the value of which does not change
WWFE-II-Part A_dispatched Dec09 9 of 77 29/12/2009, 8:07
when altering the coordinate system. Their application is advantageous for an optimal
formulation of so-called scalar Failure Conditions. Due to Ref. [Has80] and others, in
general, failure conditions can be at most a function of the stress invariants under the rotation
of the x2, x3 -axes around the x1-axis or fibre axis, respectively.
The material symmetry-based UD invariants ( I 4 , I 5 from [Boe85]) used in the FMC are:
I1   1 , I2   2   3 ,
I 3   312   21
2
, I 4  (  2   3 )2  4 23
2
, (3a-e)
I 5  (  2   3 )(  312   21
2
)  4 23 31. 21 .
(for comparison the invariant set in [Has80] may be viewed.)
The invariant I3 represents the square of the axial shear stresses and I4 the square of the
transverse stresses. The invariant I5 considers the physical difference of  21 ( 2 ) to  31 ( 2 ) .
The formalistic sensitivity of I 5 in the design to shear stresses is suppressed if a ‘main axes
transformation’ around the  1 -axis is performed (see Fig.1), leading to  23  0 or - in other
words - if a transformation in the quasi-isotropic  -plane is accomplished via
(  1 , 2 , 3 , 23 , 31 , 21 ) → (  1 , 2p , 3p , 0 , 31p , 21p ) , utilizing the superscript p for principal.
In addition, the second term in I 5 , namely 4 23 31 21 , may be deleted because this
combination is very seldom of importance. The same is valid for the second term in I 4 (in TC
11 however essential).
The choice of the invariants is physically based on Beltrami’s strain energy density
formulation, see [Bel1885, Cun04b].
Fig. 2 defines the positive angle of fibre orientation in each lamina. It is measured from the
x-direction to the fibre-parallel direction. For further explanation, the in-plane state of stress is
given for the lamina coordinate system as well as for the laminate coordinate system.

2.1.2. Stress-strain relationship and micro-mechanical formulae

For accurately including pressure effects the tri-axial macro-mechanical stress–strain
relationship (Hooke) has to be employed as well as micromechanical formulas.
The stress-strain relationship reads, following [VDI 2014],
 1  ||  || 
 0 0 0 
 E|| E E 
  || 1   
 1   E E E
0 0 0   
1
   ||   

 2    ||   1   2
{}= [S] {} or  3   E 0 0 0    . (4)
E E  3
   ||   
 23   1
0
  23
0   
 13   G   
     13 
 12   ( symm.)
1
0   12 
 G|| 
 1 
 
 G|| 

Note: In the above compliance matrix, the subscript  is used - according to the chosen
model, the transversely-isotropic UD material - to denote both transverse direction and
through-thickness direction. Maxwell-Betti: ||  E   ||   E|| , G  E  /(2  2  ) .

WWFE-II-Part A_dispatched Dec09 10 of 77 29/12/2009, 8:07

 For imposing pressure-related effects on the matrix and in consequence on the UD material
micromechanical formulas had to be applied together with the data. However, the organizers
could not provide data with associated micro-mechanical formulas. Therefore, the author took
those, given in the VDI 2014, p. 29, which are valid for a non-creeping filament (fibre):
E||  E|| f  V f  E m  (1  V f )  E f  V f ,

1  0.85  V f
2
Em
E   , (5a-d)
1  m (1  V f )1.25  V f  E m /( E  f  (1   m ))
2 2

G m  (1  0.4  V f
0 .5
)
G ||  ,
(1  V f )1.5  V f  G m / G || f

 ||   || f  V f   m  (1  V f ) ,

with Gm  E m /( 2  2  m ) . The suffixes f and m stand for filament (or fibre) and matrix.
Eq.(5) may be used for the regime 0.3  V f  0.65 if they have been verified by a top-down-
bottom-up approach or in other words by a two scale simulation on the upper scale which is
the lamina in the actual micro-mechanical case. Such an approach links two scales or
structural levels and the micro-mechanical model utilized in this approach is to be validated
on the upper level. Hence, the non-measurable fibre properties can be computed backward
from the macro-mechanical lamina test data. For instance for Eq.(5b), just the constituent
properties  m and E m remain to be measured and E  f is then to be resolved by inserting the
measured E  . Therefore, the provided micromechanical properties later to be used to describe
pressure-dependent changes of the constituent matrix (fibre behaviour is constant) might not
be applicable due to missing of the corresponding micro-mechanical equations.

2.2. FMC-based failure conditions and their interaction

2.2.1 Derivation of failure conditions
Failure conditions should exhibit besides a sound mechanical basis some numerical
advantages: homogeneity in the stress terms, possessing stress terms of a low degree, show
simplicity, be numerically robust and allow for rapid computation.
In engineering practice, due to all the model uncertainties such as e.g. property scatter, the
simplest strength condition which still describes the physical phenomenon should be applied.
This always reduces the number of model parameters to be determined and additionally the
numerical effort. In total, 5 strengths and 2 material friction-related parameters (described
later) will be necessary when applying FMC-based UD failure conditions.
Using the FMC methodology for UD material is first to propose a set of equations
describing the five failure modes of each lamina (ply) by one failure function each, and
secondly to combine these equations in a suitable manner to predict failure in a lamina in
order to consider the superposition of all modes. The set of failure functions F is:
I1  I1
FF 1 : F||  , FF 2 : F||  ,
R||t R||c
I2  I4 
I 2 b I 4
IFF 1 : F  , IFF 2 : F  ( b  1 )  , (6a-e)
2 Rt Rc Rc

WWFE-II-Part A_dispatched Dec09 11 of 77 29/12/2009, 8:07

 I 33 / 2 I2  I3  I5
IFF 3 : F||  3
 b|| 3
.
R|| R||

Furthermore, as abbreviation I 2  I 3  I 5  I 235 will be used.

In the equations above, R denotes an average (typical) strength value that is to be used in
stress and deformation analysis when dealing with test results.
t c σ 
The superscripts , stand for tensile, compressive. The superscripts and mark the type of
fracture failure whether it is caused by a tensile stress  (NF) or a shear stress  (SF), e.g.
due to a compressive normal stress  ||c or a transverse normal stress  c . Whether a failure
may be called a SF or a NF, depends on the envisaged size scale. An example is IFF2: It is
macro-mechanically a SF and micro-mechanically a NF of the matrix.
When idealising a material as a crystal that possesses a distinct material symmetry (isotropic
or anisotropic body) then one can utilize the information: “The number of symmetries
determines the number of strengths and elastic properties”. However, such a material model is
an ideal one. In order to formulate strength conditions for a real material missing parameters
are to be determined. These are in the case of the brittle UD material above two parameters
b|| and b , representing internal friction of the material, which are necessary for the
application of the material friction-affected strength modes IFF2 and IFF3.
The basic strength in a failure mode condition controls the (size) volume of the partial failure
surface (body) whereas the friction parameter controls the shape of the partial failure surface.
A failure condition of a specific mode contains stress terms that imply an interaction
between the various stresses (stress tensor components) which are active in this mode. If these
stresses (from FE output e.g.), inserted into the failure functions F, are fracture stresses it
 
leads to a fracture condition because F mod e (  fr , R mod e )  1 is fulfilled.

2.2.2. Determination of the friction-related b-parameters

In the Eqs.(6) the five average strength values as anchor points have been still introduced.
So, besides the always design-mandatory five strengths the two friction-related model
parameters (b|| , b ) are to be determined from multi-axial test data in the associated ‘pure
domain of validity’ of the respected mode by a trial & error fit of the course of test data, or
statistically (minimum of the error squares) or can be estimated by experience from test curves
of similar UD composites. For IFF2, a fracture angle measured in the uni-axial Rc
compression test may be alternatively employed. This procedure is executed in the Annex of
this paper.
As a simplifying procedure, one statistically based calibration point for each of the two modes
delivers, after inserting its coordinates into the IFF conditions (for F|| from a point
 2
c

,  21
||
, see [Cun04a,b], and for F from (  2c ,  3c )), and after a resolving the two failure
conditions of Eq.(6), namely F||  1 and F  1 the determining equations for the b-
parameters
4 || 4
R||  21 1  ( 2c   3c ) / Rc
b||  and b  . (7, 8)
2 2c ||   21
||2
 R|| ( 2c   3c ) / Rc  ( 2c   3c ) 2 / Rc

Above parameters are obtained by a curve fit.

WWFE-II-Part A_dispatched Dec09 12 of 77 29/12/2009, 8:07
The author’s actual experience with own and WWFE-I test data leads to the conclusion:
Bounds for typical epoxy–based GFRP, CFRP and AFRP are: 0.1  b||  0.45 and
1.0  b  1 .6 . A value b||  0 means that there is no increase of shear resistance due to
friction in the compression domain. A value b = 1 means 'no material friction' is existing in
the quasi-isotropic -plane. In the 2D case, b is not active, see Ref.[Cun03]. For pre-
dimensioning is recommended b = 1 as it provides a safe lower bound, and b||  0.15 .
Mind: The author should have better denoted in the past the term ( b  1 ) as the friction-
related curve parameter because this leads to 0 at zero friction.

2.2.3. New developments in the FMC theory since WWFE-I

In reference [Cun06]) an effective modification of one IFF condition is highlighted. It
consists in the replacement of the so-called ‘in-plane shear failure mode condition IFF3’ by a
numerically advantageous formulation. In that paper, a very satisfying semi-validation of the
3D IFF3 condition, reduced to 2D, could be achieved when judging it versus the 2D
experiments provided by the WWFE-I and other sources. Therefore, this condition shall be
introduced as WWFE-II condition.
The driving reason was a numerical problem in the ‘old’ formulation that probably no
intersection is achieved of the applied stress vector with the mode curve IFF3. This was
tackled by a query, that requires  21  max  21 , see Appendix B of [Cun04a].
The numerical problem above is just a problem of the mode interaction zone (see section on
mode interaction). Instead of a query it can be bypassed when, at first, for F|| the usual
principle of proportional stressing (all stresses of the actually given state of stress or stress
vector are equally factored) id abandoned, that interprets so-called mode reserve factors as
mode stretch factors of the actual stress vector. This stretching ends when the associated mode
failure curve is met.
So, when establishing the interaction zone, the description can be modified if the physical
basement is not violated, I 235 is kept in order to describe the different physical effects of  2
and  3 in the possible combinations with shear stresses. Now, instead of factoring the full
state of stress in Eq.(9), that means factoring each single stress in the invariants used, just the
mode driving shear stress(es)  21 , 31 is(are) factored. And not for instance the shear stress  23
(has a normal stress effect; remind transformation).

Unfortunately, with this approach the failure condition includes a reserve factor with power 2
and 3
3 2
I 33 / 2  f Re||s I 235  f Re||s
3
 b|| 3
 1 (9)
R|| R||
with f Remods e as reserve factor of the IFF3 mode.
The determination of a mode reserve factor requires the solution of an unpleasant third order
equation. Therefore, the approach for establishing the interaction domain will be modified in
order to obtain the numerical advantageous powers 4 and 2

WWFE-II-Part A_dispatched Dec09 13 of 77 29/12/2009, 8:07

 4 2
I 3  f Re||s I 235  f Re||s
2

4
 b|| 3
 1 . (10)
R|| R||
Of course, the parameter b|| is now slightly different to the former one. The rationale behind
the creation of Eq.(10b) is that this procedure has no effect because this condition is used then
instead of Eq.(6e) as the new mapping function or curve fitting function, respectively.
The procedure takes the former unpleasant non-intersection problem completely away, no
queries are necessary anymore. Just a quadratic equation needs to be solved anymore.
As it is a measure, which is also applicable in linear and non-linear analysis and which is
more advantageous for the coming investigations, a more general quantity shall be introduced.
It is the so-called ‘mode stress effort’ which is - if linearity is valid - the inverse of a material-
based mode reserve factor. The stress effort has the advantage that it is always material-based
and can be applied for linear and nonlinear analyses, and further that the correct original
definition of the reserve factor (load-based factor, by which one can increase a Design Load)
can be kept. Especially in the case of residual stresses instead of a material-based reserve
factor f a material stress effort Eff (linearly: inverse of f) is employed, see section 2.2.5 .

2.2.4. Limit of macro-mechanical fibre failure descriptions

In general, the fibre failure mode FF1 cannot be described by a homogenized (smeared)
macro-mechanical stress value  1 . Thus, the engineering-like macro-mechanical modelling
has to be replaced by an accurate micro-mechanical one. Praiseworthy, this can be
approximately well formulated by a macro-mechanical quantity which is the FEA-computed
macro-mechanical strain  1 . Reminding reference [Cun04], it follows
FF1: F|| : I1 =  1  V f  1 f  V f   1  E1 f   1  E|| (11)
with  1 f as tensile filament (fibre) stress which is proportional to the strain and responsible
for fracture. Nevertheless shall be recalled: Within the FMC no fibre properties are required!
Above reformulation, Eq.(11), is necessary if multi-axial compression is applied which
means that the Poisson effect is active and if FF1 may occur even in case of zero applied
external stress  1 because bi-axial compression may cause fibre fracture.
Also in the case of FF2 this effect has to be considered.

2.2.5. Stress effort of modes and corresponding equivalent stress

The stress effort of each mode is derived by resolving the equations:
[  1 ] / Eff || [  1 ] / Eff ||
1 , 1 ,
R||t R||c

[(  2   3 )  (  2   3 ) 2  4   23 ] / Eff
2

1 , (12a-e)
2 Rt

[ b  (  2   3 )2  4  23  ( b  1 )  (  2   3 )] / Eff 

2

1 ,
Rc
(  31
2
  21
2 2
) 2 2   21
2
 2 3   31
2
 4 23 31 21
 b||  1
Eff 4||  R|| Eff 2||  R||
4 3

WWFE-II-Part A_dispatched Dec09 14 of 77 29/12/2009, 8:07

for the respective effort.
Mind: Fracture stresses are acting stresses divided by the stress effort Eff in order to fulfil
the failure condition F = 1.
An equivalent stress  eq is always positive such as the strength. It includes all actual load
stresses and the residual stresses (from curing etc.) that are acting together in a given mode.
The vector of the modes' equivalent stresses reads

 mod e
eq    ||
eq
||
,  eq ,  eq ,  eq ,  eq
||
 T
. (13)

Employing the mode strength R mod e , its equivalent stress  eqmod e , and Eq.(13) - according to
the general equation Eff mod e   eqmod e / R mod e - the following set of formulas for the stress
effort of each of the 5 modes can be provided and its relationship to the associated equivalent
stress:

FF 1 Eff ||   1 / R||t   eq

||
/ R||t with  1   1t  E|| ,

FF 2 Eff ||    1 / R||c ||

  eq / R||c with  1   1c  E|| , (14a-e)

IFF 1 Eff   [(  2   3 )   2  2 2   3   3  4 23 ] / 2 Rt   eq / R ,
2 2 2 t

IFF 2 Eff   [( b  1 )  (  2   3 )  b  2 2  2 2 3   3 2  4 23 2 ] / Rc   eq / Rc

2 3
||
 {[b||  I 235  ( b||  I 235  4  R||  ( 312   21 ) ] /( 2  R|| )}0.5   eq|| / R||
2 2 2 2
IFF 3 Eff

with I 235  2 2   21
2
 2 3   312  4 23 31 21 . (15)
Above stresses include the nonlinearly load-dependent load stresses {}L and the equally
nonlinearity dependent residual stresses {}R .
Note: 1) Each failure mechanism is affected by an associated typical stress state. The failure
mechanism with the highest stress effort will dominate the failure. The mode effort has to
become zero if the mode driving stress is zero!. 2) Due to IFF the curing stresses decay in
parallel to the degradation. 3) The not design driving stresses of a mode might increase or
decrease the stress effort of the design driving one. This is pronounced by  eq . 4) Not a
Mises equivalent stress exists only). There are others, too.

2.2.6. Interaction of failure modes

Due to the fact that the full failure surface consists of five parts an interaction of these
partial surfaces has to be executed. Cuntze models these failure mode interactions by a simple
probabilistically based ‘series spring model’ approach [Cun04, Rac87]. Such a model
describes the lamina failure system as a series failure system which fails whenever any of its
elements fails. Each mode is one element of the failure system and is seen to be independent
of the others. The series spring model was proposed as engineering approach because it
approximates the results of a time-consuming probabilistic interaction calculation on the safe
side.
By this method, the interaction between FF and IFF modes as well as between the various IFF
modes acts - in these ‘mixed’ failure domains - as a rounding-off procedure linked to the
determination of the desired values for the resultant stress effort Eff . This effort

WWFE-II-Part A_dispatched Dec09 15 of 77 29/12/2009, 8:07

automatically takes into account the interactions between all the affected modes by summing
up all the proportionate mode stress efforts according to Eq.(16)
5


m
 ) m  ( Eff )  ( Eff )  ...  ..
mod es mod e 1 m mod e 2 m
Eff ( Eff in general (16a,b)
1

Eff m
 ( Eff || ) m  ( Eff || ) m ( Eff 
) m  ( Eff 
) m  ( Eff || m
) for UD

= 1 = 100% , if failure.
In other words, the interaction equation includes all mode stress efforts and each of them
represents a portion of load-carrying capacity of the material. In practice in thin laminae, at
maximum, 3 modes of the 5 modes will physically interact. Considering 3D-loaded thick
laminae, there, all 3 IFF modes might interact.
For the application of the Eq.(16) a value for the interaction exponent m is to be provided.
Usually, the value of m is obtained by curve fitting of test data in the interaction zone.
Mapping experience shows 2.5 < m < 3 , e.g. for CFRP. The mode interaction exponent m is
also termed rounding-off exponent, the size of which is high in case of low scatter and vice
versa. As a simplifying engineering assumption, m is always given the same value, regardless
of the distinct mode interaction domain! As with other interaction equations also for m it is
valid: a lower value chosen for the interaction exponent is more on the safe side.
Rounding-off, by employing an interaction equation in mode interaction domains of
adjacent mode failure curves (2D) or of partial failure surfaces is leading again to a pseudo-
global failure curve or surface. In other words, a ‘single surface failure description‘ is
achieved such as with Tsai/Wu [Tsa71], however, without the well-known shortcomings.
If a unidirectional fracture stress (i.e.  2fr as transverse strength value Rt ) is inserted into
the equation above, then a point on a failure curve or on the 3D-failure surface, described by
F  1 or Eff = 1, is determined. A failure surface is the result of optimally mapping the
course of multi-axial test data and, therefore, has the attribute ‘50% survival probability’.

Of interest is not only the interaction of the fracture surface parts in the discussed mixed
failure domains or interaction zones of adjacent failure modes, respectively, but further failure
in a multi-fold failure domain (superscript MfFD) such as in the (  2t ,  3t ) -domain. Here, the
associated mode stress effort acts twofold. It activates failure in two directions and may be
engineering-like considered by adding a multi-fold failure term, proposed by Awaji in
[Awa78] for isotropic materials, which can be applied to UD material in the transversal
(quasi-isotropic) plane as well
Eff m  ...  ...  ...  ...  ( Eff  ) m  ( Eff MfFD ) m (17)

with Eff MfFD  ( 2t   3t ) / 2 Rtt , Rtt  Rt / m 2 .

Eq.(17) practically represents a biaxial tensile strength Rtt . The effect above, denoted joint
failure probability, is inherent to brittle materials. The development of the oriented flaws and
their growth is driven by across acting principle stresses in 3D states of stress. This is valid in
a 2D manner for UD material consisting of the usual matrix materials: in the  -plane the
UD principal stresses act perpendicular to the fibre direction and both stresses  2 and  3
have statistically the same effect under tensile loading.
Into Eq.(16) all mode efforts have been inserted in order to practically have the full set of

WWFE-II-Part A_dispatched Dec09 16 of 77 29/12/2009, 8:07

conditions in just one equation combined during the numerical analysis instead of dealing with
each single failure condition. However, when automatically inserting the FEA stress output
  ( 1 ,  2 ,  3 , 23 , 13 , 12 ) T into all 5 effort equations some efforts may become negative
which mechanically means zero effort. One can solve this problem by using the absolute
values of the several parts of the envisaged failure condition. The following simple 2D
example ‘in-plane loading’ demonstrates that all mode stress efforts which are not activated
can be made zero
( 1   1 ) 1  1 2  2  21  2   2
Eff m
( t
)m  ( c
)m  ( t
)m  (  R2 D ) m / 2  ( )m , (18a)
2R || 2R || 2R  R || 2 R c

R 2 D  b||   2  (b||   2 ) 2  R|| .

2
with the radicand
Eq.(18a) is obtained after inserting   (  1 , 2 , 12 )T into Eqs.(14) and then summing up the
single portions. A not essential mode effort (is a risk measure) which does not drive the design
is automatically sorted out.
Instead of applying the procedure above one may also formalistically take the Macauly
brackets (≡ Föppl symbols). They describe a discontinuous function and are defined here by
0 , Eff mod e  0
Eff   mod e mod e  .
mod e

Eff , Eff  0 
For completion, the 3D formulation of Eq.(16b) is given again (for the Effs see Eqs.(14))

Eff m
 Eff   Eff  Eff   Eff   Eff 
|| m || m  m  m  || m . (18b)
This equation can be developed similarly to Eq.(18a) and written.

2.2.7. Failure conditions for delamination

Delamination conditions are just a subset of the FMC conditions set and are thereby
captured by the Eqs.(14). They are given here in a separate manner because other researchers
present special delamination conditions.
With regard to the 3D nature of the IFF conditions, both, IFF1 ( F := transverse tensile
failure; inter-laminar stresses  3t , 32 , 31 may cause cracking) and IFF2 ( F := wedge failure;
intra-laminar stresses such as  2c , 21 cause cracking and may initiate a local 3D state of stress
activaing  3 ) can also serve as conditions for the assessment of ‘onset of delamination’
which is – in general – called laminate failure. One or two modes will be the design driving
ones in the critical local 'material' point of a composite lay-up. These are activated by the
delamination-critical stress state  la min a  ( 0 , 2 , 3t , 23 , 31 , 21 )T that includes all
interlaminar stresses. Introducing the two relevant combinations of the delamination-active
stress vector above into the Eqs.(14) delivers:
Tension/shear stressing
( Eff  ) m  ( Eff  || m
)  1 with  la min a  (0,  2t ,  3t , 23 , 31 , 21 ) T , (19a)
Compression/shear stressing
( Eff  ) m  ( Eff || ) m  1 with  la min a  (0,  2c ,  3c , 23 , 31 , 21 ) T , (19b)

and interacted, when necessary. For engineering reasons, the interaction exponent m is chosen
the same as before.

WWFE-II-Part A_dispatched Dec09 17 of 77 29/12/2009, 8:07

2.3. Investigation of the pressure effect
2.3.1. Elasto-mechanical analysis with ‘Birch stiffness increase’ effect
One can separate the pressure effect into an elasto-mechanical part and a material part
(matrix weakening) addressed later as ‘2ndTg shift effect’. The elasto-mechanical part is
automatically included in the elasticity equations but the material part from matrix weakening
has to be introduced specifically due to a different influence size and missing clearness about
this size. Matrix weakening is effective on strengths, elastic constants, ultimate failure strains beyond
a distinct hydrostatic pressure and is to be regarded when the experiments are executed at ambient
(room) temperature which is the case in the WWFE-II. The term 2ndTg shift means that the lower (2nd)
glass temperature of a polymer comes to lie above 0° C if the multi-axial pressure exceeds this distinct
value. The 2nd Tg shift alone has a direct impact on the strength value. The effect of pressure with
respect to a ‘healing’ of the ‘diffuse’ micro-cracking is included in the lamina or laminate high
pressure test data and cannot be separated as the author believes. In this context it is to be recalled:
The pressure-based lowering of the equivalent stress should not be mixed up with an increase of the
strength (mind  eq ( ) / R  1 ).

Investigating the elasto-mechanical part is to consider the pressure-dependent stiffness

increase of the polymer solid termed here the ‘Birch stiffness increase’ (finite strain theory,
[Bir38, Mar37], compressive bulk modulus). This is considered in an additional stiffness
correction directly in the usual elasto-mechanical analysis (infinite strain theory). Its increase
in matrix stiffness is introduced via the cited micro-mechanical equations.
Some insight – in the frame of the elasto-mechanical analysis - can be gained from inserting
phyd into the isotropic Hookean law shown as general formulation (x,y,z system) or as the
equivalent formulation in principal stresses (I, II, III system)
 1   
 E 0 0 0

E E
   1   
 x    1

   E
0 0 0   x 
    I   E E E   I 
  
E E
 y       y     1 
 II       II  ,
1 (20)
 z   E 0 0 0   ,

  
E E    z     E E E   
1  III      1   III 
 yz   0 0   yz 
   G     E E 
  xz 
E
 xz   1
 0   xy 
 xy  
( symm.)
G   
 1
 
 G
from which follows E  (  x   y   z )  3  p hyd  ( 1  2  ) or E  V / V  p hyd  ( 3  6  ) .
It can be recognized that the bulk modulus K  E /( 3  6 ) is ‘activated’.
When the material condenses and the volume change V approximates zero this will have
an effect on  the size of which depends on the applied material. For the epoxy matrix it can
be assumed that the matrix material becomes denser or that the little pores are reduced and
flaws are compressed. The share of the matrix as the only ‘yielding’ constituent of the UD wrt
micro-damage quasi-yielding is not fully clear. At least it depends on the ductility of the
chosen matrix material (resin system). In contrast to a yielding metal the inelasticity also
depends on some visco-elasticity from the constituent matrix and on micro-mechanical
damage. In this paper, the pressure-dependent increase of Poisson’s ratio  is applied as not
to change with pressure according to missing information on the relationship to increasing
inelasticity. How far Poisson’s ratio  might change wrt the material’s compressibility is not
WWFE-II-Part A_dispatched Dec09 18 of 77 29/12/2009, 8:07
provided. However, the sensitivity of a result to  is sometimes considered by a parameter
study.
The hydrostatic pressure-dependent increase of the moduli, measured in experiments, seems
to be mainly caused by the finite strains of the polymer. Therefore, Birch [Bir38] utilized the
finite strains in his approach (based on Murnaghan’s theory) where he still assumed a constant
 . He proposed for the matrix (suffix m deleted here) as ‘Birch finite strain modulus increase
effect’ (Birch effect) for Young’s modulus and shear modulus the growth equations
E hyd / E 0  1  ( hyd / E 0 )  (10  8 )  (1   ) , (21a)

Ghyd / G0  1  ( hyd / G0 )  (9  12 )  (2  2 ) . (21b)

Mind here: G  E /(2  2 ) . Crystalline (epoxy) polymers have a  hyd -dependence which is
higher than for amorphous polymers. For low modulus polymers at atmospheric pressure the
stiffness increase is higher than for high modulus ones. Rubber stiffness increases by two
orders of magnitude, [Hop95].
In addition, for the stress analysis some insight on the infinitesimal (squares and products of
the strains are neglected) elasto-mechanically-caused pseudo-stiffness change of an isotropic
polymer material such as the matrix is obtained by studying a simple elastic model. Essential
for the evaluation is the differentiation of the hydrostatic pressure stress and an additional
stress which may act together at the x-cross section. Of course, as basis the associated moduli
given above have to be inserted as E and G. To illustrate this pseudo-stiffness change, elastic
behaviour as well as  < 0.5 is assumed for the model. In this context and in supporting the
test data evaluation in WWFE-II, Part B, the following text parts a), b), and c) are performed:
- a) Pseudo-Youngs modulus E ps of an isotropic material:
In case of a monotonically increasing stress state   (  xadd   hyd , hyd , hyd ,0 ,0 ,0 )T ,
which is a simultaneous superposition of a uni-axial additional stress and a hydrostatic stress
state, the stress-strain equation
E  ( Iadd   Ihyd )   xadd   hyd  (1  2 )  E   xps (22a)

is obtained with the strain  xps , measured under combined loading. For the simple uni-axial
loading alone the equation reads
E   xadd   xadd . (22b)
If the normal stress is the same in both cases then  xadd  E   xadd  E ps   xps is valid and the ratio
of the two strains delivers a function for the change of the Young’s modulus
 xadd E ps  xadd 1
   (22c)
x ps
E  x   hyd  (1  2 ) 1  f hyd  (1  2 )
add

when utilizing the relation  hyd  f hyd   xadd . From this it is obvious (proportional loading) that
for tensile stress  xadd with a  hyd   p hyd the factor f hyd becomes negative. This means with
respect to E   xadd that the slope increases with increasing p hyd or in other words, a pseudo-
stiffness increase is elasto-mechanically achieved under hydrostatic compression. For
compressive stress  xadd  200 MPa with  hyd  600 MPa follows f hyd  3 , and when
setting   0.4 , an elasto-mechanically-caused stiffness decrease of 29%. With increasing 
the size of the effect is vanishing according to ( 1  2 )  0 .

WWFE-II-Part A_dispatched Dec09 19 of 77 29/12/2009, 8:07

 - b) Pseudo shear modulus G ps of an isotropic material:
In the case of  hyd with an additional shear loading  the shear stress  can be replaced by
two normal principal stresses  Iadd    and  IIadd    . This results via the principal load
stress vector   (  Iadd   hyd , IIadd   hyd , hyd )T in the equation

E  (  Iadd   Ihyd )   Iadd     IIadd   hyd  ( 1  2 )    ( 1   )   hyd  ( 1  2 ) . (23a)

For the simple uni-axial loading alone the equation reads

E   Iadd   Iadd     Iadd    ( 1   ) . (23b)
If the shear stress is the same in both cases then     G   ps
 G ps is valid and the ratio of the
two shear strains delivers a function for the change of the shear modulus which is proportional
to E, according to  hyd  f hyd   or f hyd   p hyd / 
 xadd E ps   (1   ) 1
   . (23c)
x ps
E   (1   )   hyd  (1  2 ) 1  f hyd  (1  2 ) /(1   )

The shear stress increases with phyd and with increasing  the size of the effect is vanishing.

c) Pseudo Young’s modulus E  of a UD material:

Finally the elasto-mechanically stiffening effect of hydrostatic pressure shall be shortly
described for the envisaged UD material. Here, the stress vector
 la min a  (  hyd , 2   hyd , hyd , 0 , 0 , 0 ) is to be inserted into the Eq.(20b) analogous to
add T

equation
 1  ||  || 
 

 1  E|| E E 
 1 
    || 1      . (20b)
 2       2 
   E|| E E   

 3   ||   1   3
 
 E|| E E  

Corresponding to the isotropic equation the equation

E   (  2add   2hyd )   2add   hyd  ( 1   ||    ) (24a)
is obtained and for simple transversal stressing
E    2add   2add . (24b)
with the in-plane Poisson’s ratio  || , and   as the transverse one. Again assuming the
stress, to be the same in the two cases, delivers the equation  2add   2add  E    2ps  E ps
from which a relationship, similar to Eq.(22c), follows
E ps 1
 . (24c)
E  (1  f hyd  (1   ||     ))

For f hyd   p hyd /  2add , Eq.(24c) results in case of a tensile transversal stress in a stiffness
increase.
To not insert in the analysis a wrong in-plane Poisson’s ratio the Maxwell-Betti relationship
shall be considered, here written in the symbolic VDI 2014 convention, E||  ||  E   || .

WWFE-II-Part A_dispatched Dec09 20 of 77 29/12/2009, 8:07

2.3.2. 2ndTg shift effect
Under high hydrostatic pressure, for instance the shear failure curve of a matrix (in Ref [Pae
77, Pae96, fig.5]) exhibits pretty a non-linearity. This special behaviour of the matrix beyond
a certain ‘knee’ point, Fig. 3a, is generated because the molecular motion between the cross-
links is not frozen anymore. The reason for the occurrence of this knee or kink is attributed to
the shift of the second glass transition temperature point, 2ndTg, up to ambient temperature
where the tests are performed. Therefore, this matrix weakening effect has to be considered in
test data evaluation if the matrix material is subjected to high multi-axial pressure.
The effect lowers modulus and strength as well and varies from material to material. Matrix
weakening causes a loss in stiffness and an increase of the yield capacity (remind: rock
material like e.g. marble yields under high pressure). According to this effect, also the elastic
moduli of the UD material become a function of the applied multi-axial pressure which is to
be incorporated into the stress-strain curves. However, the magnitude of the effect seems to be
not very validated: the best representative shear stress-strain curve found was in the
mentioned figure 5 from Pae. However note: the large strain there is just indicated as an
engineering strain. If this is the case, then the curves will straighten when deriving the true
stress-strain curve, and the curve maximum will lie higher and the effect becomes smaller. So,
clarification is needed.
In the context above Hoppel may be cited, [Hop95]: “In general, hydrostatic pressure
suppresses deformation mechanisms that cause an increase in volume in the polymer and does
not affect mechanisms which are constant volume processes such as with the well-known Mises
shear plasticity”. Thus, for instance a polymer material may deform at atmospheric pressure by
craze formation (means a density reduction, observed by stress whitening). A craze blunts a
micro-crack which grows from a flaw but this effect is eliminated at higher values of phyd. At
a sufficiently high pressure level the density remains constant and the porosity change,
described by I 12 , is not significant anymore. High multi-axial pressure suppresses crack
initiation and crack growth, but, since also the yield strength increases by inhibiting local
yielding in the various yield-endangered material points on the other side the brittleness
icreases. Due to suppressing the crack sensitivity the material obtains a higher deformation
capability by phyd , which means an increase in fracture strain but the final fracture might
occur more suddenly, or in other words, due to reduction of crack tip blunting at the flaw
locations the toughness may be decreased.
Considering UD material, increasing matrix stiffness E m or Gm increases the resistance of
local filament buckling and kinking under compressive loadings. Thus, the stiffness loss of the
matrix also affects the strength of the UD material due to a lower stiffening of the embedded
filaments.
For taking the effect on the lamina properties into account at first the effect on the matrix
properties has to be described. Lacking of provided information on the macro-mechanical high
pressure behaviour of the lamina the predictions have to be based on micromechanics with the
un-reinforced polymer, the matrix material. This is the only constituent which underlies
changes with increasing high pressure. Encountered are several cases:
- Effect on matrix moduli E m , Gm :
For an epoxy material, Fig. 3 depicts a bi-linear curve with a knee or inflection point or kink
at 200 MPa, according to [Shin96]. This knee value is typical for the matrix utilized.
WWFE-II-Part A_dispatched Dec09 21 of 77 29/12/2009, 8:07
From the various cited references a coarse approach can be only constructed on basis of the
gathered test results which seem to be relatively generally valid for the epoxy matrix family.
The kinked curve can be fitted in the usual high pressure domain in a bi-linear manner by the
function (series spring model again)
M M
   
M
 1  1 1
      (25)
1  a    a  a  
 Gm / Gm0   1 ef   2 3 ef 
with the fitting parameters a1 , a 2 , a 3 and the weakening exponent M. The effect on the shear
modulus can be tackled by the basic value Gm0 at room conditions times (effective beyond the
knee point) a reduction factor f 2 ndTg , termed 2ndTg-shift factor (see Fig. 3b)

M
 1  a1   ef 
G m  G m 0  f (2ndTg ) with f 2 ndTg  1/ M 1    . (26)
 a  a  
 2 3 ef 

To be inserted into this equation is an effective hydrostatic stress value

 ef  ( I , m   II , m   III ,m ) / 3 with e.g.  I ,m   1  E||  V f /(1  V f )
2
(27)
wherein the  l are the smeared principle stresses of the matrix at the envisaged critical
location within the embedded constrained lamina. Both quantities,  ef and  l are sign-
dependent. This means the pressure is inserted as a negative value and a hydrostatic tensile
stress as a positive value. Eq.(27) involves the possible hydrostatic stress  hyd and external
stresses (  1 , 2 , 3 ) applied to the matrix. Note: The term  ef is nothing else but the
factorized first invariant of the stress tensor.
Pae showed in [Pae96] that the decrease of the E- or G-modulus is similar according to Fig3a
(matrix material PR319). The effect is considered by the 2nd shift correction factor f 2 ndTg ,
depicted in Fig.3b which is taken as standard curve due to missing information for the actual
matrices used in WWFE-II.

- Effect on matrix strengths:

For the matrix strength, Pae showed in [Pae96] that the effect on strength loss is higher than
that for the moduli. As information is missing here also the factor f 2 ndTg is fully applied in the
test cases if not indicated otherwise.
- Transfer to UD stiffnesses and strengths:
The physical behaviour of the matrix has to be considered when predicting a matrix-
dominated UD material behaviour. For the UD material, the 2ndTg-effect and therefore the
reduction will differ from that which was obtained for the matrix, the effect is usually minor.
However, lacking of a provided input for the materials, in the Test Cases the strength
reduction according to Eq.(26) will be simply transferred from the matrix to the UD material
strengths. In cases where more knowledge seems to be available the reduction is indicated.

2.3.3. In-situ effect or constraint effect of embedded laminae

In-situ strengths of a lamina which is here a layer of a laminate depend on its thickness (for
this so-called ‘thin-layer effect’, see e.g. [Cam06, Fla82]) and on the location in the laminate,
that means whether the layer is fully embedded or just an outer layer that is more jeopardized

WWFE-II-Part A_dispatched Dec09 22 of 77 29/12/2009, 8:07

by fracture mechanics-based micro-cracking generating a decrease of the effective moduli.
As still briefly mentioned: When applying test data from tensile coupons to embedded
laminae, one has to consider that tensile coupon tests deliver test results of ‘weakest link type’
(series model) as numbers for the usually applied ‘isolated’ strength properties. An embedded
or - to a less extent - even the only one-sided restrained lamina where surface micro-cracks
can always easily originate, however, is of ‘redundant type’ (parallel spring model). Therefore,
besides being deformation-controlled all the material flaws in a thin lamina cannot grow freely
up to micro-crack size in thickness direction because the neighbouring laminae will act as
micro-crack-stoppers (micro-fracture mechanics problem, energy release process).
For the description of the full non-linear stress-strain curve (Fig. 4), this includes (strain-)
hardening and (strain-)softening of the embedded lamina, it is important that the peak value of
the effective stress-strain curve is higher than the strength point R of the isolated specimen
due to the change from the ‘weakest link’ behaviour to a redundant behaviour. In the IFF
conditions the classical strength values should be replaced by their respective in-situ values.
See e.g. [Cam05], in which fracture-mechanical, strain-energy-based strengths Rin|| situ ,
Rt ,in  situ were developed. For reasons of numerical simplicity this peak effect is flattened in
the proposed approach for mapping the softening branch. For the execution of nonlinear
analysis the application of such an effective stress-strain curve is very practical. It approaches
the behaviour of the lamina in the laminate regarding the stack, its position, and the thickness.
It is to be noted that after the onset of IFF only ‘smeared’ stresses can be calculated for the
micro-cracked lamina and are inserted into the failure conditions. These stresses are smeared
over some length of the cracking lamina, a length which includes a number of micro-cracks.
Therefore, the in-situ effect will be covered by the choice of the softening curve.
In non-linear analysis average values are regarded in order to perform stress analysis that
corresponds to an average structural behaviour which is – as a first step - assumed to be found
in the structure when performing a comparison between test and prediction. Therefore, when
executing a nonlinear stress analysis the stress-strain curve and its secant moduli to be utilized
in the applied non-linear program are average values.
In the strength analysis, within the Design Verification, so-called ‘A’ or ‘B’ design allowables
R  R have to be regarded, however. These have no bar over which practically means that the
designer applies a shrunk strength failure space.

2.4. Description of non-linearity

2.4.1. General
In non-linear analysis four features have to be considered in WWFE-II:
- non-linear stress-strain behaviour of the smeared lamina material (hardening, below IFF),
- the degradation-related softening above the IFF level,
- non-linear behaviour including large strains and large deformations if applicable, and
- non-linearity from weakening of the pressure level-dependent matrix behaviour.
Non-linear behaviour of well-designed composites can be only physically (inelastic lamina
behaviour) caused but rarely geometrically (laminate stack behaviour). However, the stress
strain behaviour of special composite laminates may be highly non-linear, particularly in shear
and under transverse compression in dependence of the stack. Large deformations are only to be
tackled in design if specific functional design requirements ask for a non-stiff lay-up or if a

WWFE-II-Part A_dispatched Dec09 23 of 77 29/12/2009, 8:07

structure fibre-dominated designed to a distinct load case is loaded matrix-dominated by an
unforeseen other load case. An example is an internal pressure-designed water hose under
longitudinal tension, see WWFE-I, where the fibres turn axially.
A full 3D-input in stress analysis demands for 5 elastic and 5 strength properties in case of
fibre reinforced plastics (FRP). In the 2D-case the required input is 4 elastic properties and 5
strength properties. For the non-linear stress analysis the relevant non-linear stress-strain
curves are to be provided in addition . These are the three IFF curves where mapping (fitting)
functions have to be elaborated. Such a mapping function is a mathematical description of the
full stress-strain curve which is very helpful in numerical analysis if the material behaviour
can be approximated well. It’s first branch, the hardening branch ends when the stress reaches
its strength value R mod e and thereby an initial IFF level. From that level on, that means for the
post(-initial) IFF regime or progressive damage regime, the still employed term softening is
used. Of course, damaging still begins with the material’s strain hardening analogously to
yielding of a metal below tensile strength.
In the non-linear analyses the secant modulus approach will be used.
How the multiple nonlinearities are treated is shortly presented in Annex III. The through-
thickness properties which are a little lower than the in-plane properties are not considered in
the analysis because of the consequently applied transversely-isotropic model.
The little real yield surface [Cun03, Ern08], determined by the matrix, is early changing into a
quasi-yield (damage) surface which is to be dedicated to a diffuse micro-cracking. If micro-
mechanical cracking localizes then discrete micro-cracking occurs and IFF starts.

2.4.2. Mapping of hardening and softening (degradation curve)

For the envisaged strain-hardening FRP materials the Ramberg-Osgood equation
   / E0mod e  0.002 ( / R pmod e n
0.2 ) (28a)

can be used as mapping (fitting) approach, because the well-known MIL-Handbook (now
MMPDS) applies it as an engineering mapping tool, recognized in industry for this type of
material behaviour. The Ramberg/Osgood exponent

n  n pl ( Rm ) / n Rmmod e / R pmod
0.2
e
 (29)

is estimated from strength point R mod e

m 
,  pl ( Rmmod e ) and yield point information. For
instance, a mapping of the UD shear curve by Ramberg-Osgood looks like, see [Cun04],
 21   21 / G||  0.002(  21 / R p0.2 || ) n . (28b)

The degree of the hardening non-linearity mainly affects E c and G|| . It depends on the
non-linearly behaving matrix material, the ‘2ndTg-shift effect’, and the final strain-related
‘Birch effect’ as well.
For the laminate Test Cases, beyond initial failure an appropriate progressive failure analysis
method has to be employed (a successive degradation model for the description of post initial
failure) by using a failure condition that indicates failure mode and a measure for the
material’s stress effort and damage. Final failure occurs after the laminate structure has
degraded to a level where it is no longer capable of carrying additional load. E c and G|| are
reducing gradually. For deriving unique data for the secant moduli two regimes (hardening

WWFE-II-Part A_dispatched Dec09 24 of 77 29/12/2009, 8:07

and softening) have to be distinguished, one below and one beyond  ( Rm ) . In order to
provide the non-linear analysis with the needed input, normalized stress-strain curves have
been constructed with a hardening part measured and a softening part which has to be
assumed. This is due to the fact that the designer is generally lacking of experimental
information for the degradation of the embedded lamina. An engineering modelling of this
softening part or the post initial failure behaviour of a laminate requires that assumptions have
to be made regarding the decaying properties of the actually degrading lamina or laminae. The
approach is based on the idea that the softening function is factorizing the Ramberg-Osgood
hardening function. A simple function was used to map this softening in order to later derive
the secant modulus for the non-linear analysis. It reads ( exp[ z ]  e z )
 soft  Rm /( 1exp[( a soft   ) / bsoft ])  Rm  (30)

with  termed degradation function. See also [Cun04a, Puc02, Kno03, Mat95] and Fig. 4.
Eq.(30) contains two curve parameters a soft , bsoft that are determined by the data of two
calibration points at least or curve fitting if test data is available in the softening domain,
( 0.995  Rm ,  ( Rm )) and e.g. ( 0.1  Rm ,  ( 0.1  Rm )) (31)

(applying point 1.0  Rm ,  ( Rm ) is numerically not permitted!).

A softening function models the degradation as well as the stress-strain curve of an embedded
lamina which is deformation-controlled by the laminate. Thereby, the altering micro-crack
density has to be regarded up to the characteristic damage state (CDS) where the number of
micro-cracks reaches its maximum value and beyond where the initiated micro-cracks just
increase their size. It may therefore termed an ‘effective’ curve. Without any information this
function has to be assumed. The right choice of the softening curve will reduce possible
numerical convergence problems.
If necessary for the non-linear analysis procedure, data for the secant moduli of E c and G||
may be derived from the hardening equation, Eq.(30), as
  E0
hard
E sec    . (32)
   n E  n 1
 0.002  ( ) 1  0.002  0  ( )
E0 R p 0.2 R p 0.2 R p0.2
Hence the equation for the needed secant modulus
E0 1
E sec  with  (33)
E    n 1 a soft  
1  0.002  0  ( ) 1  exp( )
R p0.2 R p 0.2 bsoft

describes the full domain by applying the degradation function  . It describes the softening-
associated degradation of the UD material. According to the definition of the secant modulus,
Eq.(32),  (equals 1 in the hardening domain) may be also directly put on E sec in the analysis
instead on  according to their linear relationship.

2.5. Post-initial failure modelling (softening branch curve)

Degradation beyond IFF, in the post-initial failure regime, is termed here softening.
Softening is treated differently.
The FMC uses an assumed softening function (evolution equation) for all three IFF curves.
This needs experience from earlier ‘top-down-bottom-up’ simulation (lamina at lower level 
two scale simulation model  laminate at higher level) of laminate measurements which prove

WWFE-II-Part A_dispatched Dec09 25 of 77 29/12/2009, 8:07

that the prediction of the laminate behaviour with the chosen softening curve is successful.
The term above is used as a technical term in parameter identification when quantities at two
scales or levels are connected and the connecting model is validated at the higher level. This is
similar to well-known micromechanics). With this engineering approach (mathematically),
besides the IFF strength point, just one reasonable point of each curve has to be ‘found’ to
obtain a 2nd equation for the determination of the two softening parameters a soft , bsoft .
As in the past, the ‘in-situ behaviour’ has attracted not too much attention in continuum
mechanics wrt an engineering treatment further work is highly recommended in order to
achieve more reliable Margins of Safety in favour of a reliable lightweight design. Meanwhile,
in the dissertation [Kno03], measured softening curves are presented for IFF1 and IFF3 which
show for IFF1 a faster decay than the function assumed by the author. The assumed functions
may be now corrected, compare e.g. [Mat95]. The carefully produced specimens in [Kno03]
were from EP/CFRP and EP/GFRP. The tests performed included micro-crack density
measurements as damage indicators. The test results were Puck model-evaluated.
On the other hand, for the quasi-brittle FRP it can be concluded that so-called cohesive
forces still exist at IFF in the fracture process zone at the micro-crack tips in the laminae.
Therefore, some researchers utilize a fracture mechanics approach applying fracture mode
assigned energy release rates Gi , cr (here index cr used instead of the usual c) as material
parameters, Ref [Alf07, Mat07]. Energy release rates (  surface energy of the newly generated
surfaces, here micro- crack surfaces) are utilized in continuum damage mechanics (CDM) to
assess the softening branch curve, such as GI , cr  GIt , cr in case of Mode I (  2t , IFF1). Of the
other standard energy release rates for the brittle UD material in case of mode II (  31 or  21 ,
IFF3 analogous) the G II ,cr - in contrast to isotropic brittle or semi-brittle material – is a real
material property like G I ,cr . Also here in the quasi-isotropic domain fully similar to the
isotropic material, Mode III tests (  13 ) do not deliver a material property but finally show
fracture plane turning and Mode I behaviour. However, the compressive mode testing in the
quasi-isotropic plane will lead -as well as in the brittle isotropic case- to a material property
GIIc , cr By the way, the application of a fracture mechanics mode approach requires a ‘far-field’
stress (brutto stress) input which can not be defined always. Using the fracture mechanics
Modes one has to superimpose the Gi , cr in a clearly defined manner.
Note 1: Keep in mind when applying energy release rates. With homogeneous isotropic brittle
materials there are 2 real energy release rates G I ,cr , G IIc ,cr , one in tension and one in
compression (see Carpinteri’s work). These two Gs possess the attribute that the crack plane
does not turn and are therefore real material properties. This is fully similar to the strengths.
G II ,cr , G III ,cr are ‘only’ helpful model parameters in fracture mechanics analyses.
Principally, in case of a homogenized (smeared) transversely-isotropic brittle UD material
there are 5 energy release rates. Two of them belong to FF and three to IFF. As there is
‘homogeneity’ in the 2-3 plane, only, the 2 FF-associated Gs cannot be measured because they
exhibit no fracture plane. All the 2 or 5 Gs above are material properties. As delamination
(structural quantity) has to be treated by the designer the associated G, as an energy release
rate of an interface between two ‘materials, phases’ (e.g. a 0° and a 90° ply) has also to be
involved in laminate investigations.
Note 2: It has to be checked whether from macro-mechanical specimen derived Gs can be
applied at micro-/meso-mechanical level. From classical fracture mechanics is known that
short cracks (below about 1 mm) behave different to a so-called technical crack.

WWFE-II-Part A_dispatched Dec09 26 of 77 29/12/2009, 8:07

 In CDM, amongst other approaches hardening and softening branches are estimated by a bi-
linear approach [Pin05] using an energy release rate, defined as area under the two branches,
for the determination of the softening branch, IFF1. This bi-linear approach is sufficient for
impact or high strain-rate analyses but is not generally sufficient for a non-linear static
analysis wrt IFF2 and IFF3.
Another approach employs a measurement of stiffness losses and crack counting methods for
the density measurements.
Also a ‘top-down-bottom-up’ approach via Representative Volume Elements (RVE) of the
constituents is pursued, see Refs. [Bus06, Ern07].
Conclusion for an arbitrary laminate: Each approach (Puck, Cuntze, Pinho etc.) requires its
associated validated test results for considering the in-situ effect to be covered by the
softening curve of the embedded lamina. There is still a lack of knowledge of this
phenomenon. But, there is some hope that the degradation behaviour is not so different within
a lamina family such as CRFP and that the laminate configuration is not so decisive. Further
experiments for data-based identification of the softening parameters of the embedded lamina
by simulation of the laminate behaviour are mandatory. Practical and reliable CDM solutions
are not yet available.
A fully accurate procedure of the non-linear behaviour would require a treatment of the
growing damage or quasi-yield IFF surface which includes a common flow rule (here,
direction of degradation growth) for this multi-fold ‘yield’ condition case (compare the
problem at a Tresca surface corner where two conditions meet).

2.6. Residual stresses and curing stresses

2.6.1 General on residual stresses
It can be distinguished between several stresses: Load stresses, hygro-thermal stresses, and
residual stresses. Residual stresses are originated by chemical shrinkage in the solid state after
the gel state and thermo-mechanical shrinkage (mismatch of the coefficients of thermal
expansion) and further, residual stresses that stem from actively or passively built-in stresses
such as pre-stressing.
Residual stresses (suffix R) are taken into account by superimposing them onto the external
load stresses (suffix L)
    L    R  . (34)
Such residual stresses in the laminae of the laminate decay with decreasing stiffness, caused
by the matrix degradation due to micro-cracking, which accompanies increasing non-linearity.
In other words, in parallel to the decay of the stiffness the non-linear analysis releases the
‘averaged, smeared’ matrix-dominated stresses. This applies for mechanical stresses as well as
for thermal load stresses. The decay of a residual stress starts marginally with matrix yielding
[Cun03] and quasi-yielding (diffuse micro-cracking) in the two matrix-dominated modes
before reaching IFF2 or IFF3. Then premature IFF-related degradation follows. From
chemistry is clear: Cross-linked thermosets merely release residual stresses under remaining
virgin conditions. Practically, the influence of the residual stress is extinguished on the steep
slope of the softening curve.
Note: Matrix-dominated deformation of a structure can be highly affected by residual
stresses (own experience when analysing the wound Mage Satellites composite motor
bottoms, [Sch88]), if the structural behaviour is matrix-dominated. In the case of fibre-
WWFE-II-Part A_dispatched Dec09 27 of 77 29/12/2009, 8:07
 dominated structures there is no substantial effect on the deformation. Pre-requisite is a
qualified curing process in order to achieve not too high residual stresses from curing that
the structure might break after curing, probably during the first handling.
For all the test cases of the WWFE-II, placed in the compressive domain, the residual stress
effectiveness is smaller than for those in the tensile domain (WWFE-I). However, it is always
to be checked whether the residual stresses might be significant.

2.6.2 Curing stresses in a matrix

At first some notions shall be given: The gel point is the point at which an infinite polymer
network first appears. It is the stage at which a liquid begins to exhibit pseudoelastic
properties and increased viscosity. This stage may be observed from the inflection point on a
viscosity-time plot [Wikipedia]. The glass transition temperature Tg , or glass transition point,
respectively, is the temperature at which an amorphous solid, such as glass or a polymer,
becomes brittle on cooling, or soft on heating. The later in the analysis used stress-free
temperature is a temperature at some distance below the curing temperature which itself is
below the ‘usual’ Tg .
Once the part starts to solidify residual stresses may build up. They are the consequence of a
constrained straining within the structural part, e.g an inclusion in the polymer. Characteristic
of the curing stresses is that they form an equilibrium state, or, they are zero over the wall’s
thickness. In a homogeneous thick epoxy polymer matrix specimen the curing stress is very
low, around 1 MPa is cited in [Sri83].

2.6.3 Curing stresses between filament and matrix and in the laminae
Chemical cross-linking leads to a volume shrinkage during the liquid-solid transition or in
the so-called glass transition domain, respectively. The cure-induced residual stresses are
generated after the polymer reaches the gel point during curing. Contractions in the solid
phase contribute to residual (built-in) stresses, see [Sch78]. Thereby, this chemical shrinking
causes physical shrinking and residual (normal) stresses when a lamina or – the more – a
laminate is cooled down from its stress-free temperature to room temperature. Mind: Curing
does not generate shear stresses.
In general, residual stresses may be reduced by means of the viscoelastic response or creep of
the matrix, respectively. Relieving (relaxation) of residual stresses through creeping is a time
consuming process. To reduce stresses more rapidly the velocity of the creep has to be
increased which can be performed by increasing the temperature and choosing an isothermal
annealing temperature somewhat below the curing temperature.
Including fibres into the polymer matrix produces filament stresses due to curing shrinkage
and according to different CTEs of the constituents, the solidifying matrix and the filament.
Between the filament and the filament-embedding matrix the matrix shrinks and produces
residual compressive stresses normal to the filament. Shrinkage of resin pockets causes tensile
stresses [Shi92a,b].
There are curing stresses of the 1st kind (lamina material level, as part of laminate;
macromechanical curing stresses) and of the 2nd kind (fibre-matrix level; micromechanical
curing stresses]). The latter are assumed to be included in the strength values. Hence forth,
residual stresses of the 1st kind will be only considered.
Curing stresses have an effect on the IFF of unflawed materials and also, naturally, on the
onset (IFF-related) and growth of still existing delamination cracks.
Mechanisms, besides the relaxation of the matrix, are stress reliefs as a consequence of micro-
cracks, and of fibre resin bond degradation. Composites consisting of matrices with lower

WWFE-II-Part A_dispatched Dec09 28 of 77 29/12/2009, 8:07

mechanical properties will allow a faster stress relief (and vice versa) by thermal or
mechanical cycling procedures.

2.6.4 Conclusions for a multi-ply laminate

The final curing stresses after curing and post-curing are the residual stresses the designer
has to tackle. As temperature drop the difference stress-free temperature minus room
temperature as effective temperature difference (Table 3) is applied in order to consider the
effect of the residual stresses of the 1st kind. Moisture may be assumed here to have a
balancing effect of 30°C.
Residual stresses may matrix-dependent relax over the time to some smaller extent and this
happens to be stack-dependent. A reference on the ‘magnitude’ of the relaxation was not
found, even after discussion with several experts. It is only agreed: Thermo-setting polymers,
in contrast to thermo-plastic polymers, exhibit cross-linking and are therefore merely affected
by a residual stress lowering annealing procedure of the cured laminate.
For the elaboration of the Part B test data the chapters 2.6 and 2.7 may be of importance.

2.7. Effects in thick composites

In order to carry higher loads, structures such as wind rotor blades or high pressure vessels
became thicker and thicker. The curing of such thick parts without obtaining too high curing
stresses became possible due to refined curing procedures. In general however remains, curing
of a thick wall normally ends up with higher curing stresses than in the case of thin walls.
When curing thick composite parts the solidification will not take place at the same time over
the wall, because – according to the natural transient thermo-dynamic curing process of the
visco-elastic polymer near the glass transition point - the temperature at the outer surface will
be different to that in the interior. Further, by this fact also the stiffness build-up varies over
the thickness with the consequence of obtaining stress distributions. But performing a long-
time curing (would be analogous to moisture conditioning) is not acceptable due to production
costs. The difference of the curing stresses will increase the more the thicker the wall is and
the faster the curing cycle is running. In [The?] was shown that the difference between surface
curing stress (exterior, tensile) and interior curing stress (interior, compressive) was becoming
of interest when the thickness was higher than 5 mm. In this not yet engineering application
the difference was just about 4 MPa.
In addition, the parts shape may change due to shrinkage and warpage. The size of the
unfavourable shape change depends on the manufacturing and curing process parameters and
on the thermo-mechanical behaviour of the part.
The manufacture of thick walls often leads to imperfections such as waviness, non-uniform
layer thickness, resin pockets, resin-lacking domains, broken filaments, and non-uniform
filament (fibre) distribution. These so-called ‘manufacturing signatures’ are impacting the
stress state locally and globally. For instance, a wound thick cylinder wall may have no pre-
tensioning any more in the inner layer after winding.
The effect of the curing-caused residual stress in high pressure applications is often of minor
importance wrt the other manufacturing signatures.

2.8. Variation of moduli E  ,G||  and Poisson's ratio  ||

The elaborations in [Kno03] indicate for CFRP that the IFF1 microcrack-related loss of E
is significant but that the IFF3 degradation causes just a marginal loss of G||  . Further, the
influence of damage on the larger Poisson’s ratio was also marginal. For the differently
critical wedge failure IFF2 data was not delivered.
WWFE-II-Part A_dispatched Dec09 29 of 77 29/12/2009, 8:07
 The alteration of the major Poisson’s ratio (here termed  || ) is linked to successive micro-
cracking which the generation of micro-spaces. Hydrostatic compression impedes the
generation of micro-spaces. Possessing no input for Poisson’s ratio and with respect to the
quality of the other affecting properties, the value is kept constant in the analyses.
With increasing micro-damage state the UD composite may become orthotropic and the
Poisson’s ratios should be treated differently,  31   21 .

3. Description of data input and Test Cases in the WWFE-II

For the 12 Test Cases (TC) material properties of the UD lamina and its constituents were
provided. Two important and widely used classes of fibres (carbon and glass) and one group of
resin systems (epoxy resins) were provided for the UD composites. Five types of fibres were
selected for the predictions, two types of glass fibres and three types of carbon fibres. They were
chosen for consistency with data which were available for particular laminates. In the same sense
five matrices are provided. This leads to the following laminae:
1.) E-Glass/MY750, 2.) S2-Glass/epoxy, 3.) AS carbon/epoxy,
4.) IM7/8551-7 carbon/epoxy, and 5.) T300/PR319 carbon/epoxy.
For analyzing the mechanical behaviour of multidirectional laminates under various states of
stresses the input data pack of the WWFE-II includes: 3D elastic constants, thermal properties
for the establishment of curing stresses, as well as failure strains and strengths, and nonlinear
stress-strain curves for the five UD laminae.
The Tables (1) to (3) show average (typical) data for the properties of the five UD laminae, of
the five epoxy resin matrices and the five types of glass or carbon fibres. Some energy release
rates GC are added for the CDM approaches. Properties for the constituents are not necessary for
the application of the FMC but necessary for the application of micro-mechanics in order to
consider the influence of the pressure-dependent stiffness of the matrix. A number in round
brackets, see e.g. Table (1), indicates that the provided table has been reworked a little.
In Ref [Kad07a], the figures 3 to 12 show typical stress strain curves of the selected laminae
under a variety of uni-axial loadings. In order to obtain a comparative view the corresponding
stress-strain curves are united in one graph each, see Figs. 5 through 9. All the data provided
for the moduli is an initial value of the virgin material.
In the WWFE-II the UD lamina material is still treated as a transversely-isotropic continuum
(confirmed by the provided input properties). This means, some relationships exist:
E 2  E3 , G12  G13 ,  12   13 , and G23  E 2 /( 2  2  23 ) . (35)
A UD lamina made of continuous fibres in a matrix was considered to be the basic building
block for the analysis of the multidirectional laminates. Five types of laminates were chosen for
the analysis and are described in Table (4) together with lay-up, layer thicknesses, and state of
stress which to be investigated. Table (4) summarises laminate type, material type and the
graphical results requested. All together, five basic example types are chosen:
(1) Pure resin matrix (TC1),
(2) 0 UD lamina (TC2-TC7),
(3) Angle ply [35/-35/35/-35]s laminate (TC8, TC9),
(4) Quasi-isotropic [45/-45/90/0]s laminate (TC10),
(5) Cross-ply [0/90/0/90]s laminate (TC11, TC12).

WWFE-II-Part A_dispatched Dec09 30 of 77 29/12/2009, 8:07

4. Comments on the input for the analysis
4.1. Comments on the provided properties and test cases
From the various matrices, Table (3), just PR913 epoxy has the necessary tensile fracture
strain to smooth out the stress concentrations around the embedded filaments. Due to a research
programme at MAN Technologie 6% should be the engineering goal. The consequence of a too
low fracture strain of the constituent matrix is an increase of the UD brittleness at ambient
conditions. Reasons for the low test values of some matrices might be a non-polished surface of
the matrix test specimen. In this latter case the provided fracture strains would not give a correct
basis for a judgement. Unfortunately for PR319, where the matrix information is needed for
tackling the TC 2 through TC 4 tasks, accurate basic values are missing. The provided
Young’s moduli do not well match with the given stress-strain curves. Some curves are
shorter than the table value indicates.
Information on the effect, which the author termed the 2ndTg shift, was not provided.
The application of the data of the constituents in Table (2) requires the micromechanical
formulas they have been determined with! Errors are only avoided if the same set of formulas
is applied according to the ‘top-down-bottom –up’ approach from the constituents up to the
macro-mechanical UD material level. As micro-mechanical formulas are not provided the
previously given formulas from the VDI guideline 2014 are taken.
A fully reliable UD data set could be not provided. Especially an information to assess a value
for the physically required friction parameters bc , b|| is missing and therefore were assumed
according to own tests.
In order to facilitate comparison all contributors use the same provided input in the Part A
analysis as a common basis. This means, Part A will partly be a real blind prediction, and
further, that the validation phase (Part B of the WWFE-II) will have to be based on more
reliable data and qualified associated test information to really make a validation possible.
There were some problems how to interpret the laminate Test Cases. Especially there were
some questions where the stress states of the TC 10 and 11 might exist in a real test rig or a
cut-out of a laminated structural wall. For these two Test Cases the author assumed an
Iosipescu (ARCAN) test rig and corresponding test specimen.

4.2. Assumptions for the predictions

All ‘full failure theory models’ require the knowledge of quite a large number of parameters.
This includes in the case of the FMC-based UD failure theory (for other theories similar)
parameters for the three parts of a failure theory (see §1.2) the requirement of:
a) Macro-scopic failure conditions of the transversely-isotropic UD material: 5 strengths
data from ‘isolated’ tests + 2 friction parameters +1 mode interaction exponent m +
fibre and matrix data (if the pressure dependent stiffness and strength properties are
not given as lamina information but as matrix information.
b) Stress-strain analysis: 5 elastic constants, 2 coefficients of thermal expansions, 2
coefficients of moisture expansion, 1 stress free temperature.
c) Non-linear analysis for each mode: 1 strain to failure, 2 hardening parameters
(Ramberg-Osgood exponent + ‘yield’ strength R0.2 ) and 1 point for the determination
of the softening parameter (the 2nd one is fixed by the strength point R ), and the
slope constant cPae (TC 3). The 2ndTg curve fitting parameters (M, a, b c).
Mind: Not all material model parameters physically, required due to material
symmetry reasons, could be supplied by the WWFE-II organizers.
WWFE-II-Part A_dispatched Dec09 31 of 77 29/12/2009, 8:07
In the context above the assumptions made are listed in the following:
* Hydrostatic pressure acts on all cross sections or front faces of the test specimen, an applied
external mechan. stress acts on top of one or two cross sections (due to assumed test rig).
* The first FF1 or FF2 is final failure. Also IFF2 (wedge failure) may cause final failure
through delamination.
* The void content is negligible (< 1%). Otherwise, the IFF2 approach would have to be
modified by an invariant representing a volume change to consider the porosity (see
[Cun04], p. 497).
* The parameters m  2.8; bc  1.21 (  cfp   50 ) , b||  0.3 were estimated and are basis set
for all laminae.
* Good fibre placement and alignment, and uniform distribution. Manufacturing signatures
such as fabrication-induced fibre waviness and wrinkles are small and do not vary in the
specimen.
* If applicable, residual stresses from the curing cycle are to be computed for the difference
‘stress free temperature to room temperature 22°C’ as an effective temperature difference.
* The stress-strain curves are average curves, which is the type one needs for test data
mapping. Considering curing or moisture stresses (here, the specimens are assumed to be
well conditioned) the graphs do not begin in the origin.
* Edge effects are assumed to not exist for the prediction models because a real validation of a
strength condition requires a smooth stress state in the failure area.
* The progressive behaviour of E||t (up to 10%) in the case of C-fibre rovings is not regarded.
* The course of each softening curve is assumed. Post-initial IFF is considered by gradually
degrading properties of the embedded lamina.
* Unless otherwise stated, the loads are monotonically increased by keeping any combination of
the following ratios x /y /z /xy /xz /yz constant. See Table (4).
* Up to initial failure the application of linear analysis is often sufficient in practice.
In the case of well-designed laminates, very often, linear analysis is pretty sufficient even
up to final failure according to the factor of safety to be put on the loading.
In the WWFE the investigation is performed with typical properties and zero factor of
safety as it is the case in data mapping..
* The Ramberg/Osgood exponent n was computed from the provided curve data (sometimes
they had to be estimated). The softening parameters a soft ,bsoft are determined from the
assumed softening curve with one point of it at a strain about two times the ‘isolated
lamina failure strain’.
* The 2ndTg shift effect - as a matrix attribute - is mainly affecting the matrix-dominated
behaviour of a lamina which means compressive and shear behaviour. In these cases the
associated reduction factor f 2 ndTg , derived from the matrix behaviour, is applied to the UD
material, too, due to missing knowledge but it is scaled for the specific Test Case.
* The hydrostatic pressure-caused stiffness increase for the matrix material is approximately
considered by applying the ‘Birch stiffness increase’ equations Eq.(21) together with the
corresponding micromechanical formula. Of course, this increase determined by the

WWFE-II-Part A_dispatched Dec09 32 of 77 29/12/2009, 8:07

 constituent matrix is gradually different for the UD material. This is considered by the
employed micro-mechanical formulas.

4.3. Special requirements to be followed by the contributors

1) WWFE-II requests in some cases, like WWFE-I, failure envelopes which shall indicate
any intermediate failure points caused by the various stages of failure which may take
place before final failure
2) The scales of the plots the organisers have supplied to all of the participants are to be
followed, if possible, in order to facilitate comparison between the various predictions
3) The magnitude of the stresses and if appropriate the type and location of failure as well as
the mode of failure at which each failure is predicted are to be tabulated in a separate
annex to the paper
4) Stress domains where the envelope is open are to be indicated.

5. Theoretical predictions
5.0. General
5.0.1. Possible failures, mainly linked to pressure
Matrix specimen: During deformation of most of the polymers (thermosets and
thermoplastics), an interaction of the failures crazing, local yielding, and micro-cracking
may be observed. Each single polymer may behave in a different manner, [Hop95].
Therefore, p hyd will differently affect the strain to failure.

UD specimen: During deformation of the UD material, longitudinally splitting and brooming

at the compressive load introduction end (friction under pressure head) may initiate the
failures. Within the compressed specimen, kink-band formation as well as fibre bundle
buckling may be observed. These failure phenomena are tried to be effectively described by
the given IFF and FF strength failure conditions.
UD laminae-composed laminate specimen: During deformation, above failure phenomena
may occur. Initial (IFF) and final failure may take place. Further, initial delamination of the
laminate or buckling of the specimen may occur. The specimen has to be designed to non-
buckling in order to avoid this structural failure in order to finally obtain UD material
failure, only. The laminate failure events do not generally end with an IFF as it is the case
for the isolated UD specimen.
A multi-directional wall may experience the full 3D state of stress. Therefore in principle, a
“ply-by-ply” analysis has to be performed for instance in case of a thick-walled cylinder under
pressure. Hyer reports in [Hye88] that the smearing process from 100 layers down to a number
of 8 layers may be permitted. The failure behaviour for the thick walls is similar if severe
manufacturing signatures can be excluded. Therefore, this finding is essential for the laminate
Test Cases TC 8 through TC 12, because the stack may be reduced for the analysis to 8 layers.
Due to the symmetrical lay-up of the envisaged “thin” 8-lamina stack just four laminae have
to be considered

5.0.2. Numerical analysis and calculation procedure

For eleven Test Cases the analyses were performed by the application of MathCad. Due to
the higher effort TC 10 was not yet analysed in Part A (foreseen for Part B) by the author.

WWFE-II-Part A_dispatched Dec09 33 of 77 29/12/2009, 8:07

Originally, for the laminate test cases a 3D finite element analysis with ANSYS was planned
as a ply-by-analysis in order to accurately calculate the 6 stresses in each lamina of the
laminate stack. Another reason was its better FE code-inherent non-linear solution
capabilities. As finite element was foreseen SOLID 185. However, applying ANSYS UserMat
the required failure conditions and the 2ndTg-shift affected non-linear stress-strain
relationships, including hardening and softening, would have to be implemented in order to
model and analyze the test specimen with its boundary conditions. Because the assessment of
the stress states needs stresses in the lamina coordinate system the so-called element stresses
from the ANSYS analysis have to be transformed into lamina stresses. Its implementation
costs too much effort even for specialists. Therefore, the author decided to solely apply
MathCad, because it has the advantage of the performance of fast parameter variations.
Again as in WWFE-I (now with some minor improvements) as non-linear tool a simple self-
correcting secant modulus approach is utilized. The solution procedure of the nonlinear
analysis is to establish static equilibrium at each load step after material properties have been
changed. For each iteration the procedure is repeated until convergence (equilibrium). By
employing the equivalent stress reached in each failure mode the associated secant modulus of
each IFF mode E sec (  eqmod e ) is determined for the hardening and the softening regime. For
additional details see Annex 2
Mind: In the sketches it is always discriminated between stress and loading. The stress
follows its positive definition and the loading its actual direction. It is marked by a
negative sign if the arrow shows in negative direction in the sketch.

5.1 Test case 1, Matrix MY 750,    (  x ,  y   z ,  z , 0 , 0 , 0 )T  (  I , II , III )T

Problem: How does tri-axial compression influence the failure behaviour if two components
of the stress vector are equal?
Task: Determination of the tri-axial fracture failure curve  x (  y   z ) , Fig.TC 1.
Assumptions and analyses:
The ‘isotropic’ FMC-based failure conditions for the occurring normal fracture (NF,
deformation poor, phenomenological) and for shear fracture (SF) read ([Cun08, Boer89])
I1  4J 2  I1 / 3
2
6 J 2  I1
NF: 0.5 1 and SF: ac  2
 bc  1 (36a, b)
Rt Rc Rc

with Rt , Rc as tensile and compressive strength, and the invariants described below. Above
equations exhibit a rotational-symmetric failure surface.
Mind: The ‘material mathematicians’ who derived the invariants have unfortunately
dedicated the same letter and cipher, I 1 , to the primary invariant in the isotropic case as
in the transversely-isotropic UD case as well. Of course, the two material symmetry-
related invariants I 1 are different.
Eq.(36a) represents the normal stress hypothesis (NF) and it displays in the principal tensile
stress quadrant the well known straight lines. The parameter  in Eq.(36b) represents the so-
called 120°-symmetry of brittle isotropic materials. Due to the fact that MY750 is not so
brittle  can be set one. The development of the isotropic failure conditions above is also
FMC-based. Following Beltrami [Bel1885, Cun04a, Cun08] a volume change is linked to the
square of the first invariant I 1 (Eq.(30) of the stress tensor whereas Huber-Mises-Hencky’s

WWFE-II-Part A_dispatched Dec09 34 of 77 29/12/2009, 8:07

(HMH) ‘shear yielding’-describing invariant J 2 , (  eqMises  3  J 2 ), is linked to the deviator
with  I ,  II ,  III as principal (normal) stresses:
I 1   I   II   III , (37)
J 2  [(  1   II ) 2  (  II   III ) 2  (  III   I ) 2 ] / 6 = 4  (  III   II   I )  f (  ) .
2 2 2

The J 2 formulation is given above in principal shear stresses, too.

If there would be a change in the volumetric strain (‘no constant volume process’) this is to be
2
captured by the failure condition. Then, I 1 must enter the failure condition of the matrix.
This term represents a volume change due to Beltrami, see [Cun08]). Because it may be
assumed for the actual matrix that practically no voids or resin pockets are in the matrix
specimen (means ‘dense’ consistency or non-porous material) then an additional part I 12 is not
to include into the Eqs.(36 a and b) to describe this feature. In a porous material situation the
SF mode is replaced by a so-called crushing mode. Eventually, after Mohr-Coulomb, the
invariant I 1 helps to consider the material’s internal friction. All these invariants are related to
an energy.
The two curve parameters in Eq.(36) are still not fixed. Inserting the compressive strength
 I   Rc into Eq (36b) delivers
 Rc
2
2 Rc
ac  2
 bc  1 , (38a)
Rc Rc
being a relationship between the two curve parameters
2 ac  ( 1  bc ) . (38b)

It remains the estimation of the still unknown ‘friction parameter’ bc . Its value is related to
Mohr’s fracture angle which is to be measured in tests (see example UD in Annex).
Considering the information above, the value for the fracture angle of this semi-brittle MY750
should not be much higher than the zero friction angle value, which is 45° for ductile
behaviour. From this follows, a fracture angle of  cfp = 47° (the suffix fp means fracture
plane) can be assumed. When determining the friction parameter from the fracture angle, then
Eq.(36b) has to be transformed into a Mohr-Coulomb formulation, analogously as shown in
the Annex for a UD material. This procedure results in the formula
bc  ( 3  C cfp  1 ) /( 3  C cfp  1 ) with C cfp  cos( 2   cfp    / 180 ) . (39)

Hence, for  cfp   47 a parameter value bc  1.53 is computed and applied for all test cases.
(Of course, bc can be also determined from the course of the failure curve in the bi-axial
compression domain. However, information on this was not provided).
Thus, the effort equations read:
I1  4J2  I1 / 3
2

Eff   0.5 , (40a)

Rt

bc  I 1  ( bc  I 1 ) 2  24  ac  J 2

Eff   , ac  0.5  (1  bc ) (40b)
2  Rc

with the invariants I 1 , J 2 from Eq.(37) which – as mentioned - should not be mixed up with
WWFE-II-Part A_dispatched Dec09 35 of 77 29/12/2009, 8:07
 σ τ
the UD invariants in Eq.(3). The superscripts and mark the fracture governing stress in the
physical fracture plane (Mohr).
It has to be checked now “How the matrix material is behaving, brittle or ductile or
mixed?”. Due to material symmetry, isotropic brittle and semi-brittle behaving materials
possess just two strengths. The additionally provided ‘shear strength’ (sometimes termed as
cohesive strength R ) is therefore a fully dependent value and its coordinates should fully lie
on the NF failure curve if the material behaves fully brittle. However, this is not true as Figs.
TC 1 indicate. R is neither located on the NF nor on the SF curve. So it may be assumed
from the provided strength values: This matrix material is semi-brittle. This is proven by the
relatively small difference between tensile strength and compressive strength. And it can be
further substantiated by inserting a stress max into the Eqs.(40). With the invariants I 1  0
and J 2  max  2 Eq.(40b) predicts max  SF  44 MPa for bc  1.53 . Inserting the invariants
above into the NF condition, Eq.(40a), yields max NF  R t  80 MPa . Therefore, the
provided R  54 MPa , is a fracture value of an interaction zone of two modes NF and SF.
From that it can be also concluded that the MY750 matrix is semi-brittle. This also proves
  1 in Eq.(36b).
To be noted: The stiffness increase upon p hyd was applied according to Birch’s formula. An
interaction of the two failure mode curves F  (NF) and F  (SF) made problems due to the
above proven fact that the MY750 matrix is semi-brittle. Then the application of the simple
interaction equation Eq.(16) can not handle this, see Figs. TC 1a,b. Therefore, as engineering
approach, a simple linear interaction line in the 3D domain (Fig. TC 1a) delivered the
requested interaction curve. Also to be assumed is a confining cap. The Figs. TC 1a,b outline
the final failure curves in the 3D domain. Fig. TC 1b depicts the situation in the 2D domain.
Mind: Matrix materials applied in FRP composites have to be relatively ductile (should
have about 6 % failure strain) in order to smooth out the stress concentrations around the
filaments. In case of very ductile materials (practically no material friction) the engineering
yield strength values for the properties R po .2 and Rco .2 are almost the same and a
compressive strength Rc is not determinable; just a yield strength Rco .2 is measurable. By
the way, a compressive strength would be also not necessary for design as the yield strength
becomes the design driver. Otherwise, the deformation under loading would exceed usual
design requirements.
Essential results:
* As the matrix material is semi-brittle a fracture failure surface or a failure envelope (curve)
can be determined which confines the consecutively growing yield surface. However, this
growing yield surface is not shown in Fig.TC 1c,d. Basically, in the positive quadrant NF
will occur and in the negative quadrant SF failure (denoted by F  and F  ), Figs. TC 1c,d.
* The matrix is assumed to be dense (no voids) and therefore can be infinitely compressed
when subjected to tri-axial compression (solid line). The consideration of matrix softening
above the ‘knee’ at about 200 MPa has a substantial influence in the high pressure regime
(dashed line). Shear fracture is activated if the difference of the stresses in a section plane
is large enough to cause fracture according to IFF2. This is possible for two stress
combinations associated to the two failure curves.
* For giving a global understanding on a multi-fold failure the NF mode-related multi-axial

WWFE-II-Part A_dispatched Dec09 36 of 77 29/12/2009, 8:07

 strengths Rtt , Rttt were estimated and depicted in Fig. TC 1d.
* Stiffness and strength of the polymer, max  x decreases with p hyd beyond 200 MPa. The
2 nd Tg shift effect causes an earlier fracture.
* Viewing Fig. TC 1 a, it can be simply drawn from the NF curve that R   NF
fr  Rt because

it does not lie on the respective point (cross x) of the abscissa which is computed from the
NF equation via I 1  0 , J 2  R as 2  J 2 / Rc  2  80 / 120  0.94 . From literature as a
thumb rule may be applied: Brittle behaviour is usually dedicated to an aspect ratio
Rc / Rt  3 . This is much higher than 0.94. Again it is proven: The matrix is semi-brittle.

5.2. Test Case 2, UD lamina T300/PR319,    (  1 ,  2 ,  3 0 , 0 ,  21 )T

Problem: Influence of hydrostatic pressure p hyd on the in-plane fracture shear stress.
Task: Determination of the fracture failure curve  21 ( p hyd ) with  1   2   3   p hyd , and
p hyd an absolute pressure value, Fig. TC 2
Assumptions and analyses:
It is usually assumed that the value for the basic strength R|| has sustained the curing stresses
and therefore is covering the influence of these residual stresses. Shear strength will increase
due to the pressure-improved adhesion of the filament-matrix interface with the suppression
of the flaw effects [Hop95]. Effects of the pressure dependence of the matrix material are
transferred to the composite because no information was provided on lamina level. So, with
the increase of p hyd the shear modulus and shear strength as well will increase but differently
high. For the matrix-dominated shear strength a reduction function f 2 ndTG , derived for the
MY750 matrix, serves as a standard reduction function and is also fully applied in TC1.
Following the little literature, Pae, the reduction in strength might be too high.
Essential results:
* The result of TC 2 is a combined multi-axial failure state of stress which may be termed
multi-axial strength. The elasto-mechanical effect of p hyd is displayed in Fig. TC 2 by the
solid curve. It shows that the fracture shear stress changes approximately linearly with p hyd
starting from atmospheric pressure level. The reduction in strength (strength weakening
effect) of the matrix material above the knee is indicated by the dashed curve. Due to the bi-
linear simple engineering fitting approach the numerical effect still begins at zero.
* There is one driving failure mode, shear failure IFF3. In the high hydrostatic compression
domain the failure curve becomes closed by FF2, see Fig. TC 2b.
* Thermal stresses as a result of a non-uniformly curing process of this isotropic material and
moisture stresses have a negligible effect under high tri-axial pressure and are not considered.
* The sensitivity to the unknown friction parameter b|| is studied by the choice of two values
(0.3, 0.4) for the basic curve (for these two curves is set f 2 ndTG  1 ). The result is: the higher
the friction, the higher the fracture shear stress.
Mind: The strength of a UD lamina, be it isolated or embedded, practically is not expected
to exceed the longitudinal compressive strength ( R||c ) of the lamina. This is clear from
Eq.(18b) where the left hand side should not exceed 1. However, the results in Fig TC 2b

WWFE-II-Part A_dispatched Dec09 37 of 77 29/12/2009, 8:07

 (and later TC 5, 6, and 8) seem to show that the UD strength under bi-axial, tri-axial
pressure exceeds the UD longitudinal strength by a large margin. This comes from the
elasto-mechanical effect (see § 2.3.1) “Poisson’s ratio reduces the filament compressive
stress  ||cf ”. Explanation for this is: Not the macro-strength but the strain-dominated
filament strength is fracture responsible. In other words, the effective compressive stress is
decreased, and this effect is automatically considered in the analysis. Further it is to be
kept in mind in general: not a single stress within a stress state is fracture responsible but a
mode-related equivalent stress, Eq.(14),  eq  R .

5.3. Test Case 3, UD lamina T300/PR319,    (  p hyd ,  2   p hyd ,  p hyd , 0 , 0 ,  21 )T

Problem: Influence of the hydrostatic pressure p hyd on the in-plane fracture shear strain.

Task: Determination of the fracture failure curve  21 (  2fr , 21fr ) .

Assumptions and analyses:
The envisaged stress situation is an applied external shear stress superimposed by a
synchronically increased hydrostatic pressure. Necessary for the establishment of the shear
strain failure curve is the relationship of the fracture shear stress  21fr and p hyd . This is given
by the failure condition. However, information how the fracture shear stress is linked to the
fracture strain is not known. The increase of  21 ( p hyd ) and of G12 ( p hyd ) is not proportional
according to the 2 nd Tg shift. Therefore many guesses are necessary. Applying for instance
micromechanics together with the R-O equation is therefore not successful due to the too
many open parameters which have to be assumed. Consequently, an engineering-like
approach was attempted by searching equivalent test results in the literature. From [Pae96]
could be concluded that an approximately linear relationship exists between fracture strain and
fracture stress, namely
 21fr   21fr ,initial  cPae  ( 21fr / R||  1) (41)

with the slope constant cPae  0.090 . This value has been applied for this Test Case and the
following ones. The 2nd Tg considering failure shear strain at  600 MPa is estimated as
 21fr , 600  15 % .
Essential results:
* The mechanical effect of p hyd leads to an almost linear stress-strain curve (Fig. TC 3).
* The influence of the 2 nd Tg shift effect is demonstrated by the difference of the solid to the
dashed curve.

5.4. Test Case 4, UD lamina T300/PR319,    (  p hyd ,  2   p hyd ,  p hyd , 0 , 0 ,  21 )T

Problem: Influence of an initial constant hydrostatic pressure on the in-plane shear stress-
shear strain curve.
Task: Determination of the stress-strain curve  21 (  21fr ) for p hyd  600 MPa . The shear is
superimposed on an initially acting hydrostatic pressure state.
Assumptions and analyses:

WWFE-II-Part A_dispatched Dec09 38 of 77 29/12/2009, 8:07

The R-O equation is well mapping the shear stress-shear strain curve at zero pressure and
room temperature. Therefore it is also taken as mapping function for the hydrostatic pressure
case with the consequence to determine the associated R-O parameters.
At first a value for the strength  21fr ,600 at phyd = 600 MPa was estimated by the used IFF3
failure condition as 193 MPa, not respecting the 2 nd Tg shift effect. In [Hine05, table 7] a
prediction is given, based on measured shear moduli Gm0 of the matrix and the micro-
mechanical approach which resulted in a high stiffness increase of 79% at  600 MPa of the
initial shear modulus due to p hyd . Pae and Shi in [Pae96, Shi92 by Figures 4 and 22] just give
an increase of 13%. As the increase in strength between these curves differs so much an
average value is utilized in the TC 4 analysis: G12600  ( 1.75  1.13 )  G||  1.4  G|| as basis for
the R-O parameter determination
600
 21600   21600 / G12600  0.002( 21600 /  0.600
2 , 21 )
n
(42)

with   
n 600  n 21600 / 0.002 / n  21fr , 600 /  0.600
2 , 21 
and 21600  ( 21fr , 600   21fr , 600 / G12600 ) , with  21fr , 600 after Eq.(41),
as plastic shear strain at -600 MPa. Its value can be extracted from TC 3 and counts, if not yet
600
regarding the 2 nd Tg shift effect,  21 = 4.7 %. With the equations and data above one
600
parameter is still unknown,  c 0.2 ,21 . According to [Shi92], an estimation is made simulating
the shape of his fig. 11. The result was  0.600
2, 21  125 MPa .

Data for the applied 2 nd Tg shift effect are given in the caption. With this information all R-O
parameters are determined and the solid curve in Fig. TC 4 is fixed. The 2nd Tg curve is
obtained by assuming similarity and a fracture point ( 165MPa, 15% ).
Essential results:
* The effect of p hyd indicates, due to Eq.(36), an increase of the initial shear stiffness
demonstrated by the deviation of the bold curve from the thin solid line which is the zero
pressure curve (Fig. TC 4).
* A very essential 2 nd Tg shift effect is depicted by the dashed curve. The course of the curve
may have to be adjusted because large strains are originated there and the engineering strain
needs to be transferred into a true strain. However, this needs further test information.

5.5. Test Case 5, UD lamina E-glass/MY750 Epoxy ,    (  1   3 ,  2 ,  3 , 0 , 0 , 0 )T

Problem, basically: How much lateral stress  2 can the UD material sustain if a fibre-parallel
stress  1 acts which is equal to a through-thickness stress  3 ?
Task: Determination of the failure curve  2 (  1   3 ) .
Assumptions and analyses:
In the tension quadrant, uni-axial and bi-axial tensile IFF1 failure will occur, see detail
zoom of Fig. TC 5.
In the negative quadrant, wedge failure ( F ), IFF2, occurs due to the fact that there is always
a difference between  3c and  2c . The initiation of a wedge failure is equal to the onset of
delamination damage. The wedge will slide and then locally cause a compressive reaction  3c

WWFE-II-Part A_dispatched Dec09 39 of 77 29/12/2009, 8:07

normal to the lamina's plane onto the adjacent laminae. This will induce delamination or
might increase an original delamination. In the case of a tube specimen loaded by a
hydrostatic compressive state of stress (no delamination possible), one can conclude also from
experience that the multi-axial strength is increased (  3c and  2c are acting in a favourable
manner) as long as the sliding friction is increased until a possible maximum. Such a
maximum is reached when the combined lateral compressive loading would lead to a FF
failure which is here FF2 ( F|| ), ‘kinking’ failure. This situation comes up far beyond 1000
MPa which means outside the actually practiced engineering domain. Since the Poisson effect
reduces the applied fibre-parallel compressive stress, due to the superimposed  3c stress, the
tensile failure mode F|| cannot be obtained.
For TC 5 as for the following Test Cases a fracture angle of  cfp =50° was taken (the value,
measured for the Annex A material might be too high). This results in a ‘friction parameter of
1 1
b    1.21 . As associated Mohr-Coulomb friction coefficient is
1  cos( 2 fp ) 1   
c

computed    0.17 . It is a value which is fully related to the chosen linear Mohr-Coulomb
approach.
Essential results:
* Two IFF2 fracture situations are depicted by the two sketches in Fig. TC 5. Two curves are
resulting from the fact that two differences of the stresses  2 and  3 may become possible.
* The observation of the FF2 stress effort, F|| , in the calculation output when uploading,
indicates that the two failure curve branches will close at values outside of the so far
envisaged domain beyond the 1000 MPa level. Additional coding work on Eq.(18b) to derive
the closed interaction (each IFF2 with FF2) failure curve was performed in this more
academically interesting domain. Fig. TC 5 shows the closed non-failure area. The ‘sharp’
corner in the negative quadrant is the result of the agreed and chosen deterministic modelling
using average input data (survivability, 50% expectation). A probabilistic treatment of the
situation in the corner, analogous to Ref.[Cun93], would exhibit that the lines of constant
survivability would round the corner (Fig. TC 5) according to increasing survivability values
versus the origin of the coordinate system. However this approach would require besides the
deterministic mechanical modelling a) a stochastic modelling of the uncertain basic variables
(design parameters, especially the R), b) the so-called logical modelling of the failure system
(for the laminae in a laminate failure system), and c) probabilistic analysis employing e.g. a
First-Order-Reliability-Method or a Monte Carlo method in order to grasp the joint-failure
probability for a criticality judgement.
*A strong influence of the unknown friction parameter b is indicated (dotted curves) for
lower pressures by the variation of its value, semi-brittle (1.21) and brittle (1.52). This
highlights the sensitivity to the input data.
* The non-depicted 2 nd Tg shift effect curves would lie marginally outside of the bold curves.

5.6. Test Case 6, UD lamina S-glass/Epoxy,    (  1 ,  2 ,  3   2 , 0 , 0 , 0 )T

Problem: How much compressive fibre-parallel stress  1c can the UD material sustain if a
lateral compressive stress  2c acts which is equal to a through-thickness stress  3c ?

WWFE-II-Part A_dispatched Dec09 40 of 77 29/12/2009, 8:07

Task: Determination of the tri-axial failure curve  1 (  2   3 ) .
Assumptions and analyses: Main aspect is: Due to the equal transversal stresses IFF2 cannot
occur. However, these two stresses cause filament straining under the activated bi-axial
pressure loading wrt to the Poisson effect which leads to two failure curves.
Essential results:
* In the positive  2   3 domain the interaction of the two FF with IFF1 results in the curve
of the upper part of Fig. TC 6.
* The solid curves in the lower part display how much external longitudinal stress (  1t or  1c )
can be applied to the material. In the regime  2c   3c  10 Rc no additional external fibre-
parallel tensile stress  1t can be carried by the UD material anymore. This proves the limited
applicability of the homogenized lamina stresses, because  1 is not the fracture stress but
 1 f   1  E f . In order to remain on composite level in the third quadrant  1 f has to be
multiplied by the fibre volume fraction V f , as an engineering approach.
* The  1 strength decreasing 2 nd Tg shift effect is considered by the dashed line. Its size is
not known to the author. However, some reduction (50% of the generally fixed value) is
considered and displayed for FF2 which is matrix-dominated. The right curve is not affected.
* Delamination is suppressed by the bi-axial compression.
* According to the fact that for one branch the strain is positive (FF1) and for the other
negative (FF2) there is no closing possible for TC 6 under the model assumptions.

5.7. Test Case 7, UD lamina A-S carbon/epoxy1,    (  1 ,  2 ,  3   2 , 0 , 0 , 0 )T

Task: See TC 6, just the UD material is different.

Essential results:
Fig. TC 7 proves that FF1 is met at a higher bi-axial pressure level  2   3 - in comparison
to TC 6 - according to the higher tensile strength.

5.8. Test Cases 8, laminate E-glass/MY750 ep., ˆ   ( ˆ x   z , ˆ y , z   p , 0 , 0 , 0 )T

Problem: Effect of the applied surface pressure  z   p (through-thickness stress) on the
size of the normal section force n y at fracture, when n x / t  ˆ x   z with t is laminate
thickness. Curing stress from effective temperature: 120°-23° = 97°:  2t  15MPa .
Task: Determination of the average laminate fracture stress ˆ y (  z  ˆ x )
Assumptions and analyses: No edge effects, [Bec94, Mit06].
Essential results:
Fig. TC 8 depicts the envelope of the tri-axial fracture state of stresses considering tensile as
well as compressive loadings of the multi-directional laminate, and regarding initial and final
failure (fracture):
- Quadrant 1: Delamination (IFF1), caused by  3t in    ( 0, 2t ,  3t ,0,0, 21 )T
- Quadrant 2: First a benign IFF1, basically caused by  2t in    ( 0, 2t , 3c ,0,0, 21 )T . Then
transition and finally interaction of IFF2 and IFF3

WWFE-II-Part A_dispatched Dec09 41 of 77 29/12/2009, 8:07

- Quadrant 3: Transition from initial failures IFF2/IFF3 to final kinking failure FF2 by altering
stress state    ( 1c ,  2c ,  3c ,0,0, 21 ) T
- Quadrant 4: Benign IFF2, caused by  2c in    ( 1 ,  2c ,  3t ,0,0, 21 ) T . Then transition and
interaction to combined initial failure IFF2 with IFF3.
nd
The 2 Tg shift effect is marginal. The failure surface is open and indicates that failure of a
dense material under hydrostatic compression (diagonal line) will not take place (physics).

5.9. Test Case 9, laminate E-glass/MY750 epoxy , ˆ   (100MPa, ˆ y , 100MPa, 0, 0, 0) T

Problem: Effect of the applied surface pressure ˆ y  n y / t on the in-plane average strains
ˆ x ,ˆ y at fracture, when ˆ x   z  100  MPa remains constant from beginning. Curing
stress from effective temperature: 120°-23° = 97°:  2t  15MPa .
Task: Determination of the stress-strain curves ˆ x ( ˆ y ) and ˆ y ( ˆ y ) .
Assumptions and FE analysis: No edge effects. Surfaces remain parallel to another.
Essential results:
Fig. TC 9 shows the graphs for the in-plane failure strains of the multi-directional laminate at
constant bi-axial compressive loading (x-z plane) with a growing loading ˆ y in the y
direction. The chosen load path leads from the interaction failure IFF2 combined with IFF3 to
the same interaction failure at the lower branch of Fig.8.

5.10. Test Case 10, laminate IM7/8551-7, ˆ   ( 0 , 0 ,  z , 0 , 0 , 0 )T

This case was not executed with respect to the available time of the author.

5.11. Test Case 11, laminate IM7/8551-7, ˆ   ( 0 , 0 ,  z , ˆ zy , 0 , 0 )T

Problem: Effect of an applied surface pressure  z   p on the laminate’s out-of-plane shear

capacity, when ˆ x  ˆ y  0 . The pressure is imposed by a pressure head. The lay-up is
simpler as for TC 10. Curing stress from effective temperature: T  -177° + 23° = -154°:
 2t  22.6MPa .
Task: Computation of the inter-laminar failure shear stress  zy   yz (  z ) over the laminate
extension. Judgement of the tri-axial stress state in the critical location, that is the centre of the
laminate. Check whether a shear stress  || in the 0°-lamina (y cross section) is the critical one
or   in the central 90°-laminae. This might change with increasing pressure.
Assumptions and analyses:
The compression loading causes – due to the Poisson effect - incompatible deformations of
the individual laminae. In order to achieve a bonded laminate, internal in-plane and out-of-
plane shear and normal stresses are built up over the laminate thickness. A non-linear laminate
analysis helps to indicate the desired critical location where the failure is initiated. There
should dominate a low stress gradient in view of the high objective: Validation of a failure
condition. Initial failure occurs in that critical lamina where the superposition of the shear
stress with the normal stresses takes the highest stress effort. In Fig. TC 11, the marked

WWFE-II-Part A_dispatched Dec09 42 of 77 29/12/2009, 8:07

internal element represents a location with a pretty low stress gradient where the material has
to counteract the two loadings shear and compression.
The assumed Iosipescu test (ARCAN test rig) matches the stress situation of the task. The
critical cross section is indicated in the sketch added to Fig. TC 11. The applied external stress
 zy generates different shear stresses in each of the plies. It causes a  23 lamina shear stress in
the 90° laminae and a  31 shear stress in the 0°laminae. Due to the constant stiffness over the
width (y coordinate) of the laminate specimen the  zy shear stress distributions  23 , 31 are
parabolic according to the indicated  zy distribution. The moment of the applied shear forces
F within a Iosipescu test apparatus is marginal. No edge effect is considered.
The curing stress  2t acts in-plane. Therefore, at the  23 axis (  z   p  0 ) two material
stress efforts are generated by  23 with  2t . The combined stress effort Eff  1 determines the
failure limit of  23 . This interaction points out the start of the transition zone, IFF1 to IFF2.
On the positive  z axis, the tensile stress delivers a  3t stress which together with the  2t
curing stress builds up a multifold failure situation. According to Eq.17 the common fracture
point may be estimated by Rtt  Rt / m 2 = 73 / m 2  57 MPa, but this reduced strength value
cannot become ‘active’ because  2t is much smaller than  3t .
There is some notching effect at the surfaces of the adjacent laminae but it cannot be
estimated. It’s the same with the fracture-mechanically driven ‘thin-layer effect’, wrt IFF1.
Essential results:
* Shear capacity is firstly increased (IFF1 level lowered) when applying a surface pressure p
and then decreased because IFF2 becomes the failure driving mode for higher pressures.
* Both the shear stresses were commonly checked which will be the failure driving one. The
outcome of this investigation was: The 90° lamina is the critical lamina indicated by the shear
stress  23 .
* For max  zt   3t an IFF1-based delamination fracture occurs with separation of the
laminae. The curing stresses act in the lamina plane. This means interaction but not in the
same plane as  3t .

* Uni-axial wedge failure IFF2 ( Eff  1 ) occurs if an isolated lamina is loaded up to
 z    185 MPa (‘isolated’ strength value). However, due to the beneficial multi-axial
c


compression the maximum stress effort Eff ( eq ,  eq )  1 will be firstly reached at
 z  450 MPa (  z  6% ). The compressive constraining of the tensioned fibre network
leads to a higher load-carrying capacity in comparison to a uni-axial compressive stress
applied only.
* Reason for the low strength capacity of the laminate under pure through-thickness shear
stress  zy is  23 which causes IFF1 in the 90° lamina (delamination fracture,  3t ) and
decreasingly stimulates wedge failure IFF2 for higher pressures.
* A benign IFF2 failure is caused at the y-ends due to the fact that the pressure head loading
does not allow a wedge failure in contrast to “outer pressure at a free surface”. A critical
(catastrophic) IFF2 would be just possible at the compressed y ends of the specimen. At the
free surface of the edges shear micro-cracking may occur.

WWFE-II-Part A_dispatched Dec09 43 of 77 29/12/2009, 8:07

*The failure curve does not run for p  0 through Rt on the y-axis due to the fact that  23
has a compressive component (contribution to IFF2) and a tensile component (to IFF1) which
interact with  3 .
* According to the distribution of the applied shear stress the given results just stand for a
local initiation of failure as onset of IFF. Naturally, there is some ‘quasi-plastic’ reserve
(analogous to Neuber’s notch theory) for the laminate when considering the  distribution.
The lower loaded vicinity around the specimen centre may take over a little loading after the
onset of the central IFF1-caused degradation. Therefore, the specimen will not directly fail
when reaching max  23 ( p ) or in other words will not be ‘scissored’.
* A practical and operational end of the laminate’s structural capability wrt serviceability,
fitness for further use, or limit of usability is reached when the filaments directly become
pressed on another. This level is reached far beyond the -450 MPa where the level of the
filament tensile effort (FF1) is still not high. Hence, the laminate may be compressed higher
and higher. But functionality ends with the disruption of the laminate structure. One can load
the squeezed disrupted laminate again, but significant functional properties such as heat
transfer might have strongly changed. The state of stresses in 90° lamina and 0° lamina are
   (185, 185, 450,  23  0, 0, 0) T and    (185, 185, 450, 0,  31  0, 0) T MPa.
* FF1 will be reached about -2000 MPa. No clear value for this final failure point ‘squeezing’
can be given. However, this level is not of interest in normal industrial design.
* The 2nd Tg effect is marginal.

5.12. Test Case 12, laminate IM7/8551-7, ˆ   ( 0 , 0 ,  z , 0 , 0 , 0 ) T

Problem: Effect of the applied surface pressure  z   p on the average fracture strains: in-
plane normal strains ˆ x ,ˆ y and through-thickness (out-of-plane) normal strain ̂ z . Curing
stress from effective temperature: T  -177° + 23° = -154°:  2t  22.6 MPa .
Task: Determination of the failure stress-strain curves ˆ x (  z ) , ˆ y (  z ) , and  z (  z ) of this
(balanced) symmetric angle-ply laminate. Consideration of Birch, 2ndTg, and T .
Assumptions and analyses:
Edge effects [Mit06] do not exist and surfaces of the test specimen remain parallel to another.
For the non-linear analysis the secant modulus of the average R-O curve is applied. It
considers the softening curve part, the equivalent stress (considers the ‘helpful’ effect of
multi-axial compression in comparison to uni-axial compression), the smeared stresses, and
the effective pressure in the matrix material (suffix I means lamina I). All these quantities
affect the degradation function  . As a simple engineering prediction - due to missing
knowledge – the stiffness-dedicated f 2 ndTg factor is also fully used to down-scale the failure
stress (  strength)  z   p . The utilized formulas read for the TC12 material:
E
E sec  
with 
1 ,
E    n 1  5.78  
1  0.002  (
eq
) 1  exp( )
R p0.2 R p0.2 0.56

eq  ( b 1)  (2 3 )  b  2  2  2 3 3  0 ,

2 2

WWFE-II-Part A_dispatched Dec09 44 of 77 29/12/2009, 8:07

  x  E p V f ) V f
   (
3

m ,1 ,  m , 2 ,  m , 3 )  (
T
,  2, I ,  z ) T and  ef    m ,l / 3 ,
1Vf
m ,l
l 1

22
 1  0.0018   ef 
f 2 ndTg  1 / 22 1    .
 1.20  0.00092   
 ef 
Essential results:
* The results of the non-linear MathCad analysis are visualized in Fig. TC 12.
* The curing stress is not high enough to generate IFF1.
* If there is no mechanical pressure head in the case of hydrostatic pressure the outer layers
experience a higher wedge failure risk. However, under the given boundary conditions, the
indicated wedge failure is not a hazardous one but a benign one. The load carrying capacity of
the squeezed material is not exhausted for further higher loading than -450 MPa but an
applicability of the laminate for further down- and up-loadings is not given anymore. So, in
the frame of the assumptions there is no termination of the curve before FF1 occurs at about a
filament failure strain of  1t  1.5% at ˆ z  15% . A practical and operational end of the
structural capability of the laminate is reached when the filaments directly become pressed on
another. The filaments become squeezed and bent and the damaged laminate cannot endure
one more loading (practical maximum fibre package is about 85%, but it cannot be fully used
due to non-ideal filament distribution). No clear value for a final failure point for ‘squeezing’
or ‘crushing’ can be given. See TC11, too.
* As the deterministic problem is fully symmetric, failure of the constrained fibres will take
place in all laminae at the same loading level. This presumes: stochastics and structural
reliability is not taken into account ([Rac87]), and altering curing stresses over the laminate
wall thickness do not exist.
Note: At a critical location of a structure, the failure situation together with the influence of the
stochastic design parameters can be probabilistically investigated, even in a non-linear analysis
case. An interaction of the failure in the 0°-lamina and the 90°-lamina, intensified by notching,
is not considered. Also, that the failure criticality of the centre failure point is much higher than
that of the surrounding larger laminate part is not regarded (this ‘quasi-plastic’ reserve lowers
the so-called joint failure probability of the failure system laminate a bit). All these effects are a
matter of the system reliability of the laminate (joint system) which consists of a system of
failing lamina elements.
* The 2ndTg effect is marginal wrt the assumptions made.

6 Discussion, some conclusions and outlook

6.1. Discussion
The author delivered results except for TC 10 as findings from MathCad codings. An
implementation of his failure theory into a FE code could not be realized wrt the high amount
of costs. Probably in Part B the analysis of the test specimens might require a FEA in order to
tackle the real boundary conditions of the test specimen in the test rig and the real stress
distributions.
Whether the wedge failure IFF2 is catastrophic or more benign depends on the stack and on
the load application. Under a pressure head device a wedge cannot ‘get out’ of the surface
plane as it is possible under hydrostatic pressure. Therefore outer plies underlie a higher
wedge failure risk than the interior ones.

WWFE-II-Part A_dispatched Dec09 45 of 77 29/12/2009, 8:07

 The results of the investigation have to be applied in a critical manner. There are a lot of
assumptions made due to missing information and due to quality of input data. Therefore the
paper is only the still mentioned blind prediction. Poisson’s ratios  ,  || ,   are therefore
applied as constants in the viewed hydrostatic pressure domains except when a ‘sensitivity to
 check’ is performed. Some input data found are contradictory.
As there is no information provided on the effect of the 2nd Tg shift it could be only coarsely
assumed on basis of non-sufficient literature results. Further, curing stresses are not always
considered because their magnitude is much smaller than the stresses due to multi-axial
pressure, and, regarding the uncertainties of the input. Therefore, the main idea of the paper is
a presentation of the procedure.
It is to be clarified whether the strains in the – for the author - only available stress-strain
curve in Ref. [Pae96] are true strains or not. If they are true strains then the 2nd Tg effect will
become smaller and has to be re-assessed.
Hydrostatic pressure heals the adverse effect of flaws (such as micro-cracks, micro-voids)
and delamination by compressing the flaws and increasing the amount of fracture energy
necessary for crack growth. However, when fracture begins, this will happen faster and hence
more ‘catastrophically’. This means, final fracture occurs in this compressed state in a more
brittle manner. Hydrostatic compression and in a minor manner through-thickness
compression alone improve both static ‘strength’ and cyclic ‘strength’ according to
 eq ( , p hyd )   eq   . This issue is of high importance, e.g. at bearings and joints.
Failure mode identification is inherent to the FMC. There are five different failure modes.
Delamination failure is covered by them, too. Initial failure of the critical lamina is caused by
an IFF and was indicated in the graphs for the benefit of design.
The Beltrami-based logic to choose invariants is not applicable for a deformation poor
behaviour such as it is still the case with the isotropic NF formulation.
The derived failure conditions predict no failure under hydrostatic pressure if non-porous
(dense) materials are applied.
Limits of smeared macro-mechanical modelling of a UD material were outlined for FF.
Failure occurs when the filament strength R|| f is reached. This is of high importance in case
of confining pressure.
The application of an own micromechanical formula set to the provided micro-mechanical
data was - as expected - not successful because these data were calculated with another set of
micromechanical formulas. Conclusion: When applying a micromechanical tool a closed top-
down-bottom-up simulation modelling is mandatory.
Sufficient for pre-dimensioning are the basic strengths. Remaining unknown material
friction-related curve parameters as b|| , b for UD and the softening curve can be estimated.
Note:
In practice, technical quantities such as design parameters are random variables (stochastic
quantities) but not constants, see e.g. [HSB02400]. Statistical methods must be used to
check whether the value of the technical quantity is within a tolerance interval (one-sided or
two-sided) or not. The limits of the statistical tolerance interval are termed statistical
tolerance limits. Exemplarily, the so-called A- and B-value strength design allowables ( R ,
no bar over) of the MIL-HDBK 5 (now, MMPDS-2) are the lower one-sided tolerance limits
L1 for a confidence level C  95% , and a specified survival probability (  reliability)
p  99% , respectively 90% (B-value). As so-called physical parameters E , , CTE etc. the
mean values (of the basic population) are usually used in order to achieve a mean structural
behaviour (means minimisation of computer runs, of output and of associated evaluations).

WWFE-II-Part A_dispatched Dec09 46 of 77 29/12/2009, 8:07

 However, for special tasks upper and/or lower tolerance limit values are taken in a second
run for the design verification.
Test data evaluation, however, means mapping of 50% expectation properties.
For the design of thick laminates it can be generally concluded that the former predictions
are usually not lying on the ‘safe’ side because the thickness direction possesses worse
properties in comparison to the in-plane properties. This is a consequence of the usual
‘manufacturing signatures’. This ‘orthotropic’ aspect has to be respected in design. However,
it may be considered engineering-like (for the transversely-isotropic UD lamina) in the 3-
direction during the strength analysis within the design verification.

6.2. Specific conclusions

Failure theory: The FMC is capable of handling tri-axial states of compression stresses
which means stress states having a significant impact on the mechanical behaviour of matrix
and FRP material as well. In general, ‘stiffness’ and ‘strength’ properties of the matrix
material increase with increasing tri-axial pressure. Thereby, the elasto-mechanical effect has
to be distinguished from a material degradation effect caused by the weakening matrix
material due to the shift of the 2ndTg up to room temperatures (where the tests run) and from
Birch’s finite strain stiffness increase formulations for E m and Gm .
When judging the received fracture stresses it is to be discriminated besides p hyd the specific
additional external stress and that specific stress which really acts on the specimen’s surfaces.
Isotropic matrix: The application of the FMC-based theory to the epoxy matrix worked. The
relationships behind strength and a uni-axial or a multi-axial stress state can be exemplarily
formulated for the mode IFF2 by
l-D:  eqfr ,   2fr ,c  max  2c  Rc , 2-D:  eqfr , (  2c , 21 ) , 3-D:  eqfr , (  2c , 21 , 3c ) .
Uni-axial tests let determine the strength and bi-axial tests on top the friction parameters if the
stress state belongs to a pure mode. In general, a distinct 3D stress state may address up to 3
modes. Therefore, tri-axial tests serve to validate the parameter m which is responsible for an
adequate interaction of the faced modes.
Lamina: The application of FMC-based theory to the laminae test cases is successful
however a lot of data had to be guessed. The interaction exponent can be chosen on the safe
side as e.g. CFRP: 2.5  m  3 . The same value for m can be approximately used for each
mode interaction domain. The determination of the highest mode effort as the design driving
one is automatically given. The computation of a resultant effort is integral part of the FMC
procedure. A combination of ‘own’ micro-mechanical formulae with the provided input was
only partly successful, as anticipated. For G||  it was fine, but data inconsistency was given
for E  because Eq.(5b) delivered just half the provided value. A factor of two had to be
applied to the Eq.(5b) values.
The point of initial failure at uni-axial compression is increased under tri-axial compressive
loading. In contrast to a non-porous matrix material a non-porous UD material may FF2-shear
fracture under very high hydrostatic compression.
Laminate: The application of the theory to laminate test cases is working. In the non-linear
cases the load can be increased until numerically final failure is reached. The (higher) final
failure level is indicated by a stress effort Eff  100% for the critical lamina by checking

WWFE-II-Part A_dispatched Dec09 47 of 77 29/12/2009, 8:07

during the linear or non-linear analysis whether the 100% are reached. The real load capacity
of a thick laminate is lower than for the computed ‘thin’ 8 lamina stack.
For the prediction of final failure the initial failure approach is not of that high concern, if
wedge failure, caused by F and followed by delamination failure, will not occur. In ‘well-
designed’ laminates failure will occur with the first FF in any single ply of the laminate. A
strength failure mode is a ‘material mode’ whereas a stability failure mode in a test case is a
‘structural mode’ which might be triggered by a strength failure mode.
Note: Higher load carrying capability, obtained under p hyd , is basically the effect of the
favourably affected equivalent stress  eq and not the result of an increase of R .

6.3. Outlook and relevance for design

* A strength condition is usually validated in a smooth maximum effort (critical stress) region,
that means in a region with a low stress gradient.
However, even such a condition can be only a necessary condition which may be not fully
sufficient for the prediction of ‘onset of fracture’. This holds e.g. for the in-situ lateral strength
of a thick lamina embedded in a multi-layered laminate, see [Fla82]. Due to the difference of
isolated and embedded data, an isolated lamina-based validation of a FRP strength failure
condition can be never more than an accepted compromise.
In the case of embedded, deformation-controlled laminae, where the micro-cracking lamina
influences the neighbouring laminae, in-situ strength properties are valid. When introducing
them within a damage considering procedure one has to bear in mind that normally a Strength
Design Verification requires direct measurable statistics-based strength parameters (such as
the statistically reduced strength design allowables) otherwise a strength demonstration is not
possible. This has to be tackled in future for the various parameters of the CDM approaches,
too, before industrial use.
* The failure process of the failure system laminate is a stochastic process which is generally
to be treated by probabilistic means. It can be approximated by the presented deterministic
procedure.
* The influence of the ‘thin-layer effect’ and of the location where the lamina is embedded in
the stack should be reflected by any approach.
* The transfer of the static failure conditions to cyclic loading (fatigue life) application or
impact, especially of thick-walled composites under tri-axial stress states, is a necessary
research task for the future. A failure mode-dedicated determination of the successively
generated damage portions in the damage accumulation might be promising.
In this context the FMC has relevance for fatigue and damage tolerance: Lifetime prediction
as an essential part of design analysis addresses two subjects, damage tolerance and fatigue.
Damage tolerance is the design-dependent ability of the structural component to withstand
damage caused by quasi-static overloading, by impact or fatigue within the foreseen
inspection interval. In this context essential is the residual strength of the composite part after
being damaged. Damage in general may be a ‘diffuse damage’ (micro-mechanical cracking,
crazes; in German: Schädigung) or a ‘discrete damage’ (delamination; in German: Schaden).
Cyclic fatigue means degradation the material experiences under variable loading.
The author proposed in 1996 a fatigue life model for UD laminae-composed originally flaw-
free laminates. This model uses a FMC failure mode–related quantity which considers the full
multi-axial stress state together with failure mode-associated UD S-N curves. According to
material symmetry restrictions the transversely-isotropic UD material possesses 5 strength
WWFE-II-Part A_dispatched Dec09 48 of 77 29/12/2009, 8:07
failure modes. Three of them belong to matrix-dominated mechanisms termed inter-fibre-
failure (IFF: transversal tension and compression, in-plane shear) and two to fibre-failures
(FF, fibre-parallel tension and compression). In the case of brittle behaving composites a
reasonable damage generator has to be fixed and a stress effort-based formulation for a
damage accumulation provided. Loading-unloading paths play an essential role. By now, the
usual procedure is to take over from fatigue of ductile materials a procedure one might call
“Stress amplitude procedure with mean stress correction”. It is questionable to simply adapt
this procedure for brittle behaving composite. The ‘mean stress correction’ should be replaced
by a ‘mode view respection’. This includes a change from  amplitude via  upper to  eqmod e .
* The given stress-strain curves of the UD-lamina are interpreted and termed mechanical load-
based macro-mechanical stress-strain curves. It is assumed that the stress-strain curves are
mean curves ( R values are given), the type one needs for test data mapping (see Fig. 5 etc.).
An edge effect (3D state of stress) is not considered, because the laminates are assumed to be
part of a 'closed' composite structure.
* Curing strategy, specifically dedicated to a composite part can lead to marginal curing
stresses. Optimal resin system-dependent curing cycles - set up in the frame of the production
cycle needs - will lead to low cure-induced residual stresses.
* The FMC may be applied to some of the upcoming orthogonal textile composites on the
macroscopic level, too [Cun08]. These are: UD formulation for the non-crimp fabrics, the
orthotropic formulation for woven fabrics (see also [Boe08]).
* Strength failure conditions are a necessary but not always a sufficient condition in failure
prediction when viewing the in-situ behaviour. Attempts to continuum-mechanically link
‘onset of fracture’ or ‘cracking’ prediction methods for structural components are actually
undergone [Leg02].
* In the case of discontinuities such as notches with a steep stress decay, a ‘toughness
combined with a characteristic length-based energy balance condition’ may form a sufficient
fracture condition, only. Here, according to Neuber, a structural length comes into the
applicable models. For instance, when tackling notch stress problems of an open hole [Kro06]
or a filled hole located in a joint one has to calibrate the load-determined laminate notch stress
in a distinct distance from the hole centre by a failure stress of the un-notched laminate which
is performed by a non-linear ply-by-ply failure analysis, for instance with the FMC UD
conditions.
* When comparing analytical predictions with test results test specimen-associated properties
have to be used or at least average (typical) stress-strain curves and average physical
properties. Also, detailed information about the test rig and test data evaluation (e.g.
premature failure due to load introduction effects) is necessary to better understand and – in
consequence – to better use the test data. Qualified conditioning and knowledge of fibre
volume content will be mandatory for performing the Part B analyses.
* The development of reliable composite structures requires reliability of design, analysis,
manufacture, and test (specimens, conditioning, test techniques, performance, evaluation,
NDI). All these requirements are also valid when validating a failure theory by small
specimen test results. Otherwise the input quality-dependent predictions yield no satisfying
results and the designer-desired reduction of the dozens of actually not fully validated failure
conditions to a few really validated ones is not possible and, in consequence, the confidence to
use fibre composites is not built up. In this context, careful analyses and careful evaluation of
the observed test results is mandatory.

WWFE-II-Part A_dispatched Dec09 49 of 77 29/12/2009, 8:07

* Robust design means tolerance of the structure to the variations the structural product
experiences during manufacture and usage. It might be defined: Robust design = 1 /
(uncertainty times complexity). This aspect should be a guideline for designing, for modelling
of test specimens and for testing, especially when validating a failure theory. In the same
manner it may be concluded for the failure conditions itself: They should be based on a few
physically interpretable parameters which can be reliably determined and assumed in case of
pre-dimensioning, if necessary.

Acknowledgement:
The contribution to the WWFE-II by the author, who retired from industry some years ago, was again
effectively supported by Dipl.-Ing. A. Freund (Institute of Lightweight Structures and Polymer
Technology, TU Dresden) by elaborating and visualizing the provided data, and by performing and
cross-checking laminate MathCad computations. The author is grateful for discussions with Dr.-Ing.
G. Ernst from TU-Hannover, Dipl.-Ing. S. Müller (TU-Chemnitz), with CADFEM (Drs. Hoermann,
Fritsch) and ANSYS (Dr. Kracht). And not to forget, without the excellent reviewer comments the
paper would have not matured enough!
Mind: The huge work on the FMC was never funded and is practically the result of one person.

References
[Alf07] Alfaiate, J., Aliabadi, M.H., Guagliano, M. and Susmel, L.: Advances in Fracture and
Damage Mechanics VI. Trans Tech Publications, 2008
[Awa78] Awaji, H. and Sato, S.: A Statistical Theory for the Fracture of Brittle Solids under Multi-
axial Stresses. Intern. Journal of Fracture 14 (1978), R13-16
[Bec94] Becker, W. and Kress, G.: Stiffness Reduction in Laminate Coupons due to the Free-edge
Effect. Comp.Science and Technology 52 (1994), 109-115
[Bel1885] Beltrami, E.: Sulle Condizioni di Resistenza dei Corpi Elastici. Rend. ist. d. sci. lett., Cl.
mat. nat.18, 705-714 (1885)
[Bir38] Birch, R.: The Effect of Pressure Upon the Elastic Parameters of Isotropic Solids, according
to Murnaaghan’s Theory of Finite Strains. Journal of Applied Physics 9 (4), 279-288.
[Boer89] de Boer, R. and Dresenkamp, H.T.: Constitutive Equations for concrete in failure state. J.
Eng. 115 (8), 1989, 1591-1608
[Boe85] Boehler, J.P.: Failure Criteria for Glass-Fibre Reinforced Composites under Confining
Pressure. J. Struct. Mechanics 13 (1985), 371
[Boe08] Böhm, R.: Bruchmodebezogene Beschreibung des Degradationsverhaltens textilverstärkter
Verbundwerkstoffe. Dissertation, 2008. Technical University Dresden, Institut für Leichtbau und
Kunststofftechnik
[Bro94] Brown T.L. and Hyer M.W.: Effects of layer waviness on the stresses and failure in
hydrostatically loaded cylinders. Jan 1994, Paper ID STP24333S, Digital library / STP / STP1185-
EB / STP24333S; 17 pages
[Bus06] Busse, G., Kröplin, B.-H. and Wittel, F.K.: Damage and its Evolution in Fiber-Composite
Materials: Simulation and Non-Destructive Evaluation. SFB 381, Stuttgart (2006), ISBN 3-
930681-90-3
[Cad78] Caddell, R.M., Raghava, R.S. and Atkins, A.G.: Pressure dependent yield criteria for
polymers. Materials Science and Engineering 13 (1978), 113-120

WWFE-II-Part A_dispatched Dec09 50 of 77 29/12/2009, 8:07

[Cam06] Camanho, P., Davila, C.G., Pinho, S.T. , Ianucci, L. and Robinson.P.: Prediction of In-situ
strengths and Matrix Cracking in Composites under Transverse Tension and In-plane Shear.
Composites, Part A 37 (2006), pp. 165-176
[Chr98] Christensen R.M.: The Numbers of Elastic Properties and Failure Parameters for Fibre
Composites. Transactions of the ASME, Vol. 120 (1998), 110-113
[Col90] Colvin, G.E. and Swanson, S.R.: Characterization of the failure properties of IM/8551-7
Carbon/Epoxy under Multi-axial Stress. ASME J. Eng. Mat. Tech. 112, 1990, 61-67
[Cun93] Cuntze, R.G.: Deterministic and Probabilistic Prediction of the Distribution of Inter-Fibre
Failure Test Data of Pre-strained CFRP Tubes composed of Thin Layers and loaded by Radial
Pressure. (comparison to test results) Wollongong. Advanced Composites '93, pp. 579-585. The
Minerals, Metals & Materials Society, 1993
[Cun04a] Cuntze R.G. and Freund A: The Predictive Capability of Failure Mode Concept-based
Strength Criteria for Multidirectional Laminates. Part A, Composites Science and Technology 64
(2004), 343-377
[Cun04b] Cuntze R.G.: The Predictive Capability of Failure Mode Concept-based Strength Criteria
for Multidirectional Laminates. Part B, Composites Science and Technology 63 (2004), 487-516
[Cun04c] Cuntze R.G.: Should we consider Matrix Yielding when Investigating the Deformation
Behaviour? 9th European-Japanese Symposium on Composite Materials. Technical University
Hamburg/Harburg, May 2004
[Cun06] Cuntze, R.G.: Efficient 3D and 2D Failure Conditions for UD Laminae and their Application
within the Verification of the Laminate Design. Elsevier, Composites Science and Technology 66
(2006), 1081-1096
[Cun07] Cuntze, R.G.: Prediction of Static 3D and In-plane Fracture Failure of UD Laminae
composed of Fibre-reinforced Plastics. 4th NAFEMS Nordic Seminar on Materials’s Modelling,
Sandvika, March 2007, Conference CD
[Cun08] Cuntze, R.G.: Strength Failure Conditions of the Various Structural Materials: Is there some
Common Basis existing?, SDHM, vol.074, no.1, pp.1-19, 2008
[DeTer99] De Teresa, S.J.: Failure of a composite lamina under three-dimensional stresses. UCRL-
JC-135664
[Ern07] Ernst, G., Vogler, M., Hühne, C. and Rolfes, R.: Multi-scale Simulation for Stiffnesses and
Strengths of Textile Composites. NAFEMS Seminar, Simulating Composite Materials and
Structures. Bad Kissingen, Nov. 2007
[Fla82] Flaggs, D.L. and Kural, M.H.: "Experimental Determination of the In Situ Transverse Lamina
Strength in Graphite Epoxy Laminates". J. Comp. Mat. Vol 16 (1982), 103-116
[Has80] Hashin Z.: Failure Criteria for Unidirectional Fibre Composites. J. of Appl. Mech. 47 (1980),
329-334
[Hine05] Hine, P.J., Duckett, R.A., Kaddour, A.S., Hinton, M.J., and Wells, G.M.: The effect of
hydrostatic pressure on the mechanical properties of glass fibre/epoxy unidirectional composites.
Composites Part A, 2005, 279-289
[Hin02] Hinton, M.J., Kaddour, A.S. and Soden, P.D.: A comparison of the predictive capabilities of
current failure theories for composite laminated, judged against experimental evidence.
Composites Science and Technology 2002 (62), 1725-97
[Hin04] Hinton, M.J., Soden, P.D. and Kaddour, A.S.: Failure criteria of fibre reinforced polymer
composites. The World-Wide Failure Exercise. Elsevier, 2004, ISBN 0-08-044475-X, 700 pages
[Hop95] Hoppel, C.R.R., Bogetti, T.A. and Gillespie, J.W.jr.: Literature review – Effects of
hydrostatic pressure on the Mechanical Behaviour of Composite Materials. J. of Thermoplastic
Composite Materials 8 (1995), 375-409

WWFE-II-Part A_dispatched Dec09 51 of 77 29/12/2009, 8:07

[HSB] Handbook ‘Fundamentals and Methods for Aeronautical Design and Analyses’. Issued by the
working group IASB. Issue 2009, Technical information library Hannover. Here of interest: sheet
04200-02: Determination of statistical tolerance limits
[Hye88] Hyer, M.W.: Hydrostatic response of thick laminated composites cylinders. J. of Reinforced
Plastics and Composites 7, 1988, No 4, 321-340
[Jel96] Jeltsch-Fricker, R.: Bruchbedingungen vom Mohrschen Typ für transversal-isotrope
Werkstoffe am Beispiel der Faser-Kunststoff-Verbunde. ZAMM 76 (1996), 505-520
[Kad07a] Kaddour, A.S. and Hinton, M.J.: Instruction to Contributors of the Second World-Wide
Failure Exercise (WWFE-II) , Part A , March 2007
[Kno03] Knops, M.: Sukzessives Bruchgeschehen in Faserverbundlaminaten. Dissertation 2003.
Aachen, Institut für Kunststoffverarbeitung
[Kro06] Kroll, L. and Kostka, P.: Notch strength of tensioned fibre-reinforced laminates with an open
hole. HSB 34114-01
[Leg02] Leguillon, D.: Strength or Toughness? A criterion for crack onset at a notch. European
Journal of Mechanics A/Solids 21 (2002), 61-72
[Mas94] Masters, J.: Fractography of Modern Engineering Materials. Composites and Metals. 2 nd
volume. ASTM STP1203,1994
[Mat95] Matzenmiller, A., Lubliner, J.and Taylor, R.L.: A constitutive model for Anisotropic Damage
in Fiber Composites. Mechanics of Materials 20 (1995), 125 – 152
[Mat07] Matzenmiller, A. and Koester, B.: Consistently Linearized Constitutive Equations of Micro-
mechanical Models for Fibre Composites with Evolving Damage. Int. Solids and Structures 44
(2007), 2244-2268
[MIL17] MIL-HDBK 17 “Polymer Matrix Composites”: Vol I “Guidelines for Characterization of
Structural Materials”; Vol II “Material Properties”; Vol. III "Utilization of Data". Dep. of Defence
(DOD), Technomic Publishing USA
[Mit06] Mittelstedt, C.: Free-edge and free-corner effects in composite laminates. -Closed-form
analytical and semi-analytical approaches. Fortschrittbericht, VDI Reihe 18, Nr.307, Düsseldorf,
VDI Verlag 2006
[Moh00] Mohr, O.: Welche Umstände bedingen die Elastizitätsgrenze und den Bruch eines Materials?
Civilingenieur XXXXIV (1900), 1524-1530, 1572-1577
[Pae96] Pae, K.D.: Influence of Hydrostatic Pressure on the Mechanical Behaviour and Properties of
Uni-directional, Laminated, Graphite fibre/Epoxy matrix, Thick Composites. Composites Part B
27B (1996), 599-611
[Pae77] Pae, K.D.: The macroscopic yielding behaviour of polymers in multi-axial stress fields. J. of
Materials Science 12 (1977), 1209-1214
[Par85] Parry, T.V. and Wronski, A. S.: The effect of hydrostatic pressure on the tensile properties of
pultruded CFRP. J. of Materials Science 20 (1985), no. 6, pp 2141-2147
[Pau61] Paul, B.: A modification of the Coulomb-Mohr Theory of Fracture. Journal of Appl.
Mechanics 1961, 259-268
[Pin05] Pinho, S.T., Davila, C.G., Camanho, P.P., Iannucci, L. and Robinson, P.: Failure Models and
Criteria for FRP under in-plane or three-dimensional state of stress including shear non-linearity.
NASA TM 213530-2005, Langley Research Center
[Puc02] Puck A. and Schuermann H.: Failure Analysis of FRP Laminates by Means of Physically
based Phenomenological Models. Composites Science and Technology 62 (2002), 1633-1662
[Puc96] Puck, A.: Festigkeitsanalyse von Faser-Matrix-Laminaten -Modelle für die Praxis. München:
Carl Hanser Verlag, 1996
[[Rac87] Rackwitz, R. and Cuntze, R.G.: System Reliability Aspects in Composite Structures. Engin.
Optim., 1987, Vol. 11, 69-76
WWFE-II-Part A_dispatched Dec09 52 of 77 29/12/2009, 8:07
[Rhe01] Rhee, K.Y.: Deformation characteristics of multi-directional carbon fibre/epoxy composites
under high pressure. Polymer composites 22, 2001, (6)
[Rhe04] Rhee, K.Y., Lee and Park: Effect of hydrostatic pressure on the mechanical behaviour of sea-
water-absorbed carbon-epoxy composite. Materials Science and Engineering A, vol. 384, issues 1-
2, Oct 2004, pp. 308-313
[Rol97] Rolfes, R., Noor, A.H. and Rohwer, K.: Efficient Calculation of Transverse Stresses in
Composite Plates. MSC-NASTRAN User Conference, 1997
[Rol07] Rolfes,R., Ernst, G., Vogler, M. and Hühne, C.: Material and Failure Models for textile
Composites. ECCOMAS. Thematic conference on Mechanical Response of Composites. Porto,
September 2007
[Sch88] Scharringhausen, J. and Cuntze, R.: FEM-Calculations on the Domes of Fibre-Reinforced
Plastic Cases for Solid-Propellant Motors. Advanced Structural Materials: Design for Space
Applications, Workshop 23-25 March 1988, ESTEC, Nordwijk, ESA WPP-004. (MAN
Technologie)
[Sch78] Schneermann, M.W. and Stenzenberger, H.D.: Final Report for a Study on evaluation of
Built-in stresses during Manufacture of Elements made of Carbon Fibre Reinforced Plastics.
Dornier System GmbH, 1978. ESTEC-Contract No. 2930/76/NL/PP(SC)
[Shi92a] Shin, E.S. and Pae, K.D.: Effects of hydrostatic pressure on the torsional shear behaviour of
graphite/epoxy composite. Journal of Composite Materials, 28 ,1992 , No 4, 462-485
[Shi92b] Shin, E.S. and Pae, K.D.: Effects of hydrostatic pressure on in-plane shear properties of
graphite-epoxy composites. J. of Composite Materials, vol 26, no. 6, 1992, pp. 828-868
[Sli97] Slight, D.W., Knight, N.F. and Wang, J.T.: Evaluation of a Progressive Failure Analysis
Methodology for Laminated Composite Structures. 38th Structure, Structure Dynamic and Material
Conference, April 1997. AIAA Paper 97-1187
[Sri83] Srinavasta A.K. and White, J.R.: Curing stresses in an epoxy polymer. J. of Applied Polymer
Science. Vol. 29, issue 6, pp 2155-2161
[The9?] Theriault, R., Osswald, T.A. and Stradins, L.: Properties of thermosetting polymers during
cure. (year and editor not indicated)
[Tsa71] Tsai, S.W. and Wu, E.M. A General Theory of Strength for An-isotropic Materials. Journal
Comp. Materials 5 (1971), 58-80
[VDI97] Cuntze, R.G., Deska, R., Szelinski, B., Jeltsch-Fricker, R., Meckbach, S., Huybrechts, D.,
Kopp, J., Kroll, L., Gollwitzer, S., and Rackwitz, R.: Neue Bruchkriterien und Festig-
keitsnachweise für uni-direktionalen Faserkunststoffverbund unter mehrachsiger Beanspruchung –
Modellbildung und Experimente. VDI-Fortschrittbericht, Reihe 5, Nr. 506 (1997), 250 pages. In
German. (New fracture criteria (Hashin-Puck action plane criteria) and Strength Design
Verifications’ for Uni-directional FRPs subjected to Mult-iaxial States of Stress –Model
development and experiments)
[VDI2014] VDI 2014: German Guideline, Sheet 3 “Development of Fibre-Reinforced Plastic
Components, Analysis”. (in German and English. Beuth Verlag, 2006)

WWFE-II-Part A_dispatched Dec09 53 of 77 29/12/2009, 8:07

Annex I: Determination of the UD friction parameters by experiment
A.1 Similarity betwen FMC formulations and Mohr-Coulomb formulations
The basic assumption, lying behind all action-plane fracture criteria (e.g. UD Puck/Hashin) is
the brittle-fracture hypothesis which goes back to O. Mohr’s “The strength of a material is
determined by the stresses on the fracture plane”. This means for the here applied linear
Mohr-Coulomb (M-C) formulation
R   n     n or  n  R     n . (A1)
Herein, the value  is an internal friction property of the UD material (usually termed friction
coefficient) and R the so-called cohesion strength which corresponds to Puck’s fracture
plane resistance [Puc96]. The values of the Mohr-parameters depend on the approach, whether
it is a linear or a parabolic one. In this context, just for information, the more often used
parabolic approach is formulated by the author as (see also approach [Puc96])
 nt  R parab  ( R parab   parab   n ) .
2

If IFF occurs in a parallel-to-fibre plane of the UD lamina, the components of the vector are
the normal Mohr stress  n and the two shear Mohr stresses nt and n1. “The third shear stress
 tl and the normal stress  t will have no influence”, when citing Mohr.

A.2 Relationship between the FMC friction parameters and Mohr-Coulomb

A.2.1 In-plane shear  n1 (  n ) IFF3

The transfer of the shear failure condition IFF3, Eq.(14e), as in-plane (2D) formulation

b||  2 2   21  ( b||  2 2   21 )2  4 R||   21  2 R||

2 2 2 4 3

to a Mohr-shaped one is directly possible because the fracture plane is known to be parallel to
the fibre direction and in consequence that the relationship (  n1 ,  n )  (  21 ,  2 ) holds.
However this is more complicated with the condition above. Therefore just for the
visualization of the procedure, instead of the in-plane equation above, an already available
very simple approach is used which has still the Mohr-Coulomb shape, Ref . [Cun06], namely
 21  R||   ||   2 . (A2)

This condition reads after introducing the Mohr stresses  n1  R||   ||   n . In this simple
case, due to  n   2 the fracture angle  fp   0 , it can be cocluded from Fig. A1, that the
cohesion strength equates the shear strength, R||  R|| . The property   is determined like
b|| from a curve fit of the IFF3 test data.
Mind: Using the 2D formulation for the determination of the friction coefficient
  is sufficient because the equation fully describes the failure mode.

A.2.2 Transverse shear  nt (  n ) IFF2

More challenging is the derivation of the friction-related parameter for the transversal or
out-of-plane shear failure mode, respectively. Here, the unknown angle is  cfp  45 .
Following Fig. A2, the adapted Mohr-Coulomb formulation reads, using  n  (cos  ) 2   2 ,
WWFE-II-Part A_dispatched Dec09 54 of 77 29/12/2009, 8:07
R as transverse cohesive strength and   as the material internal (static) coefficient of
friction,
 nt  R      n . (A3)
Transferring the FMC formulation, Eq.(A3), into a Mohr-Coulomb formulation, is to
express the invariants not any more in lamina stresses but in corresponding Mohr stresses,
Eqs.(14d, 6d),

b  (  n   t ) 2  4 nt2  ( b  1 )  (  n   t )  Rc . (A4)

Above formulation involves the still mentioned stress  t which cannot be representative for a
Mohr envelope curve. However, despite taking this as a presumption one has nevertheless to
prove that  t comes out of the equation when inserting apparent stress boundary conditions
into the former equation. These conditions are the strength points. A further necessary
condition when transferring the FMC formulation into a Mohr one is that the slope of Mohr’s
circle must be equal to the slope of Mohr’s envelope curve in the common touching point.
The transformation of the lamina stresses into Mohr stresses depicts Fig.A2. Assuming  23 =
0, which does not confine the generality because otherwise the principal stresses  i
pr
can be
taken, the transformation (author 1991, [Puc96])
 fp   fp    / 180
 1  1 0 0 0 0 0   1 
  0 c² s² 2 sc 0 0   2  c  cos(  fp ) ,
 n 
 t  0 s² c²  2 sc 0 0   3  s  sin(  fp ),
    (A5)
 nt  0  sc sc ( c²  s²) 0 0   23  C  cos( 2 fp )  c 2  s 2
 tl  0 0 0 0 c  s   31  S  sin( 2 fp )  2  s  c
    
 nl  0 0 0 0 s c   21  C2  S2  1
will lead to the relationships
 n  c 2 2  s 2 3 ,  t  s 2 2  c 2 3 , (A6a,b)

 nt  s  c  (  2   3 )  s  c  (  )  0.5  S   , (A6c)

when applying the advantageous additional theorems with Mohr stresses and when using the
2 2
abbreviation (  2   3 )   . Mind: Mohr combines  n   nt   n1 . Further holds

 n   t  c 2 ( 2   3 )  s 2  ( 2   3 )  ( c 2  s 2 )   C  , (A7a)
which is to be resolved for  t yielding
 t   n  C  . (A7b)
Then, the slope in the touching point is to be determined via the derivatives
d n / d fp  2  s  c      S    sin( 2 fp ) , (A8a)

d nt / d fp  0.5  2  C    cos( 2 fp ) . (A8b)

They deliver one common equation for the slope of the envelope curve
d nt / d n  cot an( 2 fp )  C / S . (A9)

WWFE-II-Part A_dispatched Dec09 55 of 77 29/12/2009, 8:07

An implicit differentiation of the failure function F

4b   nt 
b  (  n   t )
dF / d nt  , dF / d n  b c 1  . (A10a,b)
(  n   t )  4 nt R (  n   t )  4 nt R
2 2 c R 2 2 c
 

gives the necessary analogous second equation Dividing Eq.(A10b) by Eq.(A10a) one obtains
b  1 b  (  n   t )

d nt Rc (  n   t ) 2  4 nt Rc
2

 . (A11)
d n 4b   nt
(  n   t ) 2  4 nt Rc
2

The minus sign is due to inverse differentiation. Equating the slope equations (A9) and (A11)
yields after reformulation
 
C ( b  1 )  (  n   t )  4 nt  b  (  n   t )
2 2

 .
S  4b   nt

Into this combined equation, as the already mentioned stress boundary condition, the given
fracture strength point  2   Rc   together with the fracture angle  fp   cfp has to be
inserted. This leads via the Eqs.(A6) and (A7) to
 n   t  C cfp   and  nt  0.5  S cfp   .
Finally it is obtained, when taking the negative root,

C cfp ( b  1 )  ( C cfp   ) 2  4  ( 0.5  S cfp   ) 2  b  ( C cfp   ) ( b  1 )  ( 1 )  b  C cfp
  . (A12)
S cfp  4  b  ( 0.5  S cfp   ) 2  b  S cfp

As can be seen in the equation above: The stress  t has no influence. It is not representative
as Mohr supposes for the Mohr envelope curve.
The final step is resolving Eq.(A12) for bc which results in the relation

 2   cfp 
b  1 /( 1  C )

c
fp with C c
 cos(  ) . (A13)
180
fp

A.3 Relationship between friction parameter b and ‘linear’ Mohr-Coulomb parameters
Searching for a direct relationship between the FMC and the Mohr-Coulomb parameters
requires a comparison of coefficients (in the strength point Rc ). At first Eq.(A1) is employed
in the shape of Eq.(A3)
 nt  R      n (A14a)

with R denoting the transverse cohesive strength. Then the IFF2 failure condition Eq.(A4)
is applied leading to the equation

b  Rc  C cfp  4 nt  Rc  ( b  1 )  ( 2 n  Rc  C cfp ) .

2 2 2
(A14b)

In order to compare Eq.(14a) with Eq.(14b) both have to be firstly squared and secondly
compared wrt the coefficients. This delivers

WWFE-II-Part A_dispatched Dec09 56 of 77 29/12/2009, 8:07

 2
R  R c  [( C cfp ) 2  2  C cfp  b  2  C cfp  b  2  C cfp  1 ] /( 2  b ) . (A15a)

The comparison requires the fulfilment of two conditions for the friction coefficients
  ,1   Rc  [ 4  C cfp  4  b  4  b  C cfp  8  b  C cfp  4 ] /( 8  R  b ) and
2 2

  ,2  ( b  1 ) / b  C cfp . (A15b)

This means ‘over determined’. The latter and simpler equation   ,2 represents a slightly
smaller value and will be applied as   .

A.4 Experimental determination of a value for the friction coefficient  

The determination of the friction coefficient   is performed for the linear approach, only.
This is sufficient from an engineering point of view. From the evaluation of the test data in
[VDI97]), see also Fig. A3 and Eq.(A13), the measured values  cfp   55 ( C cfp  0.342 ),
Rc  104 MPa were computed. Hence, when applying Eq.(A13) it is obtained:
- FMC approach:
b  I 4  Rc  ( b  1 )  I 2 ,
1
 b   1.52 , and b friction  b  1  0.52 ;
1  C cfp )
- Mohr-Coulomb approach:
 n  R      n ,

    C cfp  0.342 , R  0.5  Rc  C cfp  1  36.1 MPa .

Note: Assuming  23  0 is accurate for the test specimen. The determination of the
material friction parameter b from test generates a generally valid value.
Conclusions:
 As the matrices of the Test Cases seem to be semi-brittle as data set is proposed
for the UD analyses:  cfp   50 , b  1.21 or    0.17 , R  43.6 MPa ,
 All parameter values depend on the chosen Mohr-Coulomb model (linear or
parabolic or ...).
Fig. A4 eventually visualizes the received Mohr envelope curves for the in-plane and out-of-
plane situation in case of the chosen linear model.

Annex II: Calculation procedure

Fig. A5 presents a flow chart of the non-linear calculation procedure which may be called a
self-correcting secant modulus procedure. It predicts the progressive failure analysis of the
laminate and its successive degradation. The solution procedure of the non-linear analysis
aims to establish static equilibrium at each load step after material properties have been
changed. This includes the matrix weakening and the Birch-based pressure-dependent
stiffness correction. The procedure is repeated for each iteration until convergence is reached
or total failure. A correction of the fibre angle in accordance with a possible change of the
specimen’s geometry as a consequence of large strain behaviour is considered.
WWFE-II-Part A_dispatched Dec09 57 of 77 29/12/2009, 8:07
In the computation, for a small load increment, sometimes just one iteration step is needed in
a secant modulus procedure to consider stress-redistribution, that means, load from the
degrading matrix (matrix-dominated modes) is transferred to a fibre (fibre-dominated mode)
and increases the fibre effort.
Approach: When having reached Eff  1 , this total combined stress effort value is to be kept
in the further degradation procedure when the fibres take over above additional load as far as
the ‘fibre net’ allows for it. Thereby, also the residual stresses are reduced similar to the
situation with metallic materials where increasing non-linearity reduces stiffness and residual
stress. After the first critical IFF (another might follow) and after reaching Eff  1 the
mod e
associated mode stress effort Eff crit   eqmod e / R mod e (  eqmod e decreases in the softening
domain, R mod e is kept constant) is left as a decaying fracture energy portion (numerical
damping) in the interaction equation for the benefit of a stabile non-linear analysis.
It is to be noted that the mode's equivalent stress-strain curve is viewed to be identical with the
uni-axial stress-strain curve measured. This has the effect that an equivalent mode stress from
a multi-axial stress state may be lower but also higher than the value from the single mode
dominating stress alone, e.g.  1 or  21 . A different value for the desired secant modulus is
obtained.
In general, in the laminae of a laminate multi-axial states of stress are acting which impact
in the interaction domains more than one failure mode. Adjacent failure modes are commonly
affected but the impact depends on the size of the mode effort. One may pay some attention
to a better interaction of the modes in the stress and strain analysis in the following manner:
The secant moduli E2sec and G21sec are taken from the 2(2)-curves or the 21(21)-curve
not just at the stresses  eq ,  eq or  eq|| resulting from the stress and strain analysis. Their
values are taken in the 'hardening branch' at a little higher stress as follows
 eqmod,corre   eqmod e  f trf (25a)
with the trigger factor for hardening

f trf  m Eff Eff crit

mod e
, Eff  m ( Eff crit )   ( Eff others
mod e m mod e m
) (25b)
to consider a mode-combined degradation effort. By this approach slightly lower secant
moduli are provided for the next calculation loop; lower than those that would result without
the correction by the triggering approach. The controlling parameter is the ratio of ‘total stress
effort’ Eff to ‘critical mode stress effort’. Naturally, this triggering approach is already active
before the onset of a distinct IFF.
If the corrected equivalent mode stresses have reached their strength level (IFF), the strain-
softening branch begins and a differently rapid decrease of the averaged (smeared over the
micro-cracks) equivalent stress of the IFF modes will follow. In dependence of  eqmod e and by
applying Eq.(33) the secant modulus E sec (  eqmod e ) can be determined.

Annex III: Treatment of multiple non-linearities

A challenge is the practical treatment of the multiple nonlinearities: in-plane shear,
transverse compression, through-thickness compression (considering an orthotropic lamina,
not treated here), and strain-softening.

WWFE-II-Part A_dispatched Dec09 58 of 77 29/12/2009, 8:07

In comparison to Mises yielding one would need for each mode an equivalent stress-
equivalent strain relationship with a corresponding microcracking-dominated ‘degradation
flow rule’, and in addition, an interaction description of the full IFF modes’ stress-strain
surface for commonly considering the non-linearities. A fully accurate procedure of the non-
linear behaviour would require a treatment of the growing damage failure surface or quasi-
yield IFF surface. This must include a flow rule for this multi-fold ‘yield’ condition case
(compare [Rol07] for the constituent matrix). Of course, applying a yield potential requires its
proof by measuring the several multi-axial inelastic strains.
A similar problem case is still known from the sharp corners of the Tresca yield surface for a
discontinuous isotropic situation. The two normal vectors have a different direction and are
interacted in a combined manner.
In reality, an IFF mode is seldom becoming critical at the same time as a second one,
primarily one is acting. Hence, the analysis can be ‘reduced’ from industrial point of view to
just utilizing the full stress-strain curve of the critical mode. Its stress effort governs the
degradation triggering, and its iteration-corrected value is taken to regard the combined
degradation effect. One has to check the limit of applicability, of course.

Annex IV: Influence of p hyd on the material’s load carrying capacity

Hydrostatic pressure has an influence on the tensile, the compressive, and the shear
behaviour of the used polymer matrices. The size of this influence is different for the three
loadings. In order to get a comparative view for the matrix material (TC 1) and one UD
material (TC 5) the stress efforts of each mode are depicted in the graphs Fig. A6. The figure
visualizes the different influence size of some mode efforts. For fibre-parallel (longitudinal)
loading, an additionally applied (external) uni-axial compressive stress, the superposition of
p hyd causes a reduction of the loading, however.
In this context the question “Is the increase caused by the equivalent stress  eq or really by an
increase of the strength R ?” can be answered. Neglecting special effects as Birch, 2ndTg,
considered in the calculation of the above graphs, it can be concluded that the increase or also
decrease in the material’s load-carrying capacity under uni-axially loaded (uni-axially
stressed) material test specimens, superimposed by p hyd , can be mechanically dedicated to the
change of the mode stress effort or of its associated equivalent mode stress and not to the
associated mode strength. Remind: Eff mod e   eqmod e / R mod e , it’s an increase of the load-
carrying capacity but not an increase of strength.

WWFE-II-Part A_dispatched Dec09 59 of 77 29/12/2009, 8:07

 Fig. 1. Formulations of lamina stresses and invariants for UD material

Fig. 2. In-plane lamina stresses with definition of positive fibre orientation angle of the lamina
(inter-laminar stresses are not depicted here).

WWFE-II-Part A_dispatched Dec09 60 of 77 29/12/2009, 8:07

 Fig. 3a. 2nd Tg shift effect: Dependence of the matrix modulus on phyd . Assumption for PR319:
knee point (140 MPa at ratio 1.4), final point (600 MPa at ratio 1.75, see [Pae96]
Fig. 3b. Decay function f 2 ndTg due to 2ndTg shift. M  22 , a 1  0.0018 , a 2  1.20 , a 3  0.00092 .

Fig. 4. IFF-related stress-strain curves of a UD-lamina with strain-hardening (isolated lamina) branch
and assumed strain-softening branch (embedded lamina), see [Cun04]

WWFE-II-Part A_dispatched Dec09 61 of 77 29/12/2009, 8:07

 Table (1): Mechanical properties of the UD laminae

S2-
Fibre type IM7 T300 A-S
glass
E-Glass

PR- Epoxy
Matrix 8551-7
319 1
Epoxy2 MY750

Fibre volume fraction Vf (%) 60 60 60 60 60

Longitudinal modulus E1 (GPa) 165* 129 140* 52 45.6
Transverse modulus E2 (GPa) 8.4 5.6+ 10 19 16.2
Through-thickness modulus E3 (GPa) 8.4 5.6+ 10 19 16.2
In-plane shear modulus G12 (GPa) 5.6* 1.33+ 6* 6.7* 5.83*
Transverse shear modulus G13 (GPa) 5.6* 1.33+ 6* 6.7* 5.83*
Through-thickness shear modulus G23 (GPa) 2.8 1.86 3.35 6.7 5.7
Major Poisson's ratio 12 0.34 0.32 0.3 0.3 0.28
Major transverse Poisson's ratio 13 0.34 0.32 0.3 0.3 0.28
Through-thickness Poisson's ratio 23 0.5 0.5 0.49 0.42 0.4
Longitudinal tensile strength XT (MPa) 2560 1378 1990 1700 1280
Longitudinal compressive strength XC (MPa) 1590 950 1500 1150 800
Transverse tensile strength YT (MPa) 73 40 38 63 40
Transverse compressive strength YC (MPa) 185** 125** 150** 180** 145**
Through-thickness tensile strength ZT (MPa) 63 40 38 50 40
Through-thickness compressive strength ZC (MPa) 185** 125** 150** 180** 145**
In-plane shear strength S12 (MPa) 90** 97** 70** 72** 73**
Transverse shear strength S13 (MPa) 90** 97** 70** 72** 73**
Through-thickness shear strength S23 (MPa) 57 45 50 40 50
Longitudinal tensile failure strain 1T (%) 1.55 1.07 1.42 3.27 2.81
Longitudinal compressive failure strain 1C (%) 1.1 0.74 1.2 2.21 1.75
Transverse tensile failure strain 2T (%) 0.87 0.43 0.38 0.33 0.246
Transverse compressive failure strain 2C (%) 3.2 2.8 1.6 1.5 1.2
Transverse tensile failure strain 3T (%) 0.76 0.43 0.38 0.263 0.25
Through-thickness compressive failure strain 3C (%) 3.2 2.8 1.6 1.5 1.2
In-plane shear failure strain 12u (%) 5 8.6 3.5 4 4
Transverse shear failure strain 13u (%) 5 8.6 3.5 4 4
Through-thickness shear failure strain 23u (%) 2.1 1.5 1.5 0.59 0.88
Longitudinal thermal coefficient 1 (10 /C)
-6
-1 -1 -1 8.6 8.6
Transverse thermal coefficient 2 (10-6/C) 18 26 26 26.4 26.4
Through-thickness thermal coefficient 3 (10-6/C) 18 26 26 26.4 26.4
2
Energy release rates GIc, GIIc (J/m = 1N/m) 200 240, 1500
mixed (fracture mechanics) mode to be assumed
Stress free temperature (C) 177 120 120 120 120
1h 107°C, 2h 90°C,
2h 177°C, 1.5h 30°C,
curing cycle
at 1.8°- 2h 150°C
3°C/min at 1°C/min
* Initial modulus.
** Nonlinear behaviour and stress strain curves and data points are provided
+ These values are considered to be low, compared with typical data for the same material published somewhere else or quoted
by the manufacturers. We have not attempted to change them in order to facilitate a comparison with test data in Part B.

WWFE-II-Part A_dispatched Dec09 62 of 77 29/12/2009, 8:07

 Table (2). Mechanical properties of fibres

Fibre type IM7 T300 AS S2-glass E-Glass

Longitudinal modulus Ef1 (GPa) 276 231 231 87 74
Transverse modulus Ef2 (GPa) 19 15 15 87 74
Transverse modulus Ef3 (GPa) 19 15 15 87 74
In-plane shear modulus Gf12 (GPa) 27 15 15 36 30.8
Major Poisson's ratio f12 0.2 0.2 0.2 0.2 0.2
Major Poisson's ratio f13 0.2 0.2 0.2 0.2 0.2
Transverse shear modulus Gf23 (GPa) 7 7 7 36 30.8
Longitudinal tensile strength Xf1T (MPa) 5180 2500 3500 4590 2150
Longitudinal compressive strength Xf1C (MPa) 3200 2000 3000 2450 1450
Longitudinal tensile failure strain f1T (%) 1.87 1.086 1.086 5.27 2.905
Longitudinal compressive failure strain f1C (%) 1.16 0.869 0.869 2.82 1.959
Longitudinal thermal coefficient f1 (10 /C) -6
-0.4 -0.7 -0.7 5 4.9
Transverse thermal coefficient f2 (10 /C) -6
5.6 12 12 5 4.9
Through-thickness thermal coefficient f3 (10 /C) -6
5.6 12 12 5 4.9

Table (3). Mechanical properties of various matrices

8551-7 PR913 Epoxy Epoxy
Matrix type epoxy epoxy 1 2
MY750
Elastic Modulus Em (GPa) 4.08 0.95 * 3.2 3.2 3.35
Elastic Shear modulus Gm (GPa) 1.478 0.35 * 1.2 1.2 1.24
Elastic Poisson's ratio m 0.38 0.35 0.35 0.35 0.35
Tensile strength YmT (MPa) 99 70 85 73 80
Compressive strength YmC (MPa) 130 130 120 120 120
Shear strength Sm (MPa) 57 41 50 52 54  R
Tensile failure strain mT (%) 4.4 7.3 2.65 5 5?
Compressive failure strain mC (%) 9 13.6 3.75 5 5
Shear failure strain m (%) 5.1 11.5 - - -
Thermal expansion coefficient m (10 /C) -6
46.7 60 58 58 58
* These values are considered to be low, compared with typical data for the same material published somewhere else or quoted
by the manufacturers. We have not attempted to change them in order to facilitate a comparison with test data in Part B

Fig. 5. Normal stress vs. normal strain of three matrix materials

WWFE-II-Part A_dispatched Dec09 63 of 77 29/12/2009, 8:07

 Fig. 6. Shear stress-vs. shear strain of three matrix materials

Fig. 7. Fibre-parallel (longitudinal) stress  1   || vs. fibre parallel strain  1   f   || of the five
UD materials

WWFE-II-Part A_dispatched Dec09 64 of 77 29/12/2009, 8:07

 Fig. 8. Transverse stress vs. transverse strain of the five UD materials

Fig. 9. Shear stress vs. shear strain of the five UD materials

WWFE-II-Part A_dispatched Dec09 65 of 77 29/12/2009, 8:07

 Table (4) Details of the Test Cases (lay-up matched to the figures)

Test Laminate lay-

Material Description of Required Prediction
Case up
1 Resin MY750 epoxy x vs. z envelope with y = z
2 0° T300/PR319 12 vs. 2 envelope with 1 = 2 = 3 = -phyd
3 0° T300/PR319 12 (12) vs. 2 envelope with 1 = 2 = 3 = -phyd
4(a) 0° T300/PR319 shear stress strain curve 12-12 at phyd = 600 MPa
5 90° E-glass/MY750ep. 2 vs. 3 envelope with 1 = 3
6 0° S-glass/epoxy 1 vs. 3 envelope with 2 = 3
7 0° A-S carbon/ep. 1 vs. 3 envelope with 2 = 3
8 [35/-35/35/-35]ns E-glass/MY750. y vs. z envelope with x = z
9(b) [35/-35/35/-35]ns E-glass/MY750 stress-strain curves y-x , y-y at z = x = -100 MPa
10 [45/-45/90/0]ns IM7/8551-7 zx vs. z envelope with y = x = 0
11 [0/90/0/90]ns IM7/8551-7 zx vs. z envelope with y = x = 0
12 [0/90/0/90]ns IM7/8551-7 stress-strain curves z-z , z-x , z-y with y = x = 0

(a) Please first apply 1 =2 = 3 = -600 MPa to the lamina, then shear loading till final failure takes place.
(b) Please first apply y = z = x = -100 MPa and record the resulting strain values. Then increase the
stress y (beyond -100MPa) gradually till final failure takes place. Please plot the full stress-strain curves
(y-x,...).
(c) Please indicate, laminate TC: Case a (with thermal/residual stresses considered), case b (not considered).
(d) The number n of all sub-laminates of the laminates (TC8-12) is not specified in order not to undermine
the computational capabilities of some of the models. It may vary between 8 and 100 layers (plies).
(e) Please provide predictions describing the effect of thickness and number of plies on the failure of the laminates
(f) Laminates are balanced and symmetric.
(g) Layers (plies) have the same thickness. Layer thickness may vary from 2 to 30 mm.
Author’s remarks: In Table (4) the description of the laminate lay-up has been made compliant with
the provided sketches. The output will address the y-section of the laminate. The provided input data
set is based on specimens that represent so-called isolated UD laminae. Therefore, the in-situ effect
within an embedded lamina of a laminate is to be assumed.

WWFE-II-Part A_dispatched Dec09 66 of 77 29/12/2009, 8:07

 Fig. TC 1. Tri-axial failure state of MY750 epoxy resin matrix:    (  x , y   z , z ,0 ,0 ,0 )T .
a) 3D visualization in Lode coordinates with approaches in the cap as well as the SF-NF interaction
domain; b)2D failure curve. c) 3D compressive failure stress x vs. stress y (= z); d) Details of c).
R   ( Rt , Rc )T  ( 80 , 120 )T MPa , R  54 MPa . Assumed: Rtt  70 MPa, Rttt  60 MPa ,
bc  1.53 for  fp   47 . 2ndTg shift effect (dashed line) at  200 MPa using the weakening
c

standard data set: M  22 , a 1  0.0018 , a 2  1.20 , a 3  0.00092 .

WWFE-II-Part A_dispatched Dec09 67 of 77 29/12/2009, 8:07

 Fig. TC 2. Fracture stress 21 vs stress 2 (= 1 = 3 = -phyd) for a UD T300 carbon/ PR319ep

R  ( 1378 , 950 , 40 , 125, 97 )T MPa.  ||  0.32 . m = 2.8, b||  0.3 and 0,4.
E  ( 12900, 12900, 5600, 5600, 1330 )T MPa,  23  0.5 , ( G23  1860 MPa ), 2ndTg shift
(dashed line): M  22 , a 1  0.0018 , a 2  1.20 , a 3  0.00092 . Em0  950 MPa ,
 m  0.35 , Gm0  E m0 /( 2  2  m ) ,  fr   ( 1.07 , 0.74 , 0.43, 2.8 , 8.6 )T % . Data from curve
fit: n  ( , , , 7.0 , 7.5 )T , R  ( , , , 107.7 , 74.1 )
p0.2
T
MPa.

Fig. TC 2b. Closing of the failure curve and sensitivity study on Poisson’s ratio
monotonically increasing with hydrostatic pressure.

WWFE-II-Part A_dispatched Dec09 68 of 77 29/12/2009, 8:07

 Fig. TC 3. Failure shear strain  21 (  2   p hyd ) in dependence of an increasing phyd for a UD
T300 carbon/PR319 epoxy. Properties, see TC2.

Fig. TC 4. Shear stress- shear strain curve  21 (  21 ) for  hyd  0 , 600 MPa.
UD T300 carbon/PR319 epoxy. Properties, see TC 2.
nd
2 Tg shift (dashed line).  21fr ,600  15% ,  c0600
.2 ,21  125 MPa , G12600  1872 MPa ,
 21fr ,600  193 MPa , R||  97 MPa (8.6%), R||, 0.2  91MPa ,  21fr ,6002ndTg  165 MPa
(15%).

WWFE-II-Part A_dispatched Dec09 69 of 77 29/12/2009, 8:07

 Fig. TC 5. Tri-axial failure state of stress: 2 vs. longitudinal stress 1 (= 3)
for a UD E-glass/MY750epoxy.

R  (1280, 800, 40, 145, 73)T MPa .  ||  0.28 . m = 2.8, b  1.21 (bold line); for  cfp  50 ,
b  1.52 for 55 (thin line) M  22 , a 1  0.0018 , a 2  1.20 , a 3  0.00092 .
   ( 2.81, 1.75, 0.25, 1.2, 4 )
fr
T
% , E m0  3350 MPa ,  m  0.35 , G m0  E m 0 /( 2  2  m ) .
Data from fit: E  (  ,  ,16260 , 16200, 5188 )T MPa, n  ( , , 8.5 , 7.0 , 8.8 )T ,
a  ( ,  , 0.3, 2.17 , 8.0 ), b  (, , 0.013, 0.21, 0.87) T . With Birch, 2 Tg on moduli +
nd

strength, and curing stresses

WWFE-II-Part A_dispatched Dec09 70 of 77 29/12/2009, 8:07

 Fig. TC 6. Through-thickness stress 3 (= 2) vs. fibre-parallel stress 1 , UD S-glass/epoxy2.
R   ( 1700, 1150, 63, 180, 72 )T MPa ,  ||  0.3 . m = 2.8, b  1.21. 2ndTg shift (dashed line).
 
M  22 , a 1  0.0018 , a 2  1.20 , a 3  0.00092 .  fr  ( 2.81, 1.75 , 0.25, 1.2 , 4 )T % ,
E  (  ,  ,16260 , 16200, 5188 )T MPa , E m0  3350 MPa ,  m  0.35 ,
Gm0  E m0 /( 2  2  m ) . From fit: n  ( , , 8.5 , 7.0 , 8.8 )T , a  ( ,  , 0.3, 2.17 , 8.0 ),
b  ( ,  , 0.013, 0.21, 0.87 )T , R p0.2   ( ,  , 156 , 137 , 54.6 )T MPa. With Birch, 2 Tg on
nd

moduli + strength, and curing stresses

WWFE-II-Part A_dispatched Dec09 71 of 77 29/12/2009, 8:07

Fig. TC 7. Through-thickness stress 3 (= 2) vs. fibre-parallel stress 1. UD A-S carbon/epoxy1.
R   ( 1990, 1500, 38, 150, 70 )T MPa .  ||  0.3 . m = 2.8, b  1.21 . M  22 ,
a 1  0.0018 , a 2  1.20 , a 3  0.00092 .  fr   ( 2.81, 1.75, 0.25, 1.2 , 4 )T % ,  m  0.35 ,
E m0  3350 MPa , Gm0  E m0 /( 2  2  m ) . Data from fit: n  ( , , 8.5 , 7.0 , 8.8 )T ,
E  (  ,  ,16260 , 16200, 5188 )T MPa, a  ( ,  , 0.3, 2.17 , 8.0 ),
b  ( ,  , 0.013, 0.21, 0.87 )T , R p0.2   ( ,  , 156 , 137 , 54.6 )T MPa. With Birch, 2 Tg on
nd

moduli + strength, and curing stresses

WWFE-II-Part A_dispatched Dec09 72 of 77 29/12/2009, 8:07

 Fig. TC 8. Effect of the applied surface pressure  z   p (through-thickness stress) on the
size of the normal section force n y  ˆ y / t at fracture, with n x / t  ˆ x   z . UD lamina-
composed laminate [35/-35/35/-35]s, E-glass/MY750/epoxy. R   ( 1280, 800, 40 , 145, 73 )T MPa .

 ||  0.28 , m = 2.8, b  1.21 , b||  0.3 , t k  0.25 mm , t  2 mm (laminate thickness).
Computation paths for TC 9 indicated. With Birch, 2ndTg on moduli + strength, and curing stresses

Fig. TC 9. Stress-strain curves ˆ y ( ˆ x ) and ˆ y ( ˆ y ) . Monotonically loaded ˆ x  ˆ y   z

up to -110MPa. Then remain constant:  z  ˆ x  100 MPa , ˆ y  n y / t grows. Properties and
loading, see TC8. Resulting strains at p hyd  100 MPa : ˆ  ( 0.12 ,0.18 ,0.29 )T %. With Birch,
2ndTG on moduli + strength, and curing stresses

WWFE-II-Part A_dispatched Dec09 73 of 77 29/12/2009, 8:07

 Fig. TC 10. Applied section shear load-caused maximum-thickness failure shear stress zy vs.
applied through-thickness stress z . for a [45/-45/90/0]ns , carbon/epoxy laminate, IM7/8551-7.
 x   y  0 , E m0  4080 MPa , Gm0  E m0 /( 2  2  m ) , m  0.38 , G||  5600 MPa ;
R||0.2  53.8 , R   ( 2560, 1590, 73, 185, 90 )T MPa .  ||  0.34 . m = 2.8, b  1.21 ,
b||  0.3 . t k  0.25 mm . M  22 , a 1  0.0018 , a 2  1.20 , a 3  0.00092 ,
  ( 1.55, 1.1, 0.87 , 3.2, 5 )
fr
T
% , Data from mapping of provided data, Table 1:
n  (  , 
 , 2.7 , 7.0 , 5.5 ) ; E or G  (  ,  , 8836 , 9001, 5679 )T MPa,
T

a   (  ,  , 1.04 , 5.78, 10 ), b   (  ,  , 0.045, 0.56 , 1.09 )

soft soft
T
,
R   (  ,  , 129, 144, 53.8 ) MPa.
p 0.2
T

Fig. TC 11. Thickness failure shear stress  zy vs. applied through-thickness stress  z .
[0/90/0/90]ns carbon/epoxy laminate IM7/8551-7. ARCAN test rig, loaded by shear forces F and
pressure p . Properties, see TC10. (for orientation, the strength values Rt , R|| and Rc are
indicated). max  zy  max  23  max  21  1.5  F / A . With Birch, 2ndTg on moduli + strength, and
curing stresses

Fig. TC 12. Stress-strain curves caused by a through-thickness compressive stress  z for a

[0/90/0/90]ns carbon/epoxy laminate, carbon/epoxy laminate, IM7/8551-7.  x   y  0 .
Properties, see TC10. With Birch, 2ndTg on moduli + strength, and curing stresses

WWFE-II-Part A_dispatched Dec09 74 of 77 29/12/2009, 8:07

 Fig. A 1. Stresses at shear failure IFF3

Fig.A 2. Lamina stresses and Mohr stresses at shear failure IFF2 (wedge failure).

Fig. A 3. Linear IFF2 Approach and Validation by Test Data. Glass/epoxy, [VDI 97].
Friction angle: tan     

WWFE-II-Part A_dispatched Dec09 75 of 77 29/12/2009, 8:07

 Fig. A 4. Mohr-Coulomb failure envelopes  n1 (  n ) ,  nt (  n )

Fig. A 5. Non-linear calculation scheme of the self-correcting secant modulus method.

Matrix weakening is considered in the degradation box

WWFE-II-Part A_dispatched Dec09 76 of 77 29/12/2009, 8:07

 Fig. A 6. Influence of phyd on the mode-related load-carrying capacities.
(for F|| , Eff  1 shall indicate the negative effect)

WWFE-II-Part A_dispatched Dec09 77 of 77 29/12/2009, 8:07

View publication stats