1 st cosacnet meeting neil mccartney npl materials centre national physical laboratory, uk...
TRANSCRIPT
1st CoSACNet Meeting
Neil McCartney
NPL Materials Centre
National Physical Laboratory, UK
Southampton University , 30 January 2001
Design tools for composites
Definition of design ?
Selection of materials, geometry, loading modes and limits so that products meet specified performance criteria e.g.
deflections within specification failure loads in excess of maximum expected load during service avoidance of microstructural damage ( ply cracks / delaminations ) lifetimes ( cycles / time ) in excess of specification
Design is quantitative and based on mathematical models that adequately represent behaviour
semi-empirical / phenomenological relationships analytical formulae ( from Hooke’s law to complex analyses ) finite element or boundary element analysis
Modelling issues for composites
Reliable design procedures will be based on physical modelling
The availability of high performance computers will revolutionise the design of composite structures
Realistic complex models can be used for design of materials (Virtual Testing) and components
Models must be thoroughly validated and incorporated into easy-to-use design procedures
Life prediction, durability are exceedingly complex phenomena that are very difficult to model physically
Phenomenological approaches can be useful but are not usually reliable – e.g. failure criteria for composites
Conventional failure criteria
phenomenological in nature - no physics ! based on invariance requirements
applied to stress states for composite structures where no damage has been allowed for
not easily applied to environmentally or fatigue damaged composites
Tsai - Wu (1971)f F F Fi i i ij i j ijk i j k( ) 1
Physically based models are needed !
Design issues for composites
Materials design Fibre / matrix / volume fraction selection for UD laminates Orientation / thickness selection for plies in a laminate
Prediction of elastic constants Prediction of expansion coefficients ( thermal / moisture ) Types of loading
In-plane biaxial - through-thickness - shear Out-of-plane bending ( anticlastic bending )
Damage growth and property degradation Ply cracking – delamination – fibre fracture – interface debonding
Strength predictions Durability issues – fatigue – environmental exposure Delivery of design methods to users ( Software – Web )
Delivering design tools to users
Commercial systems LAP and CoDA
UK Composite Design Toolset ( DERA, AEAT, NPL, SER Systems Ltd )
Web-based design tools – E-mail communication
Smart design manuals
CoDA
for
Component and Composite Design AnalysisVersion 3
Graham D Sims and Bill Broughton
NPL Materials Centre
National Physical Laboratory, UK
CoDA
A commercially supported package
What does CoDA do ?
CoDA has four independent, but integrated, modules that have been validated experimentally
Panels, Beams, Laminates, Materials Synthesiser
Pre-preg laminates, chopped strand mat, sandwich panels
Implementation of failure criteria
CoDA can be used to undertake preliminary analysis of sub-components with Plate or Beam geometries
CoDA can also synthesise the properties of composite materials, laminates and sandwich structures, which can be used in a seamless manner within the design modules
CoDA
CoDA
CoDA
CoDA
CoDA
UK Composite Design Toolset
A collaboration between DERA, UKAEA & NPL
Integrated toolbox comprising modules that can exchange data & resultsPC008A/15A – DERA – Micromechanics, LPT, 2D/3DGENLAM – DERA – Non-linear LPT – thermal stresses
– scissoringCCSM – Cambridge, IC, DERA – Micromechanics + LPT
– unnotched & notched failurePREDICT – NPL – progressive damage modelling in laminatesLAMFAIL – UMIST, DERA progressive damage with empirical
model – nonlinear scissoring – complex load histories
A global data base of materials properties – links to other systems
PREDICT - Design objectives
Predict properties of UD composites from properties of fibre and matrix
Predict in-plane properties of general symmetric laminates
Predict initial formation of fully developed ply cracks in a general symmetric laminates subject to general in-plane loading and thermal residual stresses
( in fatigue loading designers will want to avoid damage )
Predict progressive degradation of thermo-elastic constants as a function of applied stress or strain ( strain softening rules needed for FEA analyses )
Predict effects on damage resistance of varying orientations and thicknesses of plies in a laminate
Predict effects of temperature changes on ply crack formation
( investigate thermal cracking during manufacture, or cooling )
Designing composites from fibre and matrix level
Predicting ply properties – validation
Predicting laminate properties
Delaying damage formation during loading
PREDICT
Ply geometry and location of ply cracks
2L
Ply 1Ply 2 Ply3Ply 4
- 45o 90o 0o0o 90o - 45o45o 45o
Quasi - isotropic laminates
Comparison of axial stress in crack plane (GRP)
0
0.1
0.2
0.3
0.4
0.5
0 0.5 1 1.5 2
x (mm)
Str
ess
(GP
a)
Model
FEA
FEA by Tong, Guild, Ogin & Smith (1997)
45o - 45o 90o 0o
Ply crack
Interfacial shear stress on 0o / 90o interface (GRP)
0
0.01
0.02
0.03
0.04
0 0.2 0.4 0.6 0.8 1y / L
Str
ess
(G
Pa)
Model
FEA
FEA by Tong, Guild, Ogin & Smith (1997)
0
5
10
15
20
25
30
0 0.5 1 1.5 2
Ply crack density (/mm)
Ax
ial
Yo
un
g's
mo
du
lus
(G
Pa
)
Experiment
Model
Experiment
Model
Experiment
Model
[02/902]s
[02/908]s
[02/904]s
GRP : Predicted from fibre/matrix properties
Experimental data : Lodiero & Broughton, NPL, 2000
T)()(E
)(
)(E)(E
)()( AT
A
A
At
A
a
T)()(E)(E
)(
)(E
)()( T
T
T
A
At
T
tT
Stress - strain relations for damaged laminate
is a label denoting the presence of some form of damage in thelaminate defined by a set of other parameters
Same form as those for an undamaged laminateValidity confirmed by accurate stress analysis
Degradation of properties of laminates
Damage parameter : 1)(E
E)(D
A
A1
k, k’ and k1 are easily calculated using CLT
Thermo-elastic constants for damaged laminates :
A
1
E
)(DQP)(P
11122
tTAT
t
A
a
A
A
tTA
k'kkkk'kk'kk'kk1:Q
EEEE
1
E
1
E
1:P
Continuum damage mechanics (CDM)
S
L
NMMM
O
QPPP
1 1 0
1 1 0
0 0 1 1
1
2
3
/ ( ) /
/ / ( )
/ ( )
E d E
E E d
d
A A A
A A T
A
Stress - strain relation = Swhere
Damage parameters d1, d2 and d3 are such that
0 1 1 2 3 d ii , ,
dE
Ed
E
EdA
A
T
T
A
A1 2 31 1 1
( ),
( ),
( )
Face view of crack growth using continuum model
0
1
2
3
4
5
0 5 10 15Normalised crack length
Nor
mal
ised
cra
ckin
g s
tres
s
Bridged crack theory
Long crack asymptote
Griffith crack
Master curve for ply cracking
Defect size
Str
ess
for
pro
pag
atio
n
Unstable growth
Stable growth
Design limit
Design limit is derived for long cracks
Design limit is exact if growth is stable
Design limit is a lower bound if growth is unstable
Predictions are pessimistic
Designs will be safe
Ply crack initiation criterion :
s
E E
s
A A
21 1 0
( )
( )
Criterion for progressive discrete ply crack formation :
s
E E
s i ni
i i
A i A i
21 1
11
1
0
( ) ( )
( ) ( )
...
Criteria for ply crack formation
( ) 0
s0 is value of s at ply crack closure
s k k
s k Tt T
o
o
'
1
Potential cracking sites
Ply crack locations
Potential cracking sites are evenly spaced
Ply cracks are non-uniformly spaced
Progressive cracking methodology
Allocation of fracture energies
0
0.01
0.02
0.03
0 50 100 150 200Fracture energy of potential cracking site
(J/m2)P
roba
bilit
y de
nsity
Mean fracture energy = 150 J/m2
Standard deviation = 15 J/m2
Master curve for triaxial loading
0
s s0
Gradient
E
DA
1( )
si s0
Inelasticstrain
Damage initiation stress Gradient of unloading line Enclosed area
Area
= 0
Key features :
Apply directly to other stressand temperature states
for which = 0 :
Tks
k'ks
1o
Tt
A popular approach of damage mechanics
HomogenisedCrackedlaminate
Homogeneousdamaged plyin laminate
Degraded properties modelled semi-empirically
Classical laminate analysis
Crackedlaminate
HomogenisedCrackedlaminate
Homogeneousdamaged plyin laminate
Approach of NPL model
Out - of - plane bending
Non-symmetrical laminates
Through-thickness thermal gradients
Major problem is dealing with anti-clastic bending
Model already exists for ply cracks subject to plane strain bending
Anti-clastic bending of UD ply
Plane strain bending of cracked [ 0 / 90 / 0 / 90 / 0 ] laminate
i = 1
i = 2
i = N+1
M M
0
90
0
90
0
x
y0
- Work in progress to predict ply crack formation -
Comparison of model with FEA
-8
-4
0
4
8
12
16
0 0.2 0.4 0.6 0.8 1 1.2
x (mm)
Axi
al s
tres
s (
GP
a)
FEA (Becker (1998)
Plane strain model
GRP laminate
Ply crack
90o ply0o ply 90o ply0o ply 0o ply
Modelling laminate failure
Physical modelling of damage modes
Cross-ply laminates subject to biaxial loading
Prediction of failure strain and strength
Biaxial loadingThermal stresses
Multiple plies
Modelling failure of cross-ply laminates
Effects of ply cracking alone on laminate properties are well understood
Ply cracking affects thermo-elastic properties (strain softening) but we need to address laminate failure issues
Tensile failure is determined principally by fibre fracture
Statistical nature of fibre failure must be included
Predicting the failure of cross-ply laminates is the first step
0o 90o 0o
Modelling failure of cross-ply laminates
Biaxial loadingThermal stresses
Static failure of
fibres
fibre/matrix debonding
frictional contact
shear yielding
Fibre matrix cellBiaxial stresses
Thermal stresses
0o 90o 0o
0
0.4
0.8
1.2
1.6
2
2.4
0 0.05 0.1 0.15Strain
Stre
ss (G
Pa)
0
20
40
60
80
100
120
0 0.05 0.1 0.15Strain
Mod
ulus
(GPa
)
0.88
0.92
0.96
1
1.04
0 0.05 0.1 0.15Strain
b / L
Mechanical behaviour of fibre / matrix cell
L - b is length of debond zone
Could substitute any other model
for which a look-up table can be constructed
Monte Carlo model for progressive failure in 0o plies
Parameters :
M, N, L Element length
Repeated runs ofa simulation to determine the statistical variabilityof performance
Critical fibre stress or strain ?
In fibre tests performance of fibre in tension can be characterised by : axial fibre stress at failure axial strain at failure f = Ef f
In a composite fibre subject to triaxial loading there are both loading & Poisson ratio effects on axial fibre strain
Assume fibre strength in a composite is governed by axial fibre failure strain consistent with concept of fibre axial strain controlling the stability
of fibre defects initiating tensile failure
Stress-strain behaviour
0
0.2
0.4
0.6
0.8
1
1.2
1.4
0 0.5 1 1.5 2 2.5
Axial strain %
Axi
al s
tres
s (G
Pa)
No fibre fracture
Fibre fracture
Failure
Degradation of axial modulus
62
63
64
65
66
67
68
69
0 0.2 0.4 0.6 0.8 1 1.2 1.4
Axial stress (GPa)
Axi
al m
odul
us (G
Pa)
No fibre fracture
Fibre fracture
Failure
Growth of ply cracks during axial loading
0
0.5
1
1.5
2
0 0.2 0.4 0.6 0.8 1 1.2 1.4
Axial stress (GPa)
Cra
ck
de
ns
ity
(/m
m)
No fibre fracture
Fibre fracture
Failure
The effect of fibre fractureon properties
CFRP (Vf = 0.6)
Virtual testing is defined as the combination of high quality models, high performance computing and a user - friendly interface
Will replicate many aspects of physical mechanical testing so that engineers do not need to learn a new vocabulary
Will allow material properties to be derived from more fundamental properties, leading to inventive materials design
It will be more than just a simulation, because extensive validation and testing will have taken place, resulting in a reliable replacement for some physical testing
Virtual Testing
Virtual testing of composites over the Internet
Web site address http://materials.npl.co.uk/
Virtual testing : Composite laminates
Developed at NPL, the Internet based system enables a materials designer to ‘create’ an entirely new material and to test it
Image taken from NPL Internet Laminate Damage Simulation
The system simulates the damage caused by cracking
as load increases and predicts the subsequent degrading of
material properties
Composite Laminate Testing
The user can generate design data for damaged composite laminates
Results taken from NPL Internet version of Laminate Damage Simulation
The Future : The era of simulation
Conclusions
Reliable design methods for composite materials will be based on physical models of behaviour. Key to reliability is rigorous validation of design methods
Physical models are complex in nature and not usually amenable to simple design rules ( sometimes there is no alternative )
Damage models have good potential for application in construction sector
( e.g. bridge strengthening with CFRP )
The implementation of physically based design methods in design offices will usually involve the use of computer based techniques : Specific software packages – LAP, CoDA, Toolset Web-based access to specific design packages, NPL demonstrator Web-based access to distributed software – networking ? Integration of design, optimisation, prototype and production simulation in
virtual manufacturing