4th International Conference on

Composites Testing and Model Identification

Progressive Failure Modelling of

Composite Laminates Containing
Tapered Holes
Chun Wanga, Andrew J. Gunnionb, and Adrian C. Orificib,c

aAir
Vehicles Division, DSTO, Australia
bCRC-ACS, Melbourne, Australia

cRMIT University, Melbourne, Australia

 2

OUTLINE

Objective
Recent developments
Experiments
Predictive Models
Results and Discussion
Summary
 3

Objective
Strength Prediction capability for composites
ƒ Prediction of residual strength after damage
ƒ Optimise damage cutout
ƒ Design of conformal antenna slots
ƒ Design and certification of repairs

Wire dipole
Slot in infinite
x-z ground
plane Callus (DSTO)
 4

Some Examples of Recent Developments

ƒ Abaqus damage model (2006)

ƒ Milestone: research→engineering
ƒ Lapczyk and Hurtado (2007): Fracture
ƒ Camanho et al (2007): 38.5% accuracy for energy

tension of bolted joint

ƒ Inherent-flaw fracture mechanics
ƒ IBOLT: method of choice at LM Aero
(Eisenmann and Rousseau 2004)
ƒ Empirical correction for countersunk holes
ƒ Continuum damage mechanics
ƒ Camanho et al (2007): 10.5% accuracy for OHT
ƒ Bogert et al (2006): 21.4% accuracy for slits

Camanho et al (2007)
 5

Experiments

Three types of specimens subjected to tension

ƒ Straight through-hole (diameter=6.35mm)
ƒ Straight through-hole (diameter=50mm)
ƒ Scarfed hole (diameter=50→200mm)
ƒ Stiff and soft laminates:
ƒ [40/40/20]%
ƒ [20/40/40]%
ƒ Stacking sequences
ƒ Panel: [45/90/-45/02]3S
ƒ OHT coupons:
ƒ [45/02/-45/90]3S
ƒ [-45/902/45/0]3S
 6

Experiments

Scarfed hole

2R
2R+2t/tanα
 7

Tensile strength of stiff laminates

100
Room Temperature
0.25" hole in
2.0" hole in
[45/90/-45/0/0]3S
80 [45/90/-45/0/0]3S
Strength (ksi)

60

40
Scarfed hole in
[45/90/-45/0/0]3S

20

0
0.00 0.05 0.10 0.15 0.20 0.25 0.30
Crosshead displacement (inches)
 8

Failure modes
 9

Strains in straight-hole panel

Some ductility
1200 35000
30-ply panel
[45/90/-45/0/0]3S
Load
30000
1000 Room Temperature
50mm hole diameter
500mm wide panel
25000
800 Tested 28 May 2008

Strain (microstrain)
20000
Load (kN)

hole edge gauges 5mm from edge inside hole - right

600
farfield gauge 135 mm from hole
inside hole - left
15000

400 gauge 3
10000

200
5000
gauge 5

0 0
0 1 2 3 4 5 6 7 8
Crosshead displacement (mm)
 10

Strain in scarfed-hole panel

Suspected premature failure of strain gauges

12000
Tip failure (1st group of zeros) Failure of 2nd group of zeros Failure at edge of scarf

10000

8000
Strain (me)

6000

4000
Gauge 1
Gauge 2
Gauge 3
2000 Gauge 4
Gauge 5

0
0 100 200 300 400 500 600 700 800 900
Load (kN)
 11

Fracture Mechanics Model

Critical flaw determined from fracture energy and ply percentage

n0G0 + n45G45 + n90G90
a=
Cπσ un2
− notched

σ un − notched
σ notched =
f (a / R)

ƒ Predicted strength is identical to cohesive zone model prediction

ƒ Independent of actual bridging law or the softening behaviour
ƒ Reported to be hole-size dependent
 12

Identification of Fracture Parameters

1 1
Assume: G45 = G0 G90 = G45
5 5
800.0

700.0

600.0
Predicted strength (MPa)

500.0

400.0
1 1
G45 =300.0
G0 G90 = G45
5 5
Experimental data (stiff laminate)
200.0
Experimental data (soft laminate)

100.0 Stiff laminate

Soft laminate
0.0
0 100 200 300 400 500
2
Fracture energy (kJ/m ) G0
 13

Predictions

ƒ Open hole tension strength of quasi-isotropic laminate

ƒ Data by Mollenhauer et al (CompTest 2006)
ƒ Model does not predict layup effects
800

600
Tensile Strength (MPa)

400

200
[45/0/-45/90]S (Mollenhauer et al)
[0/45/90/-45]S (Mollenhauer et al)
Fracture mechanics

0
0 5 10 15

Hole diameter (mm)

 14

Predictions

ƒ Large straight-hole and tapered hole

ƒ Significant under-prediction of strength
ƒ Need greater critical flaw size

800

Straight-hole
600
Tensile strength (MPa)

Scarfed hole

400

200

Stiff laminate [40/40/20]%

0
0 50 100 150 200 250

Hole diameter (mm)

 15

Abaqus Damage Model

2500
Strain-softening model:
Element=0.1mm
ƒ Bazant’s crack band model 2000
Element=0.4mm

ƒ Best for square elements Element=0.6mm

Stress (MPa)
1500

1000
ƒ Shell elements: all plies have
500
identical strains at any time.
ƒ Scarfed region is modelled as 0
0 0.5 1 1.5 2
multi-stepped (one step per ply).
Strain
No bending.

Issues:
ƒ Mesh refinement
ƒ Identification of fracture energies
ƒ Predictions of through-thickness
geometry variation
 16

Mesh refinement

Straight-hole of 6.35mm diameter

Relative insensitivity to mesh refinement
750.0
Predicted OHT strength (MPa)

500.0

250.0

Stiff laminate
Abaqus/Explicit
Experimental value
0.0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Element size (mm)
 17

Identification of Fracture Energies

ƒ Ply fracture energies depend on solver

ƒ Consistency between two stacking sequences
800.0

Stiff laminate
700.0

600.0
Predicted strength (MPa)

500.0 Soft laminate

.

400.0

Hole diameter=6.35mm (0.25")

300.0 Smallest element=0.6mm
Experimental data of stiff laminate
200.0 Experimental data of soft laminate
Explicit
Explicit
100.0
Implicit
Implicit
0.0
0 50 100 150 200 250 300
2
Ply fracture energy G0T (kJ/mm )

G0C = 0.8G0T G90T = G90C = 0.003G0T

 18

Explicit versus Implicit

ƒ Implicit code suffered convergence problems and required damping

ƒ Explicit code more robust, damping not required, but requires large time
increments to avoid inertia effect
ρ
Fundamental resonant (in-plane) period T0 = 2 L
E
ρ
Time increment: ∆T = h L
E

Total time: many times of the fundamental period

T0 2 L
Number of increments: N ≈ N
∆T h
• Independent of density
• Element size h needs to be small fraction of critical flaw size (e.g., h = 0.1 a)
• N=?
 19

Abaqus/Explicit

ƒ Run time versus error due to inertia effect

ƒ Error less than 2% requires loading duration about 30 times the fundamental period
¾ Any disturbance resulting from damage progression reverberates 60 times
0.08

Abaqus/Explicit

0.06
Error

0.04

0.02

0.00
0 20 40 60 80 100

Ratio of run time to fundamental period T/T0

 20

Failure path

ƒ Straight hole

0 degree ply

Crack inclined at 11 degrees Crack inclined at 26 degrees

 21

Failure path
0 degree ply

ƒ Scarfed hole

Crack inclined at 11° Crack inclined at 22°

45 degree ply 90 degree ply

 22

Scarfed and straight-hole panels

ƒ Abaqus model provided an improvement but under-predicted strength by

20% and the angle of fracture path.
800

600 Abaqus/explicit prediction

Tensile strength (MPa)

400

Fracture mechanics prediction

200

Stiff laminate [40/40/20]%

0
0 50 100 150 200 250

Hole diameter (mm)

 23

Quasi-isotropic laminates

ƒ Over-prediction of strength for large holes

ƒ Using fracture energies “backed-out” from stiff laminate data
ƒ Stacking sequence effect not predicted

800

Quasi-isotropic laminate
700
Tensile Strength (MPa)

600 ±5% error band

500

400 [45/0/-45/90]S (Mollenhauer et al)

[0/45/90/-45]S (Mollenhauer et al)
Abaqus/Explicit model

300
0 5 10 15
Hole=2.54mm (element=0.2mm)
Hole diameter (mm)
 24

Damage Initiation Model

Difference in fracture path due to incorrect damage initiation model?

ƒ Need alternative failure criterion to model off-axis plies

Hole=12.7mm (G0t=160 kJ/m2)

45 deg ply 0 deg ply

(element=0.2mm) (element=0.2mm)
 25

Alternative Failure Criteria

ƒ Modified strain-invariant (Wang, C.H., Chapter 8, Multi-scale Modelling of

Composite Material Systems, 2005)

ε vM
(f)
≥ ε c( f )
ƒ Fibre tensile fracture (shear failure)
σ 1( f ) ≤ σ c( f )
ƒ Fibre compression failure (micro-buckling)

ƒ Matrix shear failure

ƒ Matrix dilatation fracture

 26
140

Stress-invariant theory 120

DFVLR tension
DFVLR compression
MBB test
Prediction

In-plane Shear Stress (MPa)

100

80

ƒ Comparison with published data 60

(Wang 2005) 40

20

0 T300/BSL914C
(00 unidirectional lamina)

-1500 -1000 -500 0 500 1000 1500 2000

Longitudinal Stress (MPa)

1000 120
E-Glass/LY556
Experimental data
(00 unidirectional lamina)
Initial failure
Final failure 100
Experimental data
500 Prediction

80

τxy (MPa)
σyy (MPa)

0 60

40

-500

20
AS4/3501-6
(0/90/45/-45)
-1000 0
-1000 -500 0 500 1000 -200 -150 -100 -50 0 50

σxx (MPa) σyy (MPa)

 27

Damage Progression Model

ƒ Predicted maximum strain is less than measurement

ƒ Stress-softening law may need modification
0.030

0.025
Maximum strain at hole edge

0.020

0.015

0.010
Straight hole (50mm dia)

0.005 Abaqus/Explicit
Strain gauge 1
Strain gauge 2
0.000
0.000 0.002 0.004 0.006 0.008

Gross strain
 28

Conclusions

ƒ Fracture mechanics (inherent-flaw) showed promises at dealing

with stacking sequence, but failed to predict effects of hole size
and through-thickness tapering.
ƒ Abaqus damage model under-predicted strength of cutouts larger
than those in calibration coupons.
ƒ Comparison of prediction with experimental data suggests
alternative damage initiation model and damage progression
model.
ƒ Optimisation techniques may be required to back-out material
properties.
ƒ Improved solution method needs to be developed to improve
computational efficiency.