Stiffness and Failure Modeling of Discontinuous Long Fiber Composites
Presented By: Moncef Salmi September 2014
Agenda
3 Copyright 2014 e-Xstream engineering
Agenda
• Description of DLF Composites • Problem & challenges • Proposed Modeling Solution • DOE Study of the Microstructure
Influence on the DLF Properties • Conclusion & perspectives • References
Description of DLF Composites Introduction & Typical Applications
5 Copyright 2014 e-Xstream engineering
What are DLF Composites ?
• DLF = Discontinuous Long Fiber composites • Compression molding process = Up to 100% of inclusions or strands • Specificities : adaptable for complex geometries, fast production /
high through-put, high variability of the mechanical properties
[1]
6 Copyright 2014 e-Xstream engineering
Typical Applications: Aero, Sports & Leisure
HexMC window frame
Different DLF aero parts: Brackets, Fittings, Gussets, Clips/Cleats, Pans
[3]
[1]
Sports & leisure
[1]
7 Copyright 2014 e-Xstream engineering
Typical Applications: Automotive
Door hinge
High performance cars suspension arms
[4] [5]
Problem & Challenges
9 Copyright 2014 e-Xstream engineering
Problem & Challenges
• Problem – DLF composites show a varying microstructure
• Not well understood prior to manufacturing parts – Macroscopic properties (stiffnesses & strengths) highly depend on the underlying
microstructure • The varying microstructure explains the varying macroscopic properties
The first building block consists in modeling the relationship between the DLF microstructures and the corresponding macroscopic properties.
The ultimate objective is to simulate the structural behavior of DLF parts based on composite properties numerically computed via DOEs at RVE level that will deliver a complete understanding of the DLF properties (mean, variance, A&B values...).
• Technical challenges to develop a specific DLF modeling tool – Handle up to 100% volume fraction (VF) – Be able to reproduce the key characteristics of DLF composites
• Mean orientation of the strands • Strands entanglement • Damage and failure occurring at the interfaces between the strands
Proposed Modeling Solution
11 Copyright 2014 e-Xstream engineering
DLF characteristics
Detailed modeling of the microstructure using RVEs
Automated execution of DOEs at RVE level Numerical characterization of DLF composites covering various microstructures
Assignation of the measured properties on FEMs
Numerical analysis of structural DLF part. Use DLF properties from DOE characterization. Statistical FEA available (mean, A & B basis)
Global Modeling Strategy FEM decomposition
and property assignment
Focus of this presentation
12 Copyright 2014 e-Xstream engineering
Digimat DLF Modeling solution at RVE Level
• A voxel based modeling method was developed. It • enables filling the RVE with up to 100% chopped tapes • is quick and robust both for geometry and mesh generation • reproduces physical phenomenon such as entanglement
between tapes
13 Copyright 2014 e-Xstream engineering
Digimat DLF Modeling solution at RVE Level
• Numerical features – Voxel mesh – Stacking algorithm – Probabilistic laws – FEA interfaces – User interface – Post processing tools
Stacking algorithm Voxel mesh
Full field response Post-processing tools
14 Copyright 2014 e-Xstream engineering
Digimat DLF Modeling solution at RVE Level
• Physical features – RVE dimensions – Chip dimensions – Orientation – Waviness & entanglement – Additional phases
• e.g. Resin rich areas – Failure
Entanglement
Failure • Cohesive elements
• Resin rich pokets
Preferential direction
Random alignment
Strand Orientation
15 Copyright 2014 e-Xstream engineering
Example of DLF Modeling / CPU Cost
DOE Study of the Microstructure Influence on the DLF Properties
17 Copyright 2014 e-Xstream engineering
DLF Characteristics in a Structural Component In a structural component, • The strands close to the part’s surfaces/edges may show a preferential tape orientation
• The strands away from the part’s surfaces/edges may show some in-plane randomness
• Resin rich pockets may exist in different locations of the suspension arm Different local volume fractions of strands
[5]
OT = (0.5,0.5), VF = 100% OT = (0.7,0.3), VF = 100%
OT = (0.5,0.5), VF = 90% OT = (0.5,0.5), VF = 95%
18 Copyright 2014 e-Xstream engineering
DOE Study
• Description
– Part of interest: High performance car suspension arm [5]
– Work achieved: • DOE study of the microstructural influence over the DLF stiffnesses and strengths • Objetive of the study:
– Analyze various local DLF behaviors in a structural part, which relate to specific local microstructures
– Evaluate the most influencing microstructural parameters • Parameters covered in the sensitivity analysis
– In-plane orientation of the strands – Volume fraction (VF) of the strands
– Strand: Carbon / Epoxy with 45% of fillers, linear behavior
– Next steps • Assign DOE properties on the structural FEM • Evaluate the global failure of the structural part from a macroscopic failure criteria
identified via the RVE studies • Evaluate the local failure mode by using a global/local approach with sub modeling
19 Copyright 2014 e-Xstream engineering
• Several techniques are proposed
– Model Resin Rich Zones (RRZ) at the tips of the strands • Record the stress evolution in those areas and trigger DLF failure from an identified
threshold in the RRZ
– Model strand interfaces and strength via • Classic solid elements • Cohesive zone elements
Strength Modeling Solutions at RVE Level
20 Copyright 2014 e-Xstream engineering
• Reference experimental data: – c11 = c22 = 38 GPa [6],
assumed to correspond to an in-plane random orientation and a 100% volume fraction
• Increasing the in-plane tape alignment – Increases the axial-1 stiffness while decreasing the transverse-2 stiffness – Does not significantly impact the out-of-plane stiffness, nor the in-plane shear stiffness
0.00E+00
1.00E+10
2.00E+10
3.00E+10
4.00E+10
5.00E+10
6.00E+10
7.00E+10
8.00E+10
0.45 0.5 0.55 0.6 0.65 0.7 0.75
Stiff
ness
(Pa)
OT(A11)
c11
c22
c33
c12
c13
c23
Results: Study of Stiffness vs. Strand Orientation
21 Copyright 2014 e-Xstream engineering
Results: Study of Strength vs. Strand Orientation
• Reference experimental data: – s11 = s22 = 300 MPa [6],
assumed to correspond to an in-plane random orientation and a 100% volume fraction
• The resin rich pocket method is used for evaluating the DLF strengths • Increasing the in-plane tape alignment
– Increases the axial-1 strength while decreasing the transverse-2 strength – Does not significantly impact the out-of-plane stiffness, nor the in-plane shear stiffness
0.00E+00
1.00E+02
2.00E+02
3.00E+02
4.00E+02
5.00E+02
6.00E+02
7.00E+02
0.45 0.5 0.55 0.6 0.65 0.7 0.75
Stre
ngth
(MPa
)
OT(A11)
s11
s22
s33
s12
s13
s23
22 Copyright 2014 e-Xstream engineering
Results: Study of Stiffness vs. Volume Fraction
0.00E+00
5.00E+09
1.00E+10
1.50E+10
2.00E+10
2.50E+10
3.00E+10
3.50E+10
4.00E+10
0.88 0.9 0.92 0.94 0.96 0.98 1 1.02
Stiff
ness
(Pa)
VF
c11
c22
c33
c12
c13
c23
• Reference experimental data: – c11 = c22 = 38 GPa [6],
assumed to correspond to an in-plane random orientation and a 100% volume fraction
• Inserting resin rich areas in the DLF, i.e. reducing the strand volume fraction, decreases all the stiffness properties
– In absolute values, more obvious on the c11 & c22 stiffnesses – Drop of ~10 to 25% on all stiffness components
23 Copyright 2014 e-Xstream engineering
Results: Study of Strength vs. Volume Fraction
0.00E+00
5.00E+01
1.00E+02
1.50E+02
2.00E+02
2.50E+02
3.00E+02
3.50E+02
0.88 0.9 0.92 0.94 0.96 0.98 1 1.02
Stre
ngth
(MPa
)
VF
s11
s22
s33
s12
s13
s23
• Reference experimental data: – s11 = s22 = 300 MPa [6],
assumed to correspond to an in-plane random orientation and a 100% volume fraction
• The resin rich pocket method is used for evaluating the DLF strengths • Inserting resin rich areas in the DLF, i.e. reducing the strand volume fraction,
decreases all the strength properties – Drop of ~5 to 25% on all strength components
Conclusion & Perspectives
25 Copyright 2014 e-Xstream engineering
• A global strategy is proposed to model DLF RVEs and DLF parts.
• A voxel method combined to a stacking algorithm is adopted for generating RVEs of DLF composites representative of real-life microstructures.
• A DOE / sensitivity analysis of the DLF properties for different in-plane strand orientation and different volume fractions was presented
– Measures quantitatively the property variation as a function of the underlying microstructure
• One strength modeling strategy has been tested so far, others will be evaluated to tackle the behavior at the interfaces between strands.
• The proposed RVE modeling solution is powerful as it can cover all of those properties: Stiffness, strength, thermal conductivity, CTE, cure shrinkage coefficients, moisture diffusion coefficients, moisture expansion coefficients…
• Next steps: – Improve the strength evaluation via other modeling techniques such as cohesive zone modeling – Automate the execution of DOEs / sensitivity analyses – Automate the assignation of the identified properties onto FE models to achieve FEA accounting for local
microstructural characteristics – Introduce a notion of statistics in the property assignment process – Extend process simulation software to compression molding of DLF parts
Conclusion & Perspectives
References
27 Copyright 2014 e-Xstream engineering
References
[1] Hexcel ATMAS project, 2009 [2] JEC innovation report, June 2012 [3] Characterisation of discontinuous carbon fiber/epoxy systems for
aeorospace applications, part1, Feraboli et al [4] High performance TPC: Rotorcraft door hinge, Clemens Dransfeld, IFK [5] Lamborghini « forged composite ®» technology for the suspension arms of
the Sesto Elemento, Feraboli et al [6] www.hexcel.com