D. Huang (Ph.D)1 , F. Abdi (Ph.D)1, S. DorMohammadi (Ph.D)1, M. Lee (Ph.D)1, H.K. Baid (Ph.D)1, Y. Song2, U. Gandhi (Ph.D)2

Crush Simulation of Compression Modeled Chopped Fiber Box Section by a De-homogenized Multi Scale Computational

Methodology

1

1AlphaSTAR Corporation, Long Beach, CA 90804And

2Toyota Corporation Technical Center, Ann Arbor Michigan, USA

ADVANCES INTHERMOPLASTIC COMPOSITES

SPE Automotive CompositesConference & Exhibition (ACCE),

6-8 September 2017Detroit, Michigan

Agenda

• Motivation

• Methodology: Chopped Fiber FE Analysis Process Flowchart•De-Homogenized vs. Homogenized Approach

• Case Study: Toyota Crush Tube Analysis• Multi-Scale Material Modeling

• Injection Molding: flow, cross flow, 3 point bend Validation

2

Validation • Compression Molding: flow, cross flow, 3 point bend

Validation • Orientation Tensor Mapping and Implementation•Multi-Scale Progressive Failure Analysis (MS-PFA)

• Conclusion

Motivation

Approach: Crush modeling of chopped fiber composite crushed tube was testvalidated using GENOA software in 3 distinct integrated steps:

a) Characterizing Material Properties - of composite materials composed ofchopped fibers using MCQ-Chopped and validating against Toyota

Objective: Qualify De-homogenized Multi-scale Crush modeling with tests using abuilding block validation Strategy

Problem: Currently FEM and/or Homogenized Multi-scale Crush modeling can notproduce accurately test observed Load – displacement (L-D), and acceleration- time

3

chopped fibers using MCQ-Chopped and validating against ToyotaPolypropylene (GF-40) coupon test data

b) Mapping and Transformation - of statistical average tensor orientation fromunstructured Moldex3D detailed model to LS-DYNA FE solver• GENOA platform software algorithm was used to perform 3D models data

management and visualize the mapping error between two dissimilar meshes.c) De-Homogenized Multi-Scale Progressive Failure Dynamic Analysis (MS-

PFDA)– Expected Output: L-D curve, damage and fracture evolution, contributing failure

mechanisms

De-Homogenized vs. Homogenized Approach

Multi-Scale Modeling of composite constituents fiber, matrix, and interface

Recognizing Effect of Defects agglomeration,

fiber waviness,

interphase,

resin rich,

Ply Material Properties

De-Homogenized Homogenized

HomogenizedHomogenizedHomogenized

Advantagesor

4

resin rich,

void shape/size

Multi-Scale Nano-micro Damage mechanics

Design Parameters Variation fiber length, fiber shape

•Micromechanics•Reverse Engineering

HomogenizedHomogenizedHomogenized

Fiber Matrix Interphase

Material Modeling: De-Homogenized Orientation Elastomer: Determine Angle Orientation Through Thickness vs. Test

Orientation Tensor through Thickness Orientation Angle through Thickness

5

Ref: Galib H. Abumeri, M. Lee, “A Computational Simulation System for Predicting Performance of Chopped Fibers Reinforced Polymer Composites”. ERMR-2006-Elastomer-Reno

Test Measured Orientation

MCO Chopped stress-strain curves CRGF15 (7.4vol%) vs. experimental data

Crush Test Set UpCompression Molding

Impact Crush Test Set-up

6

Final Impact Test Condition of Crushed Tube Part

Orientation Tensor Mapping (OTM)Flow Process of OTM to simple structural mesh

Donor Mesh

Un-Structural Donor Mesh

Input• Moldex3D generated

shell/solid un-structural mesh in GENOA/ABAQUS/LS-DYNA format

• Moldflow generated shell/solid un-structural mesh in GENOA/ABAQUS/LS-

Structural Receiver Mesh

Input• Finite Element generated

structural mesh in ABAQUS/LS-DYNA/ GENOA format

Output• Second order orientation

tensor values for shell/solid structural

Step 1Perform Mesh Mapping of Tensors

MCQ Material Model

Input• Aligned layer ply properties

from MCQ-Chopped software (stiffness, strength and non-linear ply stress-strain curve)

• Fiber/matrix properties from MCQ-Composite software (stiffness, strength and non-linear fiber/matrix stress-strain curves)

Final FE ModelsOutput

• High/Low (solid/shell) Fidelity FE mesh

• Ply lay up orientation through thickness defined for every element

• Properties defied either in terms of ply or fiber/matrix (de-

Step 2Perform Transform of Orientation Materials

and Layup

Mapped Receiver Mesh

7

GENOA/ABAQUS/LS-DYNA format

• Second order either nodal or element based orientation tensor components

shell/solid structural mesh

strain curves) or fiber/matrix (de-homogenized approach)

• Material non-linearity can be also included

• Models can be exported in GENOA/ABAQUS/LS-DYNA format

FE Low/High Fidelity ModelMapped Orientation Tensor Receiver

Methodology: Mapping and TransformationDetermine Ply Angle Through Thickness – De-Homogenization Approach

Orientation Through Thickness for Each Element2 mm Laminate PART

•Mapping from Un- structured mesh to structured mesh• using 2nd order orientation tensor (statistical stiffness averaging ) from Moldex3D or Moldflow

•Determine effective Chopped fiber orientation through-the-thickness•Step 1: Obtain oriented stiffness from aligned layered stiffness properties

• Use 4th order tensor transformation •Step 2: Generate layup using aligned layer ply to satisfies Oriented E11/E22 within threshold

8

Orientation Through Thickness for Each Element2 mm Laminate PARTOrientation Tensor Mapping

Compression Molding: Crush Tube Analysis

Load Displacement Curves

De-Homogenization Approach: Simulation results matches well with testCoupon Stress-Strain Curves

Flow direction Cross Flow Direction

9

Damaged Part at 40 (s)Acceleration Vs. Time

Injection Molding: 3 Point Bending ValidationComparison of Load vs. Displacement from Test & GENOA and ABAQUS Simulation

Load-Displacement Validation Curve

No

rma

lize

d L

oa

d

Flow-Test Cross-Flow-Test

Flow-MCQ-GENOA Cross-Flow-MCQ-GENOA

Coupon level damage (red) and damage type (index) from GENOA GUI

Through-thickness damage from GENOA GUI

10

Ref: H.K. Baid, F. Abdi, M. C. Lee, Uday Vaidya, “Chopped Fiber Composite Progressive Failure Model under Service Loading”, SAMPE 2015

0 2 4 6 8 10N

orm

aliz

ed

Lo

ad

Displacement [mm]

Injection Molding - Chopped FiberDe-Homogenization Approach with LS-DYNA, Thermoplastic (Wt = 40%)

Load Displacement CurvesExplicit chopped fiber crush tube simulation

30 (ms) 40 (ms) Deformation

11

Acceleration Vs. Time

Coupon Stress-Strain CurvesFlow direction Cross Flow Direction

Deformation Vs. Time

Ref: Frank Abdi, Saber DorMohammadi, Raghuram Mandapati, Harsh. K. Baid, Mike Lee, Umesh Gandhi, “Impact Crush Modeling of Chopped Fiber Reinforced Polymers', Michigan State University in East Lansing, Michigan from Sept. 28-30, 2015.

Tensor Orientation Prediction vs TestTest

A11

A22

A11

A22

Prediction

Schematic of Specimen

Prediction of Angle-Thickness

12

A33A33

Prediction of Angle-Thickness

Building Block Validation Strategy

1. Moldex3D Orientation Tensor –• Mapping/transformation of stiffness orientation (stiffness) from solid

13

• Mapping/transformation of stiffness orientation (stiffness) from solid(unstructured) to shell (structured) mesh.

2. Material Characterization and Qualification –• Multi-Objective Optimization performed to match coupon strength and stiffness

tests in flow and cross flow directions.• Homogenized and de-homogenized random chopped fiber properties were

generated: a) stiffness, b) strength; c) Poisson's ratio.3. Finite Element Model Generation and Analysis –

• FE model of crush tube (LS DYNA FEM single layered mesh)4. Multi-Scale Progressive Failure Analysis

• stress, strain, displacement,• Damage and fracture evolution: when, where, and why damage/fracture

MCQ Chopped Fiber Flow Process

Particle Shape & Aspect Ratio

Matrix/Ply NonlinearityObtained from Material and Aligned layer non-linearity Input

Chopped Tensor Orientation Through Thickness

Manufacturing Defects

Fiber Waviness Void Shape

AgglomerationInterphase

Vendor provided constituent Material Properties

Elastic Properties(1) Stiffness

5 ASTM Tests Results In – Non Linearity Out

0.7

0.8

0.9

1.0

Orie

ntat

ion

Test-A11 Test-A22 Test-A33MCQ-A11 MCQ-A22 MCQ-A33

14

Test Validation: Progressive Failure Design Failure EnvelopeMaterial Uncertainty

Chopped Mechanics

Through Thickness (1) Stiffness(2) Strength

0102030405060708090

100

0.00 0.01 0.02 0.03 0.04

Stre

ss [M

Pa]

Strain [mm/mm]

Test-Flow Test-45-Deg Test-Cross-FlowMCQ-Flow MCQ-45-Deg MCQ-Cross-Flow

SIG

YY

SIGXX

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

Orie

ntat

ion

Normalized Thickness [z/H]

Aligned Layer Nonlinearity

OUTPUTReverse Engineered Aligned Layer SS Curve

INPUTFlow SS Curve

Obtain reverse engineered aligned layer SS curve•Use Flow SS Curve from Test and Reverse Engine aliened Layer Stress-Strain curve

1515

PREDICTCross Flow SS Curve

Fiber Properties Matrix Properties

Particle Properties and Fabrication Variables

Input• Material Type: Poly Propylene, with 40% Wt Glass fiber•Fiber/Matrix Properties,• Particle Properties and Fabrication Variables • Orientation Distribution

Material Characterization of Chopped FiberInput: MCQ Analytical Model (No-FEM)

1616

Output: Determine Aligned Layer Mechanical Properties

Modulus Poisson’s Ratio Strength

Material Characterization of Chopped Fiber

17

Coupon Flow/Cross Flow (S-S curve)

Compression Molding: Material Characterization (MCQ)

Prediction Vs. test (Un-notched Coupon)Coupon Flow Direction

Damage Through-the-Thickness

18

Coupon Cross Flow DirectionDamage Through-the-Thickness

Compression Molding: 3 Point Bending ValidationMS-PFA (GENOA+ABAQUS Solver)

3 Point Bending (L-D) Flow Damage Types

19

Overall Damage location for Flow

Damage Index Cross Flow Damage Types

1. Input: Linear (Ply, Fiber/Matrix)2. Input: Non-Linear (Ply, Fiber/Matrix)3. Elem Removal Solid/Shell

• (Partial laminate damage)4. Plug In: UMAT/VUMAT

• LS_DYNA• ABAQUS (Implicit, Explicit)• Multi-Processor (SMP, MPP)

Capabilities & Model Assumptions

Capabilities1. Strain Rate Effect2. Linear: Fiber/Matrix or Ply Input3. Elemt Removal Criteria: Partial Laminate Damage (Ply Damage)4. 2 Level De-Homogenized Approach:

a) Fiber/Matrix Levelb) Ply Level to Orientation Tensor Mapping

Model Assumption

20

• Multi-Processor (SMP, MPP)5. Tensor Orientation Mapping (solid, shelll)

• Mold Flow, MoldeX6. Tensor Orientation Prediction (from MCQ)7. Prediction of Dehomogenized properties

• Orientation, and thickness

Material UMAT Options (Fiber Matrix or Ply) with FEM

35

Compression Molding (L-D) Injection Molding (L-D)25

Damage Index shown with LS-PREPOST as History Variable (0 means no damage)

Time = 8msTime = 0.6ms

21

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Displacement (mm)

Impa

ct F

orce

(KN

)

Test-Y13019

Test-Y13018

Test-Y13017

Test-Y13013

Test-Y13012

LSDYNA-GENOA 2017

LSDYNA-GENOA 2017 FM 0

5

10

15

20

25

0.00 0.05 0.10 0.15 0.20 0.25 0.30

Displacement (m)

Load

(KN

)

TEST 1

TEST 2

LSDYNA-GENOA 2017PLYLSDYNA-GENOA 2017FM

Time = 4 ms

GENOA+LS-DYNA

Damage Pattern & Failure type – GENOA GUI

All Damage, Contributing Failure Mechanisms

22

Methodology

23

Chopped Fiber FE Analysis Process Flowchart

Static and Fatigue Material Calibration and Validation

Fiber Orientation/Ply Thickness Determination

Static/Fatigue/Impact

Calibrated Fiber/Matrix/Ply Properties

Orientation Tensor MappingAnd

De-Homogenized Element Material Modeling

Donor Mesh, Donor Orientation Tensor

and Receiver Mesh

24

FE ModelStatic, Fatigue, Impact

Static/Fatigue/Impact Loading conditions

and boundary conditions

Multi-Scale Progressive Failure Analysis (MS-PFA)

Output Detailed Damage/Fracture Evolution

Displacement, Stress, Strain etc. contour plots

Material Modeling

Orientation Tension Mapping

Finite Element Analysis

Orientation Distribution Determination (ODD)Obtain reverse engineered angle orientation & thickness

Input: ODD Setup

Output: Equivalent Laminate Layup

This equivalent laminate layup is generated assuming that E11 and E22 obtained from this laminate layup will be within 5 % of the initial

values assumed

25

Output: Equivalent Laminate Layup

SchematicOrientation Angles

Equivalent Laminate (Theory)

Experimental

1

Approach: Ply Layup

3

Equivalent

E only depends on fiber angle to desired direction E3D = fn(θ), E3D = E2D

replace real 3D chopped fiber by a ‘virtual’ 2D equivalency

Orientation Tensor Determination (O.T.D.), and Equivalency

Paper Physics Approach

Introduce AlignedVirtual Plies

Introduce AlignedVirtual Plies

Lamina Macro-mechanics: Q, QLamina Macro-

mechanics: Q, Q

Morri-Tanaka: E, vAgrawal: S

Morri-Tanaka: E, vAgrawal: S

26

Ref: K. Jayaraman and M.T. Kortschot, Correction to the Fukuda–Kawata Young’s Modulus and the Fukuda–Chou Strength Theory for Short Fiber-Reinforced Composite Materials, 1996, Journal of Materials Science, 31 (8), 2059–2064.

Equivalent

Equivalent Laminate(in-plane)

1

EquivalentAverage 1

Equivalent

θ

3

1

θ

2nd-order Orientation Tensor

Laminate Analogy Approach

Classic Laminate Theory: A, B, D

Classic Laminate Theory: A, B, D

mechanics: Q, Qmechanics: Q, Q

Progressive Failure Analysis

Progressive Failure Analysis

Moldex3D Model Structured Mesh - Shell Structured Mesh – Solid

Orientation Tensor Mapping (OTM)Flow Process of Orientation Tensor Mapping (OTM) to simple structural mesh

27

Low-Fidelity Model

High-Fidelity Model

Model Mapping: Donor and Receiver MeshOrientation Tensor through thickness compares well between Moldex3D & FE Mesh

Moldex3D GENOA Mapping

28

Multi Scale Multiple Failure CriteriaDamage, and Fracture Mechanics based Unit Cell

damagecriteria

Delamcriteria

MATRIX1. Micro crack Density (TT) ,LT2. Matrix: Transverse tension3. Matrix: Transverse compression4. Matrix: In-plane shear (+)5. Matrix: In-plane shear (-)6. Matrix: Normal compression

FIBER7. Fiber: Longitudinal tension

DELAMINATION15. Normal tension16. Transverse out-of-plane shear (+)17. Transverse out-of-plane-shear (-)18. Longitudinal out-of-plane shear (+)19. Longitudinal out-of-plane shear (-)20. Relative rotation criteria

14. INTERACTION*• MDE (stress) or SIFT (strain)

29

*Options: Tsai-Wu, Tsai-Hill, Hashin, User defined criteria, Puck, SIFT, **Honeycomb: Wrinkling, Crimpling, Dimpling, Intra-cell buckling, Core crushing. *** Environmental: Recession, Oxidation (Global, Discrete), aging, creep

Ref: C. Chamis, F. Abdi, M. Garg, L. Minnetyan, H. Baid, D. Huang, J.Housner, F. Talagani,” Micromechanics-based progressive failure analysis prediction for WWFE-III composite coupon test cases”. Journal of Composite Materials Part A 47(20–21) 2695–2712, 2013

7. Fiber: Longitudinal tension8. Fiber: Longitudinal compression9. Fiber Probabilistic10. Fiber micro buckling11. Fiber crushing12. Delamination

20. Relative rotation criteria• Edge Effect

13. Strain limit

FRACTURE21. LEFM :VCCT (2d/3d) 22. Cohesive: DCZM (2d/3d)

23. Honeycomb**24. Environmental***

•De-Homogenized Multi-Scale Modeling Methodology (Analytical) • Effect of Defects: void shape/size/distribution, fiber waviness, resin rich• Inclusion effect• Fiber architecture• Failure Mechanisms: Translaminar, Interlaminar

• Conform to FE Standards• Integrated with ABAQUS (Implicit, Explicit), LS-DYNA

•Chopped Fiber • Material characterized Vs. limited Coupon tests• Fiber Content Vs. Fiber Length

Conclusion

30

• Fiber Content Vs. Fiber Length• Manufacturing Process: Injection Molding, Compression Molding, SMC, Mu Cell

•Service Loading Validation • Static, fatigue, impact, Crush

• Methodology allows simulation of entire manufacturing Process, • Residual stress, Deformation• Delamination lamination initiation location• Contributing failure type• Location of damage and fracture