DOT/FAA/AR-08/52 Characterization and Structural

Air Traffic Organization

Operations Planning
Behavior of Braided Composites
Office of Aviation Research
and Development
Washington, DC 20591

January 2009

Final Report

This document is available to the U.S. public

through the National Technical Information
Services (NTIS), Springfield, Virginia 22161.

U.S. Department of Transportation

Federal Aviation Administration
 NOTICE

This document is disseminated under the sponsorship of the U.S.

Department of Transportation in the interest of information exchange. The
United States Government assumes no liability for the contents or use
thereof. The United States Government does not endorse products or
manufacturers. Trade or manufacturer's names appear herein solely
because they are considered essential to the objective of this report. This
document does not constitute FAA certification policy. Consult your local
FAA aircraft certification office as to its use.

This report is available at the Federal Aviation Administration William J.

Hughes Technical Center’s Full-Text Technical Reports page:
actlibrary.act.faa.gov in Adobe Acrobat portable document format (PDF).
 Technical Report Documentation Page
1. Report No. 2. Government Accession No. 3. Recipient's Catalog No.
DOT/FAA/AR-08/52
4. Title and Subtitle 5. Report Date
CHARACTERIZATION AND STRUCTURAL BEHAVIOR OF BRAIDED January 2009
COMPOSITES
6. Performing Organization Code

7. Author(s) 8. Performing Organization Report No.

Ajit D. Kelkar* and John D. Whitcomb**
9. Performing Organization Name and Address 9 10. Work Unit No. (TRAIS)

*Department of Mechanical Engineering **Department of Aerospace Engineering

North Carolina A&T State University Texas A&M University
626 McNair Hall, 1601 E. Market Street 3141 TAMU
Greensboro, NC 27411 College Station, TX 77843
11. Contract or Grant No.
DTFA03-01-C00033
12. Sponsoring Agency Name and Address 13. Type of Report and Period Covered
U.S. Department of Transportation
Federal Aviation Administration Final Report
Air Traffic Organization Operations Planning
Office of Aviation Research and Development
Washington, DC 20591
14. Sponsoring Agency Code
AIR-100
15. Supplementary Notes
The Federal Aviation Administration Technical Monitor was Curtis Davies.
16. Abstract
The growing interest in small business jets in the general aviation industry is the motivation of this present research. The major
objective in the small business jet industry is to reduce costs while keeping takeoff weights below 12,500 lb (5670 kg), which is a
requirement of the Federal Aviation Administration. The overall objective of this research is the performance evaluation and
modeling of biaxial braided composites manufactured using vacuum-assisted resin transfer molding.

Biaxial braided composites with different braid angles were manufactured using carbon braids and two different resin systems (vinyl
ester and epoxy). Static tension and tension-tension fatigue tests were performed. It was concluded that the Sigmoidal function
accurately represents the stress-fatigue life curve of braided composites. It was observed that the braid angle has great influence on
mechanical properties but has less effect on endurance limits. The variation in ultimate tensile strength between the specimens
caused large scatter in the fatigue data. A new approach based on statistical analysis techniques for conducting fatigue tests is
recommended. A new approach was developed to model stiffness degradation curves. This model was proven to be very efficient in
all the three major stages of stiffness degradation.

A computational micromechanics strategy was developed to model 2x2 braids. Since the tow cross-section along the towpath is not
uniform, direct finite element mesh generation for the model is difficult. A mapping technique was developed to generate the finite
element mesh for various 2x2 biaxial braids from the previously developed mesh for the twill weave. This resulted in substantial
time savings. By exploiting symmetry operations such as mirroring, rotation, or a combination of the two, the analysis region was
greatly reduced.

The analyses for different braids have shown that the peak stresses in the tow mainly occur at the undulating region and along the
edges of the tow. Stress distribution in braids was also compared with those in equivalent laminates. A considerable volume of the
tow (10%-45% for the range of parameters studied) had stresses larger than an equivalent lamina. The severity of stresses in a braid
as compared to those in an equivalent lamina depends upon braid geometric parameters. Braid angle changes the stress distribution
in the tow considerably. The severity of peak stresses seems to be increasing linearly with an increase in waviness ratio.
17. Key Words 18. Distribution Statement
Composites, Braiding, VARTM, Fatigue, Analytical modeling This document is available to the U.S. public through the National
Technical Information Service (NTIS) Springfield, Virginia 22161.
19. Security Classif. (of this report) 20. Security Classif. (of this page) 21. No. of Pages 22. Price
Unclassified Unclassified 181

Form DOT F 1700.7 (8-72) Reproduction of completed page authorized

 ACKNOWLEDGEMENTS

The authors want to thank the Federal Aviation Administration for funding this work,
especially to Mr. Curtis Davies, who coordinated all the activities. The authors acknowledge
the efforts of our students Xiaodong Tang, Jitendra Tate, and Deepak Goyal who helped to
make this project a success.

iii/iv
 TABLE OF CONTENTS

Page

EXECUTIVE SUMMARY xvii

1. INTRODUCTION 1-1

1.1 Motivation 1-1

1.2 Braided Fabric 1-1
1.3 Review of Earlier Work 1-7

1.3.1 Resin Transfer Molding and VARTM Manufacturing Processes 1-7

1.3.2 Braided Composites Under Static Loading 1-8
1.3.3 Fatigue 1-10
1.3.4 Finite Element Modeling 1-16

1.4 Research Objective 1-17

1.5 Related Documentation 1-18

1.5.1 Journal Articles 1-18

1.5.2 Conference Proceedings 1-19
1.5.3 Theses 1-20

2. LOW-COST MANUFACTURING OF BRAIDED COMPOSITES 2-1

2.1 Introduction 2-1

2.1.1 Selection of Fibers and Fabrics 2-1

2.1.2 Selection of Resin 2-2

2.2 Steps in VARTM Processing 2-5

2.2.1 Mold Preparation and Vacuum Bagging 2-5

2.2.2 Formulation and Degassing of Resin 2-8
2.2.3 Resin Impregnation and Curing 2-9

2.3 Vacuum-Assisted Resin Transfer Molding Processing Using DM 411-350

and EPON 9504 2-10

2.3.1 Vacuum-Assisted Resin Transfer Molding Using DM 411-350

Vinyl Ester Resin 2-11

2.3.2 Vacuum-Assisted Resin Transfer Molding Using EPON 9504

Epoxy Resin 2-12

2.4 Overall Fiber Volume Fraction 2-21

v
 2.4.1 Ignition Method 2-21
2.4.2 Areal Weight Method 2-22
2.4.3 Density Method 2-22

2.5 Fiber Volume Fraction in the Tow 2-23

2.6 Summary of Section 2 2-24

3. PERFORMANCE EVALUATION 3-1

3.1 Introduction 3-1

3.1.1 Static Tensile Tests 3-1

3.1.2 Tension-Tension Fatigue Tests 3-1
3.1.3 Specimen Geometry 3-2
3.1.4 Dynamic/Fatigue Testing System 3-3
3.1.5 Test Matrix 3-3
3.1.6 Tensile Test Results of Neat EPON 9504 Epoxy Resin 3-4

3.2 Tensile and Fatigue Test Results of Braided Composites Manufactured

Using Braided Tubes 3-6

3.2.1 Carbon/Vinyl Ester Braided Composites 3-6

3.2.2 Carbon/Epoxy Braided Composites 3-7
3.2.3 Braided Tube Findings 3-9

3.3 Effect of Fiber Volume Fraction and Braid Angle Variations on UTS 3-10

3.3.1 Effect of Fiber Volume Fraction on UTS 3-10

3.3.2 Effect of Braid Angle on UTS 3-11

3.4 Tensile and Fatigue Test Results of Braided Composites Manufactured

Using Slit Sleeves 3-12

3.4.1 Carbon/Epoxy Braided Composites (Braid Angle 30° ±1°) 3-12

3.4.2 Carbon/Epoxy Braided Composites (Braid Angle 45° ±1°) 3-13
3.4.3 Slit Braided Sleeve Findings 3-15

3.5 Modeling of the S-N Diagram 3-18

3.6 Other Observations 3-23

3.6.1 Failure Pattern 3-23

3.6.2 Viscoelastic Behavior of Braided Composites 3-24
3.6.3 Frequency and Temperature Effect 3-27

4. STIFFNESS DEGRADATION MODELING 4-1

vi
 4.1 Importance of Stiffness Degradation Study 4-1
4.2 Representation of Tensile Modulus and Stiffness Degradation Curves 4-1
4.3 Stiffness Degradation Curves as a Tool to Predict the Endurance Limit 4-3
4.4 Stiffness Degradation Model to Predict Residual Stiffness 4-6
4.5 Stiffness Degradation Study Findings 4-15

5. COMPUTATIONAL MICROMECHANICS 5-1

5.1 Introduction 5-1

5.2 Idealized Geometry of the 2x2 Braid 5-1
5.3 Analysis Steps 5-3
5.4 Finite Element Models 5-4
5.5 Governing Differential Equations 5-14

5.5.1 Kinematics 5-15

5.5.2 Compatibility Equations 5-15
5.5.3 Kinetics 5-16
5.5.4 Constitutive Equations 5-16

5.6 Material Systems 5-17

5.7 Tape Laminate Approximation 5-19

6. PARAMETRIC STUDIES: EFFECT OF VARIOUS PARAMETERS ON

EFFECTIVE ENGINEERING PROPERTIES 6-1

6.1 Introduction 6-1

6.2 Results of Parametric Studies 6-1
6.3 Effect of Various Parameters on Effective Properties 6-1
6.4 Deviation of the Properties From Reference Laminate Values 6-6
6.5 Complementary Braids 6-8
6.6 Comparison of Predictions With Experimental Data 6-9
6.7 Summary of Effective Property Results 6-11

7. PROGRESSIVE FAILURE ANALYSIS 7-1

7.1 Introduction 7-1

7.2 Description of Multiscale Analysis 7-1
7.3 Typical Finite Element Models 7-2

7.3.1 Fiber/Matrix Model 7-3

7.3.2 Tow Architecture Model 7-3
7.3.3 Laminate Model 7-4

7.4 Material Properties for Failure Analysis 7-5

7.5 Failure Criteria, Degradation Model 7-5

vii
 7.6 Progressive Failure Analysis Findings 7-6

7.6.1 Fiber/Matrix Scale 7-6

7.6.2 Tow Architecture Scale 7-7
7.6.3 Laminate Scale 7-11

7.7 Summary 7-13

8. STRESS DISTRIBUTIONS 8-1

8.1 Introduction 8-1

8.2 Configurations Analyzed 8-1

8.3 Stress Distribution in Braids 8-2

8.4 Volume Distribution Curve 8-3

8.5 Comparison of Stress Distribution in Braids With Equivalent Laminates 8-4

8.6 Effect of Matching the Loading on Stress Distribution = Effect of Braid

Angle and Interlacing 8-6

8.7 Effect of Waviness Ratio 8-9

8.8 Summary 8-11

9. CONCLUSIONS 9-1

10. BIBLIOGRAPHY 10-1

APPENDICES

A—Braid Angle Change (in the Case of Braided Tubes) Does Not Change Fiber
Volume Fraction

B—Recommended Method of Fatigue Tests

C—Braided Tubes and Slit Sleeves Data Sheet

viii
 LIST OF FIGURES

Figure Page

1-1 Different Weave Patterns of Textile Fabric (Plain, Twill, Satin, and Basket) 1-2

1-2 Schematic of Braiding Process and Working Model of Braider 1-3

1-3 Biaxial Braid and Triaxial Braid 1-4

1-4 Biaxial Braided Tubes on Reel 1-4

1-5 Braiding Pattern Types (a) Diamond Braid (1/1), (b) Regular Braid (2/2), and
(c) Hercules Braid (3/3) 1-5

1-6 Aircraft Applications of Braided Composites (a) Jet Engine Stator Vane and
(b) Wing Flap 1-6

1-7 Effect of Braid Angle on In-Plane Moduli 1-8

1-8 Effect of Braid Angle on Shear Moduli 1-9

1-9 Effect of Braid Angle on Poisson’s Ratio 1-9

1-10 Fatigue Damage Mechanisms in Composite Laminate 1-11

1-11 Modulus Decay Under Fatigue Loading in Woven Composites 1-12

2-1 Schematic for VARTM 2-1

2-2 Steps in Fabric Lay-Up 2-6

2-3 Degassing Chamber 2-9

2-4 Resin Impregnation of DM 411-350 Resin System 2-10

2-5 Resin Impregnation of EPON 9504 Resin System 2-10

2-6 Vacuum-Assisted Resin Transfer Molding Using DM 411-350 Resin System 2-12

2-7 Vacuum-Assisted Resin Transfer Molding Using EPON 9504 Resin System 2-13

2-8 Graph of Viscosity vs Time and Temperature 2-14

2-9 Surface Defects—Processing at 70ºF (21ºC) 2-14

2-10 Fiber/Matrix Interface 2-15

2-11 Separation of Plies (Disbond) After Using of Surfactant 2-16

ix
2-12 Viscosity of EPON Resin System at Various Temperatures 2-17

2-13 Surface Defects—Processing at 80ºF (27ºC) 2-18

2-14 Micrographs of Braided Tows Using SEM 2-23

2-15 An SEM Micrograph of Fibers 2-24

3-1 Tensile Test Specimen Geometry 3-2

3-2 Dynamic/Fatigue Testing System 3-3

3-3 Plan of Test Specimens From Panel 3-4

3-4 Specimen Geometry for Neat Resin Coupon 3-5

3-5 Stress-Strain Diagram of EPON 9504 Epoxy Neat Resin 3-5

3-6 Stress-Strain Behavior of Braided Composites Using Flattened Braided Tubes

(Braid Angle 25° ±2°) 3-8

3-7 An S-N Diagram for Braided Composites Using Flattened Braided Tubes Showing
Large Scatter in the Fatigue Data (Braid Angle 25° ±2°) 3-9

3-8 Ultimate Tensile Strength Decreasing Exponentially With an Increase in

Braid Angle 3-11

3-9 Stress-Strain Behavior of Carbon/Epoxy Braided Composites (Braid Angles

25° ±2°, 30° ±1°, and 45° ±1°) 3-14

3-10 S-N Diagram for Braided Composites With Slit Sleeves Showing Less Scatter
in the Fatigue Data (Braid Angles 30° ±1° and 45° ±1°) 3-14

3-11 Modulus of Angle-Ply Laminates as a Function of Orientation Angle 3-16

3-12 Poisson’s Ratio of Angle-Ply Laminates as a Function of Orientation Angle 3-17

3-13 Modulus of Braided Composites as a Function of Braid Angle 3-17

3-14 Poisson’s Ratio of Braided Composites as a Function of Braid Angle 3-18

3-15 Linear Representation of S-N Diagram for Carbon/Vinyl Ester and Carbon/
Epoxy Material System 3-19

3-16 Sigmoidal (Boltzmann) Fit for Carbon/Epoxy Material System (Braid Angles
25° ±2°, 30° ±1°, and 45° ±1°) 3-20

3-17 Failure Patterns of Static Tensile- and Fatigue-Loaded Carbon/Vinyl Ester

Specimens (Braid Angle 25° ±2°) 3-23

x
3-18 Failure Patterns of Static Tensile- and Fatigue-Loaded Carbon/Epoxy Specimens
(Braid Angle 30° ±1°) 3-24

3-19 Creep of Carbon/Epoxy Braided Composite (Braid Angle 30° ±1°) 3-25

3-20 Stress Relaxation of Carbon/Epoxy Braided Composite (Braid Angle 30° ±1°) 3-26

3-21 Stiffness Degradation of Carbon/Epoxy Braided Composites (Braid Angle 30° ±1°)
at Different Frequencies (Applied Stress 60% of UTS) 3-26

3-22 Hystersis Loops of Carbon/Epoxy Braided Composites (Braid Angle 25° ±2°)
(Applied Stress 50% of UTS, 10-Hz Frequency, and R = 0.1) 3-27

3-23 Stress and Strain vs Time With a Phase Shift Carbon/Epoxy Braided Composites
(Braid Angle 30° ±1°) (Applied Stress 50% of UTS, 29th Cycle) 3-27

3-24 Rise in Temperature of Carbon/Epoxy Braided Composites (Braid Angle 30° ±1°)
at 10-Hz Frequency Until Failure at Fatigue Load of 60% of UTS 3-29

3-25 Stiffness Degradation of Braided Composites (Braid Angle 30° ±1°) at Different
Frequencies at Fatigue Load of 60% of UTS 3-30

4-1 Representation of Tensile Modulus 4-2

4-2 Stiffness Degradation Over Entire Fatigue Life for Carbon/Epoxy Braided
Composites at 75% of UTS (Braid Angle 25° ±2°) 4-3

4-3 Stiffness Degradation Curves for First 400 Cycles 4-4

4-4 Smoothened Stiffness Degradation Curves for First 400 Cycles 4-5

4-5 Linear Representation of First 50 Cycles of Stiffness Degradation Curves 4-5

4-6 Slope of Initial Region of Stiffness Degradation Curve vs Applied Stress as

Percentage of Ultimate Tensile Strength 4-6

4-7 Flow Chart Representing General Strategy for Stiffness Degradation Modeling 4-7

4-8 Fatigue Secant Modulus, E(n) vs Number of Cycles, n 4-9

4-9 Normalized Fatigue Modulus E(n)/E(N) vs Cycle Ratio (n/N) 4-10

4-10 Linear Relationship Between Static and Fatigue Secant Modulus 4-12

4-11 Comparison Between Experimental and Theoretical Stiffness Degradation

Curves (Carbon/Epoxy Braided Composite, Braid Angle = 25° ±2°, Applied
Stress = 70% of UTS) 4-12

xi
4-12 Tracking of Fatigue Secant Modulus 4-14

5-1 Architecture of 2x2 Braid 5-2

5-2 Twill Weave 5-3

5-3 Analysis Steps 5-4

5-4 Schematics of Braids 5-5

5-5 Simple and Symmetric Stacking of Mats 5-6

5-6 A Coarse Finite Element Mesh for Half of the Unit Cell 5-7

5-7 Typical Finite Element Models for Flattened Cross-Sections 5-12

5-8 Finite Element Model for Lenticular Cross-Section With Nodes and
Elements Labeled 5-13

5-9 A 20-Node Brick Element 5-14

5-10 Distribution of Fibers Inside the Tow 5-18

5-11 Finite Element Mesh for Calculating Tow Properties 5-18

5-12 An Equivalent Laminate Configuration 5-20

6-1 Effective Longitudinal Modulus E11 as a Function of Braid Angle and Waviness
Ratio for Carbon/Epoxy Material 6-2

6-2 Normalized Effective Engineering Properties as a Function of Braid Angle and

Waviness Ratio for Glass/Epoxy Material System With Lenticular Cross-Section 6-3

6-3a Normalized Effective Engineering Properties as a Function of Braid Angle and

Waviness Ratio for Carbon/Epoxy Material System for Tow With Flattened
Cross-Section 6-4

6-3b Normalized Effective Engineering Properties as a Function of Braid Angle and

Waviness Ratio for Tow With Lenticular Cross-Section 6-5

6-4 Deviation of 3D Finite Element Results From the Laminate Results for Glass/
Epoxy Material System With Lenticular Cross-Section 6-6

6-5 Deviation of 3D Finite Element Results From the Laminate Results for Carbon/
Epoxy Material System With Lenticular Cross-Section 6-7

6-6 Deviation of 3D Finite Element Results From the Laminate Results for Carbon/
Epoxy Material System With Lenticular Cross-Section 6-7

xii
6-7 Difference in Properties of Complementary Braids 6-9

7-1 Overview of Multiscale Modeling of Textile Composites 7-1

7-2 Finite Element Mesh for Reference Unit Cell Model 7-4

7-3 Finite Element Model at the Laminate Scale 7-5

7-4 Effective Properties of the Fiber/Matrix Scale as a Function of Fiber

Volume Fraction 7-7

7-5 Progressive Failure Response and Effective Properties as Function of Damage

State for the Reference Model 7-8

7-6 Failure States at Given Loading Levels for Reference Model 7-9

7-7 Comparison of Progressive Failure Responses for Various Unit Cell Models 7-10

7-8 Reponses of Braid Laminate Containing Variation in Tow Geometric Properties 7-12

8-1 Three-Dimensional Stress State in Tow for ±25° Braid With WR = 1/3 (<σxx> = 1) 8-2

8-2 Effect of Braid Angle on σ33 Stress Concentration (Uniaxial loading (<σxx> = 1),
WR = 1/3) 8-3

8-3 σ33 Volume Distribution in ±30° Braid Tow With WR = 1/3 When <σxx> = 1
was Applied 8-4

8-4 Comparison of Stress Volume Distribution in a ±45° Braid With That in an

Equivalent Tape Laminate 8-5

8-5 Effect of Braid Angle on σ22 and σ13 Volume Distribution (<σxx> = 1, WR = 1/3) 8-7

8-6 Effect of Braid Angle on σ22 and σ13 Volume Distribution When <σij> in the Tow
are Matched (WR = 1/3) 8-8

8-7 Phase Shift 8-9

8-8 Effect of Waviness Ratio on Stress Volume Distribution, BA = ±45° 8-10

8-9 Variation of Peaks With Waviness Ratio ±45° Braid Under <σxx> =1 8-11

xiii
 LIST OF TABLES

Table Page

1-1 Fatigue Performances of Laminated Composites 1-14

2-1 Properties of AS4 Carbon Fiber 2-2

2-2 Properties of DM 411-350 and EPON 9504 2-4

2-3 Vacuum-Assisted Resin Transfer Molding Process Parameters for DM 411-350

and EPON 9504 Resins 2-11

2-4 Experimental Data for VARTM-Processed Panels Using EPON 9504 Resin 2-19

2-5 Overall Fiber Volume Fraction for Braided Composites (Braid Angle 45° ±1°) 2-23

3-1 Test Matrix 3-4

3-2 Tensile Test Results of Carbon/Vinyl Ester Braided Composites

(Braid Angle 25° ±2°) 3-6

3-3 Fatigue Test Results for Carbon/Vinyl Ester Braided Composite

(Braid Angle 25° ±2°) 3-7

3-4 Tensile Test Results of Carbon/Epoxy Braided Composites

(Braid Angle 25° ±2°) 3-7

3-5 Fatigue Test Results for Carbon/Epoxy Braided Composite

(Braid Angle 25° ±2°) 3-8

3-6 Tensile Test Results for Carbon/Epoxy Braided Composites

(Braid Angle 30° ±1°) 3-12

3-7 Fatigue Test Results for Carbon/Epoxy Braided Composite

(Braid Angle 30° ±1°) 3-12

3-8 Tensile Test Results for Carbon/Epoxy Braided Composites

(Braid Angle 45° ±1°) 3-13

3-9 Fatigue Test Results for Carbon/Epoxy Braided Composite

(Braid Angle 45° ±1°) 3-13

3-10 Comparisons of Tensile Properties of Braided Composites With Different

Material Systems and Braid Angles 3-16

3-11 Comparison of UD Laminates’ and Braided Composites’ S-N Curve 3-19

xiv
3-12 Comparison of Linear Model and Sigmoidal Model Results With
Experimental Results 3-20

3-13 Parameters in Sigmoidal (Boltzmann) Curve for Braided Composites for

Different Material Systems and Braid Angles 3-21

3-14 Fatigue Life Predicted by Linear and Sigmoidal Model at Different Fatigue
Loads for Carbon/Epoxy Braided Composites 3-22

3-15 Effect of Frequency on Fatigue Life of Carbon/Epoxy Braided Composites at

Braid Angle 30° ±1° and Fatigue Load of 60% of UTS 3-28

3-16 Effect of Frequency on Fatigue Life of Carbon/Epoxy Braided Composites at

Braid Angle 45° ±1° and Fatigue Load of 60% of UTS 3-28

4-1 Fatigue Secant Modulus at 1st and Nth Cycle for Carbon/Epoxy Braided Composites
(Su = 95.74 ksi, Eo = 7.870 Msi, ES = 5.409 Msi, and Braid Angle 25° ±2°) 4-10

4-2 Static and Fatigue Secant Modulus for Carbon/Epoxy Braided Composites
(Su = 95.74 ksi, Eo =7.870 Msi, ES = 5.409 Msi, and Braid Angle 25° ±2°) 4-11

∂u1 ∂u1 ∂u 2 ∂u 3
5-1 Multipoint Constraints for , , , or Loading
∂X 1 ∂X 2 ∂X 2 ∂X 3
⎛ ∂ui 1 ∂u ⎞
⎜ where = ∫ i dV ⎟ 5-8
⎜ ∂x j V V ∂x j ⎟
⎝ ⎠

∂u1 ∂u2 ⎛ ∂ui 1 ∂u ⎞

5-2 Multipoint Constraints for or Loading ⎜ where = ∫ i dV ⎟ 5-10
∂X 3 ∂X 3 ⎜ ∂x j V V ∂x j ⎟
⎝ ⎠

5-3 Runtimes on a Personal Computer for One Case 5-14

5-4 Material Properties for Fiber, Matrix, and Tow 5-19

6-1 Experimental Results for Carbon/Vinyl Ester 6-10

6-2 Comparison With Experimental Results for Carbon/Vinyl Ester, BA 25° ±1° 6-10

6-3 Comparison With Experimental Results for Carbon/EPON 6-10

7-1 Parameters for Defining Various Unit Cell Models 7-3

7-2 Degradation Factors for All Damage Modes 7-6

8-1 Severity of Stresses in Braid Compared to an Equivalent Tape Laminate 8-6

xv
 LIST OF ACRONYMS AND SYMBOLS

2D Two-dimensional
3D Three-dimensional
BA Braided angle
cp Centipoise
CV Sample coefficient of variation, in percent
DGEBA Diglycidyl ether of biphenol A
DM Derakane momentum
EVA Ethylene vinyl acetate
FAA Federal Aviation Administration
kg Kilogram
PC Personal computer
RTM Resin transfer molding
S-N Stress-fatigue life
SEM Scanning electron microscope
Tg Glass transition temperature
TGMDA Tetraglycidyl methylene dianiline
UD Unidirectional
UTS Ultimate tensile strength
VARTM Vacuum-assisted resin transfer molding
VOC Volatile organic content
WR Waviness ratio

xvi
 EXECUTIVE SUMMARY

The growing interest in small business jets in the general aviation industry is the motivation of
this present research. The major objective in the small business jet industry is to reduce costs
while keeping takeoff weights below 12,500 lb (5670 kg), which is a requirement of the Federal
Aviation Administration. The overall objective of this research is the performance evaluation
and modeling of biaxial braided composites manufactured using vacuum-assisted resin transfer
molding (VARTM).

The natural conformability of biaxial braids makes them more cost competitive than woven
fabric. The VARTM process has proven to be low in cost compared to resin transfer molding.
Thus, the combination of biaxial braids and the VARTM process is likely to considerably reduce
overall costs. Before the braids can be used confidently in the primary structures, it is necessary
to understand the performance of biaxial braided composites under various loading conditions
and especially under fatigue.

Biaxial braided composites with different braid angles were manufactured using carbon braids
and two different resin systems (vinyl ester and epoxy). Static tension and tension-tension
fatigue tests were performed. It was concluded that the Sigmoidal function accurately represents
the stress-fatigue life curve of braided composites. It was observed that the braid angle has great
influence on mechanical properties but has less effect on endurance limits.

It was observed that the variation in ultimate tensile strength between the specimens caused large
scatter in the fatigue data. A new approach based on statistical analysis techniques for
conducting fatigue tests is recommended.

The different mechanisms of fatigue failure of composites are reflected in the stiffness
degradation, which can be measured nondestructively. A new approach was developed to model
stiffness degradation curves. This model was proven to be very efficient in all the three major
stages of stiffness degradation.

A computational micromechanics strategy was developed to model 2x2 braids. Since the tow
cross-section along the towpath is not uniform, direct finite element mesh generation for the
model is difficult. A mapping technique was developed to generate the finite element mesh for
various 2x2 biaxial braids from the previously developed mesh for the twill weave. This resulted
in substantial time savings. By exploiting symmetry operations such as mirroring, rotation, or a
combination of the two, the analysis region was greatly reduced.

This work presents an investigation of the effect of various parameters such as braid angle,
waviness ratio (WR), material properties, and cross-section shape on the effective engineering
properties of the 2x2 braids. Extensive parametric studies were conducted for two material
systems: (1) glass fiber/epoxy matrix (S2/SC-15) and (2) carbon fiber/epoxy matrix (AS4/411-
350). Equivalent laminated materials with angle plies and a resin layer were also analyzed to
compare the difference in predictions from full three-dimensional finite element analyses of the
2x2 braided composites. The predictions are also compared with experimental results for a
carbon/epoxy material system.

xvii
The effect of variation in braid parameters on the progressive failure behavior of a 2x2 braided
composite laminate was studied. A bottom-up multiscale finite element modeling approach was
employed that sequentially considered the fiber/matrix scale, the tow architecture scale, and the
laminate scale.

The analyses for different braids have shown that the peak stresses in the tow mainly occur at the
undulating region and along the edges of the tow. Stress distribution in braids was also
compared with those in equivalent laminates. A considerable volume of the tow (10%-45% for
the range of parameters studied) had stresses larger than an equivalent lamina. The severity of
stresses in a braid as compared to those in an equivalent lamina depends upon braid geometric
parameters. Braid angle changes the stress distribution in the tow considerably. The severity of
peak stresses seems to be increasing linearly with an increase in WR.

xviii
1. INTRODUCTION.

1.1 MOTIVATION.

The motivation for the present research was to meet the challenges of small business jet
applications for which the Federal Aviation Administration (FAA) (Title 14 Code of Federal
Regulations Part 23, Appendix C, Single Pilot Certification) requires a takeoff weight of
12,500 lb (5670 kilogram (kg)) or less. Typically, small business jets have a seven-passenger
capacity, maximum cruise speed of 530 mph (465 knots or 853 kilometers per hour (km/h)) and
a maximum range of 1875 miles (3000 km). Their cost ranges from 3.5 to 4.5 million U.S.
dollars. The major objectives in the small business jet industry are to reduce costs while keeping
weight below 12,500 lb (5670 kgf). Reduced weight is possible only if the primary and
secondary structures are made of lightweight composite materials. Thus, competitive costs
depend on the selection of fabric, resin, and manufacturing methods. The main design feature of
small business jets today is a composite fuselage built with automated fiber placement
techniques (Aerospace Technology Magazine aerospace-technology.com/projects/raytheon
.premier1/, September 6, 2008)

Biaxial braided fabric, with its natural conformability, can fit over any complex shape. Thus,
there is no need for cutting, stitching, and fiber placement as required for woven fabric. This
ability reduces part count and makes biaxial braided fabric useful for primary structures like the
fuselage. Braided composites have been proven to be cost-competitive. It was recently proven
that carbon braids are extremely efficient for single-part airfoil sections like wing flaps (Swain,
2000). Vacuum-assisted resin transfer molding (VARTM) is a low-cost manufacturing process
with the capability of manufacturing complex parts with higher fiber volume fractions than those
from hand lay-up.

Braided composite manufactured using the VARTM process is one of the major candidates for
small business jet applications. Although braided composites have an enormous number of
applications, the interest of the present research is in small business jet applications. The
primary focus is on the manufacturing of braided composites using VARTM, performance
evaluation, and analytical modeling.

1.2 BRAIDED FABRIC.

The commonly used composites are tape laminates and textile composites. Tape laminates have
good in-plane properties. Textile composites, which include woven, braided, and knitted fabrics,
are also important when considering out-of-plane properties. Textile composites generally have
better dimensional stability, out-of-plane properties, and impact and delamination resistance.
Figure 1-1 shows the different orthogonal weave patterns of textile fabric (Whitcomb and Tang,
1999).

Braiding is an ancient textile process that is simple and highly productive. Braided fabric is used
for many secondary structures such as stiffeners, wing spars, floor beams, fuselage frames,
ducting, and housings in aerospace applications. There is a growing interest in using this
technique to manufacture composite preforms for primary structures like wing flaps, horizontal
stabilizers, and fuselage especially for small business jet applications (Swain, 2000).

1-1
1-2

Figure 1-1. Different Weave Patterns of Textile Fabric (Plain, Twill, Satin, and Basket) (Whitcomb and Tang, 1999)
Figure 1-2(a) shows the manufacturing of planar braids consisting of a braider head and a
traversing mechanism. Yarns are intertwined and laid on the mandrel. In biaxial braiding, two
sets of yarn carriers move along intersecting undulating circular paths in opposite directions to
form tubular interlaced fabric. Introducing axial yarns in the lengthwise direction forms a
triaxial braid (Kumar and Wang, 1997). Figure 1-2(b) displays the carriers on the working
model of a braider (Courtesy of A&P Technology, Inc.). Figure 1-3 shows the biaxial and
triaxial braids. Braid architecture resembles a combination of filament winding and weaving.
Fibers in braided tubes have continuity between the ends of the part and are mechanically
interlocked.

(a)

(b)

Figure 1-2. (a) Schematic of Braiding Process (Kumar and Wang, 1997) and (b) Working Model
of Braider (Courtesy of A&P Technology, Inc.)

1-3
 Braid Angle

Biaxial Braid Axis Triaxial Braid

(a) (b)

Figure 1-3. (a) Biaxial Braid and (b) Triaxial Braid (Byun, 2000)

Biaxial braided tubes have natural conformability, also referred to as Chinese finger trap, which
refers to the diameter reduction when the tubes are pulled along lengthwise. The braid tube fits
over the complex components easily. Therefore, cutting, stitching, and fiber placement are not
required as in woven fabric. The tube form of braid is supplied on reels, as shown in figure 1-4.
The introduction of axial yarn in triaxial braid locks the diameter and stops the braid’s natural
conformability. Braided tubes are cut along the braid axis and the edges are fused or stitched to
create slit sleeves (http://www.braider.com). As the slit sleeves are fused at the edges, they
maintain a constant braid angle during processing. The braided composites’ mechanical
properties are sensitive to the braid angle. Therefore, slit sleeves were used in the latter part of
this research.

Figure 1-4. Biaxial Braided Tubes on Reel

1-4
Full coverage of the braid is desirable to obtain high-fiber volume fraction in the final
composites. A relationship exists between the circumferences, braid angle, yarn width, and the
number of yarns in the braid for full fabric cover (Seobroto and Ko, 1989). The number of
carriers on the machine limits the number of yarns.

Biaxial braids are classified as diamond braid (1/1), regular braid (2/2), and Hercules braid (3/3)
depending on the interlacement as shown in figure 1-5 (Kumar and Wang, 1997). The braided
tubes are specified at ±45° orientation. Fiber orientation is the angle measured from the axis of
the braid to the axis of the bias yarns. This angle is also called the braid angle, the fiber angle, or
the bias angle and is usually denoted by θ (figure 1-3). Fiber orientation typically ranges from
15° to 75°. When a biaxial braid tube is used for a component of varying cross-sections, the
braid angle, thickness, and areal weight (yield) vary from point to point.

(a)

(b)

(c)

Figure 1-5. Braiding Pattern Types (a) Diamond Braid (1/1), (b) Regular Braid (2/2), and
(c) Hercules Braid (3/3) (Kumar and Wang, 1997)

Advantages of braided composites can be summarized as:

• Braided structures are impact resistant. Since all the fibers in the structure are involved
in the load distribution, the braid absorbs a large amount of energy as it fails. These
braids are used in fan blade containment in commercial aircraft and energy-absorbing
crash structures in Formula One racing cars (A&P Technology, Inc., http://www
.braider.com/index_hi.html).

1-5
• Since braids are woven on the bias, they provide very efficient reinforcement for parts
that are subjected to torsional loads such as drive shafts (A&P Technology, Inc.,
http://www.braider.com/index_hi.html).

• Braided composites greatly improve interlaminar shear properties. There is virtually no

delamination observed when braided composites are subjected to fatigue loadings (Tate
and Kelkar, 2003a,b).

• Braided composites have better fatigue life. They have outperformed unidirectional
laminates for jet engine stator vanes in fatigue strength (A&P Technology, Inc.,
http://www.braider.com/index_hi.html).

• It has been proven that braiding can be used to improve performance and reduce the
manufacturing cost of composite structures. Braiding has been compared with other
manufacturing methods such as filament winding, manual and automated tape lay-up, roll
wrapping (tape winding) and pultrusion in terms of versatility, component
reproducibility, composite quality/structural integrity, design flexibility, damage
tolerance, repairability, microcracking resistance, joints/attachments, resistance to
thermally induced twist, and lower manufacturing costs (Munjal and Maloney, 1990).

• Braided composites exhibited better fatigue strength in comparison to compression-

molded unidirectional prepreg laminates for jet engine stator vanes. Vanes molded with
the use of braided composites also reduced costs (A&P Technology, Inc.,
http://www.braider.com/index_hi.html).

Braided composites have a wide range of applications in the aerospace, industrial, medical, and
recreational industries. Figure 1-6 shows the aircraft applications of braided composites (wing
flap and jet engine stator vane).

(b)
(a)

Figure 1-6. Aircraft Applications of Braided Composites

(a) Jet Engine Stator Vane (Courtesy of A&P Technology) and
(b) Wing Flap (Courtesy of Raytheon Aircraft Company)

1-6
1.3 REVIEW OF EARLIER WORK.

A systematic literature review of earlier research in the areas of manufacturing, performance

evaluation, and modeling is presented in the following sections.

1.3.1 Resin Transfer Molding and VARTM Manufacturing Processes.

There are many processes available for composite manufacturing: hand lay-up, wet lay-up,
autoclave, resin transfer molding (RTM), and VARTM. Autoclave and RTM are the most
popular processes in the aircraft industry.

Autoclave processing is used extensively for manufacturing high-performance aerospace parts.

It consists of a pressure vessel that is capable of applying heat and pressure (up to 7 bar ~ 100
psig) simultaneously. The major difficulty in the autoclave process is the high capitalization
cost, which arises from stringent pressure code regulations.

RTM is a widely accepted process in the aerospace industry due to its cost-effectiveness and the
quality of the parts it produces. During the RTM process, a mold is loaded with the fabric
material and then closed. The resin is then injected into the mold. The mold with the fabric is
often placed under vacuum pressure so that all the entrapped air is removed (Strong, 1989). A
considerable amount of research has been performed to study various aspects of RTM. Hess
(1990) explained how this method can be used effectively for manufacturing braided composite
structures for naval structures and helicopter components. High-quality graphite/epoxy braided
composite tubes for space structures are successfully produced by RTM (Munjal, et al., 1990).
Automated tow placement and stitching of dry textile composite preform followed by RTM has
been investigated as a cost-effective manufacturing process for obtaining damage tolerant
fuselage and wing structures for transport aircraft (Deaton, et al., 1992). Kobayashi, et al. (1992)
fabricated T-shaped braided graphite/epoxy composite truss joints. Beck (1993) explained the
finer points of an RTM process for fabricating structures. Lee, et al. (1997) suggested different
ways to make this process more reliable. Many researchers have used this process to
manufacture high-quality braided composites. Swain (2000) explained how structurally sound
aircraft components can be fabricated by using carbon fiber braided sleeves with epoxy resin.

Principally, VARTM and RTM are similar. However, there is no external pressure applied to the
mold in VARTM. As the name suggests, VARTM is a resin transfer molding with the use of a
vacuum. The dry preform is placed in the mold and vacuum bagged. A low-viscosity resin is
then drawn into the mold through the aid of the vacuum. The pressure differential between resin
at atmospheric pressure and vacuum is a driving force for resin impregnation. The VARTM
process is explained in detail in section 2. This process is becoming very popular because of the
following advantages:

• Simple low-cost tooling

• Practicality of very large and complex components
• Higher-fiber volume fractions than those composites produced from hand lay-up
• Reduced environmental concerns than those composites produced from hand lay-up

1-7
VARTM is a comparatively new process and many researchers are investigating its various
aspects. Nguyen, et al. (1997) evaluated this process with little modifications for large-scale
composite ship structures. Pike, et al. (1996) used this process for producing structural laminates
with integrated armor for ground combat vehicle hulls. Advances in VARTM were studied by
Beckwith and Hyland (1998) and the relationship between various forms of the RTM have been
reported. Kelkar, et al. (1999) compared the performance of composites manufactured using the
RTM and the VARTM processes for composite integral armor vehicle applications. It was
determined that the moduli, ultimate tensile strength, and fatigue life of RTM and VARTM
panels are comparable if the fiber volume fractions remain the same. Heider, et al. (2000)
studied this process for complete automation, and sensing and control for manufacturing large-
scale composite parts. Rigas, et al. (2001) have explained the cost-effectiveness of this process
in manufacturing ballistic, multifunctional, composite material systems, and structures.

1.3.2 Braided Composites Under Static Loading.

Naik, et al. (1995) studied the effects of braiding parameters, such as yarn size, braid angle, and
the percentage of axial yarns, in two-dimensional (2D) triaxial braided composites. The
researchers concluded that stiffness properties are not a function of yarn size but are strongly
influenced by braid angle and the percentage of axial yarns. Increasing braid angle reduces the
longitudinal modulus while increasing the transverse modulus. Shear modulus and the in-plane
Poisson’s ratio increase with an increase in braid angle. Shear modulus peaks at 45º and then
declines, whereas in-plane Poisson’s ratio peaks at 30º before it declines. Figures 1-7 through
1-9 show the effect of braid angle on in-plane moduli (Ex, Ey, and Ez), shear moduli (Gxy, Gxz, and
Gyz), and Poisson’s ratio (νxy, νxz, and νyz). In these figures, the subscripts L and T are used to
indicate x and y directions, respectively.

Figure 1-7. Effect of Braid Angle on In-Plane Moduli (Naik, et al., 1995)

1-8
 Figure 1-8. Effect of Braid Angle on Shear Moduli (Naik, et al., 1995)

Figure 1-9. Effect of Braid Angle on Poisson’s Ratio (Naik, et al., 1993)

Minguet (1995) compared the mechanical properties of graphite/epoxy tape laminates and 2D
triaxial braided composite in tension in both longitudinal and transverse directions. The
longitudinal modulus of both material forms was found to be quite similar, but the transverse
modulus of the braids was lower. The longitudinal strength of the braids was lower than that of
the laminates, while the transverse strength was significantly lower. The crimp in the tows of the
braid may have been the main cause of the reduced properties.

1-9
Smith and Swanson (1996) conducted a series of biaxial tests to measure the strength of 2D
triaxial braided carbon specimens and compared the measured strength with that of laminates
fabricated with similar materials and fiber orientations. Failures controlled by the axial fibers
were found to have reasonably similar values of ultimate strength for braids and laminates, with
the braid specimens having somewhat lower values because of irregularities in the axial fibers
related to the various braid designs. However, when failures were controlled by the fibers in the
braid direction, the strength of the braid specimens was much less than that of the laminates.

Falzon and Herszberg (1998) conducted tension, compression, and shear tests to evaluate the
mechanical performance of 2D triaxial braided carbon/epoxy composites. The performance of
braided composite was compared with unidirectional (UD) tape laminates. It was found that
braided composites have a comparable tension and compression stiffness to the laminate.
However, there was considerably reduced tension and compression strength attributed to the
fiber damage and fiber tow waviness. Tension tests on the fiber tows revealed a 20% reduction
in tension strength resulting from fiber damage during the braiding process.

Chen, et al. (1998) demonstrated the sensitivity of strength to a small degree of misalignment
(±1° to ±5°) in braid angle in 2D triaxial braided composites. Clemente, et al. (1998) showed
that 2D triaxial braided composite structures exhibit very effective energy absorption compared
to traditional materials.

The literature review indicates that most of the work was done on 2D triaxial braided
composites, and very little work was done in the area of performance evaluation and modeling of
biaxial braided composites.

1.3.3 Fatigue.

1.3.3.1 Fatigue of Multidirectional and Woven Composites.

The fatigue mechanism in composites is much more complex than that of metals. The damage
may be in one or more forms, such as a failure in a fiber-matrix interface, a matrix cracking,
delaminations, and final fiber breakage. Both matrix cracking and delamination reduce stored
energy and stiffness. Detectable damage can be found very early in the fatigue life. The damage
causes a reduction of elastic properties such as stiffness. There is always a correlation between
damage and stiffness reduction. Since the reduction of stiffness (or damage) occurs long before
the composite is in danger, the definition of fatigue failure in composite materials may change
from one application to another.

Fatigue in multidirectional tape laminates is a well-understood phenomenon. The damage takes

place in three stages, as shown in figure 1-10. The three main stages in the fatigue life are cross-
ply cracks with large initial stiffness reduction, initiation and propagation of delamination of
plies with slow stiffness reduction, and complete delamination and fiber failure (Reifsnider,
1990).

1-10
 Figure 1-10. Fatigue Damage Mechanisms in Composite Laminate

Fujii and Amijima (1993) proposed the three-stage failure mechanism model for woven
composites as measured by modulus decay, as shown in figure 1-11.

• Stage I: Debonds (between fiber and the matrix) in the weft occur near the crossover
points. The number of matrix cracks also increases rapidly. At a cycle ratio (n/N) of 0.1,
debonds in the weft do not progress very much and the matrix crack density does not
increase significantly. This stage is termed the meta-characteristic damage state. The
modulus decreases rapidly.

• Stage II: In this stage, debonds occur in the warp direction. The debonds in warp and
meta-delamination mainly cause gradual modulus decay in an almost linear fashion. This
stage constitutes more than 75% of the total fatigue life.

• Stage III: Final failure occurs with the breakage of fibers. However, the final stage is
sometimes hardly distinguishable. The stiffness decreases rapidly again during the last
few cycles before the specimen fails.

1-11
 Figure 1-11. Modulus Decay Under Fatigue Loading in Woven Composites

1.3.3.2 Fatigue of Braided Composites.

Although braided fabric resembles woven fabric, it is inappropriate to apply the damage
mechanism concepts of woven composites to braided composites. This is because the woven
fabrics have orthogonal weaving, that is, warp and fill tows run 0º and 90º, but braided fabrics
tows run at θ/-θ angles. In the case of braided composites, very little work is performed on
fatigue damage, the relationship between internal damage, and macroscopic properties, such as
stiffness, interlaminar stresses, and the delamination of the braided composites. Some additional
work performed is discussed in the following paragraphs.

Two discontinuous blade-stiffened specimens of three-dimensional (3D) braided composites

(0º/±17º) and 3D orthogonal woven composites (0º/90º/90º) were evaluated under tension-
tension fatigue (S/Su = 0.6, 5-10 Hz, and R = 0.1). It was found that both specimens survived one
million cycles, but the woven panel lost its stiffener. The residual strength of braided and woven
composite specimens was 88.6% and 90.9% of their static strength (Shivkumar, et al., 1999).

Carlos (1994) studied fatigue behavior of 3D triaxially (0°/±30º) braided carbon/epoxy

composites in tension-tension fatigue at 10-Hz frequencies with an R ratio of 0.1. He reported
±5º variations in braid angle when the specimens were manufactured using braided tubes and
compression molding. The specimens survived one million cycles at the applied stress level of
50% of the ultimate tensile strength. There was a large scatter in fatigue data, which may have
resulted from the variation in braid angle.

Portanova and Deaton (1995) evaluated fatigue (5 Hz and R = 0.1) resistance of multiaxial
braided, 3D graphite/epoxy composites in both unnotched and postimpacted conditions. Damage

1-12
initiation and growth were documented using radiography and ultrasonic through-transmission
(C-scans). Stiffness and strength degradation were measured as a function of applied cycles.

The effect of elevated (107ºC) and lower (-40ºC) temperatures on the fatigue behavior of RTM
manufactured triaxially braided (30º/0º/30º) glass/vinyl ester was determined. Results showed a
significant decrease in tensile and fatigue strength at 107°C and a modest increase at -40ºC
(Houston and Chernenkoff, 1992).

Burr, et al. (1995) conducted tension-tension fatigue tests on 2D triaxial braided composites.
The first types of damage that occurred were splits in the braider bundles and cracks in the resin
rich areas. Once a sufficient density of this type of damage occurred, the axial bundles began to
disbond from the surrounding constituents and continued to do so until ultimate failure.

The literature review clearly indicates that there is a need to understand the behavior of biaxial
braided composite under tension-tension fatigue loading.

1.3.3.3 Fatigue Modeling of Composite Materials.

Fatigue damage modeling is difficult because there are many damage mechanisms (matrix
cracking, fiber/matrix debonding, delamination, and fiber breakage) and because the test
specimens are not heterogeneous. It is impossible to produce specimens with identical
microstructural features. Therefore, a material damage model including all the damage states is
difficult to establish.

Degrieck and Paepegem (2001) provided an excellent review of major fatigue models. These
models can be classified in three major categories: fatigue life models using stress-fatigue lift
(S-N) curves, phenomenological models for residual stiffness/strength, and progressive damage
models using variables related to damage (transverse matrix cracks, delamination size).

1.3.3.3.1 Fatigue-Life Model.

Fawaz and Ellyin (1994) proposed a semilog linear relationship between applied cyclic stress, S,
and the number of cycles to failure, N, for laminated composites.

S = Su (m log N + b)

where Su is the mean static strength, and m and b are constants. Low values of m and high values
of b indicated higher fatigue strength. This is the most popular linear model used in
conventional laminated composites. The values of constants m and b for different material
systems are presented in table 1-1 (Maurice, 2000).

1-13
 Table 1-1. Fatigue Performances of Laminated Composites

Material System R m b
E-glass/ductile epoxy–UD 0.1 -0.1573 1.3743
T300 carbon/ductile epoxy–UD 0.1 -0.0542 1.0420
E-glass/brittle epoxy–UD 0.1 -0.1110 1.0935
T300 carbon/brittle epoxy–UD 0.1 -0.0873 1.2103
E-glass/epoxy [0/90] s 0.05 -0.0815 0.9340

1.3.3.3.2 Phenomenological (Stiffness Degradation) Model.

The fatigue damage in composites can be evaluated by measuring the residual stiffness or
strength. Measurement of the residual strength is a destructive test and, thus, cannot be used to
predict or track the fatigue damage. On the other hand, stiffness can be measured
nondestructively. Therefore, the stiffness degradation model is much more applicable to the
practical design of composite structures (Lee, et al., 1996). There is an enormous number of
models developed by researchers. Only the major stiffness degradation models relevant to
braided composites are discussed in detail in the following paragraphs.

Hwang and Han (1986a, 1986b) introduced the concept of the fatigue modulus defined as the
ratio of applied stress and resultant strain at a specific cycle. The fatigue modulus degradation
rate was assumed to follow a power function of the number of fatigue cycles.

dF
= − Acn C −1
dn

where A and c were material constants. Further they assumed that applied cyclic stress σa varied
linearly with resultant strain in any arbitrary loading cycle, so that

σ a = F ( ni ) × ε( ni )

where F(ni) and ε (ni) were the fatigue modulus and strain at loading cycle ni, respectively. After
integration and introduction of the strain failure criteria, the fatigue life N was calculated as:

N = [ B(1- r )]1/ c

σa
where r = was the ratio of the applied cyclic stress to the ultimate static strength, and B and c
σu
were material constants.

The variable D has been used in residual stiffness models to describe stiffness loss, as
⎛ E ⎞
D = ⎜1- ⎟ , where Eo is the initial tangent modulus.
⎝ Eo ⎠

1-14
1.3.3.3.3 Progressive Damage Models.

Hwang and Han (1986a) proposed three cumulative damage models based on the fatigue
modulus F(n) and the resultant strain. Model III demonstrated better agreement with
experimental data than the first two models (I and II). In the third model, residual stiffness D
was defined as:

r ⎡ F0 ⎤
D= ⋅⎢ − 1⎥ .
1 − r ⎣ F ( n) ⎦

Fatigue modulus F(n) was the ratio of applied stress to corresponding strain at nth cycle. F0 was
the modulus defined for static strength to corresponding strain.

Failure occurred when

m
D= ∑ ΔDi = 1
i =1

where ΔDi was the amount of damage accumulation during fatigue at stress level ri and m was
the number of load sequences until final failure.

Whitworth (1987) proposed a residual stiffness model for graphite/epoxy composites:

a a
⎛ E(N * ) ⎞ ⎛ S ⎞ *
⎜⎜ ⎟⎟ = 1 − H ⋅ ⎜⎜1 − ⎟⎟ N
⎝ E (0) ⎠ ⎝ R(0) ⎠

where N* = n/ N was the ratio of applied cycles to the fatigue life N, S was the applied stress
level, R(0) was the static strength, E(0) was the initial modulus, and a and H were parameters
independent of the applied stress level.

This residual stiffness model was used by Whitworth (1990) to propose a cumulative damage
model, where the damage function was defined as:

⎡ H × (1- S )a ⎤ n
D=⎢ a ⎥× N
⎣ 1- S ⎦

S
where S = was the normalized applied stress range and a and H were parameters.
R (0)

When D = 0, no cycles were applied and E = E(0). When D = 1, the residual modulus equaled
the failure stiffness Ef.

1-15
Recently, Whitworth (1998) proposed a new residual stiffness model, which followed the
degradation law:

dE * ( n) -a
= m -1
dn ( n + 1) ⎡⎣ E * (n) ⎤⎦

where E*(n) = E(n)/E(N) was the ratio of the residual stiffness to the failure stiffness E(N), n was
the number of loading cycles, and a and m were parameters that depended on the applied stress,
loading frequency, and so forth. The residual stiffness E(n) could be expressed in terms of the
static tensile strength Su with the introduction of the strain failure criterion. A statistical
distribution of the residual stiffness could then be obtained, if the static ultimate strength could
be represented by a two-parameter Weibull distribution.

Yang, et al. (1990) developed a residual stiffness model for fiber-dominated composite
laminates:

dE ( n)
= − E (0)Qνn ν −1
dn

where Q and ν were two parameters correlated by a linear equation. Experimental data revealed
that ν could be written as a linear function of the applied stress level. The researchers also
derived a statistical distribution of the residual stiffness. They observed that this model was not
immediately applicable to matrix-dominated composite laminates because the stress-strain curve
was no longer linear. Yang, et al. (1992) extended the model for matrix-dominated composites
by replacing the modulus E(n) by the fatigue modulus F(n). The latter was defined as the
applied stress level S divided by the corresponding strain at the nth cycle. They derived an
expression relating the fatigue modulus F(0) with the initial stiffness E(0) through the modeling
of the nonlinear stress-strain response. They proved that this new damage law was a particular
case of fiber-dominated composites. The model for matrix-dominated behavior was applied to
the fatigue behavior of [±45°]2S graphite/epoxy laminates.

The stiffness degradation diagram of woven composites (figure 1-11) shows three distinct stages.
There is a rapid stiffness decline in the first stage; the second stage shows gradual stiffness decay
in an almost linear fashion; and the stiffness decreases rapidly again during the third stage.
Almost all of the models discussed above do not predict all three stages.

The linear fatigue-life model on a semilog scale and the stiffness degradation models discussed
above may not represent the exact behavior of biaxial braided composites. A different approach
is developed for braided composites and discussed in sections 3 and 4.

1.3.4 Finite Element Modeling.

A thorough understanding of the behavior of the 2x2 braids is required to exploit the advantages
these materials offer. To achieve this, computational micromechanics analyses were performed
to predict the effective engineering properties and the 3D stress state.

1-16
First, the finite element meshes for braids were generated. For conducting parametric studies,
meshes were required to be generated. Earlier (Tang, 2001), a strategy was described for
generating the finite element meshes for a wide variety of weaves using a general-purpose
preprocessor Meshweaver. Since the 2x2 braid architecture is similar to that of the twill weave,
a mapping technique was used to generate the braid meshes from the meshes of the twill weave.
Solid models of the braids were generated for thorough understanding of its architecture and its
effect on the effective properties and the stress distribution. In any finite element analysis, it is
useful to minimize the analysis region to save computational time and computer memory. The
boundary conditions were imposed in such a way that periodicity and symmetries (Tang and
Whitcomb, 2003) that exist in the microstructure of the braids could be exploited to minimize the
analysis region. Boundary conditions that include a number of multipoint constraints were
derived using a technique given in Tang and Whitcomb, 2003.

Obtaining effective engineering properties is the first order of concern for any structural analysis.
Parametric studies were conducted to obtain effective properties. The effect of various
parameters like braid angle, waviness ratio (WR), material system, and stacking sequence was
analyzed. Thus, the dominant characteristics that determine the behavior of the 2x2 braids were
identified. The effective property results produced by this 3D finite element analysis were
compared with an equivalent laminate analysis, because laminate theory codes are widely
available and understood. It was investigated whether the effective properties of the braids can
be predicted using the laminate theory and how much error there is in doing so. Also, such
comparisons revealed relative performance.

Practical textile composites do not have perfectly uniform and periodic microstructures, and thus,
their behavior cannot be completely characterized by conducting periodic analysis on single unit
cells. Numerous cases were run (in addition to the 450 cases run for parametric studies for
effective properties) to characterize progressive failure analysis of 2x2 braids with
microstructural variation, including the variation of braid angle, WR, and fiber volume fraction
in the fiber tow.

The stress distribution in braided composites is complex even for simple uniaxial loading.
Interlacing the tows creates a complex load path that results in full 3D stress distributions. The
location and magnitude of peak stresses depend on the particular stress component and vary with
various braid architecture parameters such as braid angle and degree of waviness. Finite element
analyses for different braids were performed to find peak magnitudes and locations. The stress
distribution in braids was compared with those in equivalent laminates.

1.4 RESEARCH OBJECTIVE.

In summary, the objective of the present research was to manufacture braided composites using a
low-cost VARTM process and to study their performance in fatigue loading. The research also
developed an analytical model for stiffness degradation model to predict residual stiffness. The
detailed 3D finite element model that considered undulations and fiber continuity was developed.
The mesh generation strategy, finite element models, results of effective properties, progressive
failure analysis, and stress distribution is presented. The research focused on the following.

1-17
• The design and modification of VARTM process parameters to manufacture braided
composites using two different resin systems (i.e., vinyl ester and epoxy).

• The studies of the effects of braid angles on mechanical properties.

• The study of fatigue behavior of a biaxial braided composite in tension-tension fatigue

loading and the effects of braid angles on fatigue performance.

• The development of a stress-fatigue life diagram (S/Su-N) and stiffness degradation

diagram related to fatigue performance.

• The comparison of the fatigue performance of two resin systems (i.e., vinyl ester and
epoxy).

• The development of a stiffness degradation model for braided composites.

• The definition of the idealized tow architecture to analyze the microstructure of the 2x2
braids and generation of meshes for 3D finite element analysis.

• Parameters such as braid angle, WR, tow cross-section, stacking sequence, and material
properties were studied as to their influence on the effective engineering properties of the
2x2 braids.

• The comparison of experimental results with 3D finite element model.

• The comparison of the difference in predictions of the 3D finite element analysis and the
simple 3D laminate analysis.

• The prediction of the 3D stress state in the tow and determination of the effect of the
braid angle and WR on location and magnitude of the peak stresses.

• The characterization of the progressive failure behavior of 2x2 braids with

microstructural heterogeneity.

1.5 RELATED DOCUMENTATION.

1.5.1 Journal Articles.

1. Jitendra S. Tate and A.D. Kelkar, “Effect of Braid Angle on Fatigue Performance of
Biaxial Braided Composites,” Composites Part A: Applied Science and Engineering,
2005.

2. Ajit D. Kelkar and Jitendra S. Tate, “Low Cost Manufacturing of Textile Composite
Using VARTM Process,” Journal of Manufacturing Technology and Research, Vol. 1,
July 2004.

3. Jitendra S. Tate and Ajit D. Kelkar, “Stiffness Degradation Modeling of Biaxial Braided
Composites,” Composites Part B: Engineering, 2005.

1-18
4. Jitendra S. Tate and Ajit D. Kelkar, “VARTM Manufacturing of Carbon Biaxial Braided
Composites using EPON 9504 Epoxy Resin System,” Journal of Advanced Materials
(SAMPE), 2005.

5. Tang, X., Whitcomb, J., Kelkar, A.D., and Tate, J., “Progressive Failure Analysis of 2x2
Braided Composites Exhibiting Multiscale Heterogeneity,” Journal of Composites
Science & Technology, 2005.

6. Goyal, D., Tang, X., and Whitcomb, J.D., “Effect of Various Parameters on Effective
Engineering Properties of 2x2 Braided Composites,” Journal for Mechanics of Advanced
Materials and Structures, 2004.

7. D. Goyal and J.D. Whitcomb, “Analysis of Stress Concentrations in 2x2 Braided

Composites,” Journal of Composite Materials, 2004.

1.5.2 Conference Proceedings.

1. Jitendra S. Tate, Ajit D. Kelkar, and Ronnie Bolick, “Performance Evaluation of Notched
Biaxial Braided Composites,” ASME 2004 International Mechanical Engineering
Congress and Exposition, Anaheim, California, IMECE2004-59883, November 13-19,
2004.

2. Jitendra S. Tate and Ajit D. Kelkar, “Effect of Braid Angle on Fatigue Performance of
Biaxial Braided Composites,” The 3rd Conference on Fatigue of Composites, Kyoto,
Japan, September 13-15, 2004.

3. Jitendra S. Tate, Ajit D. Kelkar, and Ronnie Bolick, “Tension and Fatigue Behavior of
Slit Sleeve Braided Composites,” 11th International Conference on Composites/Nano
Engineering (ICCE11), South Carolina, August 11-14, 2004.

4. Jitendra S. Tate, Ajit D. Kelkar, and Vinaya A. Kelkar, “Failure Analysis of Biaxial
Braided Composites Under Fatigue Loading,” The 15th European Conference of
Fracture, Stockholm, Sweden, August 11-13, 2004.

5. Ajit D. Kelkar, Jitendra S. Tate, and John D. Whitcomb, “Durability of VARTM

Manufactured Textile Composites,” International Conference on Advances in Structural
Integrity, Bangalore, India, July 14-17, 2004.

6. Jitendra S. Tate, Ajit D. Kelkar, and John Rice, “Feasibility Study of VARTM
Manufacturing of Carbon Biaxial Braided Composites Using EPON 9504 Epoxy Resin
System,” JISSE-8, 8th International SAMPE Symposium & Exhibition, Vol. 1, ISBN
4-9900028-8-1, Tokyo, 12-21 November 2003, pp. 1145-1148.

7. Ajit D. Kelkar and Jitendra S. Tate, “Fatigue Behavior of VARTM Manufactured Biaxial
Braided Composites,” ASME 2003 International Mechanical Engineering Congress and
RD & D Exposition, Washington, DC, USA, November 15-21, 2003, n 43850.

1-19
8. Ajit D. Kelkar, Jitendra S. Tate, and Ron Bolick, “Introduction to Low Cost
Manufacturing of Composite Laminates,” ASEE 2003 Annual Conference & Exposition,
Nashville, Tennessee, USA, June 22-25, 2003, Session 2003-2526.

9. Ajit D. Kelkar and Jitendra S. Tate, “Effect of Fatigue Loading on the Stiffness
Degradation of VARTM Manufactured Biaxial Braided Composites,” 9th International
Conference on the Mechanical Behavior of Materials (ICM9), PALEXPO Congress
Center, Geneva, Switzerland, May 25-29, 2003, pp. 46.

10. Ajit. D. Kelkar and Jitendra S. Tate, “Low Cost Manufacturing of Textile Composites
Using Vacuum Assisted Resin Transfer Molding,” Proceedings of the 20th All India
Manufacturing Technology, Design and Research Conference, Ranchi, India, December
2002, pp. 712-716.

11. Kelkar, Ajit D. and Jitendra S. Tate, “Low Cost Manufacturing of Biaxial Braided
Composite Laminate,” Ninth Annual International Conference on Composite
Engineering, San Diego, California, USA, July 2002, pp. 373-374.

12. X. Tang, J. Whitcomb, D. Goyal, and A.D. Kelkar, “Effect of Braid Angle and Waviness
Ratio on Effective Moduli of 2x2 Biaxial Braided Composites,” 44th
AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics, and Materials
Conference, Norfolk, Virginia, AIAA Paper 2003-1475.

13. Xiaodong Tang, John D. Whitcomb, Ajit D. Kelkar, and Jitendra S. Tate, “Progressive
Failure Analysis of 2x2 Braided Composites Exhibiting Multiscale Heterogeneity,” 18th
American Society for Composites Technical Conference, Gainesville, Florida, USA,
18-22 October 2003.

14. Xiaodong Tang, John D. Whitcomb, and Deepak Goyal, “Micromechanics Modeling of
2x2 Braided Composites,” ASME 2002 Conference.

15. D. Goyal and J.D. Whitcomb, “Analysis of Stress Concentrations in 2x2 Braided
Composites,” International Conference on Computational and Experimental Engineering
and Sciences, 2004.

16. J.D. Whitcomb and D. Goyal, “Effect of Braid Design on Stress Concentration
Magnitude and Location,” ASME Symposium on Textiles, 2004.

1.5.3 Theses.

1. Jitendra S. Tate, May 2004, “Performance Evaluation and Modeling of Braided

Composites,” Ph.D. dissertation, Department of Mechanical Engineering. North Carolina
A&T State University, Greensboro, North Carolina.

2. D. Goyal, “Analysis of 2x2 Braided Composites,” MS thesis, Department of Aerospace

Engineering, Texas A&M University, College Station, Texas, August 2003.

1-20
2. LOW-COST MANUFACTURING OF BRAIDED COMPOSITES.

2.1 INTRODUCTION.

There are various methods that can be used to manufacture composite laminates. These methods
include wet lay-up, autoclave processing, filament winding, pultrusion, RTM, and VARTM.
VARTM is a comparatively new process and is proven to be cost-effective compared to RTM.
The braided composite laminates in the present research are manufactured using the VARTM
process.

During VARTM, dry fabric is placed into a tool and vacuum bagged in conjunction with the
resin distribution line, the vacuum distribution line, and the distribution media. A low-viscosity
resin is drawn into the fabric through a vacuum. Resin distribution media ensures resin
infiltration in the through-the-thickness direction. The key to successful resin infiltration of the
fabric is the design and placement of the resin distribution media, which allows complete wet-out
of the fabric and eliminates voids and dry spots. Properly designed and properly placed resin
distribution media eliminate race tracking and resin leakage around the fabric (Seeman, 1990 and
1994). The schematic for the fabrication is shown in figure 2-1.

Distribution Media Mold Release Film

Peel Ply Sealant Tape
Fabrics
Resin Distribution Line Vacuum Distribution Line

Vacuum Bag

Mold

Figure 2-1. Schematic for VARTM

The parameters of the VARTM process are currently designed by a trial and error method.
Therefore, a series of experiments are required to design a suitable distribution media and to
determine the proper location of the resin line and the vacuum line. Attempts have been made to
develop simulation tools for this process (Loos, et al., 2001, ccm.udel.edu, University of
Delaware). Accurate predictions by simulation are possible only if all the resin and fabric
properties are known, which is usually not the case. Although the experimental approach is
time-consuming and expensive, there are currently no alternative methods to optimize the
VARTM process parameters.

2.1.1 Selection of Fibers and Fabrics.

Carbon fibers are used in aircraft applications because of their high specific strength (strength-to-
density ratio) and high specific modulus (modulus-to-density ratio) compared to glass fibers.
Moreover, carbon fibers have better moisture resistance and a very low coefficient of linear
expansion. However, carbon fibers do have some shortcomings. They are expensive, brittle, and

2-1
have a low-impact resistance and a low elongation. The very low coefficient of linear expansion
complicates the processing. Carbon fibers are surface treated to improve the fiber/matrix
interfacial bonding and, in turn, the interlaminar shear strength. Organic coatings known as size
are also added to carbon fibers to improve fiber/matrix bonds and to protect fibers during
handling and processing. The major manufacturers of carbon fibers are Hexcel, Inc., Amoco
Performance Products, Inc., Grafil, Inc., and Union Carbide, Inc.

The fabrics used in the present research were 2x2 biaxial carbon braided tubes and slit sleeves
manufactured by A&P Technology, Inc. The fibers used for both braided tubes and slit sleeves
were AS4 manufactured by Hexcel, Inc. The properties of AS4 fiber as supplied by the
manufacturer are listed in table 2-1 (Abdallah, 2002).

Table 2-1. Properties of AS4 Carbon Fiber (Hexcel, Inc.)

Ef11, Ef22 = Ef33, Gf12 = Gf13, Gf23, ρf , Su ,

Msi Msi Msi Msi g/cm3 ksi εu ,
(GPa) (GPa) (GPa) (GPa) νf12 = νf13 νf23 (lb/in3) (MPa) (%)
33 2.4 3.6 0.99 0.2 0.25 1.79 602 1.82
(227.53) (16.55) (24.82) (6.89) (0.0645) (4150)

Note: Diameter of fibers = 7 µm

The selected braided tubes are suitable for high-load applications and are the preferred choice of
small aircraft manufacturers. When the diameter of braided tubes is 2 in. (50.8 mm) and the
braid angle is 45°, the areal weight is 19.9 oz/yd2 (675 g/m2). Areal weight of braided fabrics is
a function of braid angle. Slit sleeve is a special braid manufactured with a constant braid angle
throughout the fabric. Slit sleeves look like woven fabrics, but their tows are at [θ/-θ] angles.
Braided tubes are manufactured at required braid angles and are then slit to create slit sleeves.
The edges are fused with a small amount of ethylene vinyl acetate (EVA) adhesive. The
adhesive maintains a constant braid angle even during handling. Slit sleeves with braid angles of
30° and 45° were used in the present research. The numbers of fibers in a tow in the present
research were 12,000 (generally specified as 12k tow size).

2.1.2 Selection of Resin.

Resin plays a very important role in polymer matrix composites. Although the loads are mainly
carried by fibers, modulus, failure strain, and resin/matrix adhesion play a dominant role in the
performance of composites. Resin also determines the type of fabrication process, the service
temperature, and the flammability and corrosion resistance of the composite. Since the focus of
the present research was a material system for small business jet applications, different resins
systems were considered. Aircraft applications require resins with high-glass transition
temperature (Tg), high-damage tolerance, high-impact resistance, and high-fatigue life.
Physically, glass transition is a reversible change in amorphous polymers from (or to) a viscous
to (or from) a hard and brittle phase. For most applications, it is a practical measure of the upper
use (service) temperature of composites. The value of Tg depends on the test method. Usually,
resin manufacturers provide the value of heat distortion temperature found by ASTM D 648.
Sometimes Tg is also determined as the temperature at which the Young’s modulus begins to

2-2
decline. Tg defined this way has been found equivalent to the heat distortion temperature (Juska
and Puckett, 1997).

From a VARTM processing point of view, the viscosity of the resin should be 100 to 1000
centipoises (cp) (Smith, 2001). Low-viscosity resin is not able to impregnate fabric thoroughly
and leaves dry fabrics and voids. Low-viscosity resins also have high concentrations of reactive
diluents, which tend to vaporize under vacuum and create voids. On the other hand, high-
viscosity resin is not able to flow through the fabric and gels in the fabric. Most of the resins
available in the market for VARTM processing have a viscosity of 350 to 500 cp at 77ºF (25ºC).
Gel time is also an important parameter in processing. Gel time is the period of time from the
initial mixing of the resin to the point when gelation occurs, as defined by the specific test
method. Many resin manufacturers specify gel time as the time required for mixed resin to
double its viscosity. However, from the processing point of view, the time when resin is unable
to flow through fabrics is important.

The most commonly used thermoset resins for aerospace applications are vinyl esters and
epoxies. Thermoset resins have excellent adhesion, high thermal stability, high chemical
resistance, and less creep than thermoplastics. Their low viscosity prior to cure aids in the
complete wet-out of the fabric. Vinyl esters have a higher failure strain than polyesters, which
improve mechanical properties, impact resistance, and fatigue life. The formulation of vinyl
esters is complicated and consists of a catalyst, a promoter, and a gel-time retardant. A catalyst
is a substance that promotes or controls the curing of resin without being consumed in the
reaction. A promoter is a chemical additive that accelerates the curing process. The curing of
thermoset resin is an exothermic reaction. It dissipates heat to the surroundings. Gel-time
retardant is a chemical additive that absorbs free radicals once the exothermic reaction has
begun. It retards the curing process so that molding is complete before the resin gels. Many
vinyl esters are available on the market that have similar mechanical properties to those of
epoxies. Most of the vinyl esters also cure at room temperature but still have a high heat
distortion temperature of 194ºF (90ºC) without postcure (Juska and Puckett, 1997).

Epoxy resins dominate the aerospace composite market. Their high curing temperature of
approximately 350ºF (177ºC) boosts their Tg to 302ºF (150ºC). Epoxies also have high fracture
toughness, which makes them superior in fatigue. Epoxies have a low cure shrinkage compared
to vinyl esters, so there is less possibility of a component cracking during cure. Epoxies have a
simple formulation (two-part systems) consisting of an epoxy and a curing agent. The epoxy
determines the mechanical properties, and the curing agent decides the cure temperature. Tg is
governed by both the epoxy and the curing agent. Common epoxy chemistries are tetraglycidyl
methylene dianiline (TGMDA), diglycidyl ether of biphenol A (DGEBA), and phenol-
formaldehyde novolac epoxy. TGMDA epoxies have higher mechanical properties and a higher
Tg, and DGEBA epoxies have higher failure strain and lower water absorption (Juska and
Puckett, 1997).

Two resin systems were selected that meet the requirements of VARTM processing and have
comparatively high Tg. They were Derakane Momentum 411-350 vinyl ester, (DM 411-350),
manufactured by Dow Chemical Company, Inc. and EPON 9504 epoxy, manufactured by
Resolution Performance Products, Inc., The DM 411-350 resin system is extensively used in

2-3
adverse chemical environments. Applications include chemical processing, pulp and paper, and
the food and beverage industry. DM 411-350 vinyl ester resin systems have also been
investigated for sandwich structures used in aerospace applications (Smith, 2001). EPON 9504
epoxy resin systems have high tensile strength and elongation, which impart superior composite
properties. The flexural fatigue performance of fiberglass laminates made from the EPON 9504
resin system manufactured by VARTM is much higher than that of laminates made with vinyl
ester, modified vinyl ester, or isophthalic polyester resins. Therefore, the use of the EPON 9504
resin system in the fabrication of fatigue-resistant structures is significantly more economical
than the use of vinyl ester or polyester resins (Resolution Performance Products, Inc.). Smith
(2001) investigated DM 411-350 for VARTM processing and provided properties of the resin.
The experimental properties of EPON 9504 were evaluated from static tensile tests on neat resin
coupons. Table 2-2 lists the properties of both resins.

Table 2-2. Properties of DM 411-350 and EPON 9504

DM 411-350 EPON 9504

Property Mfr. Value Exp. Value Mfr. Value Exp. Value
Tensile Strength, ksi (MPa) 11.45 (79.0) --- 11.10 (76.5) 9.42 (64.94)
Tensile Modulus, Msi (GPa) 0.45 (3.1) 0.49(3.4) 0.465 (3.2) 0.43 (2.96)
Tensile Elongation, % 6.5 --- 7.8 2.83
Elongation
at UTS
Poisson’s Ratio --- 0.35 --- 0.38
Glass Transition Temperature --- (220) 104 230-276 ---
ºF (ºC) (110-135)
DMA
Heat Distortion Temperature 215 (102) --- 199 (93) ---
ºF (ºC) ASTM D648
Viscosity of mix at 350 --- 350 350
77ºF (25ºC), cp
Flexural Strength, ksi (MPa) 17 (117) --- 18.4 (126.8) ---
Flexural Modulus, Msi (GPa) 0.45 (3.1) --- 0.47 (3.24) ---
Density of cured resin, g/cm3 1.13 (0.041) 1.12 (0.040) 1.13 (0.041) 1.16 (0.042)
(lb/in3)
Gel time at 77ºF (25ºC), 40 to 60 90 60 100
minutes Time when Time to Time when
resin gelled double resin gelled
preventing the viscosity preventing
flow flow

2-4
 Table 2-2. Properties of DM 411-350 and EPON 9504 (Continued)

DM 411-350 EPON 9504

Property Mfr. Value Exp. Value Mfr. Value Exp. Value
Formulation Resin (100) + Resin (100) + curing agent
(Recommended by Promoter (0.02 - 0.13%) + (26%) by weight
manufacturer) Retardant (0.015-0.0175%) +
Catalyst (1-1.5%) by weight
Chemical Description Resin—bisphenol-A epoxy Resin—bisphenol-A
resin epoxy resin (modified)
Promoter—CoNap 6% Curing Agent—Non-MDA
Retardant—2,4 Pentanedione (methylene dianiline)
Catalyst—MEKP proprietary polyamine
Manufacturer Dow Chemical Company Resolution Performance
Products

UTS = Ultimate tensile strength

MEKP = Methylethyl ketone peroxide
MDA = Methylene dianiline
DMA = Dynamic mechanical analysis

Both resins have a viscosity of 350 cp at a temperature of 77ºF (25ºC), which makes them
suitable for processing at this lower temperature. Other mechanical properties of the resins are
almost identical. However, EPON 9504 has a higher Tg of 230º-275ºF (110º-135ºC) and a higher
elongation (7.8%) than that of DM 411-350. The price of EPON 9504 is also much lower than
the price of DM 411-350.

VARTM was successfully implemented in the beginning for biaxial carbon braided tubes using
DM 411-350 at a room temperature of 77ºF (25ºC). EPON 9504 experienced difficulties
processing at room temperature. Typical steps in the VARTM processing are discussed in the
next section, and details pertaining to the VARTM processing of DM 411-350 and EPON 9504
are discussed in the following sections.

2.2 STEPS IN VARTM PROCESSING.

Typically, the VARTM process involves the following steps (Kelkar and Tate, 2002 and 2003c):

1. Mold Preparation and Vacuum Bagging

2. Formulation and Degassing of Resin
3. Resin Impregnation and Curing

2.2.1 Mold Preparation and Vacuum Bagging.

In VARTM there is a typical sequence of vacuum bagging. The sequence of lay-up from bottom
to top is mold, mold surface protection film, bottom release fabric (also called bottom peel ply),
fabrics, top release fabric (also called top peel ply), resin distribution media, vacuum and resin

2-5
distribution lines, and vacuum bag. The vacuum bag is sealed using sealant tape. This
procedure is depicted in figure 2-2.

Mold Surface

Mold Release Film

Bottom Release
Fabric
Fabric Preform
(Slit Sleeves) Fabric Preform (Tubes)
(a)

Resin Transfer Media

Top Release Fabric

(b)

Vacuum Distribution
Line

Vacuum Bag

Sealant Tape

Resin Distribution
Line
(c)

Figure 2-2. Steps in Fabric Lay-Up

2-6
The purpose of each of these items is as follows.

• In figure 2-2(a):

Mold Surface: The mold used for the fabrication is a plywood plate lined with
Formica™ at the top surface. A heated metal plate (usually aluminum) is also popular.

Mold Release Film (Surface Protection): Two layers of sand paste wax are applied to the
surface. A 25-µm polyester film is then used to protect the mold surface. The wax and
polyester film facilitates easy removal of the panel from the mold surface after the
fabrication process is completed.

Bottom Release Fabric (Peel Ply): This is a porous nylon fabric, which leaves an
impression on the part suitable for secondary adhesive bonding (like tabbing) without
further surface preparation. Its use is optional.

Fabric Preforms: The fabrics used in this research were 2x2 biaxial braided tubes and slit
sleeves. Braided tubes were flattened to a width of 1.75 in. (44.5 mm). A braided tube
with a width of 1.75 in. (44.5 mm) provided a braid angle of ~25º. Two tubes were
stacked above each other, creating four layers. Precaution was taken so that the braid
angle (25º) was undisturbed and the width of 1.75 was (44.5 mm) maintained. This
method is referred to as a collapse method. Four slit sleeves were also stacked above
each other to obtain four layers.

• In figure 2-2(b):

Top Release Fabric (Peel Ply): This is the same material as the bottom release fabric. It
is laid on top of the braided fabrics to facilitate the flow of resin through it. It also leaves
an impression on the part suitable for secondary bonding without further surface
preparation.

Resin Transfer (Distribution) Media: The distribution media is a polyethylene mesh laid
on top of the top release fabric. This helps maintain an even distribution of resin and
facilitates the flow of resin through the thickness of the panel.

The use of distribution media is a patented technology termed as Seemann Composite

Resin Infusion Manufacturing Process that was invented by W. H. Seemann. Seemann
also patented different patterns of distribution media and the placement technique for
these patterns (Seemann, 1990 and 1994). Distribution media control the flow of resin
through the thickness. Resin flows quickly through the media and then remains in the
mesh pockets. It then travels through the thickness.

• In figure 2-2(c):

Resin and Vacuum Distribution Lines: Spirally cut, high-density polyethylene tubes are
used for this purpose. These lines are laid above the distribution media at two sides of
the fabric lay-up and can run along the length or along the width. One end of the resin

2-7
 line is closed and the other end is connected to the resin supply through the flow control
device (if used). The vacuum line is closed at one end and connected to the vacuum
pump through the vacuum gage.

A breather material acts as a distributor medium for the air and escaping volatiles and
gases. It is placed over the resin distribution media and the resin and vacuum lines. It
also acts as a buffer between the vacuum bag wrinkles and the part surface. It is a highly
porous material composed mostly of fiberglass, polyester felt, and cotton. The use of a
breather is optional. It was not used in this research.

Vacuum Bag and Sealant Tape: The vacuum bag is made from 25-µm nylon film. The
film is placed completely over the mold area and sealed firmly using a special sealant
tape. The sealant seals the vacuum bag and helps maintain a uniform vacuum throughout
the molding process.

The other equipment used in the processing included a vacuum pump, flow control devices
(optional), a vacuum gage, a degassing chamber, a temperature and humidity gage, and a stop
watch. Flow control devices like valves, clamps, and peristaltic pumps are used with certain
material systems. These devices deliver a controlled amount of resin according to the unit time
in the mold. Thus, the resin has a chance to flow through the thickness and complete wet-out of
the fabrics is ensured. A peristaltic pump delivers a fixed amount of resin in the mold per unit of
time. The quantity of resin (e.g., cm3/min) is dependent on the pump speed. The pump speed is
selected according to the fabric-resin system and the thickness of the panel.

Once the fabrics and other relevant materials are laid down in the required sequence, the entire
mold is sealed with sealant and a vacuum bag. The vacuum pump is then used to maintain the
lowest possible vacuum pressure throughout the process. Bag leaks are the most common
problems that occur in VARTM. One of the reasons for leaks is a damaged vacuum bag. A
vacuum bag is typically made of nylon film. The moisture level in the surrounding environment
affects the nylon film. Dry and brittle film can cause cracking when handled frequently.
Another common reason for bag leak is foreign material entrapped between the vacuum bag and
the sealant tape. Once the leaks have been removed and the vacuum bag is completely sealed,
the vacuum pump remains running for at least 1 to 2 hours to achieve a good vacuum in the bag.
The typical vacuum achieved is in the order of 0.5 torr. The vacuum pump is then shut off, and
the vacuum line is clamped. If the bag remains tight and holds almost the same vacuum after 1
to 2 hours, the mold is ready for resin impregnation.

The vacuum plays a vital role in the VARTM process. The pressure differential between the
atmosphere and the vacuum provides the driving force for infusing the resin into the mold. The
vacuum also removes all of the air from the mold before and during the introduction of resin
(TPI Composites, Inc., http://www.tpicomp.com/technology).

2.2.2 Formulation and Degassing of Resin.

DM 411-350 was mixed with a catalyst, a promoter, and a gel-time retardant, and EPON 9504
was mixed with a curing agent (EPI-CURE 9554) in the proportions suggested by the resin
manufacturer. The resin had to be free from entrapped air and gases that could cause voids

2-8
before being added to the mold. After the mixing of all the ingredients, the resin container was
kept in the degassing chamber to maintain a vacuum of approximately 5-10 torr, as shown in
figure 2-3. The vacuum in the chamber removed all the entrapped air and gases out of the resin.
This was a crucial step in the VARTM process and had to be performed very carefully to ensure
high-quality composite panels. Degassing resin for too short a period of time could not ensure
complete removal of the entrapped air and gases. If the resin was degassed for too long a period
of time, some of the ingredients (mainly styrene) in the resin could evaporate during processing.
This would change the final formulation of resin and also create voids. Five to ten minutes was
sufficient to remove all the entrapped air and gases.

Figure 2-3. Degassing Chamber

2.2.3 Resin Impregnation and Curing.

The resin impregnation process was different for DM 411-350 and EPON 9504. The resin was
injected into the mold at a very slow rate for DM 411-350. The flow of resin was controlled with
the help of a peristaltic pump along with an ON-OFF timer. The resin was injected in the mold
until the whole panel was soaked. Figure 2-4 displays the resin impregnation setup. Panels
remained in the mold for 18 to 24 hours at room temperature for curing, which is termed the
green cure.

A peristaltic pump was not used for EPON 9504. The mold and resin were heated to 120ºF
(49ºC) for 20 hours in the oven, and impregnation was completed in the oven at 120ºF (49ºC).
Figure 2-5 displays the resin impregnation setup for EPON 9504. After the impregnation, the
oven temperature remained at 120ºF (49ºC) for 2 hours before it was turned off. Panels were
kept in the mold for 16 hours for the green cure.

Postcure is the final and most important step in VARTM processing. Although the degree of
cure increases with time at room temperature, postcuring at elevated temperatures accelerates the
process and achieves an ultimate heat distortion temperature and optimal mechanical properties.
The manufacturer of the resin recommends the postcure cycle according to the type of curing
agent in the resin system.

2-9
 Temperature and
Humidity Gage

Resin Flask

Peristaltic Pump

Figure 2-4. Resin Impregnation of DM 411-350 Resin System

Vacuum line is
placed below the
top peel ply and
Resin next to the
distribution braided fabric
medium is cut
½ inch from the Resin
other two sides distribution
medium is cut
¾ inch from the
vacuum line side

Figure 2-5. Resin Impregnation of EPON 9504 Resin System

Postcure was performed at the temperatures and time recommended by the resin manufacturer.
The braided composites of DM 411-350 were kept in the oven at 187ºF (86ºC) for 8 hours and
then cooled in the oven to room temperature. The braided composites made of EPON 9504 were
kept in the oven at 201ºF (94ºC) for 1 hour and then cooled in the oven to room temperature.

2.3 VACUUM-ASSISTED RESIN TRANSFER MOLDING PROCESSING USING

DM 411-350 AND EPON 9504.

The typical steps in VARTM processing for both resins were the same, but with a few different
parameters. This section explains the details pertaining to both resin systems. Table 2-3
summarizes the differences. Extensive study on EPON 9504 was required to reach a particular
combination of process parameters such as time, temperature, and the design of resin distribution
media.

2-10
 Table 2-3. Vacuum-Assisted Resin Transfer Molding Process Parameters for
DM 411-350 and EPON 9504 Resins

DM 411-350 EPON 9504

Processing temperature, 77 (25) 120 (49)
ºF (ºC)
Mold heating, No 120 (49)
ºF (ºC)
Resin temperature during 77 (25) 120 (49)
mixing, ºF (ºC)
Peristaltic pump with Yes No
ON-OFF timer
Distribution media Cut short in length from Cut 0.5 in. (12.7 mm) short in
edges of fabrics length from edges of fabrics
and 0.75 in. (19 mm) short
near vacuum line
Vacuum distribution line Above distribution media Below distribution media and
placement below top peel ply, and
adjacent to braided fabrics
Processing time for panel 30 minutes 10 minutes
of size
22 in. x 9.5 in.
(56 cm x 24 cm)

2.3.1 Vacuum-Assisted Resin Transfer Molding Using DM 411-350 Vinyl Ester Resin.

2.3.1.1 Resin Distribution Media and Flow Control Devices.

The placement and design of resin distribution media play dominant roles in the complete wet-
out of the fabric. Its permeability and placement determine the boundary conditions for the flow.
Distribution media for DM 411-350 was placed so that the entire fabric was covered in the width
direction and cut 0.5 in. (12.7 mm) short in the length direction, as shown in figure 2-6. Both
the resin distribution line and the vacuum distribution lines were placed on top of the distribution
media. Resin flowed across the width through the distribution media in the in-plane direction
and seeped through the vacuum line. This prevented complete wet-out of the fabrics. Thus, the
design required a flow control device such as a peristaltic pump that could deliver a controlled
amount of resin in the mold. There was a fixed amount of resin entering the mold so that resin
would remain in the mesh pockets of the distribution media and then flow through the thickness.
The flow of resin was further controlled by use of an ON-OFF timer. A cycle of 30 seconds on
and 30 seconds off was used based on previous experience with the resin. The use of the
peristaltic pump and the ON-OFF timer significantly increased the impregnation time. It took
almost 40 minutes to impregnate a panel 22 in. long by 9.5 in. wide (56 by 24 cm).

2-11
 Distribution Media

Fabric
Vacuum Line

Resin
0.5 in
Vacuum
Gage
Flow
Control
Vacuum
Pump
Resin Line

Figure 2-6. Vacuum-Assisted Resin Transfer Molding Using DM 411-350 Resin System

2.3.1.2 Processing Temperature and Gel Time.

The viscosity of the resin is a function of its initial temperature. When resin is mixed with a
catalyst or curing agents, the viscosity initially declines and then increases along with the
temperature. When the viscosity reaches a certain value, the resin is unable to flow and gels in
the mold. According to the manufacturer, the gel time for DM 411-350 at 77ºF (25ºC) is 40 to
60 minutes. Gel time can be increased by experimenting with the formulation of the resin. Use
of the formulation in table 2-2 increased gel time to 90 minutes (Smith, 2001). It is very
important to design a VARTM process in such a way that the total time required for degassing
and resin impregnation is less than the gel time. Processing for DM 411-350 was performed at a
room temperature of 70ºF (21ºC) with the aid of a flow control device. Degassing and
impregnation took 40 minutes, which was less than the gel time.

The VARTM process was appropriate for the DM 411-350 resin with the above-mentioned
design of resin distribution media and the use of a flow control device. The quality of composite
panels was good and a fiber volume fraction as high as 57% was achieved. The major drawback
of this design was the higher processing time and the complications involved with the peristaltic
pump and the ON-OFF timer. This design would be expensive for a mass production
environment or the production of large components. Large components typically require the
laying of multiple resin and vacuum lines to ensure complete wet-out of the fabrics.

2.3.2 Vacuum-Assisted Resin Transfer Molding Using EPON 9504 Epoxy Resin.

2.3.2.1 Resin Distribution Media.

The use of the peristaltic pump with the ON-OFF timer increased impregnation time
considerably. Therefore, experiments with EPON 9504 used a proper design of the distribution
media and other process parameters to achieve complete wet-out of the fabric.

2-12
The distribution media for EPON 9504 was cut 12.7 mm (0.5 in.) short in the length direction
and 0.75 in. (19 mm) short near the vacuum line, as shown in figure 2-7. The resin distribution
line was placed on top of the distribution media, but the vacuum distribution line was placed
below the top peel ply and next to the fabrics. Since the distribution media did not touch a
vacuum line, the resin would not run through the vacuum line. It traveled to the edge of the
distribution media in the width direction and then remained in the lay-up. There was no need to
use a flow control device with this design. Resin was forced to remain in the lay-up because of
the flow resistance created at the edges of the distribution media. The driving force created by
the vacuum alone was sufficient for complete wet-out of the fabric. The impregnation in this
case was very fast. It took only 10 minutes to impregnate a 22-in.-wide by 9.5-in.-long (56- by
24-cm) panel.

Distribution Media

Fabric

Vacuum Line

0.5 in Vacuum
Resin

0.75 in Gage

Vacuum
Pump
Resin Line

Figure 2-7. Vacuum-Assisted Resin Transfer Molding Using EPON 9504 Resin System

2.3.2.2 Processing Temperature and Gel Time.

A viscosity study for EPON 9504 was performed in a laboratory where the room temperature
was 70ºF (21ºC). The viscosity changed from 310 to 395 cp in 100 minutes, and the temperature
rose from 70ºF (21ºC) to 93ºF (34ºC). A graph of viscosity versus time and temperature is
shown in figure 2-8. The viscosity led to the conclusion that VARTM processing of this resin is
also possible at a room temperature of 70ºF (21ºC).

2-13
 90 370

360
85
Temperature 350

Temperature (degrees F)
Viscosity
80 340

Viscosity (cP)
330
75
320

70 310

300
65
290

60 280
0 10 20 30 40 50 60 70 80 90 100
Time (min.)

Figure 2-8. Graph of Viscosity vs Time and Temperature

The quality of composite panels, however, was not satisfactory when processing was performed
at room temperature of 70ºF (21ºC). Random shiny spots (air pockets) were observed on the top
surface of the cured composite panels as displayed in figure 2-9. All of the entrapped air and
vapors (of volatile contents) in the mold were not removed by the vacuum alone. Some
accumulated at the top surface of the composite. Premature opening of the mold could cause this
problem, but this was not the case here.

Figure 2-9. Surface Defects—Processing at 70ºF (21ºC)

2-14
Random air pockets at the surface are simply voids at the surface. A void is any physical and
mechanical discontinuity occurring within a material or part and may be 2D (e.g., disbands,
delaminations) or 3D (e.g., vacuum-, air-, or gas-filled pockets). Voids are incapable of
transmitting structural stresses and create stress concentrations that lead to crack initiation. Both
void and dry spots indicate incomplete wetting of the fabrics in the resin.

The important parameters responsible for complete wetting of the fabric by the resin are use of a
flow control device, fiber/matrix adhesion (type of sizing component on the fiber), volatile
content in the size and the resin, wetting and air release agent, and temperature of the resin (in
turn viscosity) and the mold. An investigation was performed to determine the problem
parameter/s. It should be noted that other parameters such as the permeability of the fabric, the
fiber architecture of the fabric, and the fabric’s crimp are fixed once a particular fabric is
selected. The following sections describe the investigation efforts in chronological order.

2.3.2.2.1 Flow Control Device.

Flow control is very often useful to achieve complete fiber saturation and a void-free fill.
Although a flow control device was not used, it was determined that the use of a peristaltic pump
with an ON-OFF timer would not help solve the problem of random air pockets at the surface.

2.3.2.2.2 Fiber/Matrix Adhesion.

Fiber/matrix adhesion is not only critical in processing, but is also critical to the performance of
composites. Fiber/matrix debonding is one important damage state in composites under loading.
Fiber/matrix adhesion may affect the wetting of fiber with a matrix. A fiber/matrix interface is
shown in figure 2-10. Adhesion is a complex phenomenon that depends on compatibility of the
resin with the size of carbon fibers and braids used during braiding. The manufacturers of
carbon fibers and braids confirmed that the size used was compatible with the EPON 9504 epoxy
resin system.

Resin
Size
Fiber Surface layer

Figure 2-10. Fiber/Matrix Interface

2.3.2.2.3 Volatile Content.

Styrene is used as a cross-linking agent in vinyl ester resins. The typical percentage in vinyl
esters is 35%. Styrene has a volatile organic content (VOC), and there are restrictions on its
emission. Other volatiles such as water and alcohol are used in sizing or resin formulation.
There is a possibility of evaporation of these volatiles at a very low vacuum. If vapors are not
pushed out of the mold during processing, a tacky, air-inhibited composite surface may be
created. The VOC was not a major issue in the processing of EPON 9504.

2-15
2.3.2.2.4 Wetting and Air Release Agent.

This problem was reported to the EPON 9504 manufacturer. The manufacturer’s engineer
suspected that the carbon fabrics might be aged. There is always a possibility of damage in the
sizing of fibers if a fabric is improperly stored or old. The braid manufacturer confirmed that the
supplied lot was not old. The manufacturer further suggested the use of an air release agent,
BYK A530, or a wetting agent, BYK W390, to improve wetting. BYK Chemie, Inc.,
manufactures BYK A530 and BYK W390. The suggested percentage to be used was 0.5% by
weight based on the total resin and curing agent. The manufacturer claimed that the very low
percentage would not affect mechanical properties. An air release agent or a wetting agent
reduces surface tension between the resin and the fabric, which improves wetting and spreading.

Fluorad Fluorosurfactant FC-4430, manufactured by 3M, Inc., was available in the laboratory.
This surfactant could also work according to the EPON manufacturer. This surfactant is popular
in many industrial and commercial applications, including paints, resins, adhesives, and inks.
One panel was manufactured using FC-4430 with a concentration of 250 ppm. While the
problem of random air pockets was resolved, the plies separated easily after the postcure of
composite panels, as shown in figure 2-11. This demonstrated an adhesion failure between two
plies, which is known as disbond.

Figure 2-11. Separation of Plies (Disbond) After Using of Surfactant

The molecules of the surfactant are composed of a soluble group and an insoluble group. The
insoluble end tends to push out of the liquid and the soluble end tends to be pulled into the liquid.
The combined action lowers the surface energy of a resin allowing it to better wet and flow over
a surface (3M Company, http://www.3m.com). It was concluded that the accumulation of an
insoluble group at the interface of the two adjacent plies caused the separation. There was no
further investigation of the surfactant, wetting agent, or air release agent because there was no
strong database available to explain the effect of these agents on the performance of the
composites.

2-16
2.3.2.2.5 Temperature of Resin and Mold.

During the experiments, the resin solidified at the bottom of the container in the laboratory
environment at ~70ºF (21ºC). The top resin in the container may have been nonhomogeneous
(i.e., with an incorrect proportion of ingredients); therefore, this resin was heated at 80ºF (27ºC)
for a few hours in the oven to achieve a homogeneous resin. The resin was then used for
impregnation at room temperature. The problem remained unresolved.

The cross-linking (reactivity) of the thermoset resins was governed by a curing agent and its
temperature, as shown in figure 2-12.

160ºF
140ºF
Viscosity, cp

120ºF

Time, minutes

Figure 2-12. Viscosity of EPON Resin System at Various Temperatures

(Courtesy of Resolution Performance Products, Inc.)

In the next panel, both resin and mold were heated at 80ºF (27ºC) overnight (approximately 20
hours) in the oven and impregnation was then performed at a room temperature of 70ºF (21ºC).
Random air pockets were still observed, but their quantity was reduced.

It was concluded that the sudden temperature drop from 80ºF (27ºC) to 70ºF (21ºC) could have
aggravated the problem. Thus, for the next panel, both the resin and mold were heated at 80ºF
(27ºC) and processing was performed at 80ºF (27ºC). The quantity of random air pockets was
significantly reduced, as shown in figure 2-13. Therefore, efforts were continued in this
direction.

2-17
 Figure 2-13. Surface Defects—Processing at 80ºF (27ºC)

The series of composite flat panels were fabricated to achieve better quality panels. Table 2-4
explains the various experimental combinations. Different process parameters were varied to
observe their effects on the quality of the panel. The better quality panel was one that had a
higher fiber volume fraction, uniform thickness, and almost a zero void content.

2-18
 Table 2-4. Experimental Data for VARTM-Processed Panels Using EPON 9504 Resin

Serial No. Parameter Panel 1 Panel 2 Panel 3 Panel 4

1 Duration of heating ~20 hours/ ~20 hours/ No ~8 hours/
mold and temperature 80ºF (27ºC) 120ºF (49ºC) 120ºF (49ºC)
2 Duration of heating ~20 hours/ ~20 hours/ ~20 hours/ ~20 hours/
resin and temperature 80ºF (27ºC) 120ºF (49ºC) 120ºF (49ºC) 120ºF (49ºC)
3 Duration of heating 4 hours/ 4 hours/ 1 hour/ 1 hour/
curing agent and 80ºF (27ºC) 80ºF (27ºC) 90ºF (32ºC) 80ºF (27ºC)
temperature
4 Temperature at the 80ºF (27ºC)/ 115ºF (46ºC)/ 110ºF (43ºC)/ 89ºF (32ºC)/
time of mixing 80ºF (27ºC) 80ºF (27ºC) 90ºF (32ºC) 77ºF(25ºC)
(resin/curing agent)
5 Temperature after ------ 110ºF (43ºC) 105ºF (41ºC) 84ºF (29ºC)
mixing
6 Degassing 15 min/ 15 min/ No *10 min/
time/vacuum 10-15 torr 10-15 torr Degassing 10-15 torr
7 Temperature after 85ºF (29ºC) 118ºF (48ºC) 116ºF (47ºC) 93ºF (34ºC)
degassing
8 Number of times 1 3 ----- 3
bubbles broken
9 Double bagging Yes Yes No No
10 Bottom peel ply Yes No No No
11 Impregnation time <10 min <10 min <15 min <22 min
12 Green cure ~2 hours/ ~2 hour/ 16 hours/ ~2 hour/120ºF
80ºF (27ºC) 120ºF (49ºC) Room (49ºC) followed by
followed by 8 followed by 8 temperature 8 hour ramp down
hour ramp hour ramp
down down
13 Postcure 200ºF (93ºC)/ 200ºF (93ºC)/ 200ºF (93ºC)/ 200ºF (93ºC)/
1 hour 1 hour 1 hour 1 hour
14 Inclination of mold Yes Yes No Yes
15 Control on flow of No No No **Yes
resin
16 Observations ----- Exothermic# Exothermic# ----
17 Thickness near resin 0.157 (3.98)/ 0.164 (4.17)/ 0.170 (4.32)/ 0.147 (3.73)/
line/vacuum line, 3.53 (0.139 0.172 (4.37) 0.152(3.86) 0.131 (3.33)
inch (mm)
18 Overall fiber volume 0.47 0.45 0.46 0.53
fraction (density
method)
17 Problems Random air OK Random air Top surface tows
pockets at top pockets at top not impregnated
surface surface fully

2-19
Table 2-4. Experimental Data for VARTM-Processed Panels Using EPON 9504 Resin
(Continued)

Serial No. Parameter Panel 5 Panel 6 Panel 7

1 Duration of heating ~20 hours/ ~20 hours/ ~20 hours/
mold and temperature 120ºF (49ºC) 120ºF (49ºC) 120ºF (49ºC)
2 Duration of heating ~20 hours/ ~20 hours/ ~20 hours/
resin and temperature 120ºF (49ºC) 120ºF (49ºC) 120ºF (49ºC)
3 Duration of heating 1 hour/ 1 hour/ 1 hour/
curing agent and 80ºF (27ºC) 80ºF (27ºC) 80ºF (27ºC)
temperature
4 Temperature at the 104ºF (40ºC)/ 103ºF (39ºC)/ 114ºF (46ºC)/
time of mixing 82ºF (28ºC) 87ºF (31ºC) 87ºF (31ºC)
(resin/curing agent)
5 Temperature after 100ºF (38ºC) 100ºF (38ºC) 110ºF (43ºC)
mixing
6 Degassing 13 min/ 10 min/ 10 min/
time/vacuum 10-15 torr 10-15 torr 10-15 torr
7 Temperature after 113ºF (45ºC) 110ºF (43ºC) 114ºF (46ºC)
degassing
8 Number of times 2 2 2
bubbles broken
9 Double bagging No No No
10 Bottom peel ply Yes Yes Yes
11 Impregnation time <10 min <10 min <10 min
12 Green cure ~2 hours/120ºF ~2 hours/120ºF ~2 hours/120ºF (49ºC)
(49ºC) followed (49ºC) followed followed by 8-hour
by 8-hour ramp by 8-hour ramp ramp down
down down
13 Postcure 200ºF (93ºC)/ 200ºF (93ºC)/ 200ºF (93ºC)/
1 hour 1 hour 1 hour
14 Inclination of mold Yes Yes Yes
15 Control on flow of No No No
resin
16 Observations ----- ----- Exothermic#
17 Thickness near resin 0.157 (3.98)/ 0.134 (3.40)/ 0.136 (3.45)/
line/vacuum line, 0.139 (3.53) 0.131 (3.33) 0.131 (3.33)
inch (mm)
18 Overall fiber volume 0.54 0.58 0.52
fraction (density
method)
17 Problems Random air OK OK
pockets at
bottom surface
Note: *Resin was degassed separately for 15 minutes and bubbles were broken three times.
**The resin line was partly clamped and then opened slowly.
#
Exothermic reaction burned the resin flask when resin was in the flask for 15 minutes.
Panels 1-5 were manufactured using biaxial braid sleeves or tubes, and panels 6-7 were manufactured
using biaxial braid slit sleeves.

2-20
The proper combination of process parameters was finally achieved. This combination
consistently produced composite panels of good quality. The combination is explained as
follows:

• The resin distribution media was cut 0.75 in. (19 mm) short from the vacuum line and
0.5 in. (12.7 mm) short along the width side. The vacuum distribution line was placed
below the distribution media and the top peel ply next to the fabrics.

• The mold and the entire resin container were heated at 120ºF (49ºC) for approximately
20 hours overnight in the oven, as shown in figure 2-5.

• The curing agent EPI-CURE 9554 was heated at 80ºF (27ºC) for approximately 1 hour.

• Mixing was completed when the resin was between 110ºF (44ºC) and 120ºF (49ºC), and
the curing agent was between 77ºF (25ºC) and 80ºF (27ºC).

• Mixed resin was carefully degassed for 8 to 10 minutes at approximately 5 to 10 torr.

• The impregnation was performed with the mold at 120ºF (49ºC).

• The composite panel was kept in the mold at 120ºF (49ºC) for 2 hours and then the oven
was turned off. The composite panel remained at room temperature for 16 hours, which
is commonly known as the green cure.

• The composite panels were postcured at 199ºF (93ºC) for 2 hours, as recommended by
the manufacturer.

2.4 OVERALL FIBER VOLUME FRACTION.

It is very important to evaluate the overall fiber volume fraction ( V f Overall ) in the composite
panels after manufacturing. Since the fibers are the main load carrying elements in the
composites, their percentage has a direct effect on mechanical properties of the composites.
Various methods are available to determine the overall fiber volume fraction in composite
panels:

• Ignition Method (ASTM D 2584-68)

• Areal Weight Method (ASTM D 792-86)
• Density Method

2.4.1 Ignition Method.

The resin is burned off in a high-temperature oven. The ash is rinsed from the remaining fiber
(using acetone or alcohol) and the fiber is dried and weighed. The volume of the fiber is
calculated by dividing the mass of the fiber by the density of the fiber material. This method
cannot be used with carbon fiber because carbon oxidizes at elevated temperatures.

2-21
2.4.2 Areal Weight Method.

The fiber volume fraction is determined from the areal weight of the reinforcing fabric and the
volume of the composite using the following relationship.

(V ) Fiber (n ∗ W ∗ A) / ρ f n ∗ W
V f Overall = = =
(V )Composite A∗t t∗ρf

where

(V) Fiber = Volume of the fiber material in the specimen

(V) Composite = Volume of the composite specimen
n = Number of layers or plies in the composite specimen
W = Areal weight of the fabric
A = Cross-sectional area of the composite specimen
t = Thickness of the composite specimen
ρf = Mass density of fiber material

The areal weight changes in the case of braided fabric according to the braid angle. It is very
difficult to maintain a constant braid angle in the VARTM process when braided tubes are used
as fabric. It is also necessary to know the accurate relationship between areal weights and the
braid angle. Further, this method may produce wrong answers due to errors in the braid angle
measurement and the areal weight-braid angle relationship.

2.4.3 Density Method.

The fiber volume fraction is determined from the densities of the composite, assuming that voids
are negligible or less than 1% (Daniel and Ishai, 1994). The density of braided composite,
postcured resin, and carbon fibers are found by using the techniques explained by
ASTM D 792-86. The expression for fiber volume fraction based on the density of the
composite is:

ρc − ρm
V f Overall =
ρ f − ρm

where ρ f , ρ m , ρ c = Densities of fiber, matrix, and composite.

This method is easy to implement and, therefore, was used to calculate the overall fiber volume
fraction of the braided composites manufactured in this research. Typically, five 0.5- by 1-in.
(12.7- by 25.4-mm) specimens or smaller were cut from each panel and their densities were
evaluated. The densities of carbon fibers, cured DM 411-350 vinyl ester resin, and cured EPON
9504 epoxy resin was 1.79 g/cm3, 1.13 g/cm3, and 1.134 g/cm3, respectively. The sample
calculations pertaining to the carbon/epoxy braided composite panel (braid angle 45º) are
presented in table 2-5. The overall fiber volume fraction was calculated using the above
equation. It was observed that the average overall fiber volume fraction for this panel was 0.53.

2-22
In general, the composite panels manufactured by the VARTM process provide a fiber volume
fraction of 0.5.

Table 2-5. Overall Fiber Volume Fraction for Braided Composites (Braid Angle 45° ±1°)

Density of the
Specimen Specimen
No. (g/cm3) Overall Fiber Volume Fraction
1 1.488 0.54
2 1.462 0.50
3 1.485 0.54
4 1.476 0.52
5 1.468 0.52
Average 0.53

The VARTM-processed composite panels exhibited almost 12% higher thickness near the
vacuum line compared to the thickness near the resin line. The effects of thickness and fiber
volume fraction on mechanical properties are discussed in the next section.

2.5 FIBER VOLUME FRACTION IN THE TOW.

The fiber volume fraction in the tow ( V f Tow ) is also an important characteristic in finite element
modeling. Micrographs were used to observe the shape of the tow, and a software package
(Image-Pro) was used to measure the tow area. Figure 2-14 shows the scanning electron
microscope (SEM) micrograph of tows. The diameter of the carbon fiber was measured as 7 µm
using the SEM micrograph shown in figure 2-15. The fiber volume fraction in tow varied from
0.66 to 0.72 in the braided composites. Therefore, an average value of 0.69 was used for the
further calculations.

Figure 2-14. Micrographs of Braided Tows Using SEM

2-23
 Figure 2-15. An SEM Micrograph of Fibers (Fiber diameter 7 μm)

The sample calculations for the tow fiber volume fraction are listed below.

• Average braid angle = 25º

• Average tow area = 0.7361 mm2
• Area of fiber (φ = 7 μm) = π/4* (7*10-3 )2 = 3.848*10-5 mm2
• Area of fiber at angle 25º = 3.848*10-5/cos (25º) = 4.246*10-5 mm2
• Area of 12000 fibers (12K/tow) = 0.5095 mm2
• Fiber volume in the tow fraction = 0.5095/0.7361 = 0.69

2.6 SUMMARY OF SECTION 2.

The following conclusions were drawn from the experimental study regarding VARTM
processing in general and braided composites in particular:

1. When EPON 9504 resin was formulated at 120ºF (49ºC), its gel time reduced to
approximately 25 minutes. This fact limited the size of the component and added further
costs to the processing.

2. The use of a flow control device in the case of DM 411-350 and the heating of the entire
mold for a prolonged period of 20 hours in the case of EPON 9504 increases the cost of
processing.

3. The VARTM processed composite panels varied in thickness, which in turn caused fiber
volume fraction variation. The thickness of the panels near the vacuum line was greater
than that at the resin line. Thickness variation was due to the drop in pressure
differential. Initially, the pressure differential was at a maximum level and later dropped
as resin flowed in the mold.

2-24
4. It is sometimes claimed that the use of two vacuum bags (termed double bagging)
improves fiber volume fraction. However, there was no effect on fiber volume fraction
from the double bagging in this research. However, double bagging definitely helped to
maintain a leak-proof bag.

5. The use of a bottom release fabric is actually optional. However, if it is not used, then
the final bottom surface is shiny. This shiny surface needs to be roughened using
sandpaper before tabbing. Even with a properly roughened surface, the tabs slipped in
the grip in this study. This meant that the shear strength of the adhesive between the tab
and the specimen became less than the applied load. Therefore, to avoid the problem of
tab slipping, the use of top and bottom peel plies is strongly recommended.

6. Although VARTM is a simple process, as far as equipment is concerned, it is quite

complicated and has many variables. Designing the proper VARTM process for a
particular material combination is a challenge. Some of the important variables are listed
below:

a. Processing temperature and viscosity of resin

b. Temperature of mold
c. Fabric architecture (type of weave and size)
d. Degassing
e. Proper placement of resin distribution media
f. Placement of resin and vacuum distribution lines
g. Use of flow control device

7. It was recently reported by Bolick and Kelkar (2004) that EPON 9504 produces panels of
good quality if processing is completed in a room where temperature is continuously
maintained at 80ºF (27ºC). The fabric used in this study was a carbon plain-woven fabric
with a 3 K tow size, manufactured by BGF Industries, Inc. It is interesting that the same
fibers in braided fabric with a 12 K tow size have a wetting problem with the same resin.
Therefore, it is concluded that the size of tow and fabric crimp may have an effect on
wetting.

8. Further investigation regarding the use of an air release agent or wetting agent is highly
recommended to successfully perform processing at a room temperature of 77ºF (21ºC).

The braided composite panels were cut into test coupons according to ASTM standards. Section
3 discusses the performance evaluation of the braided composites in static and tension-tension
fatigue loading.

2-25/2-26
3. PERFORMANCE EVALUATION.

3.1 INTRODUCTION.

Section 2 discussed the low-cost VARTM manufacturing process for biaxial braided composites
in detail. This section discusses the performance evaluation of the braided composites under
static and tension-tension fatigue loading. The effect of the braid angle on mechanical properties
and fatigue life was also studied. All the static and fatigue tests were performed according to the
ASTM standards discussed below.

3.1.1 Static Tensile Tests.

Static tensile tests were performed according to ASTM D 3039/D 3039M, “Standard Test
Method for Tensile Properties of Fiber Resin Composites.” The in-plane tensile properties like
ultimate tensile strength (UTS) or strain at UTS (Su), longitudinal tensile modulus, and Poisson's
ratio were evaluated using the ASTM D 3039/D 3039M standard. The axial extension was
measured by an extensometer and transverse strain was measured by a strain gage. All static
tensile tests were conducted in the displacement control mode with a cross lead rate of
0.05 in./min (1.27 mm/min).

3.1.2 Tension-Tension Fatigue Tests.

Tension-tension fatigue tests were performed according to ASTM D 3479/D 3479M, “Standard
Test Method for Tension-Tension Fatigue of Polymer Matrix Composite Materials.” This test
method is limited to unnotched test specimens subjected to constant-amplitude uniaxial in-plane
loading that is defined in terms of a test control parameter (defined by Procedure A and B of
ASTM D 3479/D 3479M).

• Procedure A (Load Controlled)—A system in which the test control parameter is the load
(stress) and the machine is controlled so that the test specimen is subjected to repetitive
constant amplitude load cycles.

• Procedure B (Displacement Controlled)—A system in which the test control parameter is

the strain in the loading direction and the machine is controlled so that the test specimen
is subjected to repetitive constant-amplitude strain cycles.

All tension-tension fatigue tests were conducted according to Procedure A. Spectrum loading
(variable amplitude) is also often used to simulate real-life loading conditions. In constant-
amplitude load-controlled tests, stress is applied as a percentage of UTS. The stress was applied
from 80% of UTS and reduced in steps of 10% until specimens survived 1 million cycles.
Typically, the stress level at infinite life is referred to as the fatigue limit or endurance limit, but
for most engineering purposes, infinite life is usually considered between 1 million and
10 million cycles. In this research, endurance limit refers to the stress level at 1 million cycles.

3-1
The other test parameters selected were sinusoidal waveform, 10-Hz frequency, and 0.1 stress
ratio. Stress ratio (R) is defined as

σ min Pmin
R= =
σ max Pmax

where:

σmin = Minimum stress and corresponding load is Pmin

σmax = Maximum stress and corresponding load is Pmax

The test methodology in the present research is to define methodology to accelerate testing time.
Accelerated testing consists of test methods that deliberately shorten (in a measured way) the life
of the tested product or accelerate the degradation of the product’s performance. Most
accelerated testing is performed on materials or products to characterize their degradation
mechanisms (e.g., fatigue, creep, cracking, wear, corrosion/oxidation, and weathering (Battat,
http://amptiac.alionscience.com/pdf/2001MaterialEase16.pdf)). In accelerated fatigue testing,
the frequency is limited to 5-10 Hz for polymer composite materials because of their viscoelastic
nature. These materials may show dramatic change in fatigue life and different failure
mechanisms at lower temperatures. The FAA recommended a frequency of 10 Hz and an R ratio
of 0.1 was recommended by the FAA. Most of the primary structures of aircraft, such as the
fuselage, experience tension-tension fatigue. The fuselage is pressurized with air when the plane
reaches a high altitude and the pressure is released when the plane reaches the ground. When the
fuselage is pressurized it stretches, and when pressure is released the fuselage contracts. Thus, it
experiences tension-tension fatigue.

3.1.3 Specimen Geometry.

The geometry of the test specimens was identical for both tension and fatigue tests, as shown in
figure 3-1. The length and width of the specimens were 10 in. (254 mm) and 1 in. (25.4 mm),
respectively. The thickness varied between 0.14 in. (3.55 mm) and 0.16 in. (4 mm), depending
upon the braid angle. All the specimens were tabbed to avoid failure in the grip.

10 in.

2 in.

1 in.

Figure 3-1. Tensile Test Specimen Geometry

3-2
The specimens that met the requirements of the ASTM D 3039/D 3039M and the
ASTM D 3479/D 3479M standard were selected. The critical requirements of the specimens
were a width tolerance of ±1% and a thickness tolerance of ±4%.

3.1.4 Dynamic/Fatigue Testing System.

All the tests were performed on a Dynamic/Fatigue Testing System, as shown in figure 3-2. The
capacity of the load frame was 490 kN (110 kips), and the capacity of the load cell was 245 kN
(50 kips). The accuracy of the load cell was ±8.9 N (2 lb). The hydraulic grips in which the
specimen was clamped can apply a maximum pressure of 10 ksi (69 MPa) with a grip capacity of
300 kN (67.5 kips). The machine was capable of conducting static tensile, static compression,
flexural or bending, and fatigue tests. It was controlled by the use of Max software developed by
the INSTRON Corporation.

Load cell

Fixed
head

Hydraulic
Moving
power
supply
Tower Load frame
console

Figure 3-2. Dynamic/Fatigue Testing System

3.1.5 Test Matrix.

Table 3-1 presents the test matrix used for the experimental study. A sufficient number of
specimens needed to be tested for statistical relevance. Since the manufacturing of composites,
preparation of specimens (cutting and tabbing) and fatigue testing take a considerable amount of
time and effort, only the minimum numbers of specimens necessary for statistical considerations
were used.

3-3
 Table 3-1. Test Matrix

Tensile Test Fatigue Test

Material System Braid Angle Specimens Specimens
Neat Resin EPON 9504 - 2 -
Carbon/Vinyl Ester 25 5 15
Carbon/Epoxy 25 5 15
Carbon/Epoxy 30 5 15
Carbon/Epoxy 45 5 15
Frequency Effect 30 -- 5
Stiffness Degradation Study 25 -- 5
Total 22 70

Once the biaxial braided composite panels were postcured, specimens were cut according to the
plan shown in figure 3-3. The side view of this figure shows the thickness variation in the
composite panels processed by VARTM. The issues related to thickness variations are discussed
in subsequent sections. Ten tension/fatigue specimens and five overall fiber volume fraction
specimens were typically cut from each panel.

Specimens for
overall fiber
volume fraction
Vacuum Line t2

Specimens for
tension/fatigue
tests

Resin Line t1

Figure 3-3. Plan of Test Specimens From Panel (Thickness varies from t1 to t2)

3.1.6 Tensile Test Results of Neat EPON 9504 Epoxy Resin.

The properties of DM 411-350 vinyl ester resin and EPON 9504 epoxy resin are presented in
table 2-2. Previous research of the DM 411-350 vinyl ester resin data, found that tensile
properties were readily available (Smith, 2001). The tensile properties of EPON 9504 epoxy
resin were evaluated by static tension tests on neat resin coupons according to ASTM D 638.
The resin was mixed with the curing agent in the recommended proportion and degassed. Then,
it was poured into a 12-in.-wide by 12-in.-long (25.4- by 25.4-cm) steel mold and kept for almost
16 hours at a room temperature of 77ºF (25ºC) that is termed green cure. After the green cure,
the panel was kept in the oven for postcure at 200ºF (93ºC) for 1 hour as recommended by the

3-4
manufacturer. After postcuring, the dog bone specimens were cut and tested according to
ASTM D 638. This ASTM D 638 standard recommends selecting a test speed to produce
rupture in 1/2 to 5 minutes. The selected speed of testing was 0.2 in./min (5 mm/min). The axial
and lateral strains were measured using an extensometer and a strain gage, respectively. The
geometry of the test specimens is shown in figure 3-4, and the stress-strain diagram is presented
in figure 3-5. The average tensile strength, axial modulus, and Poisson’s ratio were 9.42 ksi
(64.94 MPa), 2.96 Msi (0.43 GPa), and 0.38, respectively.

Figure 3-4. Specimen Geometry for Neat Resin Coupon (ASTM D 638)

10000

8000
Stress, psi

6000

4000

2000

0
0.00 0.02 0.04 0.06 0.08 0.10
Strain, in/in

Figure 3-5. Stress-Strain Diagram of EPON 9504 Epoxy Neat Resin

3-5
3.2 TENSILE AND FATIGUE TEST RESULTS OF BRAIDED COMPOSITES
MANUFACTURED USING BRAIDED TUBES.

In the first stage of the research, biaxial braided composite panels were manufactured using
braided tubes and two different resin systems, DM 411-350 vinyl ester and EPOPN 9504 epoxy.
Braided tubes were flattened and their width was maintained at 1.75 in. (44.5 mm) to ensure a
braid angle of approximately 25°. Two braided tubes were stacked on top of each other to obtain
four layers. The braid angle of all the specimens was measured on a micrograph using Image-
Pro software, as shown in figure 1-3. It was found that the braid angle varied from 24° to 32°.
This could be due to increased flattening of the braided tubes after application of the vacuum in
VARTM. Specimens whose braid angle fell in the range of 25° ±2° were selected. All the
selected specimens met the requirements of the ASTM standard of width and thickness tolerance
as discussed in section 3.13.

3.2.1 Carbon/Vinyl Ester Braided Composites.

The overall fiber volume fraction in the carbon/vinyl ester braided composites panel was
0.52 ±0.04. Table 3-2 lists the results of tensile tests for five specimens. The “s” represents the
sample standard deviation, and CV is the sample coefficient of variation in percentage terms. Ex
was computed based on a linear regression fit to the stress-strain data over the 0.1% to 0.3%
strain range. Experimental error was calculated using the student-t distribution at a confidence
level of 95%.

Table 3-2. Tensile Test Results of Carbon/Vinyl Ester Braided Composites

(Braid Angle 25° ±2°)

UTS Ex
Specimen ksi MPa Msi GPa % εut νxy
1 99.11 683.31 10.35 71.33 1.11 1.22
2 88.84 612.52 9.94 68.52 0.98 1.17
3 92.62 638.59 9.24 63.71 1.16 1.16
4 106.40 733.52 10.18 70.18 1.24 1.11
5 100.70 694.22 10.48 72.25 1.04 1.02
Average 97.53 672.37 10.04 69.22 1.19 1.14
s 6.91 47.62 0.49 3.38 0.10 0.08
CV 7.08 7.08 4.88 4.88 8.15 6.66
Experimental ±8.58 ±59.12 ±0.61 ±4.20 ±0.13 ±0.10
Error

It was assumed that the properties in table 3-2 represented the entire panel. Fatigue tests were
conducted to obtain the stress level at 1 million cycles. The test parameters were those explained
in section 3.1.2. Three specimens were tested at each stress level. The fatigue test data are
tabulated in table 3-3.

3-6
 Table 3-3. Fatigue Test Results for Carbon/Vinyl Ester Braided Composite
(Braid Angle 25° ±2°)

% of UTS 1 2 3
85 146 1416 5560
75 1887 16590 570
65 3854 98830 2540
55 34451 13430 11607
45 1000000* 107340 39733
40 1000000* 1000000* 1000000*

* No failure, run-out

3.2.2 Carbon/Epoxy Braided Composites.

The overall fiber volume fraction in the carbon/epoxy braided composites panel was 0.51 ±0.05.
Table 3-4 lists the tensile test results, and table 3-5 presents fatigue test data. The stress-strain
diagram and the stress-fatigue life diagram (S/Su-N) for both carbon/vinyl ester and carbon/epoxy
composites are displayed in figures 3-6 and 3-7, respectively.

Table 3-4. Tensile Test Results of Carbon/Epoxy Braided Composites

(Braid Angle 25° ±2°)

UTS Ex
Specimen ksi MPa Msi GPa % εut νxy
1 94.780 653.41 7.440 51.29 1.860 1.460
2 98.780 681.00 7.930 54.67 1.990 1.440
3 96.550 665.62 7.850 54.12 1.960 1.460
4 85.800 591.50 7.000 48.25 1.680 1.400
5 102.800 708.70 9.120 73.52 1.350 1.470
Average 95.740 660.03 7.870 54.26 1.770 1.440
s 6.314 43.53 0.792 5.46 0.263 0.027
CV 6.600 6.60 10.060 10.06 14.850 1.880
Experimental ±7.840 ±54.04 ±0.980 ±6.78 ±0.330 ±0.030
Error

3-7
 Table 3-5. Fatigue Test Results for Carbon/Epoxy Braided Composite
(Braid Angle 25° ±2°)

% of UTS 1 2 3
80 516 434 274
70 986 693 839
60 5575 2258 1225
50 4807 4372 3470
45 24135 1000000* 1000000*
40 1000000* 1000000* 1000000*

* No failure, run-out

o
Braid 25 Vinyl Ester
100
o
Braid 25 Epoxy
Stress, ksi

80

60

40

20

0
0.000 0.004 0.008 0.012 0.016 0.020
Strain, in/in

Figure 3-6. Stress-Strain Behavior of Braided Composites Using Flattened Braided Tubes
(Braided Angle 25° ±2°)

3-8
 1.1 0
Braid Angle 25 Epoxy
0
Braid Angle 25 Vinyl Ester
1.0
0.9
0.8
S/Su

0.7
0.6
0.5 Runout

0.4
0.3
0 1 2 3 4 5 6
Log 10 (N)

Figure 3-7. S-N Diagram for Braided Composites Using Flattened Braided Tubes Showing
Large Scatter in the Fatigue Data (Braid Angle 25° ±2°)

3.2.3 Braided Tube Findings.

• It was found that the UTS for both the carbon/vinyl ester and the carbon/epoxy braided
composites were almost identical at a braid angle of 25º. However, carbon/vinyl ester
composites were stiffer than carbon/epoxy composites. Carbon/epoxy composites had
excessive nonlinear behavior and a higher strain at UTS. Poisson’s ratio was found to be
greater than 1 for the braided composites because of their anisotropic nature.

• The CV values in tables 3-2 and 3-4 generally demonstrate that UTS and failure strain
measurements exhibited less repeatability and reproducibility than stiffness parameters.

• The student-t distribution indicated an experimental error of ±8.58 ksi (±59.12 MPa) in
the case of the carbon/vinyl ester composites and ±7.84 ksi (±54.04 MPa) for the
carbon/epoxy composites. This was a very significant error and was mainly due to the
heterogeneity of the material as opposed to the tensile test itself. Increasing the sample
size and studying the repeatability of the manufacturing process and material behavior
can reduce this error.

• Fatigue data of braided composites exhibited large scatter, especially for carbon/vinyl
ester composites. The stress was applied as a percentage of UTS in load-controlled
fatigue tests. Thus, the experimental error in UTS was the main cause of large scatter in
the fatigue data.

3-9
• Since the failure strain of carbon/epoxy composites was greater than that of the
carbon/vinyl ester braided composites, one might expect higher fracture toughness
(damage tolerance) and better performance in fatigue and impact for these composites.
However, the carbon/vinyl ester braided composites exhibited better fatigue performance
at higher loads. Both composites survived 1 million cycles at a fatigue load of 40% of
UTS.

Fibers are the major load carrying members in composites. Thus, the fiber volume fraction has a
direct effect on mechanical properties including UTS. The literature review also indicates that
braid angle variation drastically changes in-plane tensile properties including UTS. The next
section discusses the effect of fiber volume fraction and braid angle variation on UTS.

3.3 EFFECT OF FIBER VOLUME FRACTION AND BRAID ANGLE VARIATIONS

ON UTS.

The large scatter in the fatigue data was mainly due to the scatter of UTS. The UTS was directly
affected by variations in fiber volume fraction and braid angle. The main causes of the effects
were the limitation of the VARTM process and misalignment of braided tubes in the mold during
processing.

3.3.1 Effect of Fiber Volume Fraction on UTS.

Even with the proper design and placement of distribution media and the proper placement of
vacuum and resin lines, the thickness of VARTM processed composites varied from the resin
line to the vacuum line. The thickness increased by 12% at the vacuum line compared to the
thickness at the resin line. The thickness of the panel was a function of the instant of time when
a particular location was impregnated with resin. Thickness variation was also due to the drop in
pressure differential. When resin enters a mold, the pressure differential is at a maximum, and as
resin flows forward in a mold, the pressure drops. Therefore, VARTM manufactured flat panels
are thicker at the vacuum line. Thickness variations within a panel are one of the limitations of
the VARTM process. The greater thickness at the vacuum line results in a local resin rich area.
Thus, fiber volume fraction (Vf) varies from location to location. Vf is at a maximum near the
resin line and at a minimum near the vacuum line.

According to simplified micromechanics equations (Chamis, 1984), the longitudinal tensile

strength of unidirectional lamina is

Slt ~ Sft * Vf

where:

S1t = Longitudinal tensile strength of unidirectional lamina,

Sft = Tensile strength of fiber, and
Vf = Fiber volume fraction

The carbon/vinyl ester and carbon/epoxy braided composites manufactured by VARTM

exhibited fiber volume fractions of 0.52 ±0.04 and 0.53 ±0.05, respectively. This variation in Vf

3-10
may alter UTS as much as ±9%. It should be noted that the above equation is not applicable to
braided composites. This is because tows in the braided composites, unlike laminates, are
mechanically interlocked and have undulations. The purpose of the above equation is to show
the effect of Vf on UTS. Thickness variation (in turn Vf variation) in VARTM was inevitable and
could not be controlled to less than ±6% in this research.

3.3.2 Effect of Braid Angle on UTS.

When braided tubes were used to manufacture braided composites in the VARTM process, the
braid angle varied from 24º to 32º within the panel. The braid angle variation was caused by the
misalignment of braided fabric. This misalignment could occur while stacking braided tubes one
above the other, or it could occur at high vacuum (0.5 torr) during vacuum bagging. Almost all
the in-plane properties were affected by the braid angle, including UTS. As the braid angle
increased, UTS decreased exponentially, as shown in figure 3-8. When the braid angle increased
from 25º to 30º, UTS dropped from 95.74 ksi (660.03 MPa) to 64.09 ksi (441.84 MPa). The 5º
decline in the braid angle caused a decline of approximately 33% in the UTS. Braid angle must
be controlled because it has a significant effect on UTS.

100

80
UTS, ksi

60

40

20

25 30 35 40 45
Braid Angle, degrees

Figure 3-8. Ultimate Tensile Strength Decreasing Exponentially With an Increase

in Braid Angle

Braid angle can be controlled if slit sleeves are used instead of braided tubes. As discussed in
section 2, the slit sleeve is a special braid manufactured with a constant braid angle throughout
the braided fabric. Braided tubes are manufactured at required braid angles and then slit to
produce slit sleeves. The edges are fused with a small amount of EVA adhesive. The adhesive
maintains a constant braid angle even during handling. Slit sleeves with fixed braid angles of 30º
and 45º were used for further study. A change in the braid angle in braided tubes does not affect
the fiber volume fraction of braided composites (appendix A).

3-11
3.4 TENSILE AND FATIGUE TEST RESULTS OF BRAIDED COMPOSITES
MANUFACTURED USING SLIT SLEEVES.

In a later phase of the research, biaxial braided composite panels were manufactured using slit
sleeves with 30° and 45° braid angles and EPOPN 9504 epoxy resin. Four slit sleeves were
stacked to obtain four layers. Braid angle variation with the slit sleeves was ±2° within each
panel. Specimens were selected whose braid angle fell at 30° ±1º and 45° ±1º. All the selected
specimens met the requirements of ASTM standards for width and thickness tolerance, as
discussed in section 3.1.3.

3.4.1 Carbon/Epoxy Braided Composites (Braid Angle 30° ±1º).

The overall fiber volume fraction was 0.53 ±0.05 for 30º slit sleeves. Table 3-6 shows the results
of tensile tests, and table 3-7 presents fatigue test data.

Table 3-6. Tensile Test Results for Carbon/Epoxy Braided Composites

(Braid Angle 30° ±1°)

UTS Ex
Specimen ksi MPa Msi GPa % εut νxy
1 66.97 461.74 6.02 41.51 2.42 1.33
2 64.74 447.69 5.74 39.57 2.61 1.29
3 62.48 430.73 5.07 34.95 2.48 1.31
4 66.03 455.21 5.36 36.95 2.60 1.42
5 60.23 415.23 5.39 37.16 2.19 1.33
Average 64.09 441.84 5.52 38.05 2.46 1.34
s 2.737 18.87 0.369 2.54 0.171 0.050
CV 4.27 4.27 6.67 6.67 6.95 0.04
Experimental ±3.40 ±23.42 ±0.46 ±3.15 ±0.22 0.06
Error

Table 3-7. Fatigue Test Results for Carbon/Epoxy Braided Composite

(Braid Angle 30° ±1°)

% of UTS 1 2 3
80 450 475 392
70 578 910 660
60 1880 1910 2240
50 6020 3688 4562
45 5333 12836 7770
40 1000000* 1000000* 1000000*

*No failure, run-out

3-12
3.4.2 Carbon/Epoxy Braided Composites (Braid Angle 45° ±1°).

The overall fiber volume fraction was 0.52 ±0.03 for 45º slit sleeves. Table 3-8 shows the tensile
test results, and table 3-9 presents fatigue test data. The stress-strain diagram and the stress-
fatigue life diagram (S/Su–N) epoxy carbon/epoxy composites are shown in figures 3-9 and 3-10,
respectively.

Table 3-8. Tensile Test Results for Carbon/Epoxy Braided Composites

(Braid Angle 45° ±1°)

UTS Ex
Specimen ksi MPa Msi GPa % εut νxy
1 25.31 174.48 2.15 14.82 2.16 0.70
2 24.32 167.68 2.03 13.99 2.08 0.68
3 24.62 169.73 2.07 14.27 1.98 0.69
4 24.03 165.66 2.63 18.13 2.06 0.69
5 30.9 213.03 2.67 18.41 2.03 0.71
Average 25.84 178.14 2.31 15.93 2.06 0.69
S 2.87 19.79 0.32 2.17 0.07 0.013
CV 11.11 11.11 13.62 13.62 3.40 1.88
Experimental ±3.56 ±24.57 ±0.40 ±2.69 ±0.09 ±0.02
Error

Table 3-9. Fatigue Test Results for Carbon/Epoxy Braided Composite

(Braid Angle 45° ±1°)

% of UTS 1 2 3
80 369 1533 967
70 3507 29475* 2339
60 8236 9031 139329*
50 1000000+ 1000000+ 1000000+

* Not included in the graph

+
No failure, run-out

3-13
 100
o
Braid Angle 25
80

Stress, ksi
60
o
Braid Angle 30

40
o
Braid Angle 45
20

0
0.000 0.008 0.016 0.024
Strain, in/in

Figure 3-9. Stress-Strain Behavior of Carbon/Epoxy Braided Composites

(Braid Angles 25° ±2°, 30° ±1°, and 45° ±1°)

0
1.1 Braid 45 Epoxy
0
Braid 30 Epoxy
1.0
0.9
0.8
S/Su

0.7
0.6
Runout
0.5
0.4
0.3
0 1 2 3 4 5 6
Log 10 (N)

Figure 3-10. S-N Diagram for Braided Composites With Slit Sleeves Showing Less Scatter in
the Fatigue Data (Braid Angles 30° ±1° and 45° ±1°)

3-14
3.4.3 Slit Braided Sleeve Findings.

Table 3-10 presents comparisons of the tensile properties for different material systems. The
following are a few of the conclusions from the tensile test data:

• Experimental error in the UTS was reduced after the use of slit sleeves. The
experimental error values were ±3.4 ksi (±23.42 MPa) and ±3.56 ksi (±24.57 MPa) for
braided composites with braid angles of 30º and 45º, respectively.

• Fatigue data had less scatter with less error in the UTS. Thus, it can be concluded that
the braid angle variation within a specimen changes the UTS among the specimens and
created large scatter, especially in load-controlled fatigue tests. The use of slit sleeves is
highly recommended for the fatigue study of braided composites to control braid angles
within a specimen.

• Classical laminate theory is well developed and accepted for multidirectional laminates.
The biaxial braided composites resemble angle ply laminates [±θ]s. Therefore, it is
useful to compare the trend in Ex and νxy of biaxial braided composites with that of angle-
ply laminates. Figure 3-11 shows Young’s modulus, and figure 3-12 shows the Poisson’s
ratio of UD laminate and an angle ply laminate as a function of fiber orientation. As the
angle increases, longitudinal modulus decreases and increasingly becomes a matrix-
dominated property. Poisson’s ratio increases with an angle up to 25º and then decreases.
Poisson’s ratio is greater than 1 with angles in the range of 15º to 35º. Similar behavior
was observed in the case of braided composites. Ex was a fiber-dominated property and
νxy was a matrix-dominated property. The higher Poisson’s ratio of carbon/epoxy
composites could be solely due to the higher Poisson’s ratio of the epoxy itself. As the
braid angle increased from 25º to 45º, Ex decreased, as shown in figure 3-13. The
Poisson’s ratio of the braided composites peaked at 25º and later decreased as the braid
angle increased form 25º to 45º, similar to angle ply laminates as shown in figure 3-14.
Poisson’s ratio was greater than 1 at angle between 25º and 30º and less than 1 for 45º
angles. When the braid angle increased from 25º to 45º, the percentage drop in the UTS,
modulus, and Poisson’s ratio was 75%, 70%, and 96%, respectively.

3-15
 Table 3-10. Comparisons of Tensile Properties of Braided Composites With Different Material
Systems and Braid Angles

Carbon/
Property Vinyl ester Carbon/Epoxy Carbon/Epoxy Carbon/Epoxy
Braid Angle 25° ±2° 25° ±2° 30° ±1° 45° ±1°
Braided Tubes Braided Tubes Slit Sleeves Slit Sleeves
UTS, ksi (MPa) 97.53 95.74 64.09 24.84
(672.37) (660.03) (441.84) (178.14)
εut , % 1.19 1.77 2.46 2.06
E, Msi (GPa) 10.04 7.87 5.52 2.31
(69.22) (54.26) (38.05) (15.93)
νxy 1.14 1.44 1.34 0.69
Fiber Volume Fraction* 0.52 ±0.04 0.51 ±0.05 0.53 ±0.05 0.52 ±0.03
Thickness*, in. 0.162 ±0.009 0.155 ±0.009 0.140 ±0.007 0.129 ±0.005

*These values vary among different panels and not within a single tensile test specimen.

Figure 3-11. Modulus of Angle-Ply Laminates as a Function of Orientation Angle

3-16
Figure 3-12. Poisson’s Ratio of Angle-Ply Laminates as a Function of Orientation Angle
(Daniel and Ishai, 1994)

10
9
8
7
Modulus, Msi

6
5
4
3
2
1
20 25 30 35 40 45 50
Braid Angle, degrees

Figure 3-13. Modulus of Braided Composites as a Function of Braid Angle

3-17
 1.6

1.4

Possion's Ratio
1.2

1.0

0.8

0.6
20 25 30 35 40 45 50
Braid Angle

Figure 3-14. Poisson’s Ratio of Braided Composites as a Function of Braid Angle

• For carbon/epoxy braided composites, the endurance limit at 1 million cycles was 40% of
the UTS for braid angles of 25º and 30º, whereas it was 50% of the UTS for braid angles
of 45º. The fatigue strength values for 25º, 30º, and 45º angles were 38.3 ksi (264 MPa),
25.6 ksi (176.5 MPa), and 12.9 ksi (88.94 MPa), respectively.

• Fiber volume fraction is mainly dependent on the manufacturing process. The

composites manufactured with the VARTM process typically yield a fiber volume
fraction of 0.5. Since slit sleeves were manufactured individually for 30º and 45º angles,
their percentage cover (spacing between two tows) could differ. This could result in a
small variation in the fiber volume fraction (appendix A).

• All braided composites manufactured in this research had four layers. As the braid
angles increased, the thickness of the composite decreased. The thickness variation in
braided composites is quite a complicated issue mainly due to braid angle variation
within panels and the limitation of the VARTM process.

• UTS is sensitive to braid angle and fiber volume fraction. Statistically, UTS can be
expressed as a function of braid angle and fiber volume fraction. This is explained in
appendix B.

3.5 MODELING OF THE S-N DIAGRAM.

There are many models to represent the S-N diagram. The most popular is the simple linear
model represented as a straight line on a semilog plot. The equation of the line is represented as

S = Su (m log N + d)

3-18
The lower the value of the slope (m) and the higher the value of the y-axis intercept (d), the
higher the value of the fatigue strength. Figure 3-15 shows the linear fit of an S-N diagram for
braided composites. Table 3-11 displays a comparison of UD laminates’ and braided
composites’ S-N curves.

0
Braid 45 Epoxy
0
1.1 Braid 30 Epoxy
0
Braid 25 Epoxy
1.0 0
Braid 25 Vinyl Ester
0.9
0.8
S/Su

0.7
0.6
Runout
0.5
0.4
0.3
0 1 2 3 4 5 6
Log 10 (N)

Figure 3-15. Linear Representation of S-N Diagram for Carbon/Vinyl Ester and Carbon/Epoxy
Material System

Table 3-11. Comparison of UD Laminates’ and Braided Composites’ S-N Curve

Material R m d
T300 carbon/ductile epoxy-UD (Amateau, 2003) 0.1 -0.0542 1.0420
T300 carbon/brittle epoxy-UD (Amateau, 2003) 0.1 -0.0873 1.2103
BASF G30-500 carbon/epoxy-3D Triaxial Braid 0.1 -0.0770 1.000
(Carlos, 1994)
AS4/Vinyl Ester Biaxial Braid 25º 0.1 -0.0977 1.0221
AS4/Epoxy Biaxial Braid 25º 0.1 -0.1018 0.9861
AS4/Epoxy Biaxial Braid 30º 0.1 -0.1117 0.9935
AS4/Epoxy Biaxial Braid 45º 0.1 -0.08531 1.0028

Table 3-11 shows that UD laminates have superior endurance limits (at 1 million cycles)
compared to those of the braided composites’ material system. The linear model did not
represent the fatigue of the braided composites. It either underpredicted or overpredicted the
endurance limit. The trend of fatigue data can be best represented by a Sigmoidal (Boltzman) fit,

3-19
as shown in figure 3-16. The R-square values and the predicted fatigue strength at 1 million
cycles are tabulated in table 3-12.

0
1.3 Braid 45 Epoxy
0
Braid 30 Epoxy
1.2 Braid 25
0
Epoxy
0
1.1 Braid 25 Vinyl Ester
1.0 II
0.9
0.8
S/Su

III
0.7 I
0.6
0.5
Critical Zone
0.4
0.3
0 1 2 3 4 5 6
Log 10
(N )

Figure 3-16. Sigmoidal (Boltzmann) Fit for Carbon/Epoxy Material System

(Braid Angles 25° ±2°, 30° ±1°, and 45° ±1°)

Table 3-12. Comparison of Linear Model and Sigmoidal Model Results With
Experimental Results

Applied Stress for N ≥106

2
R Value (% of UTS)
Material System Biaxial Exp. Linear Sigmoidal
Braid (Braid Angle) Linear Sigmoidal (%) (%) (%)
AS4/Vinyl Ester (25º) 0.8603 0.8917 40 44 40
AS4/Epoxy Biaxial (25º) 0.9007 0.9659 40 37 42
AS4/Epoxy Biaxial (30º) 0.8913 0.9874 40 32 40
AS4/Epoxy Biaxial (45º) 0.8893 0.9057 50 49.1 49.7

Sigmoidal curves (S-shaped) are frequently used in biology to describe growth patterns. The
growth in fatigue life with reduction in the applied fatigue load can be described using this curve
(mirrored S-shape) for of braided composites. The equation for the Sigmoidal (Boltzmann)
curve is given by

A1 − A2
y= + A2
1 + e( x − xo )/ dx

3-20
where:

S
y=
Su

x = log10 ( N )

A1, A2, xo, and dx are the parameters in the equation. The value of y varies from A1 at the top to
A2 at the bottom. The dx describes the steepness of the curve and the larger values denote more
shallow curves. The xo indicates the value of x when y is halfway between A1 and A2. The values
of these four parameters for all material systems are listed in table 3-13. The higher values of A2
and dx indicate better fatigue life.

Table 3-13. Parameters in Sigmoidal (Boltzmann) Curve for Braided Composites for Different
Material Systems and Braid Angles

Material System A1 A2 xo dx
AS4/Vinyl Ester 1.0304 0.30124 3.7700 1.2297
Biaxial Braid 25º
AS4/Epoxy 1.0011 0.41909 2.9160 0.50788
Biaxial Braid 25º
AS4/Epoxy 1.0023 0.40005 2.8982 0.47307
Biaxial Braid 30º
AS4/Epoxy 1.0046 0.48598 3.1882 0.72782
Biaxial Braid 45º

The S-N diagram exhibited three major stages when represented in this manner.

• Stage I: In this stage, fatigue life was almost constant and less than 100 cycles until the
applied load was 80% of the UTS. This stage was termed low-cycle (high-stress) fatigue.

• Stage II: In this stage, when the applied load decreased from 80% to 50%, there was no
considerable increase in fatigue life in the laminates. The number of cycles to failure
typically increased to 7000, except for the carbon/epoxy braid (braid angle 45°±1º). The
middle zone is very important in designing composite structures. This stage was termed
midcycle fatigue.

A linear model showed a large error in this region. For example, for the carbon/epoxy
braid (braid angle 25° ±2°) at an applied load of 50% of the UTS, a linear model
predicted a fatigue life of 59,573 cycles, whereas a Sigmoidal model predicted 6,952
cycles. Experimentally, the average fatigue cycles at this load varied from 3,470 to 4,807
cycles. Other examples are shown in table 3-14. Damage accumulation was very fast in
the mid stage.

3-21
• Stage III: Failure cycles jumped from 7000 to 1 million in this stage when applied stress
was decreased by only 10% from 50% to 40%. This indicated that major damage occurs
between the applied loads of 40% and 50% of the UTS. This stage was termed high-
cycle (low-stress) fatigue.

For braided composites, the transition from Stage II to Stage III was very critical. The
major damage occurred in this critical zone. If the component was subjected to overload
for a small period of time in this zone, catastrophic failure could occur. A higher factor
of safety had to be used. Even with this problem, braided composites are preferred
because of their suitability for complex components and their low cost compared to
woven composites.

Even with its complexity, the Sigmoidal (Boltzmann) model is superior because it
captures fatigue life in the critical zone and also in all fatigue stages.

Table 3-14. Fatigue Life Predicted by Linear and Sigmoidal Model at Different Fatigue Loads
for Carbon/Epoxy Braided Composites

% of 25º 30º 45º

UTS Exp. Linear Sigmoidal Exp. Linear Sigmoidal Exp. Linear Sigmoidal
80 274- 68 391 392- 54 377 369- 238 18
516 475 1533
70 693- 647 894 578- 425 798 2339- 3543 2787
986 910 3507
60 1225- 6205 2092 1880- 3333 1695 8236- 56675 12883
5575 240 9031
50 3470- 59573 6952 3688- 26187 4592 1M 783067 625437
4807 020
45 24135- 184590 23940 5333- 73403 10840 --- --- ---
1M 2836

It is possible to predict fatigue strength at the endurance limit from a careful study of the stress-
strain diagram of the static tension test. It is very interesting to note that stress-strain diagrams
were linear up to 0.6%, 0.5%, and 0.4% strain for braided composites with braid angles of 25º,
30º, and 45º, respectively. The corresponding stresses for braided composites with braid angles
of 25º, 30º, and 45º were approximately 40 ksi (275 MPa), 25 ksi (172 MPa), and 7.5 ksi (52
MPa) and were 42%, 39%, and 29% of the UTS, respectively. The stress and strain were
directly proportional to each other in the linear zone. After this zone, the stress-strain curve
became nonlinear. Nonlinearity indicates a continuous decline in the slope (or, instantaneous
modulus). The increase in strain was much greater than the increase in stress. Thus, the
inflection point where the stress-strain curve became nonlinear could be used to predict the range
of fatigue strength at the endurance limit (infinite cycles). These predictions were not valid in
the fatigue experiments in this research.

Researchers have used the kinks in stress-strain diagrams to predict fatigue strength at 1 million
cycles. The basis for these predictions was the stress and strain. From the CV values in tables

3-22
3-2 and 3-4, it is clear that failure strain and UTS measurements have less repeatability and
reproducibility (at least for carbon/vinyl ester composites), while the modulus provides the
highest. Therefore, the more reliable models in the prediction of fatigue strength are the stiffness
degradation (or modulus reduction) models. These models are explained in detail in section 4.

3.6 OTHER OBSERVATIONS.

This section elaborates the observations related to failure patterns of braided composites under
tension-tension fatigue loading, viscoelastic behavior, and effect of frequency and temperature
on fatigue life.

3.6.1 Failure Pattern.

Irrespective of the resin system and the braid angle, braided composites did not exhibit any
noticeable matrix cracks either at the surface or at the edges until 90% of the fatigue life, when
tested at 10-Hz frequency. Failure occurred very quickly in the last 10% of the fatigue life.
Figures 3-17 and 3-18 show the failure patterns of braided composites in static and fatigue
loadings for both the carbon/vinyl ester and the carbon/epoxy resin system. The failure in tensile
loading was fiber dominated and as the applied fatigue load increased the fatigue failure
resembled the tensile failure.

Tensile 80% of UTS 70% of UTS 60% of UTS 50% of UTS

Figure 3-17. Failure Patterns of Static Tensile- and Fatigue-Loaded Carbon/Vinyl Ester
Specimens (Braid Angle 25° ±2°)

3-23
 Tensile 85% of UTS 75% of UTS 65% of UTS 55% of UTS 45% of UTS

Figure 3-18. Failure Patterns of Static Tensile- and Fatigue-Loaded Carbon/Epoxy Specimens
(Braid Angle 30° ±1°)

3.6.2 Viscoelastic Behavior of Braided Composites.

All materials deviate from Hooke’s law of linear elasticity, (e.g., by exhibiting viscous-like as
well as elastic characteristics). Viscoelastic materials are materials whose stress and strain
depend on time. Some phenomena in viscoelastic materials are (Lakes, 1998):

• If the stress is held constant, the strain increases with time (creep).

• If the strain is held constant, the stress decreases with time (stress relaxation).

• The effective stiffness depends on the rate of application of the load.

• If cyclic loading is applied, hysteresis (a phase lag) occurs, leading to a dissipation of

mechanical energy.

• Acoustic waves experience attenuation.

• Rebound of an object after an impact is less than 100%.

• Frictional resistance occurs during rolling.

The focus of this research was not directed toward the study of the viscoelastic behavior of
braided composites. Limited tests were conducted on carbon/epoxy braided composites (braid
angle 30° ±1°) to study the creep, stress relaxation, and frequency effect on fatigue life and the
stiffness degradation of braided composites.

Figure 3-19 displays the creep of a carbon/epoxy braided composite (braid angle 30° ±1°).
When a load of 2 kips was held constant, strain increased with time. It should be noted that there
was only a 0.91% increase in the strain in 1020 seconds (17 minutes), which was not very
significant. Figure 3-20 shows the stress relaxation of a carbon/epoxy braided composite (braid

3-24
angle 30° ±1º). When a displacement of 0.0338 in. was held constant, the load began to drop.
The load dropped by 4.4% in 1110 seconds (18 minutes). Figure 3-21 shows the stiffness
degradation of a carbon/epoxy braided composite (braid angle 30° ±1º) at two different
frequencies, 5 and 10 Hz, at an applied stress of 60% of the UTS. The stiffness at 10 Hz
declined from 4.62 Msi to 4.11 Msi (i.e., an 11% decrease), whereas at 5 Hz, stiffness declined
from 4.89 Msi to 1.09 Msi (i.e., a 77% decrease). These declines occurred in the same amount of
time, 138 seconds. Thus, the rate of loading (frequency) had a dominating effect on the damage
rate. Figure 3-22 shows stress-strain curves for the 1st, 10th, 20th, and 30th cycles for
carbon/epoxy braided composites (braid angle 25° ±2°) at a fatigue load of 50% of the UTS.
There was no significant hystersis loss at this low stress level and in the early fatigue life. Figure
3-23 shows stress and strain versus time during a phase shift. As discussed earlier, viscoelastic
materials such as carbon/epoxy braided composites exhibit hysteresis loss (a phase lag), which
leads to a dissipation of mechanical energy in the form of heat. This fact was confirmed by
measuring the surface temperature of the specimens at different frequencies.

0.00467

0.00466

0.00465
Strain, in/in

Applied
AppliedLoad = 3 Kips
Load = 2 kips
0.00464

0.00463

0.00462
0 200 400 600 800 1000 1200
Time, s

Figure 3-19. Creep of Carbon/Epoxy Braided Composite (Braid Angle 30° ±1°)

3-25
 5 E = 4.11 Msi

4
5 Hz

Modulus, Msi
3

10 Hz
2
E = 1.09 Msi

0
0 25 50 75 100 125 150 175 200
Time,s

Figure 3-20. Stress Relaxation of Carbon/Epoxy Braided Composite

(Braid Angle 30° ±1°)

2.10

2.08

2.06
Load, kips

2.04 Applied Displacement = 0.0338 in

2.02

2.00

1.98
0 200 400 600 800 1000 1200
Time, s

Figure 3-21. Stiffness Degradation of Carbon/Epoxy Braided Composites (Braid Angle 30° ±1°)
at Different Frequencies (Applied Stress 60% of UTS)

3-26
 50

40 st
1 Cycle
th
10 Cycle

Stress, ksi
30 th
20 Cycle
th
30 Cycle
20

10

0
0.000 0.002 0.004 0.006
Strain, in/in

Figure 3-22. Hystersis Loops of Carbon/Epoxy Braided Composites (Braid Angle 25° ±2°)
(Applied Stress 50% of UTS, 10-Hz Frequency, and R = 0.1)

60000 0.008
S train 0.007
50000
0.006
40000

Strain, in/in
Stress, psi

0.005
30000 0.004
0.003
20000
0.002
10000 S tress 0.001
0 0
5.5 5.6 5.7 5.8 5.9 6 6.1
Tim e, s

Figure 3-23. Stress and Strain vs Time With a Phase Shift Carbon/Epoxy Braided Composites
(Braid Angle 30° ±1°) (Applied Stress 50% of UTS, 29th Cycle)

3.6.3 Frequency and Temperature Effect.

As discussed in detail in the literature review, frequency has a dominant effect on the fatigue life
of polymer matrix composites. A limited number of tests were conducted to understand the
effects of frequency. Tables 3-15 and 3-16 present the test results.

3-27
Table 3-15. Effect of Frequency on Fatigue Life of Carbon/Epoxy Braided Composites at Braid
Angle 30° ±1° and Fatigue Load of 60% of UTS

Room Maximum Temperature Maximum

Frequency Temperature on the Surface Temperature Rise Fatigue Life
(Hz) (ºF) (ºF) (ºF) (cycles)
2 76 88.6 12.6 19,980
5 77.4 96.2 18.8 2,439
10 76 100.2 24.2 1,535

Table 3-16. Effect of Frequency on Fatigue Life of Carbon/Epoxy Braided Composites at Braid
Angle 45° ±1° and Fatigue Load of 60% of UTS

Room Maximum Temperature Maximum

Frequency Temperature on the Surface Temperature Rise Fatigue Life
(Hz) (ºF) (ºF) (ºF) (cycles)
2 77 86 9 110,406
5 77 88.4 11.4 124,736
10 74.7 96 21.3 8,236

At 5 and 10 Hz, there was no noticeable damage, such as matrix cracking and delamination at the
edges. But at 2 Hz, a small amount of matrix cracking and delamination was observed at the
edges. The surface temperature of each braided composite specimen reached a maximum at all
frequencies and then remained almost constant during the remainder of the test.

The effect of frequency on laminates is dependent on the lay-up. This was also true for braided
composites. When the braid angles were 30º and 45º, the composite was matrix-dominated. For
30º braided composites, fatigue life was sensitive from 2 to 5 Hz, and was almost the same at
5 and 10 Hz. For 45º braided composites, fatigue life was almost the same at 2 and 5 Hz and was
sensitive from 5 to 10 Hz. However, the fatigue life of 30º and 45º braided composites at 10-Hz
frequency declined by almost 1/14 compared to fatigue life at 2 Hz.

As frequency increased, internal friction between tows increased, and generated heat caused a
rise in temperature. There was almost a 24ºF (13.4ºC) temperature increase at 10 Hz, as shown
in figure 3-24. The reduction in fatigue life could also have been due to an internal friction
between tows and a rise in temperature from the internal friction. The maximum temperature at
10 Hz was 100.2ºF (38ºC) when a test was performed at 77ºF (25ºC). Therefore, a test was
conducted at an elevated temperature of 101.5ºF (39ºC) at 5-Hz frequency for the 45º braided
composite (matrix dominated lay-up) to determine the effect of temperature alone. Fatigue life
was not reduced and doubled (214,839 cycles) in comparison to a similar test performed at 77ºF
when the specimen survived 124,736 cycles.

3-28
 105
100
95

Temperature, F
90
85
80
75
70
65
60
0 50 100 150 200
Time, s

Figure 3-24. Rise in Temperature of Carbon/Epoxy Braided Composites (Braid Angle 30° ±1°)
at 10-Hz Frequency Until Failure at Fatigue Load of 60% of UTS

More study is needed for a more reliable understanding of the frequency and temperature effects.
The following observations about braided composites were made based on the limited numbers
of tests.

• The creep, stress relaxation, hystersis (a phase lag), and frequency effect on stiffness
degradation and fatigue life demonstrated that carbon/epoxy braided composites exhibit
viscoelasticity.

• Internal friction generated heat within the specimen. This heat would not dissipate
quickly to surrounding areas and created a temperature gradient. This temperature
gradient was responsible for the reduction in fatigue life.

• The higher temperature of surrounding areas had no effect on fatigue life. In fact, it
eliminated the temperature gradient and could have improved fatigue life.

• Figure 3-25 shows that the stiffness degradation mechanisms at 5- and 10-Hz frequencies
were unchanged (Normalized En/E1 and n/N). This is important in the issue of
accelerated fatigue.

• The size of the tow in this study was 12 K (12,000 fibers/tow). The amount of heat
generation from internal friction could be dependent on tow size.

• The tubular structure of braided composites has fiber continuity from end to end of the
part. Flat specimens of braided composites were used in this study. It is essential to test
the tubular specimens for fatigue before drawing any final conclusions about the fatigue
life of braided composites.

3-29
 Fatigue Secant Modulus E(n), Msi
5 10 Hz

5 Hz
3

0
0.0 0.2 0.4 0.6 0.8 1.0
Cycle Ratio, n/N

Figure 3-25. Stiffness Degradation of Braided Composites (Braid Angle 30° ±1°) at Different
Frequencies at Fatigue Load of 60% of UTS

The VARTM process is prone to thickness variation. The change in thickness causes a change in
fiber volume fraction. The change in fiber volume fraction changes the UTS of the specimen,
and the change in UTS from specimen to specimen causes large scatter in the fatigue data. In the
case of braided composites, the braid angle variation within a specimen causes a dramatic
variation in the UTS. Thus, the fiber volume fraction and the braid angle are the two most
important parameters to be monitored in the case of braided composites. A specific monitoring
method for these parameters was developed and is explained in appendix B.

Stiffness degradation is always related to the damage in composites. Stiffness degradation

curves can be effectively used to predict fatigue life, modulus at a particular cycle, and
endurance limits. The next section discusses the efforts related to stiffness degradation
modeling.

3-30
4. STIFFNESS DEGRADATION MODELING.

4.1 IMPORTANCE OF STIFFNESS DEGRADATION STUDY.

Aircraft components are subjected to a variety of loads termed environmental stress factors. The
most important factors contributing to the reduction of component lifetime are mechanical loads
(static and fatigue), moisture (hygrothermal effects), and temperature. Other factors include
oxygen, ultraviolet radiation, and ozone. The net effects of these factors are physical aging,
chemical aging, and damage (Gates, 2002).

The fatigue damage mechanism of composites is highly complex and may be in one or more
forms such as fiber/debonding, matrix cracking, delaminations, and fiber breakage. Some of
these damage mechanisms may interact simultaneously. Damage during fatigue is reflected in
strength, stiffness, and fatigue life reduction. Different damage mechanisms, such as matrix
cracking and delamination, reduce stored energy, which in turn reduces stiffness. There is
always a correlation between damage and stiffness reduction. Thus, stiffness reduction is the
only parameter that can be monitored to evaluate the useful life of a component. Stiffness
reduction is often referred to as stiffness degradation or modulus reduction in the literature. All
of these terms are used interchangeably in this research.

The evaluation of residual strength and fatigue life involves destructive testing, whereas stiffness
can be measured using nondestructive techniques such as ultrasound. It is well-known that the
wave speed of sound in a material is related to its stiffness. The presence of defects (e.g., voids,
cracks, micro damages, etc.) changes the effective stiffness of a material. When a wave is
propagated through a material, the change in stiffness is manifested as a change in the sound
velocity. Furthermore, the defects act as wave scatterers. As a result, the defect population also
manifests itself in an attenuation of the wave that passes through the material. Many researchers
have used the changes in wave speed and attenuation due to damage as a nondestructive tool for
stiffness measurement.

In the aircraft industry, many components are designed in such a way that the maximum strain
does not exceed 3000 με (i.e., 0.3%). The typical life of the primary components of an aircraft
(e.g., skins, wings, and fuselage) is 60,000 flying hours. The components experience 100,000
load cycles during their lifetime (Gates, 2002). Accelerated fatigue tests are conducted in the
laboratory. These accelerated tests deliberately shorten the life of a specimen or accelerate
degradation mechanisms (e.g., fatigue, creep, wear, etc.). Degradation is usually achieved by
performing constant-amplitude fatigue tests at higher frequencies. However, frequency is
limited to 10 Hz for polymer matrix composites because of their viscoelasticity.

4.2 REPRESENTATION OF TENSILE MODULUS AND STIFFNESS DEGRADATION

CURVES.

There are different ways to define modulus in the literature, as shown in figure 4-1. EO is the
initial tangent modulus and is usually calculated according to ASTM D 3039. This is a very
convenient way to define modulus if the stress-strain response is linear as in the case of the fiber-
dominated lay-ups of laminates. As laminates become matrix-dominated, the stress-strain

4-1
response becomes nonlinear. In these materials, ES is defined as a static secant modulus, which
is the UTS divided by the corresponding strain

Su
ES =
εu
120000

100000

Eo

80000 Es
S= 80% of UTS
A B
Stress, psi

(Es)0.8

60000

40000

20000

o0
0.00E+00 5.00E-03 1.00E-02 1.50E-02 2.00E-02 2.50E-02

Strain, in/in

Figure 4-1. Representation of Tensile Modulus

In fatigue studies, a specimen is subjected to applied maximum stress (S) and is cycled for a set
number of cycles (viz., 1000) before the fatigue test is completed. The specimen is then tested
for tension. The load is applied until the stress reaches the applied fatigue stress (S) and the
initial tangent modulus is then calculated (En). The specimen is cycled again for the next set of
cycles and the procedure is repeated. Ten readings are typically taken at each applied stress (S).
This is a very laborious method of evaluating modulus.

Fatigue secant modulus is a much more convenient way to define modulus. The axial strain is
measured continuously during the fatigue test using an extensometer or a strain gage. The
fatigue secant modulus E(n) is defined at each cycle as applied fatigue stress (S) divided by the
corresponding strain at the nth cycle. Thus, E(N) is fatigue secant modulus at failure cycle (N).

S
E( N ) =
ε fatigue

A different concept of static secant modulus is introduced in this study. The secant modulus is
defined at a particular stress level (e.g., 80% of Su). Thus, ES(0.8) is 0.8 times Su divided by the
corresponding strain on the tensile stress-strain diagram.

(0.8* Su )
( ES )0.8 =
(ε)at 0.8*Su

This static secant modulus can be represented as the slope of line OA in figure 4-1. The static
secant modulus is denoted as ( ES ) S in the present discussion. Typically, stiffness degradation
Su

4-2
curves are presented as normalized curves. The numbers of fatigue cycles (n) are normalized
with the number of failure cycles (N). Thus, the x axis is always represented as n/N. However,
there are two ways to normalize the modulus. If the fatigue modulus is calculated by stopping
the test after n cycles (i.e., En), En is normalized with an initial tangent modulus in the static
tensile test (EO). Thus, the y axis is represented as En/EO. The fatigue secant modulus cannot be
normalized with EO. This is because the fatigue secant modulus at the first cycle E(1) may be
larger or smaller than EO for nonlinear materials. It should be noted that for nonlinear materials,
E(1) usually differs depending on the applied stress (S). As applied stress (S) increases, E(1)
decreases. Therefore, the fatigue secant modulus E(n) is normalized with the fatigue secant
modulus at the first cycle, E(1). Thus, the y axis is E(n)/E(1). The following section explains
how stiffness degradation curves can be used to predict the endurance limit (fatigue stress at
1 million cycles).

4.3 STIFFNESS DEGRADATION CURVES AS A TOOL TO PREDICT THE ENDURANCE

LIMIT.

Figure 4-2 displays the stiffness degradation curve for the entire life at of carbon/vinyl ester
braided composites (braid angle 25° ±2°) at an applied stress of 75% of the UTS. The number of
cycles (n) were normalized with the number of cycles at failure (N), and the fatigue secant
modulus E(n) was normalized with the fatigue secant modulus at the first cycle E(1). Thus, the
x axis was n/N and the y axis was E(n)/E(1). This curve exhibited a rapid reduction in the
modulus in the first few cycles (Stage I). Later, there was a gradual, almost linear, modulus
decay (Stage II), and the modulus decreased rapidly (Stage III) until the specimen failed (Kelkar
and Tate, 2003c). This behavior for braided composites was different than that for woven
composites, as shown in figure 1-11. In the case of woven composites, the maximum decrease in
fatigue secant modulus until failure was 25% of the fatigue secant modulus at the first cycle. In
braided composites, the decrease in modulus until failure could have been as high as 60% of the
modulus in the first cycle.

Stage I
1.0
Stiffness Reduction Ratio, E(n)/E(1)

Stage II
0.8

0.6
Stage III

0.4

0.2

0.0
0.0 0.2 0.4 0.6 0.8 1.0
Cycle Ratio, n/N

Figure 4-2. Stiffness Degradation Over Entire Fatigue Life for Carbon/Epoxy Braided
Composites at 75% of UTS (Braid Angle 25° ±2°)

4-3
After careful observation of the stiffness degradation curve, it was determined that Stage I lasted
only about 5% of the fatigue life. Thus, it was possible to predict the endurance limit by
conducting fatigue tests for only approximately 5% of the fatigue life. For this study, constant-
amplitude fatigue tests according to ASTM D 3479 were performed at a frequency of 1 Hz, with
an R ratio of 0.1 and applied stress (S) ranging from 45% to 85% of the ultimate tensile strength.
The material system used was carbon/vinyl ester (braid angle 25° ±2°). The load and strain data
were collected for every cycle for the first 400 cycles.

The fatigue secant modulus E(n) was calculated at every cycle. E(n) was then normalized with
E(1). The curves representing normalized stiffness reduction E(n)/E(1) versus the number of
cycles (n) are presented in figure 4-3. The smoothened stiffness degradation curves are
displayed in figure 4-4. All of the stiffness degradation curves indicate that the initial portion of
the curve was linear. The linear segments of the stiffness degradation curves are presented in
figure 4-5. The slope of each curve was computed. Figure 4-6 is a representation of the absolute
values of the slopes versus the applied stress as a percentage of the UTS. The applied stress,
where the slope of an initial region approached zero, represented the endurance limit of the
material. From the graph in figure 4-6, it is evident that the endurance limit for the material
system was 40% of the UTS, which is the same value obtained in the S-N studies (table 3-3).
Thus, stiffness degradation studies in Stage I can be used to predict the endurance limit of a
material.

1.000
45% of UTS
Stiffness Reduction Ratio, E(n)/E(1)

0.975 55% of UTS

65% of UTS
0.950

0.925
75% of UTS
0.900

0.875
85% of UTS
0.850
0 50 100 150 200 250 300 350 400
Number of Cycles, n

Figure 4-3. Stiffness Degradation Curves for First 400 Cycles

4-4
 1.00
45% of UTS

Stiffness Reduction Ratio, E(n)/E(1)

0.98 55% of UTS

0.96 65% of UTS

0.94

0.92
75% of UTS
0.90

0.88
85% of UTS
0.86

0 100 200 300 400

Number of Cycles, n

Figure 4-4. Smoothened Stiffness Degradation Curves for First 400 Cycles

1 45% of UTS
55% of UTS
Stiffness Degradation , E (n)/ E(1)

65% of UTS
0.98
75% of UTS
0.96

0.94

0.92
85% of UTS
0.9
0 10 20 30 40 50
Number of Cycles, (n)

Figure 4-5. Linear Representation of First 50 Cycles of Stiffness Degradation Curves

4-5
 2.00E-03

1.60E-03

Slope of Initial Linear Region 1.20E-03

8.00E-04

4.00E-04

0.00E+00
40 50 60 70 80 90
Percentage of Ultimate Strength

Figure 4-6. Slope of Initial Region of Stiffness Degradation Curve vs Applied Stress as
Percentage of Ultimate Tensile Strength

4.4 STIFFNESS DEGRADATION MODEL TO PREDICT RESIDUAL STIFFNESS.

There are many models available to predict the residual stiffness in a fatigue study. The typical
strategy is represented in the flow chart shown in figure 4-7. This is a stiffness degradation
model developed by Whitworth (1998) for carbon/epoxy angle-ply [±35]2s composite laminates
subjected to tension-tension fatigue (R = 0.1) at a frequency of 10 Hz. The model is discussed
here due to its similarity for braided composites and angle-ply laminates. In short, the fatigue
secant modulus E(n) is represented as a function of the number of fatigue cycles n and the
S
fraction of the applied stress . An appropriate failure criterion is then used to define unknown
Su
parameters in the equation of E(n). For example, a failure criterion for linear materials is often
defined as follows: “…failure occurs when the fatigue strain in the fatigue test reaches to
ultimate strain at UTS in the static tensile test.” (O’Brian and Reifsnider, 1981)

4-6
 Residual Stiffness
1/ m
⎡ ⎛ ES ⎞ ⎤
m

E (n) = E ( N ) ⎢ − h *ln(n + 1) + ⎜ ⎟ ⎥ ………… (1)

⎣⎢ ⎝ E ( N ) ⎠ ⎦⎥
h and m are independent of stress level

Modified failure criteria for Nonlinearity

1/ m
⎡ ⎛ ES ⎞ ⎤
m

E (n) = E ( n) ⎢ −h *ln(n + 1) = ⎜ ⎟ ⎥ ……… (2)

⎢⎣ ⎝ E (n) ⎠ ⎥⎦

1/ C2
⎡ S ⎤
∴ E ( N ) = ES ⎢ ⎥ ………… (3)
⎣ C1Su ⎦
E(N) is substituted in equation of E(n)

Final Formula
1/ m
⎡ ⎛ S ⎞
m / C2
⎤
E (n) = ES ⎢ −h *ln(n + 1) * ⎜ ⎟ + 1⎥ …… (4)
⎢⎣ ⎝ C1 S u ⎠ ⎥⎦
⎛ S ⎞
Thus, E (n) = f ⎜⎜ n, ⎟⎟ ………………………. … (5)
Predictive Tool for ⎝ Su ⎠
Residual Stiffness
E(n)

When Failure occurs n = N and E(n) = E(N)

Using equations (3) and (4), N is written as
⎧⎪ 1 ⎡⎛ C S ⎞ m / C2 ⎤ ⎫⎪
N = exp⎨ ⎢⎜ 1 u ⎟ − 1⎥ ⎬ − 1
⎪⎩ h ⎢⎣⎝ S ⎠ ⎥⎦ ⎪⎭
⎛ S ⎞
Predictive Tool for Thus, N = f ⎜⎜ ⎟⎟
Failure Cycles, N ⎝ Su ⎠

Figure 4-7. Flow Chart Representing General Strategy for Stiffness Degradation Modeling
(Whitworth, 1998)

4-7
Thus, failure occurs when the fatigue secant modulus E(N) reaches a value represented by the
slope of line OB in figure 4-1. The value of E(N) is calculated by using an appropriate criterion
and substituting it back into the general equation of E(n) to compute the failure cycle, N.

The general equation of the model developed by Whitworth (1998) is

1/ m
⎡ ⎛ ES ⎞ ⎤
m

E (n) = E ( N ) ⎢ − h *ln( n + 1) + ⎜ ⎟ ⎥
⎢⎣ ⎝ E ( N ) ⎠ ⎥⎦

The failure criterion for linear materials was modified to consider the effect of nonlinearity as
C2
S ⎛ E( N ) ⎞
= c1 ⎜ ⎟
Su ⎝ ES ⎠

E(N) was calculated from this failure criterion and substituted into the general equation of E(n)
as

1/ m
⎡ ⎛ S ⎞
m / C2
⎤
E (n) = Es ⎢ −h *ln(n + 1) * ⎜ ⎟ + 1⎥
⎢⎣ ⎝ C1Su ⎠ ⎥⎦

Failure occurred when E(n) = E(N). Thus, the number of cycles at failure, N could be computed
as

⎧⎪ 1 ⎡⎛ C S ⎞m / C2 ⎤ ⎫⎪
N = exp ⎨ ⎢⎜ 1 u ⎟ − 1⎥ ⎬ − 1
⎪⎩ h ⎣⎢⎝ S ⎠ ⎥⎦ ⎪⎭

This model and many other models do not use valuable information available from the S-N
diagram and the nonlinear stress-strain curves obtained from static tensile tests. A different
approach to stiffness degradation modeling was planned and is discussed below.

• First, the number of cycles at failure can be predicted accurately from a properly modeled
S-N diagram. The S-N diagram has a natural failure criterion, which can be stated as,
“failure occurs when the applied stress (S) reaches to the residual strength of the
specimen.” The best model of the S-N diagram for braided composites is the Sigmoidal
(Boltzmann) curve. It was decided that the number of failure cycles N would be
computed from the S-N diagram model and used in the stiffness degradation model.

• Second, the failure strain criteria used in different models relate fatigue strains at failure
cycles with either ES or EO. However, ES or EO may not accurately represent nonlinearity
in a stress-strain curve, and therefore, the static secant modulus at each stress level was
defined and denoted as ( ES ) S as discussed earlier. Thus, nonlinearity could be captured
Su

4-8
 based on the stress level. There is a definite relationship between strain on the stress-
strain curve at a stress level S and fatigue strain at failure cycle at the same stress level.
The stress-strain curve for carbon/epoxy braided composites is nonlinear in nature. It
was determined that fatigue strain at the failure cycle (εfatigue)N followed a nonlinear
relationship with respect to the applied stress. Thus, there was a simple linear
relationship between the fatigue secant modulus E(N) and ( ES ) S . This linear
Su
relationship could be used as the failure criteria. The relationship could be easily
modified to consider the effects of frequency and temperature.

• Finally, a stiffness degradation model was developed based upon failure cycle N from the
S-N diagram and the fatigue secant modulus at the failure E(N) of the failure criterion.

Constant-amplitude fatigue tests according to ASTM D 3479 were performed at a frequency of

10 Hz, with an R ratio of 0.1 on carbon/epoxy braided composites (braid angle 25° ±2º). The
applied stress (S) levels were 80%, 70%, 60%, 50%, and 45% of UTS. The load and strain data
were collected for every cycle for the entire fatigue life. The fatigue secant modulus was
calculated at every cycle. Figure 4-8 displays stiffness degradation curves represented as the
fatigue secant modulus E(n) versus the number of cycles (n). It should be noted that the fatigue
secant modulus at the 1st cycle and the Nth cycle decreased with an increase in the stress level.
Figure 4-9 shows the normalized fatigue secant modulus E(n)/E(N) versus the cycle ratio (n/N).
Table 4-1 lists the values of E(1) and E(N) for carbon/epoxy braided composites (braid angle
25 ±2º). It should be noted that the rate of modulus decay that can be represented by the slope of
n n
the stiffness degradation curve is undefined at = 0 and = 1.
N N

Figure 4-8. Fatigue Secant Modulus, E(n) vs Number of Cycles, n

4-9
 Figure 4-9. Normalized Fatigue Modulus E(n)/E(N) vs Cycle Ratio (n/N)

Table 4-1. Fatigue Secant Modulus at 1st and Nth Cycle for Carbon/Epoxy Braided Composites
(Su = 95.74 ksi, Eo = 7.870 Msi, Es = 5.409 Msi, and Braid Angle 25° ±2°)

Applied Stress Failure Cycles E(1) E(N)

as % of UTS N Msi GPa Msi GPa
45 935359 9.783 67.44 4.582 31.59
60 5325 9.140 63.01 3.831 26.41
70 824 7.470 51.50 3.366 23.21
80 424 6.541 45.10 2.325 16.03

Therefore, the rate of stiffness reduction was assumed to have the following relationship:

d (E* ) −a
= *
d ( n ) n .ln( n* )
*

where

E ( n) n
E* = and, n* =
E(N ) N

Integrating the above equation

E ( n) = E ( N ) * {− a *[ln < ln( n* ) >]} + c …………………. 0 < n* < 1

4-10
and modifying the equation

E (n) = E ( N )*{a *[ln < −b *ln(n* ) >]} + c ………………. 0 < n* < 1

This is a general equation for E(n), and it is applicable for n* values between 0 and 1. The
purpose of this model was to track residual fatigue secant modulus E(n). There was no need to
compute E(N) because it was already known from the failure criterion. However, this model
( N − 1)
could predict E(n) at failure if n* = was substituted into the equation.
N

In the above equation, the fatigue secant modulus at failure E(N) was unknown. It could be
found by using a suitable failure criterion. As discussed earlier, the secant modulus can be
defined at a particular stress level as ( ES ) S and can be related to the fatigue secant modulus at
Su
failure, E(N). Table 4-2 lists the values of the stress level and the corresponding strain values in
the static tension test and the fatigue test. Figure 4-10 displays the linear relationship between
( ES ) S and E(N). This linear relationship was used as the failure criteria and can be written as
Su

E( N ) = A * (ES ) S + B
Su

Table 4-2. Static and Fatigue Secant Modulus for Carbon/Epoxy Braided Composites
(Su = 95.74 ksi, Eo = 7.870 Msi, ES = 5.409 Msi, and Braid Angle 25° ±2°)

Secant
Static Tensile Modulus at Fatigue Secant
Strain at Stress Strain in Stress Level Modulus at
Level Fatigue at ( ES ) S Failure
Applied Stress Failure E(N)
(ε s ) S Su

as % of UTS εf
Su Msi GPa Msi GPa
45 0.005963 0.009402 7.225 49.81 4.582 31.59
60 0.008342 0.014995 6.886 47.47 3.831 26.41
70 0.010131 0.019910 6.615 45.60 3.366 23.21
80 0.012248 0.032943 6.253 43.11 2.325 16.03

4-11
 5

Fatigue Secant Modulus, Msi

4

3 y = 2.2819x - 11.865
R2 = 0.99
2

0
6.2 6.4 6.6 6.8 7 7.2 7.4
Static Secant Modulus, Msi

Figure 4-10. Linear Relationship Between Static and Fatigue Secant Modulus

For carbon/epoxy braided composites (braid angle 25° ±2°) under stated fatigue test parameters
(10-Hz frequency, R ratio 0.1), the values of A and B were evaluated as 2.2819 and -11.865,
respectively. The parameters a, b, and c from the equation were evaluated by using nonlinear
regression analysis on the data from all the tests. These values were 0.2532, 0.06452, and
2.8228, respectively.

Figure 4-11 displays a comparison between the experimental and the theoretical curves of the
residual fatigue secant modulus at a stress level of 70% of the UTS. At this stress level, the
predicted fatigue life from the Sigmoidal model of the S-N diagram was 894 cycles, and the
fatigue secant modulus E(N) from the failure criteria was 3.229 Msi (22.26 GPa). Thus, there
was a very good agreement between the theoretical and the experimental stiffness degradation
curves.

Figure 4-11. Comparison Between Experimental and Theoretical Stiffness Degradation Curves
(Carbon/Epoxy Braided Composite, Braid Angle = 25° ±2°, Applied Stress = 70% of UTS)

4-12
 ⎛S ⎞
If the stress level ⎜ ⎟ and the number of cycles (n) are known, then the residual fatigue secant
⎝ Su ⎠
modulus E(n) can be very accurately predicted. The main advantage of the model is that it
represents all three stages of a stiffness degradation curve. The parameters a, b, and c in the
equations depend mainly on the stress level and the frequency. However, because of the limited
availability of data, parameters a, b, and c were assumed to be independent of the stress level and
the frequency. The model could be modified with more data.

Once the residual fatigue secant modulus is calculated, it can be compared with the modulus
measured by nondestructive techniques. Thus, if measured stiffness is less than predicted
stiffness, there is more damage than expected, or the component has experienced overload. In
this case, the residual fatigue life would be less than (N-n). This stiffness degradation model
could be used as a tool for tracking the residual modulus. Figure 4-12 explains the procedure for
tracking the fatigue secant modulus with this model.

4-13
 S
and n are known
Su

S
Compute static secant modulus at stress level , which
Su
is defined as ( ES ) S from stress- strain curve
Su

Compute fatigue secant modulus at failure using failure criterion

E( N ) = A*( ES ) S + B
Su

Compute number of failure cycles from S- N diagram

⎧ ⎡ ⎤⎫
⎪ ⎢ ( A − A2 ) ⎥ ⎪⎪
⎪
N = ⎨dx * exp ⎢ 1 − 1⎥ ⎬ + X = X 0
⎪ ⎢ ( − A2 ) ⎥ ⎪
S
⎪⎩ ⎢ S ⎥⎪
⎣ u ⎦⎭
A1, A2, X0 and dx are dependent on frequency and braid angle
Calculate cycle ratio n
N ( )
Compute residual stiffness
{ [ ] }
E (n) = E ( N ) a * ln − b * ln(n * ) + c

( N)
n* = n The parameters ‘a’,’b’, and ‘c’ are dependent on stress level,
frequency and braid angle. Note that E ( n) = f ( n* )

If Measured E(n) < Predicted E(n)

Then,
There is more damage than expected and residual life <(N-n)
Else,
Damage is according to expectation and residual life ≥(N-n)

Figure 4-12. Tracking of Fatigue Secant Modulus

4-14
4.5 STIFFNESS DEGRADATION STUDY FINDINGS.

The stiffness degradation study findings are as follows:

• The stiffness degradation curve typically has three distinct stages. In the first stage, the
modulus decreases rapidly. In the second stage, there is gradual modulus decay almost
linearly, and in the third stage, the modulus decreases rapidly. In braided composites, the
fatigue secant modulus loss can be as high as 60% of the fatigue secant modulus in the
first cycle. In woven composites, the fatigue secant modulus decrease is 25% of the
fatigue secant modulus in the first cycle.

• It is possible to predict the endurance limit (stress level at 1 million cycles) by testing
specimens for only a few hundred cycles. The stress level at which the stiffness remains
almost constant indicates the endurance limit.

• The developed stiffness degradation model agreed well with experimental results in all
three stages of the stiffness degradation. The parameters a, b, and c from the equations
could be modified to consider the effect of the stress level and the frequency.

• It should be noted that the proposed model predicts the fatigue secant modulus E(n).
However, most of the nondestructive techniques quantify the modulus that is best
represented by the initial tangent modulus, EO. Therefore, the fatigue secant modulus,
E (n), needs to be related to the initial tangent modulus, EO.

4-15/4-16
5. COMPUTATIONAL MICROMECHANICS.

5.1 INTRODUCTION.

In this section, the idealized geometry of 2x2 braids and different steps involved in the analysis
of 2x2 braids will be discussed briefly. The finite element models used to obtain effective
properties and stress distribution will be shown and the governing differential equations will be
provided. A tape laminate configuration will also be discussed, which was used to compare the
3D finite element results for the braids with a simple laminate analysis.

5.2 IDEALIZED GEOMETRY OF THE 2X2 BRAID.

A 2x2 braid microstructure is formed by mutually intertwining two or more sets of tows (yarns)
about each other (figure 5-1). There can be different types of braids. The braid that was studied
here can be specified as a 2x2 biaxial ±θ braid. The numbers 2x2 mean that two +θ yarns pass
over and under two -θ yarns and vice versa. Biaxial means that the yarns run in two directions
(if there are yarns in the axial direction also, then the resulting structure is also called a triaxial
braid), θ is the braid angle. Figure 5-1 shows an idealized braided mat with matrix pockets
removed to show the geometry. As shown in the figure, the 2x2 biaxial braid structure has an
interlaced tow structure characterized by any given tow passing over two opposing tows, then
under two opposing tows, and repeating this pattern. A repeating unit cell can be defined in the
structure, as labeled in figure 5-1.

Figure 5-1(b) shows a typical tow taken out of the braid microstructure of figure 5-1(a). In this
figure, h is the mat thickness and λ is the wavelength of the wavy region. The WR is defined
herein as h/λ. This tow has straight as well as undulated regions. Here the straight and
undulated regions cannot be separated by planes parallel to yz plane except for the ±45° braid.
This is due to the fact that the braid tows are not orthogonal to each other (except for the ±45º
braid). Hence, different fibers in a single tow do not have the same phase angle. This means that
the different fibers of the tow do not uniformly undulate and straighten at the same x coordinate.
The phase of a fiber running at the edge of the tow is not the same as that of a fiber running in
the middle of the tow. This phase shift is = j = y * tan(θ) , where θ is the braid angle. Since the
tows are not orthogonal to each other, this causes the tow cross-section to vary in an unusual
fashion. This is illustrated in figure 5-1(b), which shows the cross-sections at different points
along the towpath. It should be noted that although the cross-sections are of different shapes, the
cross-sectional area is constant. Hence, there is no concern about disappearing material or
varying fiber volume fraction in the tow. The shape of the cross-section in the straight region of
the tow is lenticular and can be described by simple sinusoidal function as

h 2πy
z= cos( )
4 λ

5-1
 (a) Braid Angle and Coordinate System
(transparent region is half of the unit cell for symmetric stacking)

Tow cross-sections along the tow-path

z Straight
y region

Mat thickness h
Flattened tow
Cross-section

Undulated region λ/2

Straight region L1 Tow thickness
t
Undulated region λ/2
Waviness ratio: Straight region L2 w
WR = h / λ Lenticular tow
Tow width cross-section
x
(b) Tow Path, Tow Cross Sections, and WR

Figure 5-1. Architecture of 2x2 Braid

A few parameters are used to describe the braid architecture: the braid angle, which is measured
by ±θ and the tow WR, which is defined as h/λ. The overall fiber volume fraction VfOverall is used
to specify the fiber volume fraction in the entire braid, while the tow fiber volume fraction VfTOW
is used to specify the fiber volume fraction in the tows. The parameter that bridges VfOverall and
VfTOW is the tow volume fraction VTOW (≥2/π), which measures how much volume the tow
occupies in the entire braid unit cell model. Note that VfOverall = VfTOW VTOW. The tow volume
fraction VTOW is related to the tow cross-section shape. Specifically, when VTOW = 2/π, there is
no straight region in the boundary of the tow cross-section and the corresponding tow cross-

5-2
section shape is called lenticular. When VTOW >2/π, the corresponding tow cross-section shape is
flattened. Given the overall fiber volume fraction and either the tow fraction or fiber volume
fraction in the tow, the other can be calculated. For a lenticular cross-section, the model
geometry requires the tow fraction to be equal to 2 / π . In these studies, the glass/epoxy material
system had a lenticular cross-section (i.e., VT = 2 / π ). The overall fiber volume fraction was
50%, so the fiber volume fraction in the tow was 78.5%. For the carbon/epoxy material system,
both lenticular and flattened cross-sections were considered (for the same overall fiber volume
fraction) to see the effect of tow cross-section shape on the effective properties. For the same
overall fiber volume fraction, the lenticular tow has a larger fiber fraction than a flattened tow.

Finite element models were built for the unit cell defined in figure 5-1. Since the tow cross-
section along the towpath is not uniform, direct finite element mesh generation for the model is
difficult. A twill weave is shown in figure 5-2. Careful examination of the tow architecture of
the 2x2 twill weave and 2x2 biaxial braid (see figures 5-1 and 5-2) reveals that a ±45° 2x2
biaxial braid is geometrically indistinguishable from a 2x2 twill rotated by 45°. Both
configurations have orthogonal tows. Inspired by this fact, a mapping technique was developed
to generate the finite element mesh for various 2x2 biaxial braids from the mesh for the twill
weave, which had been developed in previous studies (Tang, 2001). The mapping
transformation is defined in equation 5-1.

⎧ x′ ⎫ ⎡ sin θ −cos θ ⎤ ⎡ sin 2θ 0 ⎤ ⎧ x ⎫

⎨ ⎬=⎢ ⎥⎢ ⎥⎨ ⎬ (5-1)
⎩ y′⎭ ⎣cos θ sin θ ⎦ ⎣ −cos 2θ 1 ⎦ ⎩ y ⎭

where ( x, y ) and ( x′, y′) are the coordinates in the twill and braid models, respectively, and θ is
the braid angle.

Figure 5-2. Twill Weave

5.3 ANALYSIS STEPS.

Different steps involved in the analysis are shown in figure 5-3, starting with the fiber and matrix
and moving in a counter clockwise direction. It is assumed that the fibers are arranged in a
hexagonal array in the matrix. The properties of tow are calculated using micromechanics. The

5-3
tows are interlaced with each other to get a mat and the mats are stacked on the top of each other
in some kind of stacking sequence to get the required thickness. The mats impregnated with the
matrix constitute the braid under investigations. A part of this braid microstructure can be
analyzed to predict the behavior. This braid is finally used to make structural components like
the one shown in the figure.

Figure 5-3. Analysis Steps

5.4 FINITE ELEMENT MODELS.

Figure 5-4 shows a micrograph (Whitcomb and Kelkar, 2002) of a dry braid mat. Clearly shown
in the figure, there is a repeated pattern of interlacing. In micromechanics, this is referred to as
periodicity. If the periodicity is exploited, finite element models can be developed for a single
representative volume element that will behave as though it is surrounded by many other RVEs.
The unit cell of a periodic microstructure is the smallest region that can produce the whole
structure by spatially translating its copies without the use of rotation or reflection. One does not
need to model the entire microstructure. A single unit cell can be modeled to reduce the analysis
region. But in textile composites, modeling even a single unit cell can be very expensive
because of the complex geometry involved. Fortunately, textile composites often exhibit
symmetry inside a single unit cell. The analysis region can be further reduced from a single unit
cell to a smaller subcell (for example, one-half or one-fourth of a unit cell, etc.) by exploiting
symmetry operations like mirroring, rotation, or a combination of the two. Hence, the unit cell
should be chosen such that it can offer some symmetry to reduce the analysis region. Here, the
symmetries were exploited and the analysis region was reduced to one-fourth of the unit cell for
the symmetric stacking and to one-half for the simple stacking (see figure 5-5) of the braided
mats. The boundary conditions were derived using the technique described by Whitcomb,
Chapman, and Tang, 2000 and Tang and Whitcomb, 2003a. Boundary conditions are imposed
on the paired regions as labeled by the paired letters through multipoint constraints shown in

5-4
figure 5-6. For example, the displacements on the partial planes B and B are related through the
multipoint constraints, which mean the displacements on one face are slaved (dependent) to the
displacements on the other face. As clearly shown in the figure, the boundary conditions are a
bit unusual and are not intuitively obvious. Tables 5-1 and 5-2 give the complete set of boundary
conditions for the various load cases for one-fourth of the unit cell of a symmetrically stacked
mat. In the tables, u1, u2, and u3 are the displacements, and T1, T2, and T3 are the tractions in X1,
X2 and X3 directions, respectively.

Figure 5-4. Schematics of Braids

5-5
Figure 5-5. Simple and Symmetric Stacking of Mats

5-6
 B B
3 A
C
a C F
b
A F
t
E E

X2

X1
θ, braid angle
Figure 5-6. A Coarse Finite Element Mesh for Half of the Unit Cell
(The quarter unit cell model is the region in which the matrix pockets are shown transparent.
Multipoint constraints are imposed on the paired regions (e.g., A and A , B, and B , etc.).)

5-7
 ∂u1 ∂u1 ∂u 2 ∂u 3 ⎛ ∂ui 1 ∂u ⎞
Table 5-1. Multipoint Constraints for , , , or Loading* ⎜ where = ∫ i dV ⎟
∂X 1 ∂X 2 ∂X 2 ∂X 3 ⎜ ∂x j V V ∂x j ⎟
⎝ ⎠

Figure Faces (Figure 5-6) Multipoint Constraint Relationships

X3 A and A ⎛ 1 a X2 ⎞ ⎛1 a X2 ⎞ ∂
u1⎜⎜ − a + , X2, X3 ⎟⎟ = u1⎜⎜ a + , X2, X3 ⎟⎟ − 〈 u 〉 a
⎝ 2 b ⎠ ⎝ 2 b ⎠ ∂X1 1
X2 ⎛ 1 a X2 ⎞ ⎛1 a X2 ⎞ ∂
u2⎜⎜ − a + , X2, X3 ⎟⎟ = u2⎜⎜ a + , X2, X3 ⎟⎟ − 〈 u 〉 a
⎝ 2 b ⎠ ⎝2 b ⎠ ∂X2 1
X1
⎛ 1 a X2 ⎞ ⎛1 a X2 ⎞
u3⎜⎜ − a + , X2, X3 ⎟⎟ = u3⎜⎜ a + , X2, X3 ⎟⎟
⎝ 2 b ⎠ ⎝2 b ⎠

X3 B and B ⎛ a X2 ⎞ ⎛ a X2 ⎞
u1⎜⎜ , ⎟
−X2, X3 ⎟ =
⎜
−u1⎜ − , X2, X3 ⎟⎟
⎝ b ⎠ ⎝ b ⎠
X2 ⎛ a X2 ⎞ ⎛ a X2 ⎞
u2⎜⎜ , ⎟
−X2, X3 ⎟ =
⎜
−u2⎜ − , X2, X3 ⎟⎟
⎝ b ⎠ ⎝ b ⎠
X1 ⎛ a X2 ⎞ ⎛ a X2 ⎞
u3⎜⎜ , ⎟
−X2, X3 ⎟ =
⎜
u3⎜ − ⎟
, X2, X3 ⎟
⎝ b ⎠ ⎝ ⎠
5-8

X3 C and C ⎛1 a X2 ⎞ ⎛1 a X2 ⎞ 3 ∂ ∂
u1⎜⎜ a − , −X2 − b, X3 ⎟⎟ = −u1⎜⎜ a + , X2, X3 ⎟⎟ + 〈
1 1
u1 〉 a − 〈 u 〉 b
⎝ 4 b 4 ⎠ ⎝ 2 b ⎠ 4 ∂ 1
X 4 ∂X2 1
X2 ⎛1 a X2 ⎞ ⎛1 a X2 ⎞ 3 ∂ ∂
u2⎜⎜ a − , −X2 − b, X3 ⎟⎟ = −u2⎜⎜ a + , X2, X3 ⎟⎟ + 〈
1 1
u 〉 a− 〈 u 〉 b
⎝4 b 4 ⎠ ⎝2 b ⎠ 4 ∂X2 1 4 ∂X2 2
X1
⎛1 a X2 ⎞ ⎛1 a X2 ⎞
u3⎜⎜ a − , −X2 − b, X3 ⎟⎟ = u3⎜⎜ a + , X2, X3 ⎟⎟
1
⎝4 b 4 ⎠ ⎝2 b ⎠

∂u1 ∂u1 ∂u 2 ∂u 3
*Only one of the , , , or is specified for a particular analysis.
∂X 1 ∂X 2 ∂X 2 ∂X 3
The geometric parameters a, b, and t are shown in figure 5-4.
The others are unknowns that are solved as part of the solution.
 ∂u1 ∂u1 ∂u 2 ∂u 3 ⎛ ∂ui 1 ∂u ⎞
Table 5-1. Multipoint Constraints for , , , or Loading* ⎜ where = ∫ i dV ⎟ (Continued)
∂X 1 ∂X 2 ∂X 2 ∂X 3 ⎜ ∂x j V V ∂x j ⎟
⎝ ⎠

Figure Faces (Figure 5-6) Multipoint Constraint Relationships

X3 E and E ⎛ a X2 ⎞ ⎛ a X2 ⎞ ∂ ∂
u1⎜⎜ , −X2 + 2 b, X3 ⎟⎟ = −u1⎜⎜ 2 a − , X2, X3 ⎟⎟ + 2 〈 u 〉 a+2 〈 u 〉 b
⎝ b ⎠ ⎝ b ⎠ ∂X1 1 ∂X2 1
X2 ⎛ a X2 ⎞ ⎛ a X2 ⎞ ∂ ∂
u2⎜⎜ , −X2 + 2 b, X3 ⎟⎟ = −u2⎜⎜ 2 a − , X2, X3 ⎟⎟ + 2 〈 u 〉 a+2 〈 u 〉 b
⎝ b ⎠ ⎝ b ⎠ ∂X2 1 ∂X2 2
X1
⎛ a X2 ⎞ ⎛ a X2 ⎞
u3⎜⎜ , −X2 + 2 b, X3 ⎟⎟ = u3⎜⎜ 2 a − , X2, X3 ⎟⎟
⎝ b ⎠ ⎝ b ⎠
X3 F and F ⎛3 a X2 ⎞ ⎛ 1 a X2 ⎞ ∂ ∂
u1⎜⎜ a − , −X2 + 2 b, X3 ⎟⎟ = −u1⎜⎜ − a + , X2, X3 ⎟⎟ + 〈 u 〉 a+2 〈 u 〉 b
⎝ 2 b ⎠ ⎝ 2 b ⎠ ∂X1 1 ∂X2 1
X2 ⎛3 a X2 ⎞ ⎛ 1 a X2 ⎞ ∂ ∂
u2⎜⎜ a − , −X2 + 2 b, X3 ⎟⎟ = −u2⎜⎜ − a + , X2, X3 ⎟⎟ + 〈 u 〉 a+2 〈 u 〉 b
⎝2 b ⎠ ⎝ 2 b ⎠ ∂X2 1 ∂X2 2
X1
⎛3 a X2 ⎞ ⎛ 1 a X2 ⎞
u3⎜⎜ a − , −X2 + 2 b, X3 ⎟⎟ = u3⎜⎜ − a + , X2, X3 ⎟⎟
⎝2 b ⎠ ⎝ 2 b ⎠
X3 ⎛ 1 ⎞ ⎛ 1 ⎞
Top and u1⎜⎜ X1, X2, t ⎟⎟ = u1⎜⎜ X1, X2, − t ⎟⎟
X2 ⎝ 2 ⎠ ⎝ 2 ⎠
Bottom
5-9

⎛ 1 ⎞ ⎛ 1 ⎞
(Simple Stacking) u2⎜⎜ X1, X2, t ⎟⎟ = u2⎜⎜ X1, X2, − t ⎟⎟
⎝ 2 ⎠ ⎝ 2 ⎠
X1 ⎛ 1 ⎞ ⎛ 1 ⎞ ∂
⎜ ⎟
u3⎜ X1, X2, t ⎟ = u3⎜ X1, X2, − t ⎟⎟ + 〈
⎜ u 〉 t
⎝ 2 ⎠ ⎝ 2 ⎠ ∂X3 3

X3 ⎛ 1⎞
Top and T 1⎜⎜ X1, t ⎟⎟ = 0
X2,
X2 ⎝ 2⎠
Bottom ⎛ 1⎞
(Symmetric Stacking) T 2⎜⎜ X1, t ⎟⎟ = 0
X2,
⎝ 2⎠
X1 ⎛ 1 ⎞ 1 ∂
u3⎜⎜ X1, X2, t ⎟⎟ = 〈 u 〉 t
⎝ 2 ⎠ 2 ∂X3 3

∂u1 ∂u1 ∂u 2 ∂u 3
*Only one of the , , , or is specified for a particular analysis.
∂X 1 ∂X 2 ∂X 2 ∂X 3
The geometric parameters a, b, and t are shown in figure 5-4.
The others are unknowns that are solved as part of the solution.
 ∂u1 ∂u2 ⎛ ∂ui 1 ∂u ⎞
Table 5-2. Multipoint Constraints for or Loading* ⎜ where = ∫ i dV ⎟
∂X 3 ∂X 3 ⎜ ∂x j V V ∂x j ⎟
⎝ ⎠

Figure Faces (Figure 5-6) Multipoint Constraint Relationships

X3 A and A ⎛ 1 a X2 ⎞ ⎛1 a X2 ⎞
u1⎜⎜ − a + , X2, X3 ⎟⎟ = u1⎜⎜ a + , X2, X3 ⎟⎟
⎝ 2 b ⎠ ⎝2 b ⎠
X2 ⎛ 1 a X2 ⎞ ⎛1 a X2 ⎞
u2⎜⎜ − a + , X2, X3 ⎟⎟ = u2⎜⎜ a + , X2, X3 ⎟⎟
⎝ 2 b ⎠ ⎝2 b ⎠
X1 ⎛ 1 a X2 ⎞ ⎛1 a X2 ⎞ ∂
u3⎜⎜ − a + , X2, X3 ⎟⎟ = u3⎜⎜ a + , X2, X3 ⎟⎟ − 〈 u 〉 a
⎝ 2 b ⎠ ⎝2 b ⎠ ∂X3 1

X3 B and B ⎛ a X2 ⎞ ⎛ a X2 ⎞
u1⎜⎜ , −X2, X3 ⎟⎟ = u1⎜⎜ − , X2, X3 ⎟⎟
⎝ b ⎠ ⎝ b ⎠
X2 ⎛ a X2 ⎞ ⎛ a X2 ⎞
u2⎜⎜ , −X2, X3 ⎟⎟ = u2⎜⎜ − , X2, X3 ⎟⎟
⎝ b ⎠ ⎝ b ⎠
X1 ⎛ a X2 ⎞ ⎛ a X2 ⎞
u3⎜⎜ , −X2, X3 ⎟⎟ = ⎜
−u3⎜ − , X2, X3 ⎟⎟
⎝ b ⎠ ⎝ b ⎠
5-10

X3 C and C ⎛1 a X2 ⎞ ⎛1 a X2 ⎞
u1⎜⎜ a − , −X2 − b, X3 ⎟⎟ = u1⎜⎜ a + , X2, X3 ⎟⎟
1
⎝4 b 4 ⎠ ⎝2 b ⎠
X2 ⎛1 a X2 ⎞ ⎛1 a X2 ⎞
u2⎜⎜ a − , −X2 − b, X3 ⎟⎟ = u2⎜⎜ a + , X2, X3 ⎟⎟
1
⎝4 b 4 ⎠ ⎝2 b ⎠
X1 ⎛1 a X2 ⎞ ⎛1 a X2 ⎞ 3 ∂ ∂
u3⎜⎜ a − , −X2 − b, X3 ⎟⎟ = −u3⎜⎜ a + , X2, X3 ⎟⎟ + 〈
1 1
u 〉 a− 〈 u 〉 b
⎝4 b 4 ⎠ ⎝2 b ⎠ 4 ∂X3 1 4 ∂X3 2

∂u1 ∂u2
*Only one of the or is specified for a particular analysis.
∂X 3 ∂X 3
The geometric parameters a, b, and t are shown in figure 5-4.
The others are unknowns that are solved as part of the solution.
 ∂u1 ∂u2 ⎛ ∂ui 1 ∂u ⎞
Table 5-2. Multipoint Constraints for or Loading* ⎜ where = ∫ i dV ⎟ (Continued)
∂X 3 ∂X 3 ⎜ ∂x j V V ∂x j ⎟
⎝ ⎠

Figure Faces (Figure 5-6) Multipoint Constraint Relationships

X3 E and E ⎛ a X2 ⎞ ⎛ a X2 ⎞
u1⎜⎜ , −X2 + 2 b, X3 ⎟⎟ = u1⎜⎜ 2 a − , X2, X3 ⎟⎟
⎝ b ⎠ ⎝ b ⎠
X2 ⎛ a X2 ⎞ ⎛ a X2 ⎞
u2⎜⎜ , −X2 + 2 b, ⎟ ⎜
X3 ⎟ = u2⎜ 2 a − , X2, X3 ⎟⎟
⎝ b ⎠ ⎝ b ⎠
X1 ⎛ a X2 ⎞ ⎛ a X2 ⎞ ∂ ∂
u3⎜⎜ , −X2 + 2 b, ⎟ ⎜
X3 ⎟ = −u3⎜ 2 a − , X2, X3 ⎟⎟ + 2 〈 u 〉 a+2 〈 u 〉 b
⎝ b ⎠ ⎝ b ⎠ ∂X3 1 ∂X3 2

X3 F and F ⎛3 a X2 ⎞ ⎛ 1 a X2 ⎞
u1⎜⎜ a − , −X2 + 2 b, X3 ⎟⎟ = u1⎜⎜ − a + , X2, X3 ⎟⎟
⎝2 b ⎠ ⎝ 2 b ⎠
X2 ⎛3 a X2 ⎞ ⎛ 1 a X2 ⎞
u2⎜⎜ a − , −X2 + 2 b, X3 ⎟⎟ = u2⎜⎜ − a + , X2, X3 ⎟⎟
⎝2 b ⎠ ⎝ 2 b ⎠
X1 ⎛3 a X2 ⎞ ⎛ 1 a X2 ⎞ ∂ ∂
u3⎜⎜ a − , −X2 + 2 b, X3 ⎟⎟ = −u3⎜⎜ − a + , X2, X3 ⎟⎟ + 〈 u 〉 a+2 〈 u 〉 b
⎝2 b ⎠ ⎝ 2 b ⎠ ∂X3 1 ∂X3 2

X3 ⎛ 1 ⎞ ⎛ 1 ⎞ ∂
Top and u1⎜⎜ X1, X2, t ⎟⎟ = u1⎜⎜ X1, X2, − t ⎟⎟ + 〈 u 〉 t
5-11

X2 ⎝ 2 ⎠ ⎝ 2 ⎠ ∂X3 1
Bottom
⎛ 1 ⎞ ⎛ 1 ⎞ ∂
(Simple Stacking) u2⎜⎜ X1, X2, t ⎟⎟ = u2⎜⎜ X1, X2, − t ⎟⎟ + 〈 u 〉 t
⎝ 2 ⎠ ⎝ 2 ⎠ ∂X3 2
X1
⎛ 1 ⎞ ⎛ 1 ⎞
u3⎜⎜ X1, X2, t ⎟⎟ = u3⎜⎜ X1, X2, − t ⎟⎟
⎝ 2 ⎠ ⎝ 2 ⎠

X3 ⎛ 1 ⎞ 1 ∂
Top and u1⎜⎜ X1, X2, t ⎟⎟ = 〈 u 〉 t
⎝ 2 ⎠ 2 ∂X3 1
X2 Bottom
⎛ 1 ⎞ 1 ∂
(Symmetric Stacking) u2⎜⎜ X1, X2, t ⎟⎟ = 〈 u 〉 t
⎝ 2 ⎠ 2 ∂X3 2
X1
⎛ 1 ⎞
T 3⎜⎜ X1, X2, t ⎟⎟ = 0
⎝ 2 ⎠

∂u1 ∂u2
*Only one of the or is specified for a particular analysis.
∂X 3 ∂X 3
The geometric parameters a, b, and t are shown in figure 5-4.
The Others are unknowns that are solved as part of the solution.
Finite element models (figure 5-7) were developed for the unit cell shown in figure 5-4. As
discussed in section 5.2, the cross-sections of the braid tow are not uniform, so direct finite
element mesh generation was difficult and a mapping technique was used to get finite element
models of 2x2 braids from twill weave models, which were produced in earlier studies by a
building-block technique (Tang, 2001). This resulted in a substantial time savings.

Figure 5-7. Typical Finite Element Models for Flattened Cross-Sections

5-12
Depending upon the requirement, any number of mats can be stacked on the top of each other.
Figure 5-5 shows the simple and symmetric stacking of mats. The analysis region can be
reduced to one-half for the simple stacking and to one-fourth for the symmetric stacking. The
results produced by full unit cell and a part of the unit cell were compared with each other to
validate the methodology, and it was seen that the results matched for the two cases.

Figure 5-7 shows two unit cells for ±70° braid. One has very large WR (1/3) and the other has
small WR (1/9).

As discussed in section 5.2, depending on the tow volume fraction in a model, the tow cross-
section will either be lenticular or flattened. For example, for the material systems used in the
parametric studies (will follow in the next section), one material system had flattened cross-
section and the other had lenticular. The finite element models with flattened tow cross-section
are shown in figure 5-7 and ones with lenticular cross-section are shown in figure 5-8.

(a) Elements are labeled

(b) Nodes are labeled

Figure 5-8. Finite Element Model for Lenticular Cross-Section With Nodes and
Elements Labeled

5-13
For every combination of WR (braid angle and volume fraction), a different finite element model
must be generated. For these studies, approximately 450 cases were run for the two material
systems. A typical mesh used for obtaining effective properties contained 2,304 hexahedral 20-
node brick elements and 10,060 nodes for models with flattened cross-section and 720 elements
and 3130 nodes for models with lenticular cross-section. A 20-node brick element is also shown
in figure 5-9. One coarse finite element model with 48 elements and 238 nodes is shown in
figure 5-8.

Figure 5-9. A 20-Node Brick Element

All the parametric studies for calculating the effective properties were done using an inexpensive
personal computer (PC). For models with flattened cross-sections, the mesh generation time and
equation solving time were considerably higher than for models with lenticular cross-sections.
Equation solving time using a 400-MHz PC were approximately five to six times more than that
using 2.4-GHz PC. Typical runtimes are shown in table 5-3.

Table 5-3. Runtimes on a Personal Computer for One Case

Computer Model Pentium II 400 MHz CPU Pentium IV 2.4 GHz CPU
Tow Cross-Section Lenticular Flattened Lenticular Flattened
Mesh Generation Time 21 sec 01 min, 06 sec 07 sec 14 sec
Equation Solving Time 02 min, 45 sec 11 min, 31 sec 28 sec 01 min, 56 sec

5.5 GOVERNING DIFFERENTIAL EQUATIONS.

The governing differential equations are reviewed in this section (Reddy and Miravete, 1995).
The equations of motion and deformation of a solid body can be classified into following four
categories:

1. Kinematics (strain-displacement relations)

5-14
2. Compatibility equations
3. Kinetics (conservation of linear and angular momentum)
4. Constitutive equations (stress-strain relations)

An overview of the above equations is presented below.

5.5.1 Kinematics.

Kinematics involves the study of the geometric changes or deformations in the body without
considering the forces acting on it. Let ε, the strain tensor in Voight’s notation, be given by

ε = {ε xx , ε yy , ε zz , ε xy , ε yz , ε zx }

Let u denote the displacement vector and the coordinates X = (X1, X2, X3) denote the material
coordinates of the body. The coordinate system X is fixed on the given body in the undeformed
configuration and its position x at any time is referred to as the material coordinate X as

x = x (X1, X2, X3)

If one assumes the strains and rotations are infinitesimal, the strain-displacement relations can be
given by

1 ⎛ ∂u ∂u j ⎞
ε ij = ⎜ i + ⎟ (5-2)
2 ⎜⎝ ∂x j ∂xi ⎟⎠

For infinitesimal deformations, components of the infinitesimal strain satisfy ε ij << 1 . The use
of the infinitesimal strain tensor to characterize the infinitesimal deformation of the body is in
fact linearization, which means that if ε(1) is the strain corresponding to displacement field u(1)
and ε(2) is the strain corresponding to displacement field u(2), then ε(1)+ ε(2) is the strain
corresponding to displacement field u(1)+u(2).

5.5.2 Compatibility Equations.

If a displacement field is given, the Cartesian components of the strain tensor are uniquely
obtained by the strain-displacement relations of equation 5-2. But if a strain field is given for
which a corresponding displacement field u is not given, one seeks to find the solution for the
displacement field. To do that, one has to solve the strain-displacement relations for u. That
involves six independent equations and only three unknown components of u. Hence, a single
valued solution for u might not exist. But if an allowable displacement field does exist, then the
corresponding strain field is said to be compatible. The following is an example of the
compatibility equations, which ensure that a single-valued displacement field exists.

∂ 2ε kn ∂ 2ε lm ∂ 2ε km ∂ 2ε ln
+ − − =0 (5-3)
∂xl ∂xm ∂xk ∂xn ∂xl ∂xn ∂xk ∂xm

5-15
This set of compatibility equations contains 81 individual equations, out of which only 6 are
linearly independent. The compatibility equations are not needed in the finite element analysis
because the governing equations were written in terms of displacements.

5.5.3 Kinetics.

Forces acting on a body can be classified into internal and external forces. Kinetics is the study
of static or dynamic equilibrium of forces acting on a body. If one considers a given mass with
density ρ, on which some forces are acting, then from the balance of linear momentum, the
equations of equilibrium can be written as

∂σ xx ∂σ xy ∂σ xz
+ + + ρ bx = ρ ax
∂x ∂y ∂z
∂σ yx ∂σ yy ∂σ yz
+ + + ρ by = ρ a y
∂x ∂y ∂z (5-4)
∂σ zx ∂σ zy ∂σ zz
+ + + ρ bz = ρ az
∂x ∂y ∂z

where σ = {σ xx , σ yy , σ zz , σ xy , σ yz , σ zx } denotes the Voight’s stress tensor, b is the body force, and a
is the acceleration of the body. If there are no body forces and no acceleration, then the terms
containing body force b and acceleration a vanish. In the absence of any body moments, the
conservation of angular momentum leads to symmetry of the stress-tensor.

5.5.4 Constitutive Equations.

The kinematics, mechanical principles are applicable to any material body irrespective of its
constitution. The constitutive equations characterize the individual material response of a body.
They relate the dependent variable introduced in the kinetic equations to those in the kinematic
relations. Constitutive equations give a relationship between stresses and strains. In general, the
stress-strain relation for infinitesimal deformation of an elastic material is nonlinear, as stress
does not have to be linearly proportional to strain. But if components of stress are assumed to be
in linear proportion to components of stress, then the constitutive equations can be written in the
following general form.

σi = Cij ε j

5-16
where Cij is the stiffness matrix, which is symmetric and has at most 21 independent stiffness
coefficients. The inverse of stiffness matrix is the compliance matrix (Sij) and is given as:

ε i = Sij σ j
⎡ 1 −υ12 −υ13 η14 η15 η16 ⎤
⎢E E11 E11 E11 E11 E11 ⎥
⎢ 11 ⎥
⎢ −υ21 1 −υ23 η24 η25 η26 ⎥
⎢E E22 E22 E22 E22 E22 ⎥
⎢ 22 ⎥
⎢ −υ31 −υ32 1 η34 η35 η36 ⎥
⎢ ⎥
E33 E33 E33 E33 E33 E33 ⎥
where Sij = ⎢
(5-5)
⎢ η 41 η42 η43 1 η45 η46 ⎥
⎢ ⎥
⎢ G12 G12 G12 G12 G12 G12 ⎥
⎢ η51 η52 η53 η54 1 η56 ⎥
⎢ ⎥
⎢ G23 G23 G23 G23 G23 G23 ⎥
⎢η η62 η63 η64 η65 1 ⎥
⎢ 61 ⎥
⎣⎢ G31 G31 G31 G31 G31 G31 ⎦⎥

Here, E11, E22 , E33 denote the Young’s moduli in the 1, 2, and 3 material directions, respectively,
G12 is the in-plane shear modulus, and G23 and G13 are out-of-plane shear moduli. The Poisson’s
ratio νij is defined as the negative of the ratio of transverse strain in the jth direction to the axial
strain in the ith direction when stressed uniaxially in the ith direction. The shear coupling
coefficient ηij is the ratio of strain in the jth direction to the strain in the ith direction when
stressed uniaxially in the ith direction. The compliance matrix shown in equation 5-5 is fully
populated. When the analysis is done, it will be shown that some of the engineering constants
are zero for 2x2 braids and the material is orthotropic (ηij = 0) with only 9 independent constants.

5.6 MATERIAL SYSTEMS.

Two material systems were used for the parametric studies. One consists of S2 glass fibers and
SC-15 epoxy resin. The other consists of AS4 carbon and 411-350 epoxy (vinyl ester or
derakane momentum) resin. The fibers are assumed to be arranged in a hexagonal array (figure
5-10) in the matrix in the tow. The tow properties were found using 3D finite element
micromechanics analysis using an in-house finite element code alpha. One-fourth of the unit cell
was used to calculate the tow properties and is shown in figure 5-11. The material properties of
the fiber, matrix, and the tow are given in table 5-4 for both material systems. For carbon/epoxy
material system, the properties of the tow are given both for lenticular and flattened cross-
sections as the fiber volume fraction in them is different.

5-17
 Figure 5-10. Distribution of Fibers Inside the Tow

Figure 5-11. Finite Element Mesh for Calculating Tow Properties

5-18
 Table 5-4. Material Properties for Fiber, Matrix, and Tow

Glass/Epoxy Material System Carbon/Epoxy Material System

b c
Tow Tow
a
Flattened Lenticular
Property* Fiber Resin Tow Fiber Resin x Section x Section
E11 96.53 2.82 76.42 228 3.1 157.95 204.36
E22 96.53 2.82 20.18 16.54 3.1 009.027 013.4
E33 96.53 2.82 20.18 16.54 3.1 009.027 013.4
ν12 0.23 0.395 00.2654 00.2 0.35 000.2412 000.2122
ν23 0.23 0.395 00.3352 00.25 0.35 000.3749 000.2974
ν13 0.23 0.395 00.2654 00.2 0.35 000.2412 000.2122
G12 39.23 1.01 07.4 24.82 1.15 005.12 014.62
G23 39.23 1.01 04.03 06.89 1.15 003.34 005.33
G13 39.23 1.01 07.4 24.82 1.15 005.12 014.62
a
Fiber volume fraction in tow = 0.7854
b
Fiber volume fraction in tow = 0.69
c
Fiber volume fraction in tow = 0.895
*All moduli in GPa

5.7 TAPE LAMINATE APPROXIMATION.

The predictions of the effective properties of the 2x2 braids using finite element analysis were
compared with the predictions of a tape laminate. In the braid, tows run at an angle of +θ and -θ,
and mats are symmetrically stacked. Hence, one-half the laminate configuration essentially
consists of two unidirectional laminas (with properties of the tow) in +θ and -θ directions and a
third layer as matrix to account for matrix pockets in the braid, as shown in figure 5-12. The
thickness of each layer was specified such that the laminate and the braid had the same volume
fraction of resin and tows in the model. The reason for doing this laminate analysis is threefold:
(1) It will compare the performance of 2x2 braids with equivalent laminated materials. (2) It
would be helpful in investigating whether the effective properties of the braids can be predicted
using simple analysis like laminate theory (laminate theory codes are easily accessible and well
understood by designers). (3) The properties predicted by 3D finite element analysis of braids
were normalized by the laminate predictions to see the exclusive effect of interlacing. The
effective properties for the equivalent laminate were obtained using a 3D finite element model
with periodic boundary conditions. It should be noted that it is usually more convenient to use
classical laminated plate theory for the in-plane properties or the full 3D homogenization
formulas, given in several references (Sun and Li, 1988; Whitcomb and Noh, 2000), if out-of-
plane properties are needed.

5-19
Figure 5-12. An Equivalent Laminate Configuration

5-20
6. PARAMETRIC STUDIES: EFFECT OF VARIOUS PARAMETERS ON EFFECTIVE
ENGINEERING PROPERTIES.

6.1 INTRODUCTION.

Obtaining effective mechanical properties is the first order of concern in any structural analysis.
This section presents an investigation of the effect of various parameters like braid angle, WR,
material properties, and cross-section shape on the effective engineering properties of the 2x2
braids. Extensive parametric studies were conducted to determine the behavior of the 2x2 braids.
The studies were conducted for two material systems: glass fiber/epoxy (S2/SC-15) matrix and
carbon fiber/epoxy (AS4/411-350) matrix. The properties were compared with those for an
equivalent laminated material with angle plies and a resin layer. The results of parametric
studies will be discussed in detail.

The predictions of the moduli and the Poisson’s ratio were also compared with the experimental
data (Kelkar, et al., 2003a) for the carbon/epoxy material system.

6.2 RESULTS OF PARAMETRIC STUDIES.

To characterize the mechanical behavior of braided composites, the effective engineering

properties were calculated for a wide range of braid angle and WR for glass/epoxy and
carbon/epoxy material systems. For carbon/epoxy material system, the effect of tow cross-
section shape on effective properties was also determined. Then the predictions of the finite
element analyses were compared with the predictions of a simple 3D laminate analysis. All the
elements of the 6x6 compliance matrix were calculated and it was observed that all the shear
coupling coefficients were almost zero, which predicted the material to be effectively
orthotropic. The braid patterns along the longitudinal and transverse directions differ, which
caused the complementary properties to differ along these directions. The difference in
complementary properties due to unbalance in material architecture was investigated. The
results of parametric studies are presented here. Finally, the predictions are compared with the
experimental results.

6.3 EFFECT OF VARIOUS PARAMETERS ON EFFECTIVE PROPERTIES.

Figure 6-1 shows the variation of E11 with braid angle and WR for the carbon/epoxy material.
The dots in the figure are finite element data. The data were fitted to obtain the surface plot.
The outlined curve shows the values for the reference laminate. The figure shows that most of
the effect of braid angle on the E11 can be predicted using laminate theory. This is the case for
other effective properties. Hence, to reveal the exclusive effects of the braid architecture and to
filter out the braid angle effect by the laminate theory, the results were normalized by the
respective reference laminate values.

6-1
 Figure 6-1. Effective Longitudinal Modulus E11 as a Function of Braid Angle and Waviness
Ratio for Carbon/Epoxy Material (Discrete points are the data from the finite element analysis.)

Figures 6-2 and 6-3 show the normalized effective engineering properties as a function of braid
angle and WR for glass/epoxy and carbon/epoxy, respectively. Figure 6-3a shows the properties
of the carbon/epoxy material system with flattened tow cross-section, and figure 6-3b shows
those calculated with lenticular cross-section. For carbon/epoxy material system effect of tow
cross-section on effective properties was also determined. The range for the waviness change is
from 0.03 (very flat) to 0.33 (very wavy) for both materials. The range for the braid angle is
20º~70º for glass/epoxy and 15º~75º for carbon/epoxy. Figure 6-2 shows that for the
glass/epoxy material system, E11, E22, E33, and G12 decrease while G23, G13, ν23, and ν13 increase
with an increase in WR for all braid angles. The ν12 does not change much with WR for all braid
angles. Figures 6-3a and 6-3b show that for the carbon/epoxy material system, the in-plane
moduli (E11, E22, and G12) decrease while the out-of-plane shear moduli G23 and G13 increase
with an increase in WR. The out-of-plane extensional modulus E33, remains almost constant and
is greater than the corresponding laminate values for all braid angles and WR. The change in the
υ12 is small but not uniform for different braid angles. For some braid angles, υ12 decreases with
WR, and for others, it first increases and then decreases. As far as ν23 and ν13 are concerned, the
normalized values do not change much with WR as long as corresponding reference values are
not very small (i.e., close to zero). But, it should be noted that there are cases in which the
reference laminate value for the out-of-plane Poisson’s ratio is extremely small in magnitude at
certain braid angles. For example, the value of ν23 for reference laminate of carbon/epoxy
material system corresponding to ±65.25° braid was 0.0023. As a result, a small deviation will
result in an extremely large normalized value. Hence, some range of braid angle for
carbon/epoxy material with lenticular cross-section was excluded in the results presented for ν23
and ν13.

6-2
Further, it can be noted from figures 6-2 and 6-3 that even for very low WR, the values do not
converge to the values predicted by the laminate model. This is due to the fact that, although the
laminate model can account for the braid angle effect, it cannot account for the effect of material
distribution.

Figure 6-2. Normalized Effective Engineering Properties as a Function of Braid Angle and
Waviness Ratio for Glass/Epoxy Material System With Lenticular Cross-Section (The properties
are normalized by the laminate values.)

6-3
Figure 6-3a. Normalized Effective Engineering Properties as a Function of Braid Angle and
Waviness Ratio for Carbon/Epoxy Material System for Tow With Flattened Cross-Section
(The properties are normalized by the laminate values.)

6-4
Figure 6-3b. Normalized Effective Engineering Properties as a Function of Braid Angle
and Waviness Ratio for Tow With Lenticular Cross-Section
(The properties are normalized by the laminate values.)

6-5
 6.4 DEVIATION OF THE PROPERTIES FROM REFERENCE LAMINATE VALUES.

The deviation of the properties from the reference laminate values is measured by
(Eextreme–1)*100%, where Eextreme is either Emin or Emax, which are the normalized minimum and
maximum property values for the range in which the parameters vary. Therefore, Emin and Emax
correspond to the lower and upper deviations from laminate theory, respectively. The deviation
of properties from laminate theory is plotted in figures 6-4 through 6-6. The closer the extremity
of the band to zero, the smaller is the deviation from laminate theory.
100

50
87.0
81.2

35.8
S2/SC-15

35.1
S2/SC-15
WR=0.03~0.33 WR=0.03~0.11
Deviation from Laminate (%) .

BA=20~70

Deviation from Laminate (%) .

BA=20~70

25

10.9
10.9
39.5
50

-0.6
21.7

-1.0

-1.0

-2.4

-4.7
6.9
12.7
10.9

3.9

3.8

-2.6
-0.6
-1.0
-1.0

-6.6

-11.3
0

-12.7
-12.9
3.9

3.8
1.8

-16.0
-25
-8.0

-11.3

-12.7
-20.1
-21.1

-29.1
-50

-50
v12

v23

v13
E11

E22

E33

G12

G23

G13

v12

v23

v13
G12

G23

G13
E11

E22

E33

(a) For full range of WR (b) For realistic range of WR

Figure 6-4. Deviation of 3D Finite Element Results From the Laminate Results for Glass/Epoxy Material
System With Lenticular Cross-Section

6-6
 Deviation from laminate (%)
-50 -25 0 25 50 75 100 Deviation from Laminate (%) .

-50
0
50
100
E11 -30.8 -1.02
E11 -7.5 1.6
E22 -32.5 -1.03
E22 -7.8 1.6
E33 3.75 8.46

BA=15~75
E33 2.9 4.8

AS4/411-350
BA=15~75

WR=0.03~0.33
AS4/411-350
WR=0.03~0.33

G12 -33.2 -0.2

G12 -10.8 1.8
G23 4.4 155
G23 4.1 55.0
G13 4.3 164.8
G13 4.1 63.2
v12 -10.2 16.8
v12 -11.2 2.3
v23

(a) For full range of WR

(a) For full range of WR
-152 1301
v23 -567.9 39.5

v13 -160 862

v13 -517.6 36.6

6-7
Deviation from laminate (%)
-50 -25 0 25 50
Deviation from Laminate (%) .
-50 -25 0 25 50

E11 -14.2 -1.02

E11 -2.5 1.6

E22 -15.5 -1.03

E22 -2.5 1.6

System With Lenticular Cross-Section

System With Lenticular Cross-Section

E33 3.75 6.55

BA=15~75
E33 4.0 4.5
AS4/411-350
BA=15~75

WR=0.03~0.11
AS4/411-350
WR=0.03~0.11

G12 -12.8 -0.2 G12 0.1 1.8

G23 4.4 69.7 G23 4.1 28.8

G13 4.3 72.7 G13 4.1 29.3

v12 2.9 16.8 v12 -1.2 2.3

v23 -152 12.6 v23 -55.9 6.0

(b) For realistic range of WR

(b) For realistic range of WR

-160 v13 -52.2 6.8
v13 -2

Figure 6-6. Deviation of 3D Finite Element Results From the Laminate Results for Carbon/Epoxy Material
Figure 6-5. Deviation of 3D Finite Element Results From the Laminate Results for Carbon/Epoxy Material
For the full range of WR studied (0.03-0.33), the deviation of the finite element results from the
laminate results for the effective properties is shown in figures 6-4(a), 6-5(a), and 6-6(a) for both
the material systems. In general, the out-of-plane properties have higher deviation than in-plane
properties. Maximum deviation is observed for out-of-plane shear moduli for both the material
systems (ignoring the high deviation of out-of-plane Poisson’s ratios for the reason mentioned
earlier). The maximum deviation is 87% for the glass/epoxy and around 165% for carbon/epoxy
material system and is for G13. The largest deviation (absolute value) for in-plane properties is
33.2% for the carbon/epoxy and 29.1% for the glass/epoxy. In general, the glass/epoxy material
system has a larger deviation for almost all the properties (excluding v23 and v13 for the reasons
already mentioned) than for carbon/epoxy material system with flattened cross-section, but has a
smaller deviation than one with lenticular cross-section. Figures 6-5(a) and 6-5(b) show that
carbon/epoxy with lenticular cross-section has considerably more deviation than that with
flattened cross-section for all the properties. Hence, it is more difficult to predict the response of
a braid using simple laminate approximation for material system with lenticular tows that ones
with flattened tows even if the model contains the same overall fiber volume fraction.

Since out-of-plane modulus E33 has a small deviation from the laminate value, it can be predicted
easily using 3D laminate analysis. For the two material systems used in this study, the E33 is
larger than the laminate value, as shown in figures 6-2 and 6-3. In general, the E33 can be more
than, equal to, or less than the laminate value, depending on the material system used. For
example, for the material system used by Paumelle, et al., 1990, the E33 is smaller than the
laminate value (the results for that material system are not shown here).

In figures 6-4(b), 6-5(b), and 6-6(b), the deviation is shown for the reduced WR range
(0.03~0.11). As the WR range is reduced, the deviation band also shrinks. For carbon/epoxy
with flattened cross-section, the in-plane properties (E11, E22, G12, and v12) can be predicted quite
accurately with a maximum error of 2.5%. But for both the glass/epoxy and carbon/epoxy
material systems with lenticular cross-sections, the in-plane properties have larger deviation from
the laminate predictions. The deviation can be as large as 16%, which suggests that simple
laminate model may not be a good choice for approximation. So it can be concluded that the use
of laminate theory to approximate the properties of the braids is contingent upon the material
system and tow cross-section shape. For the transverse shear properties (G13 and G23), the
deviation is more than 29% for both material systems, which suggests the necessity for a more
accurate simple analytical model than 3D laminate theory.

6.5 COMPLEMENTARY BRAIDS.

If the braid angle is ±θ, then its complementary braid angle is defined here to be ±(90-θ).
Furthermore, the property pairs (E11, E22), (υ23, υ13), and (G23, G13) are defined to be
complementary of each other if the first property in any pair belongs to ±θ braid and the second
belongs to ±(90-θ) braid. The complementary properties are equal for an ordinary tape laminate.
But this is not the case for braids. This effect, which is not intuitively obvious, is due to the
difference in material architecture that exists in braids. This difference in material architecture is
called unbalance herein. This aspect of 2x2 braids was explored to see if there was an advantage
in selecting one braid over its complementary braid.

6-8
To illustrate the idea, figure 6-7 shows the axial moduli of a 2x2, -45° braid along four different
directions. It was observed that the value of E11' and E22' is exactly the same (= 26.8 GPa)
(Tang, 2001), as shown in figure 6-7. This is due to the fact that material architecture is identical
along X 1' and X 2' axes. One might expect the same for a 2x2, ±45° braid also. But, in the case of
2x2 braid, there exists a difference of approximately 3% in the values of E11 and E22, as shown in
figure 6-7 (E11 = 12.7 GPa and E22 = 12.3 GPa). This is because the interlace pattern is different
along lines marked by aa΄ and bb΄. Along aa΄, each +θ braid tow passes over -θ braid tow, but
along bb΄, one +θ tow passes over and then under a -θ tow making the interlace pattern different
along these two lines.

Figure 6-7. Difference in Properties of Complementary Braids

It was observed that the maximum difference in modulus is ~11% for the full range of WR
(0.03-0.3). For the realistic range of WR (0.03-0.11), the difference between complementary
moduli is negligibly small. The moduli have the largest differences (<4%), which is very small
for both the material systems. Hence, detailed results are skipped here.

6.6 COMPARISON OF PREDICTIONS WITH EXPERIMENTAL DATA.

A comparison was made with some of the preliminary experimental results for the carbon/vinyl
ester epoxy and carbon/EPON epoxy material systems used by Kelkar, et al., 2003.
Experimental results for carbon/vinyl ester are listed in tables 6-1 and 6-2 and carbon/EPON are
listed in table 6-3. The modulus and Poisson’s ratio available for comparison are E11 and ν12,
respectively.

6-9
 Table 6-1. Experimental Results for Carbon/Vinyl Ester

Specimen 1 2 3
Area, mm (in2)
2
129.03 (0.20) 123.95 (0.191) 128.06 (0.197)
Failure load, KN (Kips) 88.25 (19.84) 75.55 (16.99) 81.78 (18.39)
Ultimate tensile strength, MPa (Ksi) 683.31 (99.11) 612.52 (88.84) 638.59 (92.62)
% Elongation at failure 1.108 0.997 1.160
Modulus, GPa (Msi) 71.33 (10.35) 68.52 (9.94) 63.71 (9.24)
Poisson’s ratio 1.22 1.17 1.16
Braid angle 24.52 25.05 25.95

Kips = Kilo instructions per second

Table 6-2. Comparison With Experimental Results for Carbon/Vinyl Ester, BA 25° ±1°

Experiment
Finite Element
All Moduli Model Prediction Specimen Specimen Specimen
in GPa Flattened Lenticular 1 2 3
E11 64.67 77.60 71.33 68.52 63.71
E22 8.69 12.96 - - -
E33 8.26 8.52 - - -
G12 21.82 23.60 - - -
G23 2.75 2.51 - - -
G13 3.55 3.40 - - -
ν12 1.532 1.06 1.22 1.17 1.16
ν23 0.313 0.30 - - -
ν13 -0.226 -.02 - - -

Table 6-3. Comparison With Experimental Results for Carbon/EPON

Experimental FEM Experimental FEM Experimental FEM

±0.05 ±0.05 ±0.03
Fiber volume 0.51 0.51 0.53 0.51 0.52 0.51
fraction
Braid angle 25° ±2° 25º 30° ±1° 30º 45° ±1° 45º
Braided Tubes Slit Sleeves Slit Sleeves
Average E11, 54.23 57.03 38.04 40.31 15.92 13.59
Msi
Average ν12 1.44 1.53 1.34 1.469 0.69 0.797

FEM = Finite element model

6-10
The finite element results are also listed in table 6-2 along with experimental values for
comparison for carbon/vinyl ester material system. The braid angle of each specimen varies
within ±1° of 25°. The average of measured braid angles was used for the finite element results.
The finite element results are given for both flattened as well as lenticular cross-section. The
table shows that the experimental values of E11 and υ12 fall within the range of predicted results.
If one can determine the shape of cross-section in the actual material, finite element results might
be able to make a good prediction. Comparison for carbon/EPON is shown in table 6-3. Three
braid angles were used for experiments: ±25°, ±30°, and ±45°. For this material system,
experimental values of E11 and υ12 fall within the range of predicted results.

6.7 SUMMARY OF EFFECTIVE PROPERTY RESULTS.

The effective properties of the 2x2 braids were obtained using 3D finite element analysis. The
analysis showed that the material was essentially orthotropic. It was observed that in the range
of WR 0.03 to 0.11, which is more representative of most structural 2D braids, the in-plane
properties of carbon/epoxy material system with flattened cross-section can be predicted very
well by using a simple laminate model. However, for the glass/epoxy and carbon/epoxy with
lenticular cross-section, the laminate analysis can produce as large as 16% error for the in-plane
properties of the 2x2 biaxial braids. The most sensitive effective properties were found to be the
transverse properties (G13, G23, υ13, and υ23). This suggests that the simple laminate theory
cannot be used to get reasonable approximation for the transverse properties of the braid. It was
observed that the G13 and G23 can be as much as 72% greater than the laminate value, which
means a considerable increase in transverse shear modulus can be achieved using the 2x2 biaxial
braid, as compared to the equivalent angle-ply tape laminate. This can be significant for
structural applications in which higher G13 and G23 are desirable.

To predict the properties of braids using simple laminate analysis, it is important to know the
cross-section shape of the braid under investigation, as the response of a braid with flattened
cross-section can be predicted more confidently and with less error than that with a lenticular
cross-section. For the same overall fiber volume fraction for carbon/epoxy material system, the
lenticular cross-section was stiffer than a flattened cross-section.

The properties of complementary braids differ from each other due to unbalance in material
architecture that exists in braids. It was seen that for realistic range of WR (0.03-0.11) and braid
angle of 15°-75°, the different between complementary moduli was less than 4%, which is not
very significant.

6-11/6-12
7. PROGRESSIVE FAILURE ANALYSIS.

7.1 INTRODUCTION.

Periodic analysis is generally used in the modeling of textile composites, which are usually
idealized to possess perfectly uniform periodic microstructures. However, the as-manufactured
textile composites generally do not have perfectly uniform microstructures due to variations in
tow geometric properties such as waviness, cross-section shape, and fiber volume fractions.
Therefore, their behavior cannot be fully characterized by conducting periodic analysis on a
single unit cell model. This section focuses on the effect of variation in braid parameters on the
progressive failure behavior of a 2x2 braided composite laminate subjected to uniaxial loading.
A multiscale analysis approach (Ghosh, et al., 2001; Rudd and Broughton, 2000; and Oden,
et al., 1999) was used. The laminate consists of stacked 2x2 braid mats. Each mat is made of
interlaced tows and matrix pockets, and the tow contains fibers and matrix. A bottom-up
multiscale finite element modeling approach was employed that sequentially considered the
fiber/matrix scale, the tow architecture scale, and the laminate scale. The emphasis was placed
on examining the difference between the responses from the periodic analysis and the responses
from a model that considers the variation in braid parameters throughout the model. The details
of the analysis approach, the multiscale finite element models, and the numerical results are
described and the analysis results are presented and discussed.

7.2 DESCRIPTION OF MULTISCALE ANALYSIS.

The multiscale analysis method and approach used in this study is a sequential coupling method,
which individually considers three scales: fiber/matrix scale, tow architecture scale, and
laminate scale. Figure 7-1 illustrates these scales.

Textile composite structure

Textile laminate Textile mat (matrix pockets not shown)

(Re f.: P oe e t al, AIAA-97-1321)

X3

X2

X1 Model at fiber/matrix scale

Model at laminate scale for laminate behavior θ? , braid angle for tow properties
Variation: Variation:
Combination of braid angle, tow cross section Model at tow architecture scale for unit cell V f & packing pattern
and waviness ratio varies throughout (RVE) behavior
laminate. Variation:
Braid angle, tow cross section and waviness
ratio.

Figure 7-1. Overview of Multiscale Modeling of Textile Composites

7-1
The model for the fiber/matrix scale contains fibers and matrix. The fiber packing pattern and its
volume fraction can vary from model to model. In this study, only a hexagonal arrangement of
fibers was considered. The input requirements for properties at this scale are the moduli and
strengths for the fibers and pure matrix. Finite element models were built to obtain the effective
moduli of the fiber/matrix unit cell for a given fiber volume fraction, while the strength data of
the fiber/matrix unit cell are calculated using the Chamis simplified formulas (Chamis, 1984).
The results were used for the tows that were treated as a homogeneous transversely isotropic
material and were used in the unit cell models at the tow architecture scale.

The model at the tow architecture scale contains tows and matrix pockets. The construction of
the model depends on many geometric parameters required to define a specific textile, such as
waviness ratio, tow volume fraction, tow interlacing pattern, and tow braid angle. For the 2x2
braid considered in this research, a series of unit cell models were built to study the effect of
variation in the tow volume fraction, tow waviness, and braid angle. Periodic boundary
conditions were imposed. The properties of matrix pockets are the same as those used at the
fiber/matrix scale, while the properties of tows are the results coming from the analyses at the
previous scale, the fiber/matrix scale. Progressive failure analysis was conducted for each unit
cell model under uniaxial load condition. The progressive failure responses were then used in
the analysis at a larger scale, the laminate scale. To facilitate such utilization, the predicted
response was replaced by the equivalent nonlinear behavior of the unit cell model. In this way,
there was no failure information being passed onto the next modeling scale except this equivalent
nonlinear behavior, thus removing the need for a coupling between two scales. This was
accomplished as follows: At each loading step, the full set of effective properties of the unit cell
models were calculated based on the current damage state of the model. Six loading cases
needed to be considered to obtain the full set of 3D properties. This set of 3D properties was
then stored with the corresponding the applied strain. The collection of these results formed a
database and was used for the calculation of overall response at the laminate scale. It was
assumed that failure at a material point in the laminate model was governed by the axial stress
only. This results in a great simplification, which is probably justified, since the laminate is
subjected to uniaxial load only.

The analysis model at the laminate scale contained four 2x2 braid mats, stacked together with the
same orientation. Symmetry conditions were imposed on the lower surface and periodic
conditions on the upper surface. Hence, the model represented the behavior of an infinitely thick
laminate with a repeating eight mat group. To account for the variation in the braid parameters,
each finite element in the model was treated as a homogeneous orthotropic material whose
effective properties were determined from the previous modeling scale, the tow architecture
scale, for a specific combination of braid angle, tow volume fraction, and WR. The progressive
failure analysis proceeded such that at each loading step, the stiffness properties at each gauss
point of the element were updated using the predicted strain state from the laminate model and
the effective properties generated at the previous modeling scale.

7.3 TYPICAL FINITE ELEMENT MODELS.

The following describes the finite element models built for all three scales: fiber/matrix, tow
architecture, and laminate scales.

7-2
7.3.1 Fiber/Matrix Model.

As described in section 5, at the fiber/matrix scale, a hexagonal fiber arrangement was assumed
and a finite element model was built for a quarter of the full unit cell after exploiting the
symmetries (figure 5-11). Depending on the load conditions, different multipoint constraints
were imposed to the model. The fiber/matrix finite element model was used for the computation
of the effective moduli of material.

7.3.2 Tow Architecture Model.

The tow architecture models are similar to those shown in section 5-4. To study the effect of
variation in braid parameters on the predicted progressive failure behavior, a collection of finite
element meshes were built for various combinations of braid angle and tow volume fraction
(VTOW) while maintaining the overall fiber volume fraction (VfOverall = 57%) constant. The
complete list of the models is given in table 7-1. The model (B13) with braid angel BA = 28°
and VTOW = 74.33% was chosen as the reference unit cell model. Figure 7-2 shows the finite
element mesh for the reference unit cell model, which contains 480 hexahedral 20-node brick
elements and 2158 nodes.

Table 7-1. Parameters for Defining Various Unit Cell Models

VfOverall VTOW
Model (%) (%) Braid Angle Waviness Ratio
B01 57 63.76 24 0.167
B02 57 68.99 24 0.195
B03 57 74.33 24 0.236
B04 57 79.66 24 0.298
B05 57 85 24 0.404
B06 57 63.66 26 0.177
B07 57 68.99 26 0.207
B08 57 74.33 26 0.250
B09 57 79.66 26 0.316
B10 57 85 26 0.428
B11 57 63.66 28 0.200
B12 57 68.99 28 0.218
B13 57 74.33 28 0.273
B14 57 79.66 28 0.345
B15 57 85 28 0.451
B16 57 63.66 30 0.194
B17 57 68.99 30 0.236
B18 57 74.33 30 0.275
B19 57 79.66 30 0.347

7-3
 Table 7-1. Parameters for Defining Various Unit Cell Models (Continued)

VfOverall VTOW
Model (%) (%) Braid Angle Waviness Ratio
B20 57 85 30 0.507
B21 57 63.66 32 0.202
B22 57 68.99 32 0.236
B23 57 74.33 32 0.286
B24 57 79.66 32 0.360
B25 57 85 32 0.489

Figure 7-2. Finite Element Mesh for Reference Unit Cell Model

It should be noted that the variation in tow waviness was not explicitly considered due to the
findings from a previous study (Tang, et al., 2003). In that study, it was found that the in-plane
effective moduli were not very sensitive to the change of WR. Initial modeling results of this
study also showed little sensitivity of progressive failure behavior under in-plane axial loading to
small changes of tow waviness.

7.3.3 Laminate Model.

In this study, the finite element model was constructed as shown in figure 7-3. There were four
layers of braid mat along the thickness direction. The size of each element is comparable to the
size of the unit cell model at tow architecture scale. Each element is associated with one of the
unit cell models at the tow architecture scale to simulate the variation in braid angle and tow
volume fraction. Thus, the effective properties (nonlinear as described in the next few sections)
obtained at the tow architecture scale will be used for one element in the laminate model.

7-4
 Figure 7-3. Finite Element Model at the Laminate Scale

7.4 MATERIAL PROPERTIES FOR FAILURE ANALYSIS.

The material system used was carbon fiber/epoxy (AS4/411-350). Elastic behavior was assumed
for both the fibers and matrix. The basic input requirements are the engineering moduli, and
strength data for the fibers and matrix are listed below.

• Resin: 411-350 vinyl ester (Isotropic)

E11 = E22 = E33 = 3.1 GPa, ν12 = ν23 = ν13 = 0.35, and G12 = G13 = G23 = 1.15 GPa
X11 = X22 = X33 = 79 MPa and X12 = X13 = X23 = 72 MPa

• Fiber: AS4 (Transversely isotropic)

E11 = 227.53 GPa, E22 = E33 = 16.55 GPa, ν12 = ν13 = 0.2, ν23 = 0.25,
G12 = G13 = 24.82 Gpa, G23 = 6.89 GPa, and X11 = 4156 MPa

The effective moduli of tow were determined using finite element model of a hexagonal array of
fibers. Although the failure model implemented into the code considered difference of tensile
and compressive strength data, this particular study assumed identical tension and compressive
strength.

7.5 FAILURE CRITERIA, DEGRADATION MODEL.

To focus on the objective of this study, i.e., to understand the effect of variation in braid
parameters on the predicted progressive failure behavior of a braided composite laminate, no
elaborate failure study was conducted to identify the best failure criterion and properties
degradation model. Instead, all the progressive failure analyses conducted in this study used a
selected failure criterion and a degradation model. The material response was assumed to be
linear elastic until damage was predicted. The response after damage occurred was also assumed
to be linear elastic, but with degraded moduli. The maximum stress failure criterion was used to
predict damage. When a particular stress component exceeds its respective strength, the damage
mode caused by that stress component has occurred. The corresponding modified compliance
matrix S in terms of the original compliance matrix S is then given by

7-5
 ⎧ai Sii , i = j, no sum on i
Sij = ⎨ (7-1)
⎩ Sij , i≠ j

where ai are the degradation factors that are based on a modified version of Blackketter’s model
(Blackketter, 1993 and Chapman and Whitcomb, 2000) and are listed in table 7-2 (Chapman and
Whitcomb, 2000). Specifically, the degradation factors are chosen based on micromechanics
analysis of nonwoven laminates containing discretely modeled transverse matrix cracks
(Srirengan and Whitcomb, 1998).

Table 7-2. Degradation Factors for All Damage Modes

Degradation Mode of Damage in Fiber Tows Mode of Damage in Matrix

Factors σ11 σ22 σ33 σ12 σ23 σ13 σ11 σ22 σ33 σ12 σ23 σ13
a1 100 1 1 1 1 1 8 1 1 8 1 8
a2 8 8 1 8 8 1 1 8 1 8 8 1
a3 8 1 8 1 8 8 1 1 8 1 8 8
a4 8/3 8/3 1 8/3 8/3 1 8/3 8/3 1 8 1 1
a5 8/3 8/3 8/3 1 8/3 1 1 8/3 8/3 1 8 1
a6 8/3 1 8/3 1 8/3 8/3 8/3 1 8/3 1 1 8

The assumed fiber direction is x1

When the finite element method is used for the progressive failure analysis, the damage is
predicted at each quadrature point in the material coordinate system. Thus, for an element at a
certain point in the loading history, some quadrature points may or may not contain damage,
which is collectively taken into account in the element stiffness matrix calculation by degrading
the properties at each quadrature point in terms of the degree of predicted damage.

7.6 PROGRESSIVE FAILURE ANALYSIS FINDINGS.

There were multiple sets of results that were associated with each scale considered in the
multiscale analysis carried out in this study. Some were intermediate results, which transfer data
from a smaller scale to a larger scale, while others were from the entire multiscale analysis. In
the following, the results from each modeling scale will be presented and discussed.

7.6.1 Fiber/Matrix Scale.

At fiber/matrix scale, the results from the analysis were the effective moduli and strengths of the
fiber/matrix unit cell model. The effective moduli were obtained from the fiber/matrix finite
element model. Four loading cases, <σ11>, <σ22>, <σ12>, and <σ23>, were considered to get the
full 3D effective properties for a transversely isotropic material.

Figure 7-4 shows some of the effective properties of the fiber/matrix unit cell as a function of
fiber volume fraction. The axial engineering properties E11 and ν12, which vary linearly with the

7-6
fiber volume fraction and can be calculated using simple rule of mixture formulas, are not
shown.

1.6E+10 0.6

0.5
E22, E33, G12, G13 (Pa)

1.2E+10
E22=E33 0.4
G12=G13

v23
8.0E+09 G23 0.3
v23
0.2
4.0E+09
0.1

0.0E+00 0
0 0.2 0.4 0.6 0.8 1
Vf

Figure 7-4. Effective Properties of the Fiber/Matrix Scale as a Function of Fiber Volume
Fraction (The E11, ν12, and ν13 can be easily calculated using simple mixture rule.)

These results are used as the input properties for the tows contained in the unit cell model for the
tow architecture scale. Since at the tow architecture scale, the tow volume fraction varies from
model to model, the tow properties used are also different from model to model.

7.6.2 Tow Architecture Scale.

At tow architecture scale, the unit cell model is built for the given braid angle and tow volume
fraction (VTOW) while maintaining the overall fiber volume fraction (VfOverall) constant. The
reference unit cell model was built using BA = 28°, VTOW = 74.33%, and VfTOW = 76.69%. As
mentioned earlier, the input required from the previous smaller scale, fiber/matrix scale, was the
tow properties (effective moduli and strengths) for the tows contained at this scale. Therefore,
for the reference model, the tow properties were calculated at the previous scale using Vf = VfTOW
= 76.69%.

Progressive failure analyses were conducted for all unit cell models with parameters listed in
table 7-1. The load condition was the axial tensile loading <σ11>. As indicated in the
description of adopted multiscale analysis approach, no coupling between the current scale and
the next scale, the laminate scale, was considered. Consequently, the behavior of the unit cell
model for the current scale must be fully characterized before going to the analysis for laminate
scale. The precomputed progressive failure behavior at the tow architecture scale was then
recast in the form of equivalent nonlinear behavior. That is, all effective properties of the unit

7-7
cell model were expressed as function of the damage states predicted at each applied loading
step.

Figure 7-5 shows the progressive failure behavior for the reference unit cell model. It is
observed that the predicted response curve has some saw-teeth. This is typical for the failure
analysis of unit cell models of elastic material, especially when the unit cell has a large material
volume with a high stress. This results in a relatively large volume of material failing at about
the same time. Also shown are effective moduli calculated by considering the damage states
predicted at each loading step. Six loading cases were considered to obtain the full set of 3D
effective properties for the unit cell. It is clear that as the applied load increases, damage
initiates and then accumulates, as indicated by the decreasing effective moduli. To understand
how damage state causes moduli drop, figure 7-6 shows the damage state in the braided tows
(±θ) predicted at the given applied strain levels.

4.E+08 0

Moduli Drop due to Damage (%) .

G13
G12
E33
G23 -20
3.E+08
E11
<σ11> (Pa)

-40
2.E+08
-60

1.E+08
-80
σ11>
<<11>
E22

0.E+00 -100
0 0.005 0.01 0.015
<ε11>

Figure 7-5. Progressive Failure Response and Effective Properties as Function of Damage State
for the Reference Model

7-8
 <ε11>=0.92% <ε11>=0.95% <ε11>=0.98%

Failure
due to σyy

Failure
due to σxy

Failure
due to σyz

Failure
due to σxz

Figure 7-6. Failure States at Given Loading Levels for Reference Model

The progressive failure responses of models with different variation in braid angle and tow
volume fraction (see table 7-1) are compared in figure 7-7. Figure 7-7(a) shows the effect of
variation in tow volume fraction on the predicted behavior for a braid angle of 28°. It was found
that even when the models have the same braid angle (BA = 28°) and overall fiber volume
fraction (VfOverall = 57%), the variation in tow volume fraction (VTOW = 74.33 ±n·5.33%, n = 1, 2)
considerably affects the progressive failure responses. Since the variation in the tow cross-
section shape (thus the tow volume fraction) is not unusual in the manufactured textile structures,
it is beneficial to understand its effect on the progressive failure behavior of the material.

The effect of variation in braid angle (BA = 28 ±n·2°, n = 1, 2) is shown in figure 7-7(b).
Obviously, the variation in braid angle has a significant effect on the progressive failure
predictions, even with a variation as small as ±2°, which is not uncommon in realistic textile
structures.

Combined effect of variation in both braid angle (BA = 28 ±n·2°, n = 1, 2) and tow volume
fraction (VTOW = 74.33 ±n·5.33%, n = 1, 2) on the predicted behavior is shown in figure 7-7(c).
Significantly different predictions can be obtained from the periodic analysis of the unit cell
model when slightly varying braid angle and tow volume fraction from a desired braid
configuration. Understanding the effect of deviation from the desired nominal braid parameters
is significant.

7-9
 6.E+08

Tow volume fraction = 85%

4.E+08

<σ11>
2.E+08
Tow volume fraction = 63.67%

0.E+00
0 0.005 0.01 0.015

<ε11>
(a) Effect of variation in tow volume fraction
6.E+08

Braid angle = 24°

4.E+08
<σ11>

2.E+08

Braid angle = 28°

0.E+00
0 0.005 0.01 0.015

<ε11>

(b) Effect of variation in braid angle

6.E+08

Braid angle = 24°

4.E+08
and tow volume = 85%°
<σ11>
: :

2.E+08
Braid angle = 28°
and tow volume = 63.67%°
0.E+00
0 0.005 0.01 0.015

<ε: 11: >

(c) Combined effect of variations in braid angle and tow volume fraction

Figure 7-7. Comparison of Progressive Failure Responses for Various Unit Cell Models

7-10
7.6.3 Laminate Scale.

The effect of deviation from the desired nominal braid parameters on the progressive failure
behavior of a braid laminate was investigated at the laminate scale. The laminate has a desired
braid configuration, which is the same as the reference model for the tow architecture scale, with
BA = 28°, VTOW = 74.33%, and VfOverall = 57%. The deviation of braid angle and VTOW from the
nominal values was considered in the laminate model by including behavior of tow architecture
unit cell model with deviated braid parameters.

Six laminate models were built. The L11 model considered smaller deviations of tow volume
fraction (VTOW = 74.33 ±5.33%), and the L12 a larger deviation (VTOW = 74.33 ±n·5.33%, n = 1,
2). The effect of deviation of braid angle was consider in model L21 (BA = 28° ±2°) and L22
(BA = 28 ±n·2°, n = 1, 2). The combined effect of deviation of braid angle and tow volume
fraction were addressed using model L31 (BA = 28°° ±2°, VTOW = 74.33 ±5.33%) and L32 (BA =
28 ±n·2°, VTOW = 74.33 ±n·5.33%, n = 1, 2).

Each element in the finite element laminate model was randomly associated with a unit cell
model from the previous scale, which had the required braid parameters. For example, in the
L11 model, the unit cell models used are B12, B13, and B14 (see table 7-1). The equivalent
nonlinear behavior of these unit cell models, predicted by progressive failure analysis at the tow
architecture scale, was then used in the progressive analysis of the laminate model.

Figure 7-8 shows the predicted behavior of laminate models. Despite the different variations in
the braid parameters, all laminate models provided response close to that from the periodic
analysis of the reference unit cell model. Moreover, the saw-teeth exhibited in the response of
the reference unit cell model had been smoothed out. It was found that even a smaller deviation
of braid parameters, e.g., ±2° variation in braid angle, could basically eliminate the saw-teeth.
This is because when the laminate is not uniform, uniform failure can hardly occur on the entire
laminate level. This probably explains why the saw-teeth is seldom seen in actual experimental
tests of textile specimens, since highly uniform material microstructure is extremely difficult, if
not impossible, to manufacture.

7-11
 Vtow = 74.3 ± 5.33% Vtow = 74.3 ± 10.66%
5.E+08 5.E+08
L11 L12
4.E+08 4.E+08 Reference
Reference

3.E+08 3.E+08
<σ11>

<σ11>
2.E+08 2.E+08

1.E+08 1.E+08

0.E+00 0.E+00
0 0.005 0.01 0.015 0 0.005 0.01 0.015
<ε11> <ε11>

(a) Effect of deviation in to volume fraction

BA = 28 ± 2 deg BA = 28 ± 4 deg
5.E+08 5.E+08
L21 L22
4.E+08 Reference 4.E+08 Reference

3.E+08 3.E+08
<σ11>

<σ11>

2.E+08 2.E+08

1.E+08 1.E+08

0.E+00 0.E+00
0 0.005 0.01 0.015 0 0.005 0.01 0.015
<ε11> <ε11>

(b) Effect of deviation in braid angle

Vtow = 74.3 ± 5.33% BA = 28 ± 2 deg Vtow = 74.3 ± 10.66% BA = 28 ± 4 deg

5.E+08 5.E+08
L31 L32
4.E+08 Reference 4.E+08 Reference

3.E+08 3.E+08
<σ11>

<σ11>

2.E+08 2.E+08

1.E+08 1.E+08

0.E+00 0.E+00
0 0.005 0.01 0.015 0 0.005 0.01 0.015
<ε11> <ε11>

(c) Effect of deviation in both tow volume fraction and braid angle

Figure 7-8. Reponses of Braid Laminate Containing Variation in Tow Geometric Properties

7-12
7.7 SUMMARY.

Textile composite are traditionally idealized to have periodic tow architecture so that periodic
analysis can be conducted. However, the as-manufactured textile composite does not have
perfectly uniform periodic tow architecture. Therefore, understanding the behavioral differences
between the material with uniform microstructure and distorted microstructure is significant.

A multiscale analysis approach was adopted to study the progressive failure behavior of a 2x2
braid composite laminate. Three scales, fiber/matrix, tow architecture and laminate, were
sequentially considered and modeled. The basic input requirements are the material properties of
fiber and matrix, and nominal tow geometric properties such as braid angle and tow volume
fraction. The output from the analysis at a smaller scale is generally used as the input for the
analysis at a larger scale. Progressive failure analysis was conducted at the tow architecture
scale. The nonlinear (due to progressive failure) responses at this scale were then passed to the
next scale, the laminate scale, so that the progressive failure analysis at the laminate scale could
approximated by nonlinear analysis for a laminate containing a collection of different nonlinear
material points.

For the elastic material system and damage model used in this study, it was shown that the
progressive failure response from periodic analysis of the reference model (with nominal tow
architecture parameters) exhibits saw-teeth in the response curves. These saw-teeth are due to a
relatively large amount of material failing almost simultaneously. Once the variation in braid
angle and tow volume fraction was considered at the laminate scale, these saw-teeth were
smoothed out. This could explain why the saw-teeth are seldom seen in actual tests of textile
specimens, since highly uniform material microstructure is extremely difficult, if not impossible,
to manufacture.

The adopted process is actually a multilevel homogenization process. Several fundamental

assumptions were made to focus on gaining insight from the analysis. First, weak coupling
between the scales was assumed such that each scale can be treated individually and
sequentially. This assumption can be voided by increasing the model size and computation time.
Second, both matrix and fiber are treated elastic. In most cases, polymer or metal matrix exhibits
significant nonlinear behavior (e.g., due to plasticity). This can be considered by using more
elaborate material constitutive relationships in the model, such as in Tang, et al., 2001. Third,
the effectiveness of the failure criterion and property degradation model needs to be validated for
the textile architecture and material system under consideration.

7-13/7-14
8. STRESS DISTRIBUTIONS.

8.1 INTRODUCTION.

The stress distribution in braided composites is complex even for simple uniaxial loading. The
interlacing of the tows creates a complex load path that results in full 3D stress distributions.
The location and magnitude of peak stresses depend on the particular stress component and vary
with various braid architecture parameters such as braid angle and degree of waviness. Finite
element analyses for different braids were performed and it was seen that the peak stresses in the
tow mainly occur at the undulating region and along the edges of the tow. Stress distribution in
braids was compared with those in equivalent laminates.

Various techniques were used to process the stress distribution data. Stress contours give some
surface information about stress distributions. Of course, much of the information is not seen in
the contour plots. Stress volume distribution plots were used to characterize the extent of high
stress regions. It was shown that even for simple uniaxial loading, the stress state in braids is
fully 3D. The location and magnitude of peak stresses in the tow was shown.

A considerable volume of the tow (10%-45% for the considered range of parameters) had
stresses larger than an equivalent lamina. The severity of stresses in a braid compared to those in
an equivalent lamina depends upon braid geometric parameters. Braid angle changes the stress
distribution in the tow considerably. However, most of that effect is due to the orientation of the
tows in the +θ and -θ directions and can be eliminated by matching the loading on the tow of
different braids. The severity of peak stresses seems to be increasing linearly with an increase in
WR.

8.2 CONFIGURATIONS ANALYZED.

The effect of various parameters on the stress distributions in a symmetrically stacked braid was
studied. Uniaxial loading was applied along the longitudinal direction. The overall fiber volume
fraction in the model was assumed to be 50%. The range of braid angle considered in the stress
studies was ±15° to ±65°. Very low (1/20, 1/9), moderate (1/6) and very high (1/3) WRs were
considered.

The material system used consists of AS4 carbon fibers and EPON epoxy resin. The resin is
isotropic with E = 2.96 GPa and υ = 0.38. The fiber is transversely isotropic with the following
properties: E11 = 227.53 GPa, E22 = E33 = 16.55 GPa, G12 = G13 = 24.82 GPa, G23 = 6.89 GPa,
υ12 = υ13 = 0.2, and υ23 = 0.25, where the one direction is along the longitudinal axis of the fiber.
The symbols E, G, and υ refer to extensional modulus, shear modulus, and Poisson’s ratio
respectively. The fibers were assumed to be arranged in a hexagonal array in the tow, and the
properties of the tow were calculated using finite element based micromechanics. The fiber
fraction in the tow was assumed to be 0.69. The properties of the tow were determined to be
E11 = 157.9 GPa, E22 = E33 = 9.088 GPa, G12 = G13 = 4.839 GPa, G23 = 3.276 GPa, υ12 = υ13 =
0.251, υ23 = 0.4117.

8-1
8.3 STRESS DISTRIBUTION IN BRAIDS.

Figure 8-1 shows the locations of peak stresses in the tow when unit uniaxial load
(<σxx > = 1) along the longitudinal direction is applied to a ±25° braid with WR of 1/3. Even for
simple loading like this, a 3D stress state exists in the tow and any stress component could be
critical, depending upon the allowables. Figure 8-1 shows the contours for all the six stress
components with their respective ranges. The value of each stress component for an equivalent
lamina is also given. For a lamina, only in-plane stresses are nonzero and the value of each
stress component lies within the braid stress range. The out-of-plane stress values are zero
because there are no free edge effects. Moreover, the in-plane stresses in the lamina of a tape
laminate are constant, but the value of each stress component varies significantly in the braid
tow.

Figure 8-1. Three-Dimensional Stress State in Tow for ±25° Braid With WR = 1/3 (<σxx> = 1)

Three-dimensional stress distributions exist for ±45° and ±65° braid tows (not shown in figure
8-1) also, but there are certain similarities and differences in the stress distributions as one
changes the braid angle. The σ11 stress peaks in the tow are tensile for all the braid angles. In
contrast, the peaks for σ22 are compressive for ±25° and tensile for ±45° and ±65°, which is
consistent with equivalent laminates. Peaks for σ12 lie on the edge of the tow. The σ13 is the
only component whose peaks extend through the thickness of the tow, as shown in figure 8-1.
For the rest of the stress components, the peaks are only on the surface of the tow. For all the
braid configurations, there were significant tensile and compressive σ33 concentrations.
However, for a ±65° braid, tensile peaks were much larger than the compressive peaks. Figure
8-2 shows the effect of braid angle on the σ33 stress distribution. Braids with different braid
angles were stressed at the same stress level (<σxx> = 1). It is clear that stress distribution differs
considerably with braid angle. With increase in braid angle to ±65°, the peak stresses change

8-2
from tension to compression in nature in the center portion of the tow. Also, the location of peak
stresses changes from center portion of the tow to the edge of the tow. Magnitudes of tensile
peaks also change from 0.22 to 0.367 for a ±15° and ±65° braid, respectively. The ±65° braid
tow has a smaller tensile area, which is near its edges, but a larger tensile value than the others.
In summary, the location and magnitude of peak stresses vary considerably with braid angle.

Figure 8-2. Effect of Braid Angle on σ33 Stress Concentration

(Uniaxial loading (<σxx> = 1), WR = 1/3)

8.4 VOLUME DISTRIBUTION CURVE.

Stress contour plots shown in figures 8-1 and 8-2 give stress information only on the surface of
the tow. To get internal information, one has to cut the tow and show more plots. The
percentage of the tow having peak stress can be small enough not to be noticeable in the stress
contours. Also, it is possible that peak stresses are hidden in the interior of the tow. One needs
to determine what percentage of the tow exceeds a certain stress level. A volume distribution
plot quantifies the percentage volume of the material that is stressed more than a particular value.
Figure 8-3 shows a typical volume distribution plot of σ33 in the +θ tow of a ±30° braid with
WR = 1/3 when uniaxial load is applied. The σ33 in an equivalent lamina is also plotted. For the
lamina, σ33 is zero. The σ33 in the braid is normalized by the applied stress. The figure shows
that 18% of the volume of the tow has a tensile σ33. Point A, which is in the tensile region,
indicates that 10% of the volume has a σ33 that is larger than 0.037 times the applied stress.
Point B, which is in the compressive region, indicates that 24% of the volume has a compressive
stress larger than 0.1 times the magnitude of the applied stress. The information provided by a
volume distribution plot can be used in several ways. For example, the percentage of the tow
having larger stress than that in an equivalent lamina can be found; the percentage of the tow
having tensile or compressive stress can be found; the percentage of the tow having stress larger
than a particular value can be found. Also, questions like, what percentage of the tow has a
larger σ33 than 0.25 times the applied stress can be answered.

8-3
 100
Laminate
24%
80
B
% Volume of tow

60
Braid
40

20
A
10%
0
-0.35 -0.25 -0.15 -0.05 0.05 0.15 0.25 0.35
σ 33 / < σ xx > braid

Figure 8-3. σ33 Volume Distribution in ±30° Braid Tow With WR = 1/3 When
<σxx> = 1 was Applied

8.5 COMPARISON OF STRESS DISTRIBUTION IN BRAIDS WITH EQUIVALENT

LAMINATES.

Tape laminates and braids have their own advantages and disadvantages in terms of ease and
cost of manufacturing, engineering properties, and ease of analysis. Analysis of tape laminates is
easily understood by designers and engineers, but tapes have the disadvantage of hand lay-up
and high manufacturing cost. In contrast, analysis of braids is complicated, but they have an
edge in terms of manufacturing cost. Here, the severity of stresses in braids compared to those in
equivalent tape laminates is discussed. Figure 8-4 shows the volume distribution of in-plane
normal stresses in the +θ tow of ±45° braid when unit uniaxial load along the longitudinal
direction was applied (<σxx> = 1). In both the plots, the volume distribution curves correspond to
five WRs: 1/3, 1/4, 1/6, 1/9, and 1/20. The vertical straight line corresponds to the constant
stress value in the +θ lamina of an equivalent tape laminate. The stresses shown in figure 8-4 are
normalized with the corresponding absolute +θ lamina values. That is, the plots show
σ11_tow/|σ11_lamina| and σ22_tow / |σ22_lamina|. There are several differences between the stress
distribution in a braid tow and a lamina. Figure 8-1 showed that braid tows have a wide
variation in in-plane stresses. In contrast, in-plane stresses in a lamina are constant. The volume
of the tow that exceeds the value of stresses in a tape lamina gives the severity of stresses in a
tow as compared to those in lamina. Figure 8-4(a) shows the volume distribution of σ11 for a
±45° braid tow for several WRs. The figure shows that 19% of the tow (marked by arrow A) has
larger σ11 than an equivalent lamina for WR = 1/3 and the severity is maximum for this WR.
When the WR is reduced to 1/20, the severity reduces to 11.5% (marked by arrow B). The σ11
severity is smallest for the smallest WR (1/20) in the considered range of WR (1/20-1/3). Figure
8-4(b) shows the volume distribution of σ22. The figure shows that 31% of the tow (marked by
arrow A) has larger σ22 than an equivalent lamina for WR = 1/3. When the WR is reduced to
1/20, the severity reduces to 19% (marked by arrow B). For all other WRs considered, the

8-4
severity lies in between the highest WR (1/3) and lowest WR (1/20) for both the in-plane stresses
shown in figure 8-4.

100
WR=1/3 WR=1/4

80 WR=1/6 WR=1/9
% Volume of the tow

WR=1/20 Laminate
60

40
A
20
B
0
0.4 0.6 0.8 1 1.2 1.4 1.6
(a) σ11_tow / |σ11_lamina|

100
WR=1/3 WR=1/4

80 WR=1/6 WR=1/9
% Volume of the tow

WR=1/20 Laminate
60

40 A

20
B
0
0 0.5 1 1.5 2 2.5
(b) σ22_tow / |σ22_lamina|

Figure 8-4. Comparison of Stress Volume Distribution in a ±45° Braid With That in an
Equivalent Tape Laminate
(Normalized stress ( σ tow | σ lamina | ) in the tow vs percentage volume of the tow exceeding a
particular value is plotted.)

8-5
Figure 8-4 shows the stress distribution for ±45° braid only. The severity of stresses in ±25° and
±65° braids was also investigated. The results are tabulated in table 8-1, which show the severity
of stresses for the considered range of BA (±25°-±65°) and WR (1/20-1/3).

Table 8-1. Severity of Stresses in Braid Compared to an Equivalent Tape Laminate

Range of BA = ±25°-±65°, Range of WR = 1/20-1/3

% volume of tow having more severe stress than those in an equivalent lamina
WR =1/3 WR =1/6 WR =1/9 WR =1/20
BA ±25° ±45° ±65° ±25° ±45° ±65° ±25° ±45° ±65° ±25° ±45° ±65°
σ11 40 18.5 24.6 19.5 12.3 27.5 16 12.1 28.8 15 11.2 30
σ22 45 31 27.3 38.5 21 31 36 18.5 30 36 18.2 29.5
σ12 24.5 33 9.5 24.5 33 26.5 27.5 33.7 29.5 28 34 30

The severity decreases with a decrease in WR for ±25° and ±45° braids for both the in-plane
normal stresses. For ±65° braid, the severity increases with a decrease in WR for σ11 and
remains almost constant for σ22. For σ12, the severity increases with a decrease in WR for all the
braid angles. Overall, in this range of parameters, maximum volume of the tow that could have
more stresses than an equivalent lamina is 40% for σ11, 45% for σ22, and 34% for σ12.

8.6 EFFECT OF MATCHING THE LOADING ON STRESS DISTRIBUTION = EFFECT OF

BRAID ANGLE AND INTERLACING.

Figure 8-5 shows the effect of braid angle on the σ22 and σ13 stress volume distribution. The
WR = 1/3 and a unit volume averaged stress of <σxx> was applied. Also shown are the values of
equivalent tape laminates corresponding to each braid. The dotted lines show the constant
lamina stress values. There is wide variation in the stresses in the tow of all the braids. Figure
8-5(a) shows that as the braid angle changes, the σ22 stress distribution changes considerably.
The ±15° braid tow has compressive stresses in >95% of its volume, whereas ±45° and ±65°
braid tows have only tensile stresses in their entire volume. The peak values of σ22 for ±15°,
±45°, and ±65° tows are -0.14, 0.4, and 1.5, respectively. Figure 8-5(b) shows the variation for
σ13. The laminate value is zero for all equivalent laminates in this case, but braids have wide
variation. The wide variation in σ22 is mostly due to orientation effect and is expected based on
the behavior of an equivalent laminate. The orientation affects the volume distribution for other
stress components also. Hence, a technique was used to eliminate the effect of orientation. A
volume averaged stress <σxx> = 1 was applied on ±45° braid. The +45° braid tow experienced
certain amount of loading because of this. That is, the tow was subjected to the following
volume averaged stresses: <σ11> = 1.203, <σ22> = 0.087, <σ33> = -0.0026, <σ12> = -0.593,
<σ23>=0, and <σ13>=0, where the 1 direction is along the axis of the braid tow. The same
amount of loading was applied on a +25° braid tow. This was made possible by applying
multiaxial loading on the ±25° braid model. The same thing was done for other braid angles.
The stress volume distribution in the different braids was compared. The stress volume
distribution after matching the loading on tows of different braids is shown in figure 8-6. It is
interesting to see that now the stress volume distribution curves lie very close to each other
compared to when the loading on tow was not matched (figure 8-5). The σ22 ranges from tensile
to compressive for all the braids, as shown in figure 8-6(a). The peak values of all the braids are

8-6
also very close to each other for all the braids. Similarly, volume distribution curves came closer
for σ13 stress also, as shown in figure 8-6(b). The orientation effect could be eliminated similarly
for other stress components as well. The difference that still remains is attributed to the fact that
the tows are interlaced and the tow shape of different braids is different due to phase shift in the
undulation. This phase shift is shown in figure 8-7. In this figure, x, y, and z are the local
coordinates. Different fibers across the width of the tow start to undulate at a different
x distance, which means that they have a different phase angle. The phase shift is given by
Φ = y*tan(2θ°-90°), where θ is the braid angle. Due to this, different fibers of the tow do not
undulate and straighten at the same x coordinate. One fiber may have started undulating and
another may not have yet started to undulate. The overall effect of this phase shift is different
material architecture for different braids, which results in different stress distribution even if the
loading on the tow of different braids was matched.

100
BA=15
% Volume of tow

80 BA=45
BA=65

60 Laminate

40

20

0
-0.3 0 0.3 0.6 0.9 1.2 1.5
(a) σ22 volume distribution
100
BA=15
BA=45
80
% Volume of tow

BA=65
Laminate
60

40

20

0
-0.15 -0.1 -0.05 0 0.05 0.1 0.15 0.2 0.25
(b) σ13 volume distribution

Figure 8-5. Effect of Braid Angle on σ22 and σ13 Volume Distribution
(<σxx> = 1, WR = 1/3)

8-7
 100
BA=15
80 BA=45
% Volume of tow

BA=65
60

40

20

0
-0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8

(a) σ22 volume distribution

100
BA=15
80 BA=45
% Volume of tow

BA=65
60

40

20

0
-0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8

(b) σ13 volume distribution

Figure 8-6. Effect of Braid Angle on σ22 and σ13 Volume Distribution When <σij> in the Tow
are Matched (WR = 1/3)

8-8
 Figure 8-7. Phase Shift

8.7 EFFECT OF WAVINESS RATIO.

Figure 8-8 shows the effect of WR on the stress volume distribution for a ±45° braid. A unit
uniaxial load (<σxx> = 1) was applied. Three different WRs (1/3, 1/6, and 1/9) were used. The
figure shows the stress volume distribution for all the stress components. It is clear that for all
the stress components, with an increase in WR, the volume distribution curve tends to broaden in
the horizontal direction. Therefore, the severity of the peaks increases with an increase in
waviness ratio for all the stress components. The effect of waviness ratio is more pronounced for
out-of-plane stresses (σ33, σ23, and σ13) than for in-plane stresses (σ11, σ22, and σ12).

Figure 8-9 summarizes the effect of WR on the stress peaks. The figure shows the variation of
normalized stress peak values with WR for all the six stress components. The peak values
correspond to a particular percentage of the tow volume. For example, in figure 8-9(a), 5% of
the tow volume had more severe stresses than the value plotted for all the stress components.
Similarly, in figure 8-9(b), 2% of the tow volume was chosen to find the value of peak stresses.
The σij (where i, j = 1, 2, 3) peak values were normalized with the peak values corresponding to a
WR of 1/9. The results are for a ±45° braid under unit uniaxial load (<σxx> = 1). It is clear that
the peaks increase linearly with an increase in waviness ratio (except for σ12 for which there is
little variation). Again, is clear that the effect of WR is more pronounced for out-of-plane
stresses compared to in-plane stresses. Out-of-plane normal stress is most severely affected.

8-9
 100
100
80 80
60 60
40 40
20 20
0 0
0 0.5 1 1.5 2 2.5 -0.1 0 0.1 0.2 0.3 0.4
σ11 σ22
100 100
80 80
60 60
40 40
20 20

0 0
-0.3 -0.2 -0.1 0 0.1 0.2 -1.2 -1 -0.8 -0.6 -0.4 -0.2

σ33 σ12
100 100
80 80
60 60
40 40
20 20
0 0
-0.6 -0.4 -0.2 0 0.2 -0.3 -0.2 -0.1 0 0.1 0.2 0.3

σ23 σ13

Waviness ratio:

Figure 8-8. Effect of Waviness Ratio on Stress Volume Distribution, BA = ±45°

(For all the cases, stress component is plotted along the horizontal, and % volume of the tow is
plotted along the vertical. Applied stress is <σxx> = 1.)

8-10
 5
σ33
4

3 σ13
σ/σWR =1/9

2 σ23

σ22
1
σ11
σ12
0
0.1 0.15 0.2 0.25 0.3 0.35
Waviness Ratio
(a) Peak stresses correspond to 5% of the tow volume

5
σ33
4

3
σ/σWR =1/9

σ13
2 σ23
σ22
1
σ11
σ12
0
0.1 0.15 0.2 0.25 0.3 0.35
Waviness Ratio
(b) Peak stresses correspond to 2% of the tow volume

Figure 8-9. Variation of Peaks With Waviness Ratio ±45° Braid Under <σxx> = 1

8.8 SUMMARY.

The tow stress state in 2x2 braids was investigated. The effect of various parameters on the
stress state was investigated. The following observations were made:

• A complex 3D stress state, which is fully 3D, exists in the tow even for simple uniaxial
loading.

8-11
• In the considered range of parameters (WR = 1/3-1/20, BA = ±25°-±65°), a considerable
volume of the tow has more stresses than an equivalent lamina.

• The wide variation in stress volume distribution with braid angle is due to simple
orientation effects and can be eliminated by matching the loading on the tow. Some
difference that still remains can be attributed to the phase shift and interlacing effect.

• The severity of the peaks increases linearly with an increase in WR for all stress
components (except for σ12 for which there is little variation).

8-12
9. CONCLUSIONS.

Low-cost, vacuum-assisted resin transfer molding (VARTM) was successfully implemented to

manufacture 2x2 biaxial braided composites using two different resin systems with carbon braids
(i.e., DM 411-350 vinyl ester and EPON 9504 epoxy resin). The DM 411-350 vinyl ester resin
system was previously studied by Smith in 2001, and the same process parameters of VARTM
were used in the present study. The following are the conclusions concerning braided
composites manufactured by using carbon braids and the EPON 9504 epoxy resin system:

• Good quality panels can be obtained if the processing is performed at an elevated

temperature of 49ºC (120ºF). Both the mold and resin need to be heated to 49ºC (120ºF).

• The complete wet-out of the fabric was achieved by placing the resin distribution media
and the vacuum line in a specific way.

• A flow control device like the peristaltic pump was not required for this resin system.
This is in contrast to its requirement for the DM 411-350 vinyl ester resin system.

• Double bagging may prevent leakages in the vacuum bag, but it may not improve overall
fiber volume fraction.

• A small amount of surfactant in the resin may cause the delamination of plies.

• The braided composite panels exhibited variation in thickness between the vacuum line
and the resin line.

• The experimental study indicated that the overall fiber volume fraction is independent of
the braid angle. The composites manufactured by VARTM yielded a 50% fiber volume.

The performance of 2x2 biaxial braided composites was studied in detail for two different resin
systems (i.e., DM 411-350 vinyl ester and EPON 9504 epoxy resin) under static tension and
tension-tension fatigue loading (R = 0.1). The effect of the braid angle on the fundamental
mechanical properties and fatigue life was studied. The following are the conclusions of the
performance evaluation:

• The fundamental mechanical properties (i.e., Young’s modulus, Poisson’s ratio, and
ultimate tensile strength) for both carbon/vinyl ester and carbon/epoxy material systems
were very close at a braid angle of 25º.

• The axial modulus, Poisson’s ratio, and ultimate tensile strength decreased as the braid
angle increased from 25º to 45º.

• The braid angle variation within a specimen was controlled to ±1º by using slit sleeves.
A large variation in braid angle within a specimen caused large scatter in the ultimate
tensile strength and the fatigue life.

9-1
• Based on a limited number of fatigue tests, the endurance limit of the braided composites
was 40% to 50% of the ultimate tensile strength for the braid angles between 25º and 45º
for carbon/epoxy and carbon/vinyl ester material systems. At higher stress levels,
carbon/vinyl ester material systems showed better fatigue life.

• The S-N diagram of braided composites for various braid angles can be approximated
using the Sigmoidal (Boltzmann) function.

• Stiffness degradation curves exhibit a typical three-stage pattern. However, stiffness

degradation was more rapid in braided composites than in woven composites. The
decrease in fatigue secant modulus at failure can be as high as 60% of the fatigue secant
modulus in the first cycle. Thus, the damage accumulation rate was much higher for
braided composites.

• The stress level near the endurance limit is very critical. Fatigue life dropped from 1
million cycles to 5000 cycles when stress levels increased from 40% to 50% for braid
angles of 25º and 30º. In the same manner, fatigue life dropped from 1 million to 50,000
cycles when the stress level increased from 50% to 60% for a braid angle of 45º.

• The failure of braided composites always occurred suddenly in the last 10% of the fatigue
life without any visible matrix cracking or delamination of plies.

An analytical model was developed from stiffness degradation curves to predict the residual
fatigue secant modulus. The following are the conclusions concerning the modeling effort of
2x2 biaxial braided composites:

• The first stage of stiffness degradation curves, which displays a rapid decrease in the
stiffness, exists for approximately the first 5% of the fatigue life. Therefore, the
endurance limit of braided composites can be evaluated by performing fatigue tests for
only 5% of the fatigue life. The stress level at which the stiffness remains almost
constant indicates the endurance limit.

• A unique analytical model was developed based on stiffness degradation curves to predict
the residual fatigue secant modulus. This model is in good agreement for all three stages
of the stiffness degradation.

A computational micromechanics strategy was developed to model 2x2 braids. Solid models
were also created to understand the material architecture of 2x2 braids. The following are the
conclusions related to computational modeling of 2x2 braids.

• The tow x section along the length of the tow of 2x2 braids varies unusually.

• Since the tow cross-section along the tow path is not uniform, direct finite element mesh
generation for the model is difficult. A mapping technique was developed to generate the
finite element mesh for various 2x2 biaxial braids from the mesh for the twill weave,
which had been developed in previous studies. This resulted in substantial time savings.

9-2
• The boundary conditions are complex and are in terms of many multipoint constraint
relations. By exploiting symmetry operations like mirroring, rotation or a combination of
the two, the analysis region was reduced to half (for simple stacking) or one-fourth (for
symmetric stacking) of the unit cell.

The effect of various parameters like braid angle, waviness ratio (WR), material properties, and
cross-section shape on the effective engineering properties of the 2x2 braids was investigated.
The effective properties of braids were also compared with those of equivalent laminates. The
following conclusions can be drawn from these parametric studies:

• In-plane properties of carbon/epoxy material system with flattened cross-section can be

predicted very well by using a simple laminate model. However, for the glass/epoxy and
carbon/epoxy with lenticular cross-section, the laminate analysis can produce as large as
16% error for the in-plane properties of the 2x2 biaxial braids.

• The most sensitive effective properties were found to be the transverse properties (G13,
G23, υ13, and υ23). This suggests that the simple laminate theory cannot be used to get
reasonable approximation for the transverse properties of the braid.

• The G13 and G23 can be as much as 72% greater than the laminate value, which means a
considerable increase in transverse shear modulus can be achieved using the 2x2 biaxial
braid compared to the equivalent angle-ply tape laminate. This can be significant for
structural applications in which higher G13 and G23 are desirable.

• To predict the properties of braids using simple laminate analysis, it is important to know
the cross-section shape of the braid under investigation as the response of a braid with
flattened cross-section can be predicted more confidently and with less error than that
with a lenticular cross-section.

• In-plane modulus of braids is less than those of equivalent tapes.

• Moduli predictions fall within the range of experiments.

The stress distribution in a braid tow was analyzed. The effect of various braid architecture
parameters, such as braid angle and degree of waviness, on the location and magnitude of peak
stresses was studied. Stress distribution in braids was compared with those in equivalent
laminates. The conclusions are summarized below.

• A complex three-dimensional (3D) stress state, which is fully 3D, exists in the tow even
for simple uniaxial loading.

• In the considered range of parameters (WR = 1/3-1/20 and braid angle (BA) = ±25°-
±65°), a considerable volume (40% for σ11, 45% for σ22, and 34% for σ12) of the tow has
larger in-plane stresses than an equivalent lamina.

9-3
• The wide variation in stress volume distribution with braid angle is due to simple
orientation effects and can be eliminated by matching the loading on the tow. Some
difference that still remains can be attributed to the phase shift and interlacing effect.

• The severity of the peaks increases linearly with an increase in WR for all stress
components (except for σ12, for which there is little variation).

The effect of variation in braid parameters on the progressive failure behavior of a 2x2 braided
composite laminate was studied. A bottom-up multiscale finite element modeling approach was
employed that sequentially considered the fiber/matrix scale, the tow architecture scale, and the
laminate scale. The following are the conclusions from progressive failure analysis of 2x2 braids
with microstructural heterogeneity.

• The progressive failure response from periodic analysis of the reference model (with
nominal tow architecture parameters) exhibits saw-teeth in the response curves. These
saw-teeth are due to a relatively large amount of material failing almost simultaneously.
Once the variation in braid angle and tow volume fraction was considered at the laminate
scale, these saw-teeth are smoothed out. This could explain why the saw-teeth are
seldom seen in actual tests of textile specimens, since highly uniform material
microstructure is extremely difficult, if not impossible, to manufacture.

9-4
10. BIBLIOGRAPHY.

3M Company, http://www.3m.com

A&P Technology Inc. 2000, “Frequently Asked Questions,”

available at http://www.braider.com, accessed in June 2002.

Abdallah, M.G., Senior Staff Scientist, Hexcel Corporation-Composites/Carbon Fibers private

communication via email, 2002.

Amateau, M.F., Professor of Engineering Science and Mechanics, The Pennsylvania State
University, Course material for “Engineering Composite Materials EMCH471,”
http://www4.esm.psu.edu/academics/courses/emch471/Notes/Chapter18.PDF.

ASTM D 638, “Standard Test Method for Tensile Properties of Plastics,” ASTM Standards and
Literature References for Composite Materials, Philadelphia, Pennsylvania.

ASTM D 3039, “Standard Test Method for Tensile Properties of Fiber Resin Composites,”
ASTM Standards and Literature References for Composite Materials, Philadelphia,
Pennsylvania.

ASTM D 3479, “Standard Test Method for Tension-Tension Fatigue of Polymer Matrix
Composite Materials,” ASTM Standards and Literature References for Composite Materials,
Philadelphia, Pennsylvania.

Battat, http://amptiac.alionscience.com/pdf/2001MaterialEase16.pdf

Beck, W., 1993, “Designing the RTM Process and Product,” SME Technical Paper (Series), EM
1993, Publication by SME, pp. 1-13.

Beckwith, S.W. and Hyland, C.R., 1998, “Resin Transfer Molding: A Decade of Technology
Advances,” SAMPE Journal, V 34, pp. 7-19.

Blackketter, D.M., Walrath, D.E., and Hansen, A.C., 1993, “Modeling Damage in a Plain Weave
Fabric-Reinforced Composite Material,” Journal of Composites Technology & Research, 15(2):
136-142.

Bolick, R. and Kelkar A.D., North Carolina A&T State University, Greensboro, NC 27411,
private communication via email, 2004.

Burr, S.T. and Morris, D.H., 1995, “Two-Dimensionally Braided Composites Subjected to Static
and Fatigue Loading,” Virginia Polytechnic Institute and State University, Mechanics of Textile
Composites Conference, pp. 33-53 (SEE N96-25071 09-24).

Byun, J.H., 2000, “The Analytical Characterization of 2-D Braided Textile Composites,”
Composites Science and Technology, Vol. 60, pp. 705-716.

10-1
Carlos, K.G., 1994, “Fatigue Behavior of 3-D Triaxially Braided Composites,” Thesis (M.S.),
North Carolina A&T State University.

Chamis, C.C., 1984, “Simplified Composite Micromechanics Equations for Strength, Fracture
Toughness, Impact Resistance and Environmental Effects,” Report NASA TM-83696.

Chapman, C.D. and Whitcomb, J.D., 2000, “Thermally Induced Damage Initiation and Growth
in Plain and Satin Weave Carbon-Carbon Composites,” Mechanics of Composite Materials and
Structures, Vol. 7(2), pp. 177-194.

Chen, J., McBride, T.M., and Sanchez, S.B., 1998, “Sensitivity of Mechanical Properties to
Braid Misalignment in Triaxial Braid Composite Panels,” Journal of Composites Technology &
Research, Vol. 20(1), pp. 13-17.

Clemente, R., Castejon, L., and Miravete, A., 1998, “Energy Absorption of 2D Triaxial Braided
Composite Structures,” International SAMPE Technical Conference, San Antonio, Texas, USA,
Vol. 30, pp. 420-431.

Daniel, I.M. and Ishai, O., 1994, “Engineering Mechanics of Composite Materials,” Oxford
University Press.

Deaton, J.W., Kullerd, S.M., Madan, R.C., and Chen, V.L., 1992, “Test and Analysis Results for
Composite Transport Fuselage and Wing Structures,” Second NASA Advanced Composites
Technology Conference, pp. 169-193.

Degrieck, J. and Paepegem W.V., 2001, “Fatigue Damage Modelling of Fibre-Reinforced

Composite Materials: Review,” Applied Mechanics Reviews, Vol. 54(4), pp. 279-300.

Falzon, P.J. and Herszberg, I., February 1998, “Mechanical Performance of 2-D Braided
Carbon/Epoxy Composites,” Composites Science and Technology, Vol. 58(2), pp. 253-265.

Fawaz, Z. and Ellyin, F., 1994, “Fatigue Failure Model for Fibre-Reinforced Materials Under
General Loadings Conditions,” Journal of Composites Materials, Vol. 28(15), pp. 1432-1451.

Fujii, T. and Amijima, S., 1993, “Microscopic Fatigue Processes in a Plain-Weave Glass-Fiber
Composite,” Composites Science and Technology, Vol. 49, pp. 327-333.

Gates, T.M., 2002, Unclassified Presentation, Mechanics and Durability Branch, NASA Langley
Research Center.

Ghosh, S., Lee, K., and Raghavan, P., 2001, “A Multi-Level Computational Model for Multi-
Scale Damage Analysis in Composite and Porous Materials,” International Journal of Solids and
Structures, Vol. 38, pp. 2335-2385.

10-2
Heider, D., Hofmann, C., and Gillespie, J.W., Jr., 2000, “Automation and Control of Large-Scale
Composite Parts by VARTM Processing,” International SAMPE Symposium and Exhibition
(Proceedings), Long Beach, California, USA, V 45(2), pp. 1567-1575.

Hess, J.P., 1990, “Fabrication of Braided Composite Structures by Resin Transfer Molding,”
National SAMPE Technical Conference, V 22.

Houston, D.Q. and Chernenkoff, 1992, “RA Environmental Effects on the Fatigue Behavior of a
Braided ‘E’ Glass Composite Resin System Advanced Composites: Design,” Materials and
Processing Technologies, Chicago, Illinois, USA, 2-5 November. 1992 ASM International
(USA), pp. 343-353.

Hwang, W. and Han, K.S., 1986a, “Cumulative Damage Models and Multi-Stress Fatigue Life
Prediction,” Journal of Composite Materials, V 20, pp. 125-153.

Hwang, W. and Han, K.S., 1986b, “Fatigue of Composites-Fatigue Modulus Concept and Life
Prediction,” Journal of Composite Materials, V 20, pp. 154-165.

Juska, T.D. and Puckett, P.M., 1997, “Matrix Resins and Fiber/Matrix Adhesion,” Chapter 3,
Composites Engineering Handbook, Marcel Dekker, Inc., New York, pp. 101-157.

Kelkar, A.D. and Tate, J.S., 2003a, “Fatigue Behavior of VARTM Manufactured Biaxial Braided
Composites,” ASME 2003 International Mechanical Engineering Congress and RD & D
Exposition, Washington, DC, USA.

Kelkar, A.D. and Tate, J.S., 2003b, “Effect of Fatigue Loading on the Stiffness Degradation of
th
VARTM Manufactured Biaxial Braided Composites,” 9 International Conference on the
Mechanical Behavior of Materials (ICM9), PALEXPO Congress Center, Geneva, Switzerland.

Kelkar, A.D., Tate, J.S., Whitcomb, J., and Tang, X., 2003c, “Performance Evaluation and
Modeling of Braided Composites,” 44th AIAA/ASME/ASCE/AHS Structures, Structural
Dynamics, and Materials Conference, Norfolk, Virginia, USA.

Kelkar, A.D. and Tate, J.S., 2002, “Low Cost Manufacturing of Textile Composites Using
Vacuum Assisted Resin Transfer Molding,” The 20th All India Manufacturing Technology,
Design and Research Conference, AIMTDR, Ranchi, India, pp. 712-716.

Kelkar, A.D., Chaphalkar, P., and Sankar, J., 1999, “Fatigue Behavior of Resin Infusion and
Resin Transfer Molding S2-Glass Twill-Woven Composites,” 40th AIAA/ASME/ASCE/AHS/ASC
Structures, Structural Dynamics, and Materials Conference and Exhibit, St. Louis, Missouri, 12-
15 April 1999, AIAA-99-1438.

Kobayashi, H., Nakama, N., Maekawa, Z., Hamada, H., Fujita, A., and Uozumi, T., 1992,
“Fabrication and Mechanical Properties of Braided Composite Truss Joint,” International
SAMPE Symposium and Exhibition, Vol. 37, pp. 1089-1103.

10-3
Kumar S. and Wang, Y., 1997, “Fibers, Fabrics, and Fillers,” Chapter 2, Composites
Engineering Handbook, Marcel Dekker, Inc., New York, pp. 83-87.

Lakes, R.S., 1998, Viscoelastic Solids, CRC Press, First Edition 1998, pp. 1-2.

Lee, C.W., Rice, B.P., Buczek, M., and Mason, D., 1997, “Resin Transfer Process Monitoring
and Control,” SAMPE Journal 34, 1998 SAMPE, pp. 48-55.

Lee, L.J., Fu, K.E., and Yang, J.N., 1996, “Prediction of Fatigue Damage and Life for Composite
Laminates Under Service Loading Spectra,” Composites Science Technology, V 56, pp. 635-648.

Loos, A.C., Sayre J., McGrane, R., and Grimsley B., 2001, “VARTM Process Model
Development,” Proceedings of 46th International SAMPE Symposium, pp. 1049-1060.

Maurice, F.A., 2000, Professor of Engineering Science and Mechanics, The Pennsylvania State
University, Course material for “Engineering Composite Materials EMCH471,”
http://www4.esm.psu.edu/academics/courses/emch471/Notes/Chapter18.

Minguet, P.J.A., 1995, “Comparison of Graphite/Epoxy Tape Laminates and 2-D Braided
Composites Mechanical Properties,” AIAA/ASME/ASCE/AHS Structures, Structural Dynamics &
Materials Conference - Collection of Technical Papers, New Orleans, Louisiana, USA, Vol. 1,
pp. 17-26.

Munjal, A.K. and Maloney, P.F., 1990, “Braiding for Improving Performance and Reducing
Manufacturing Costs of Composite Structures for Aerospace Applications,” National SAMPE
Technical Conference, Boston, Massachusetts, USA, Vol. 22, pp. 1231-1242.

Munjal, A.K., Spencer, D.F., Rahnenfuehrer, E.W., Pickett, B.E., and Maloney, P.F., 1990,
“Design and Fabrication of High Quality Graphite/Epoxy Braided Composite Tubes for Space,”
National SAMPE Symposium and Exhibition (Proceedings), Vol. 35, pp. 2.

Naik, N.K., 1994, Woven Fabric Composite, Technomic Publishing Co., Inc.

Naik, R.A. and Masters, J.E., 1995, “Analysis of the Effects of Fiber Architecture Parameters on
the Strength Properties of 2D Triaxial Braided Composites,” American Society of Mechanical
Engineers, Aerospace Division (Publication) AD, San Francisco, California, USA, Vol. 50, pp.
145-173.

Nguyen, L.B., Juska, T., and Mayes, J.S., 1997, “Evaluation of Low Cost Manufacturing
Technologies for Large Scale Composite Ship Structures,” Collection of Technical Papers -
AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics & Materials Conference, Vol. 2,
pp. 992-1001.

O’Brien, R.K. and Reifsnider, K.L., 1981, “Fatigue Damage Evaluation Through Stiffness
Measurements in Boron-Epoxy Laminates,” Journal of Composite Materials, Vol. 15, pp. 55-70.

10-4
Oden, J.T., K. Vemaganti, and N. Moës, 1999, “Hierarchical Modeling of Heterogeneous
Solids,” Computer Methods in Applied Mechanics and Engineering, Vol. 172, pp. 3-25.

Paumelle, P., Hassim, A., and Léné, F., 1990, “Composites With Woven Reinforcements:
Calculation and Parametric Analysis of the Properties of Homogeneous Squivalent,” La
Recherche Aérospatiale, Vol. 1, pp. 1-12.

Pike, T., McArthur, M., and Schade, D., 1997, “Vacuum Assisted Resin Transfer Molding of A
Layered Structural Laminate for Application on Ground Combat Vehicles,” International
SAMPE Technical Conference, Seattle, Washington, USA, Vol. 28, pp. 374-380.

Portanova, M.A. and Deaton, J.W., 1995, “Impact and Fatigue Resistance of a [±30°/0°] 3-D
Braided Carbon Epoxy Composite,” ASTM Special Technical Publication, 1230.

Reddy, J.N. and Miravete, A., 1995, Practical Analysis of Composite Laminates, ISBN #0-8493-
9401-5, CRC Press Inc, Florida.

Reifsnider, K.L., 1990, “Damage and Damage Mechanics of Composite Materials,” Elsevier,
New York, pp. 11-77.

Rigas, E.J., Walsh, S.M., and Spurgeon, W.A., 2001, “Development of a Novel Processing
Technique for Vacuum Assisted Resin Transfer Molding (VARTM),” International SAMPE
Symposium and Exhibition 2001 a Materials and Processes, Long Beach, California, pp. 1086-
1094.

Rudd, R.E. and Broughton, J.Q., 2000, “Concurrent Coupling of Length Scales in Solid State
Systems,” Physica Status Solidi (b), Vol. 217, pp. 251-291.

Seemann, W.H., 1994, U.S. Patent 5, 316,462.

Seemann, W.H., 1990, U.S. Patent 4, 902,215.

Seobroto, H.B. and Ko, F.K., 1989, “Composite Perform Fabrication by 2-D Braiding,” 5th
Anuual ASM/ESD Advanced Composite Conference, Dearborn, Michigan, pp. 307-316.

Shivkumar, K.N., Sundareshan, M.J., and Avva, V.S., 1999, “Structural Integrity of
Discontinuous Stiffened Integrally Braided and Woven Composite Panels,” FAA report
DOT/FAA/AR-99/24, March 1999.

Smith, S., 2001, “Vacuum Assisted Resin Transfer Molding of Sandwich Structures: Material
Processing, Evaluation, Fracture Testing and Analysis,” Ph.D. thesis, North Carolina A&T State
University.

Srirengan, K. and Whitcomb, J.D., 1998, “Finite Element Based Degradation Model for
Composites With Transverse Matrix Cracks,” Journal of Thermoplastic Composites, Vol. 11(3),
pp. 113-123.

10-5
Strong A.B., 1989, “Fundamentals of Composite Manufacturing: Materials, Methods, and
Applications,” Society of Manufacturing Engineers, pp. 134-136.

Summerscales, J., 1998, Microstructural Characterisation of Fibre-Reinforced Composites,

ISBN # 0-8493-3882-4, CRC Press, Woodhead Publishing, Ltd., Cambridge, England.

Sun, C.T. and Li, S.J., 1988, “3-Dimensional Effective Elastic-Constants for Thick Laminates,”
Journal of Composite Materials, Vol. 22(7), pp. 629-639.

Swain, D., 2000, “Advanced Manufacturing Techniques Utilizing Carbon Fiber Braid and
RTM,” International SAMPE Symposium and Exhibition, Raytheon Aircraft Company, Inc.,
Long Beach, California.

Swain, D. and Abbott, R., 2000, “Advanced Manufacturing Techniques Utilizing Carbon Fiber
Braid and RTM,” International SAMPE Symposium and Exhibition (Proceedings), Long Beach,
California, USA, Vol. 45(1), pp. 277-289.

Swanson, S.R. and Smith, L.V., 1996, “Comparison of the Biaxial Strength Properties of Braided
and Laminated Carbon Fiber Composites,” Composites Part B: Engineering, Vol. 27(1),
pp. 71-77.

Tang, X. and Whitcomb, J.D., 2003a, “General Techniques for Exploiting Periodicity and
Symmetries in Micromechanics Analysis of Textile Composites,” Journal of Composite
Materials, Vol. 37, No. 13, pp. 1167-1189.

Tang, X., Whitcomb, J.D., Goyal, D., and Kelkar, A.D., 2003b, “Effect of Braid Angle and
Waviness Ratio on Effective Moduli of 2x2 Biaxial Braided Composites,” 44th
AIAA/ASME/ASCE/AHS Structures, Structural Dynamics, and Materials Conference, April 7-10,
2003, Norfolk, Virginia, AIAA-2003-1877.

Tang, X., Whitcomb, J.D., Kelkar, A.D., and Tate, J.S., 2003c, “Progressive Failure Analysis of
2x2 Braided Composites Exhibiting Multiscale Heterogeneity,” 18th ASC Technical Conference,
Gainesville, Florida, USA.

Tang, X., 2001, “Micromechanics of 2D Woven Composites,” Ph.D. Dissertation, Department of

Aerospace Engineering, Texas A&M University, College Station, Texas.

Tate, J.S., Kelkar, A.D., and Rice, J., 2003, “Feasibility Study of VARTM Manufacturing of
Carbon Biaxial Braided Composites Using EPON 9504 Epoxy Resin System,” JISSE-8, 8th
International SAMPE Symposium & Exhibition, Tokyo.

TPI Composites, Inc., http://www.tpicomp.com/technology.

Whitcomb, J.D. and Kelkar, A.D., 2002, “Modeling and Performance Evaluation of Braided
Composites,” FAA annual report of Grant No. DTFA03-01-C00033.

10-6
Whitcomb, J.D., Chapman, C.D., and Tang, X., 2000, “Derivation of Boundary Conditions for
Micromechanics Analyses of Plain and Satin Weave Composites,” Journal of Composite
Materials, Vol. 34(9), pp. 724-747.

Whitcomb, J. and Noh, J., 2000, “Concise Derivation of Formulas for 3D Sublaminate
Homogenization,” Journal of Composite Materials, Vol. 34(6), pp. 522-535.

Whitcomb, J.D. and Tang, X., 1999, “Effect of Tow Architecture on Stresses in Woven
Composites,” 40th AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics and Materials
Conference.

Whitworth, H.A., 1998, “A Stiffness Degradation Model for Composite Laminates Under
Fatigue Loading,” Composite Structure, Vol. 40(2), pp. 95-101.

Whitworth, H.A., 1990, “Cumulative Damage in Composites,” Journal of Engineering Materials

and Techology, Vol. 112, pp. 358-361.

Whitworth, H.A., 1987, “Modeling Stiffness Reduction of Graphite Epoxy Composite

Laminates,” Journal of Composite Materials, Vol. 21, pp. 362-372.

Yang, J.N., Lee, L.J., and Sheu, D.Y., 1992, “Modulus Reduction and Fatigue Damage of Matrix
Dominated Composite Laminates,” Composite Structure, Vol. 21, pp. 91-100.

Yang, J.N., Jones, D.L., Yang, S.H., and Meskini, A., 1990, “A Stiffness Degradation Model for
Graphite/Epoxy Laminates,” Journal Composite Materials, Vol. 24, pp. 753-769.

10-7/10-8
 APPENDIX A—BRAID ANGLE CHANGE (IN THE CASE OF BRAIDED TUBES) DOES
NOT CHANGE FIBER VOLUME FRACTION

The braid angle of a braided tube can be varied by pulling or compressing the braided tube
axially. There is a misconception that a change in the braid angle changes the fiber volume
fraction. This is not true because as the braid angle changes, it changes thickness as well. This
can be proven by using the following relation for fiber volume fraction based on the areal weight
method as discussed in section 2.4.2.

(n ∗W ∗ A) / ρ f ⎛ n ⎞ ⎛W ⎞
Vf = =⎜ ⎟⎟ * ⎜ ⎟
A∗t ⎜ ρf
⎝ ⎠ ⎝ t ⎠

Usually, braided tubes are specified at braid angles of 45º. When braided tubes are used in
configurations other than 45º, the areal weight (W) and ply thickness (t) change as a function of
the braid angle. This is because braid tube diameter reduces and braid tows come closer to each
other when a braid tube is pulled axially. This increases areal weight and ply thickness. Exactly
the reverse happens when braid tube is compressed axially. The percentage variation in areal
weight and the thickness of a composite with respect to braid angle is predictable and is usually
provided by the braid manufacturer. The percentage change in the areal weight (W) and the ply
thickness (t) is typically the same. Table A-1 lists the percentage variation in the areal weight
and ply thickness with respect to the braid angle documented by A&P Technology, Inc., the
supplier of braided fabric in the present research (Head, 1998). This variation is represented in
⎛W ⎞
figure A-1. Thus, in the above equation, ratio ⎜ ⎟ remains constant for any braid angle. For a
⎝ t ⎠
particular lay-up of a braided composite, the number of layers (n) and the density of the fiber
material ( ρ f ) are also constant. The ratio ⎛⎜ n ⎞⎟ remains constant. Thus, while braid angle
⎜ ρ ⎟
⎝ f ⎠
variation of a braid tube causes a variation in thickness, it does not cause variation in the fiber
volume fraction.

Table A-1. Percentage Reduction in Areal Weight and Ply Thickness

Percent Variation in the Areal

Braid Angle Weight and Ply Thickness
45 0
40 1
35 6
30 15
25 30
20 55
15 100

(Courtesy of A&P Technology, Inc.)

A-1
 120

100

Percentage Variation
80

60

40

20

0
10 15 20 25 30 35 40 45 50
Braid Angle, degrees

Figure A-1. Percentage Reduction in Areal Weight and Ply Thickness

This may not be true in the case of slit sleeves. Slit sleeves are manufactured individually for a
certain braid angle and may have a different architecture for spacing between two tows. The
spacing between tows (commonly termed cover factor by manufacturers) may increase or
decrease fiber volume fraction by only a small amount. Fiber volume fraction is mainly
dependent on the manufacturing process. Vacuum-assisted resin transfer molding typically
results in a fiber volume fraction between 0.45 and 0.55.

BIBLIOGRAPHY.

Head, A., 1998, “New Braid Design Spreadsheet Calculates Fiber Architecture and Aerial
Weight Simplifying the Design of Braid Reinforced Composites,” International SAMPE
Symposium and Exhibition (Proceedings), Anaheim, California, USA, Vol. 43(2), pp. 1269-
1274.

A-2
 APPENDIX B—RECOMMENDED METHOD FOR FATIGUE TESTS

Resin transfer molding and vacuum-assisted resin transfer molding processes produce
composites with thickness variations. When thickness changes, fiber volume fraction changes as
well. Woven composites exhibit fiber misalignment, and braided composites exhibit braid angle
variation. Thus, there is a considerable amount of heterogeneity within a specimen and also
among different specimens. Most in-plane mechanical properties are dependent on fiber volume
fraction and fiber orientation.

The average mechanical properties of tensile test specimens are commonly assumed as the
properties of both an entire panel and a large number of panels. In load-controlled fatigue tests,
stress is applied as a percentage of the average ultimate tensile strength (UTS). This may not be
a correct approach because the UTS of different specimens may differ. Fatigue tests, with
current practices, produce a large scatter in fatigue data.

Researchers have reported that the thickness variation in glass/epoxy composite laminates was
induced by processing affects, static tensile strength, and tensile modulus. However, its
influence on fatigue life is negligible (Hahn and Kim, 1976). It is actually, much more
appropriate to relate static tensile strength to fiber volume fraction rather than to thickness.

Carlos (1994) reported that a braid angle variation of ±5º in three-dimensional triaxial braided
composites caused a very large scatter in fatigue data (e.g., at a fatigue load of 80% of the UTS,
failure cycles varied from 20 to 9000 cycles). Carlos emphasized the necessity of detailed study
to determine the effect of braid angle on fatigue life.

A different approach for load-controlled fatigue tests of woven and braided composites is
proposed based on this research. Measurement of the fiber volume fraction and fiber orientation,
or braid angle, on each specimen before tabbing is recommended. The fiber volume fraction can
be measured very easily by the density method. Fiber orientation, or braid angle, can be
measured on micrographs using special software such as Image-Pro. A minimum of 15
specimens should be tested in tension tests.

UTS as a function of fiber volume fraction and fiber orientation, or braid angle, should now be
evaluated using statistical analysis. The equation may be in the form

UTS = β 0 + ( β1 *θ ) + ( β 2 *V f ) + ( β3 *θ *V f )

where

θ = Fiber misplacement angle in woven composites or braid angle in braided composites

Vf = Fiber volume percentage
β0, β1, β2, and β3 = Parameters in statistical analysis

Once this relation is established, it should be used in load controlled fatigue tests. Fiber volume
fraction and fiber orientation of each fatigue test specimen should be measured before tabbing.
Using the above relation, the UTS of each individual specimen should be computed. Stress

B-1
should be applied as a percentage of the computed UTS. Figure B-1 displays the flow chart of
the recommended procedure.

Cut Panels in Test Coupons

Accept the Specimens if;

Thickness variation ≤ ±4% (ASTM)

Width variation ≤ ±1% (ASTM)

Measure fiber volume fraction and fiber orientation, or

braid angle on each specimen

Conduct static tension tests and tabulate test results,

especially UTS

Using Statistical analysis establish relation for UTS,

UTS = β 0 + ( β 1 * θ ) + ( β 2 * V f ) + ( β 3 *θ * V f )

Measure fiber volume fraction and fiber orientation, or

braid angle on each fatigue test specimen

Compute the UTS from the above established equation

Apply stress as percentage of this computed UTS in load controlled

fatigue tests

Figure B-1. Recommended Flow Chart for Load Controlled Fatigue Tests of Woven and
Braided Composites

Table B-1 presents the tensile test results of 15 specimens of carbon/epoxy braided composites.
The fiber volume fraction was not measured for each specimen before testing, but it was
measured by the density method on a small rectangular piece cut from a failed specimen. The
braid angle was measured on micrographs using Image-Pro software. A linear regression
analysis was performed using SAS software to establish the relation for UTS as a function of
braid angle and fiber volume fraction. The statistical parameters β0, β1, β2, and β3 were
evaluated as 7.62, 5.24989, 3.331879, and -0.1707, respectively.

B-2
 Table B-1. Effect of Braid Angle and Fiber Volume Fraction on UTS for Carbon/Epoxy
Braided Composites

Specimen Fiber Volume

Number Braid Angle Fraction UTS
1* 0.0 0.00 7.62
2 26.2 0.49 85.80
3 25.1 0.46 94.78
4 25.0 0.51 96.55
5 24.9 0.47 98.78
6 23.8 0.54 102.80
7 30.8 0.51 60.23
8 30.1 0.52 62.48
9 30.0 0.54 64.74
10 29.6 0.56 66.03
11 29.4 0.54 66.97
12 45.9 0.49 24.03
13 45.6 0.51 24.32
14 45.3 0.50 24.62
15 45.5 0.51 25.31
16 44.0 0.53 30.90

*This row refers to the average UTS of neat epoxy resin coupons.

BIBLIOGRAPHY.

Carlos, K.G., 1994, “Fatigue Behavior of 3-D Triaxially Braided Composites,” Thesis (M.S.),
North Carolina A&T State University.

Hahn, H.T. and Kim, R.Y., 1976, “Fatigue Behavior of Composite Laminate,” Journal of
Composite Materials, Vol. 10, pp.176-180.

B-3/B-4
 APPENDIX C—BRAIDED TUBES AND SLIT SLEEVES DATA SHEET

The properties of braided tubes and slit sleeves are shown in the table below.

Slit Sleeve Braid Tubes

A B C T56L200X
Width 2.356 1.666 1.395
Diameter 0.750 0.530 0.444 2.000
Angle 45.0 30.00 25 45.0
Raw yd/lb 620 620 620 620
Carrier 36 36 36 96
Ends per carrier 1 1 1 1
Fiber density lb/cu in. 0.064 0.064 0.064 0.064
Part Fiber Volume 50.0% 50.0% 50.0% 50%
Outputs
One layer thickness 0.0303 0.0349 0.0398 0.0303
Percent coverage 75.7% 82.8% 75.7%
Qz/sq yd 20.1 23.2 26.4 20.1
ft/lb 36.53 44.74 46.92 13.70
ppi: picks per inch 7.6 6.2 5.9 7.6

(Courtesy of A&P Technology, Inc.)

C-1/C-2