INTERLAMINAR FRACTURE TOUGHNESS OF A QUASI-3D

COMPOSITE

By

Anthony Wente

A THESIS

Submitted to

Michigan State University

in partial fulfillment of the requirements

for the degree of

Mechanical Engineering — Master of Science

2019

ABSTRACT

INTERLAMINAR FRACTURE TOUGHNESS OF A QUASI-3D

COMPOSITE

By

Anthony Wente

Fiber reinforced polymer (FRP) composites are a heavily sought-after material for light-

weighting components in next generation vehicles due to their high specific strength and

stiffness. Traditional laminates have relatively weak interlaminar strength and are prone

to delamination, and this is especially the case when a delamination crack already exists.

Quasi-3D (Q3D) braided composites seek to solve this issue by weaving the bias tows around

the axial tows of the adjacent (above and below) plies.

The [0◦/60◦/-60◦] UD and Q3D carbon composites are investigated in this study for their

relative in-plane isotropy. Mode I and mode II interlaminar fracture toughness tests were

conducted on UD and Q3D samples.

In mode I experiments, the samples were continuously loaded to full beam split using

the double cantilever beam (DCB) method to obtain the fracture toughness throughout the

sample. The Q3D composite shows a large increase in fracture toughness once the crack

stabilizes and the interlaminar tows become engaged.

In mode II testing under end-notch flexural (ENF) test conditions, the Q3D composite

shows a significant increase in fracture toughness after a pre-crack has formed and the in-

terlaminar tows have been engaged. The 4ENF test also proves to be a good method for

creating a mode II resistance curve that could only be achieved through extensive testing

through traditional methods.

This thesis is dedicated to my parents for all of their love and support.

iii

ACKNOWLEDGMENTS

I would like to acknowledge my advisor, Dr. Sharon Xiao, for her guidance and support

through graduate school.

I am also thankful for all of those who taught and helped me

with experimental methods. These include, in no particular order, Wu Zhou, Erik Stitt,

Xinyu Mao, Shutian Yan, Italo Benedetti, and Royal Ihuaenyi. Without their help, this

thesis would not have been possible. I would also like to express my appreciation for Dr.

Danielle Zeng and Homa Torab at the Ford Motor Company for their guidance and support

throughout this project.

iv

TABLE OF CONTENTS

LIST OF TABLES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii

LIST OF FIGURES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viii

KEY TO SYMBOLS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

xi

Chapter 1

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1

2.1 Preforming

Chapter 2 Manufacturing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.1.1 Braiding Machine . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.1.2 Hand Braiding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2
Infusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.3 Volume Fraction Determination . . . . . . . . . . . . . . . . . . . . . . . . .

3.3 Mode II

Chapter 3 Fracture Toughness . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.2 Mode I . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.2.1 Experiment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.2.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.3.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3ENF . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.3.2
3.3.2.1 Experiment . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.3.2.2 Results
. . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4ENF . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.3.3.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.3.3.2 Experiment . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.3.3.3 Results

3.3.3

Chapter 4

In-Plane Characteristics . . . . . . . . . . . . . . . . . . . . . . . .
4.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.2 Tension . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.2.1 Experiment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.2.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.3.1 Experiment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.3.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

4.3 Flexural

v

7
7
8
9
11
13

16
16
17
17
20
25
25
26
26
26
29
29
32
38

42
42
42
42
45
47
47
48

Chapter 5 Hat Section . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.1 Manufacture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.2 Experiment
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Chapter 6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6.1 Summary and Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6.2 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

52
53
58

59
59
60

APPENDIX . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

62

BIBLIOGRAPHY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

67

vi

LIST OF TABLES

Table 2.1: Volume Fractions [%]

. . . . . . . . . . . . . . . . . . . . . . . . . . . . .

14

Table 3.1: Mode I Sample Dimensions [mm] . . . . . . . . . . . . . . . . . . . . . . .

20

Table 3.2: Mode I Averages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

21

Table 3.3: Q3D Mode I Architecture Advantages [%] . . . . . . . . . . . . . . . . . .

21

Table 3.4: Mode II Sample Dimensions [mm]

. . . . . . . . . . . . . . . . . . . . . .

27

Table 3.5: Mode II Averages

. . . . . . . . . . . . . . . . . . . . . . . . . . . .

29

(cid:21)

(cid:20)

J
m2

(cid:21)

(cid:20)

J
m2

Table 3.6: Q3D Mode II Architecture Advantages [%]

. . . . . . . . . . . . . . . . .

29

Table 3.7: 4ENF Sample Dimensions [mm]

. . . . . . . . . . . . . . . . . . . . . . .

35

Table 3.8: 4ENF Parameter Dimensions [mm] . . . . . . . . . . . . . . . . . . . . . .

37

Table 3.9: 4ENF Mode II Averages

. . . . . . . . . . . . . . . . . . . . . . . .

38

Table 4.1: Tensile Sample Dimensions [mm] . . . . . . . . . . . . . . . . . . . . . . .

45

Table 4.2: Q3D Tensile Properties

. . . . . . . . . . . . . . . . . . . . . . . . . . . .

46

Table 4.3: Flex Test Dimensions [mm]

. . . . . . . . . . . . . . . . . . . . . . . . . .

48

Table 4.4: Q3D Flexural Properties

. . . . . . . . . . . . . . . . . . . . . . . . . . .

51

vii

LIST OF FIGURES

Figure 1.1: Q3D unit cell with the bias tow that is braided between the layers high-
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

lighted [1]

Figure 2.1: Comparison of handwoven architecture [2]. . . . . . . . . . . . . . . . . .

Figure 2.2: Braiding machine used to create preforms used in this thesis . . . . . . .

Figure 2.3: Braided Q3D tube braided from machine . . . . . . . . . . . . . . . . . .

4

7

8

9

Figure 2.4: Flat Q3D preform crease by cutting tube length-wise . . . . . . . . . . .

10

Figure 2.5: Schematic showing layers of laminates for fracture toughness testing with
the tested architectures in the middle and the commercially available tri-
axially braided composite on the outer layers.
. . . . . . . . . . . . . . .

Figure 2.6:

Images showing how the pre-crack is generated in the Q3D composite.
The orange, polyimide film can be seen in the cured laminate.
. . . . . .

11

11

Figure 2.7: Schematic of infusion showing the different layers of consumables

. . . .

12

Figure 2.8:

Infusion process at three different points . . . . . . . . . . . . . . . . . .

13

Figure 2.9: Setup for acid digestion of the matrix to determine fiber volume fraction

14

Figure 2.10: Carbon fibers after acid digestion and oven drying . . . . . . . . . . . . .

14

Figure 3.1: The 3 different fracture modes . . . . . . . . . . . . . . . . . . . . . . . .

17

Figure 3.2: Mode I specimen at beginning of test showing the wires for loading and
markings on side on specimen . . . . . . . . . . . . . . . . . . . . . . . .

18

Figure 3.3: Test setup for mode I experiment with camera and light source

. . . . .

19

Figure 3.4: Critical dimensions for DCB test [3]

. . . . . . . . . . . . . . . . . . . .

19

Figure 3.5: Representation for nomenclature in mode I fracture toughness calculations
. . . . . . . . . . . . . . . . . . . . . . . . . . . .

from ASTMD5528 [3].

21

Figure 3.6: R-Curve comparison for all architectures. . . . . . . . . . . . . . . . . . .

22

Figure 3.7: Comparison of load-displacement curves for all architectures. . . . . . . .

23

viii

Figure 3.8: Fiber bridging crack gap of Q3D specimen. . . . . . . . . . . . . . . . . .

24

Figure 3.9: Crack interface of Q3D specimen after DCB test. The 26mm spacing of
. . . . . . . . . . . . . . . . .

the broken interlaminar tows can be seen.

25

Figure 3.10: ENF specimen during test . . . . . . . . . . . . . . . . . . . . . . . . . .

26

Figure 3.11: Specimen pre-crack conditions for NPC and PC . . . . . . . . . . . . . .

27

Figure 3.12: Representation of nomenclature for ENF test from ASTMD7905 [4]. . . .

28

Figure 3.13: ENF test load-displacement curve comparisons for three different archi-
. . . . . . . . . . . . . . . . . . . . . . . . .

tectures during the PC test

30

Figure 3.14: 3 point and 4 point bend moment diagrams

. . . . . . . . . . . . . . . .

31

Figure 3.15: UD specimen at high deflection during 4ENF test showing asymmetry .

32

Figure 3.16: Images showing relative sliding of specimen on fixture.

. . . . . . . . . .

33

Figure 3.17: 4ENF fixture with orange polyimide film to reduce friction and gray sand-
paper to increase it . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

35

Figure 3.18: Images showing no relative sliding of specimen on fixture.

. . . . . . . .

36

Figure 3.19: Representation of nomenclature for 4ENF test [5]. Not shown is d, which
. . . . . . . . . . . . . . . . .

is the spacing between the loading rollers.

36

Figure 3.20: 4ENF setup showing cameras of both sides of the specimen . . . . . . . .

37

Figure 3.21: Comparison of representative load-deflection curves from UD and Q3D
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

4ENF tests

39

Figure 3.22: Comparison of representative R-Curves from UD and Q3D specimens . .

40

Figure 3.23: Images showing relative sliding of specimen on fixture.

. . . . . . . . . .

41

Figure 4.1: Setup of tensile tests showing laser extensometer and specimen . . . . . .

43

Figure 4.2: Close-up of specimen showing the silver tape that is reflecting the laser .

44

Figure 4.3: Stress-strain curves from Q3D tensile tests . . . . . . . . . . . . . . . . .

46

Figure 4.4: Tensile failures of the Q3D1, Q3D2, and Q3D3 specimens, respectively .

47

ix

Figure 4.5: Diagram showing flexural test nomenclature . . . . . . . . . . . . . . . .

48

Figure 4.6:

Isometric view of 3 point bend setup . . . . . . . . . . . . . . . . . . . .

49

Figure 4.7: Front view of 3 point bend specimen mid-test . . . . . . . . . . . . . . .

49

Figure 4.8: Stress-strain curves from Q3D flexural tests

. . . . . . . . . . . . . . . .

50

Figure 4.9: First failure at the compression side of the Q3D4 specimen at the peak load 51

Figure 4.10: Failure at the tensile side of the Q3D4 specimen . . . . . . . . . . . . . .

51

Figure 5.1: General cross section of the hat section . . . . . . . . . . . . . . . . . . .

52

Figure 5.2: Tool used to manufacture hat section . . . . . . . . . . . . . . . . . . . .

53

Figure 5.3: Pleats in the vacuum bag of a hat section infusion . . . . . . . . . . . . .

55

Figure 5.4:

Infusion of a hat section . . . . . . . . . . . . . . . . . . . . . . . . . . .

56

Figure 5.5: The resin rich area on the left side of the part is a direct effect of preform
bridging during composite manufacture . . . . . . . . . . . . . . . . . . .

56

Figure 5.6:

Isometric view of a Q3D hat section . . . . . . . . . . . . . . . . . . . .

57

Figure 5.7: Comparison of Q3D and 2DW hat sections . . . . . . . . . . . . . . . . .

57

Figure 5.8: Q3D hat section with flat plate to be bonded for testing . . . . . . . . .

57

Figure A.1: UD mode I R-curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

63

Figure A.2: 2DW mode I R-curves . . . . . . . . . . . . . . . . . . . . . . . . . . . .

63

Figure A.3: Q3D mode I R-curves

. . . . . . . . . . . . . . . . . . . . . . . . . . . .

64

Figure A.4: UD mode II load-displacement plots

. . . . . . . . . . . . . . . . . . . .

64

Figure A.5: 2DW mode II load-displacement plots

. . . . . . . . . . . . . . . . . . .

65

Figure A.6: Q3D mode II load-displacement plots . . . . . . . . . . . . . . . . . . . .

65

Figure A.7: UD mode II 4ENF R-curves . . . . . . . . . . . . . . . . . . . . . . . . .

66

Figure A.8: Q3D mode II 4ENF R-curves

. . . . . . . . . . . . . . . . . . . . . . . .

66

x

KEY TO SYMBOLS

UD

Unidirectional

2DW

2 Dimensional Woven

Q3D

Quasi-3 Dimensional Braided

VARTM Vacuum Assisted Resin Transfer Molding

PTFE

Polytetrafluoroethylene

GI

GIC

GQ

GII

Mode I Fracture Toughness

Mode I Initial Fracture Toughness

Mode II Fracture Toughness Tentative Value

Mode II Fracture Toughness

ENF

End Notch Flexure

NPC

No Pre-Crack

PC

Pre-Crack

b

h

P

L

d

a

ao

δ

Width

Thickness (Mode I) Half Thickness (Mode II)

Load

Length of Specimen (Mode I) Half Support Span (Mode II)

Loading Span

Crack Length

Initial Pre-crack Length

Load Point Extension

4ENF

4 Point End Notch Flexure

ELS

End Load Split

xi

ONF

Over-notched Flexure

Lg

A

Extensometer Gage Length

Cross-sectional Area

xii

Chapter 1:

Introduction

The high demand for more fuel efficient vehicles has increased the necessity to use high-

specific strength and stiffness materials. This is especially true with the budding electric

vehicle divisions in the automotive industry because the range of the vehicles has been an

important issue. In this example, there are two ways to increase the range of the vehicle.

One would be to make a more efficient powertrain using higher energy density batteries

and more efficient motors. Another would be to make the vehicle’s chassis, bodywork, and

other components lighter. One method for light-weighting is to use the aforementioned

high-specific stiffness and strength materials. Carbon-fiber reinforced polymers (CFRPs)

have been proven in the aerospace, racing, and sporting industries as a prime example of

this type of material. The use of CFRPs currently range from tennis racquets to Formula 1

chassis to commercial jet fuselages. CFRPs have shown time and time again that they can

be used when high strength and stiffness coupled with a low mass is paramount.

However, one major issue with CFRPs is the failure mode of delamination under out-

of-plane impact loading. This is a major issue in the automotive industry due to high

crash-safety standards.

In a typical automotive crash, critical composite parts can expe-

rience impact loading in the through-thickness direction and cause delamination.

If the

delamination resistance of the composite can be improved, the composite can have better

specific energy absorption properties. A composite with a higher fracture toughness will also

have less crack propagation in an event with crack damage. With less damage, the com-

ponent will maintain a higher stiffness and strength compared to a comparable composite

with a lower fracture toughness. This is extremely important for the crash-safety critical

components of a vehicle like bumpers, roof headers, and frame rails where CFRPs can make

1

a big impact in mass savings.

There have been many proposed solutions for increasing the fracture toughness of lam-

inated composites, and two of the most common methods include stitching [6, 7, 8, 9] and

z-pinning [8, 9, 10]. In both of these methods, the composite preform is first compiled, and

the through-thickness reinforcement is later inserted into the preform. These reinforcements

are also typically inserted orthogonal to the fibers in the laminate: purely in the 3-direction

respective to the fibers. Stitched composites are manufactured by taking a preform and

using a stitching machine to run a needle with a high-strength yarn through the preform

[6, 7, 8, 9, 11], where z-pinning takes the preform and inserts small rods made of a metal or

composite into the preform [8, 9, 10]. Both of these methods have been shown to improve

the mode I and mode II fracture toughness with typical increases in GIC of around 75%

[6, 12] with the steady state crack growth resistance, GI, increasing by factor of 2.9-15.7

[12] in stitched composites. Similar increases are found with laminates that are improved

by z-pinning with the GI value typically increasing by a magnitude [13, 14, 15]. Z-pins have

also increased the mode II fracture toughness, GII, but not by as significant of an amount.

Cartie et.al. found a 2-5 factor increase depending on the density and diameter of the pins

being used [13].

While stitching and z-pinning may increase the fracture toughness of the composites,

it has been found that they also reduce the in-plane properties that make them desirable

initially.

In a review paper by Mouritz, Leong, and Hersberg, tensile properties can be

reduced anywhere from 30% to 45% and compressive properties can be reduced by 5% to

55% due to stitching [11]. The values for the degradation, along with whether or not the

properties are actually degraded, largely vary between the papers reviewed; however, the

majority of papers show degradation in the in-plane properties. This degradation has been

2

attributed to fiber damage/breakage and fiber angle spreading due to the insertion of the

needle in the stitching process [11]. Z-pinning also experiences fiber angle spread along

with resin rich areas around the pin [8, 10, 16]. For thick pins with a 2% pin density in

the laminate, Mouritz found that there is a range of 7%-17% decrease in Young’s Modulus

and typically a 11%-14% decrease in compressive strength depending on the layup sequence

[16]. It is also found that tensile strength is typically degraded by 5%-10% and the fatigue

performance is reduced [15].

It has been shown by by Kerber et. al. that the use of a z-binder reinforcement improves

the fracture toughness of a composite without a great reduction in the in-plane properties

[17]. These z-binder fibers are woven into the laminate along with the fibers in the warp

direction, and they are woven around each fiber in the fiber in the weft direction. This keeps

the reinforcement relatively flat; however, the z-binder reinforcement is still significantly in

the 3-direction, reducing the in-plane properties. The authors found that as the amount of

z-binders are increased, the flexural properties on the composite decrease. It was also shown

that for a 100% increase in fracture toughness, whether mode I or mode II, the blast-induced

delamination damage will be reduced by over 50% [17].

Quasi 3-dimensional (Q3D) composites are seeking to increase the fracture toughness

of a fiber-reinforced composite without degrading the desirable in-plane properties. Q3D

composites differ from other fracture toughness increase mechanisms by not using a through-

thickness reinforcement that is orthogonal to the fibers. Q3D differs from the previously

mentioned z-binder reinforcement because the through thickness reinforcement in not added

to an existing preform, and one reinforcement is not woven through all layers. Instead, the

fiber tows are braided in such a way that instead of each tow being constrained to a single

layer, it is also braided into the adjacent layers from which it originates.

In the case of

3

Figure 1.1: Q3D unit cell with the bias tow that is braided between the layers highlighted [1]

the triaxially braided composites being investigated in this study, the bias tows are braided

in into the layers adjacent to their originating layer, and the axial tows remain straight

within their respective layer. This keeps the through-thickness reinforcement used in a Q3D

preform flatter than that used in the z-binder composite. Furthermore, the through-thickness

reinforcement in a Q3D composite consists of the bias tows with no additional reinforcement

needed that displaces the fibers in the pre-existing preform where the z-binder introduces

new binder tows.

Figure 1.1 highlights a tow that bridges this interlaminar gap. To give a more specific

example, a bias tow in the third layer of a composite will not only be braided around the

axial tows of the third layer, but it will also be braided around the axial tows of the second

and fourth layers as well. These bias tows then bridge the gap between layers such that fiber

breakage is required for crack propagation instead of simply matrix failure. When a preform

of a Q3D structure is created, there are not individual layers as with traditional laminated

composites. Instead, there is a single preform where every layer is now braided together.

4

More information about Q3D composites can be found in US Patent US009273933 [18].

Because the bias tows are only braided into the adjacent layers and not through the

full laminate, they remain relatively flat. There is also no post-processing of the preform

required with needle or pin insertion, so there is no extra fiber damage, fiber spreading, or

resin-rich areas that form. This also reduces the handling of the carbon fiber because there

is one less step as opposed to other traditional delamination resistance mechanisms. Rosario

and Liu found that the Q3D architecture for biaxially woven composites have competitive in-

plane properties to its laminated counterpart while also having increased impact properties
[19]. Wu Zhou tested the in-plane tensile properties of the [0◦/60◦/-60◦] Q3D composite

in his PhD thesis and found that in the axial direction, the Q3D tensile strength exhibits

between 5-10% degradation compared to the unidirectional (UD) laminate, but the Q3D and

traditional woven (2DW) laminates have comparable tensile strengths. It is also found that

transverse to the axial direction, all three composites have nearly identical tensile strengths.

As for the tensile modulus, the UD, 2DW, and Q3D laminates were all within 5-10% of

each other [2]. Many authors have also investigated the effects of the Q3D composites with

respect to their energy absorption and damage resistance due to out-of-plane impact and

have found that the Q3D structures will absorb a similar amount of energy with a smaller

damage area [2, 19, 20] and have lower displacement, approximately the same force, and a

higher bending stiffness compared to UD and 2DW [20]. A study performed by Bayraktar,

et al showed that a Q3D composite will have a higher specific energy absorption in terms of

weight for energy absorbing structures in automotive vehicles when constant cross section

composite and steel beams are compared [21].

This thesis investigates Q3D composites and the effect that the bias tows bridging the

ply interface have on the mode I and mode II fracture toughness. Attention is especially paid

5

to the steady state growth of the crack as opposed to simply the initial values to determine

the relationship between the crack length and the fracture toughness in both the UD and

Q3D composites.

6

Chapter 2: Manufacturing

2.1 Preforming

In this study, preforms are created in two different fashions: using a braiding machine and

hand braiding. The braiding machine is used to create the preforms for the in-plane and hat

section testing, but hand braiding is required for fracture toughness testing. The fiber used

with the braiding machine is DowAksa 24k A-42 carbon fiber, and the fiber used during hand-

braiding is DowAksa 12k A-42 carbon fiber. Figure 2.1 compares the handwoven preforms

for each of the UD, 2DW, and Q3D preforms. It also shows a schematic for a comparison of

the number and relative thickness of the layers within the preform.

Figure 2.1: Comparison of handwoven architecture [2].

7

Figure 2.2: Braiding machine used to create preforms used in this thesis

2.1.1 Braiding Machine

The triaxially braided Q3D structure is made by using a large braiding machine with 480

bobbins. 120 of these bobbins are for the axial tows while the other 360 are for the bias tows.

Figure 2.2 shows the machine used to make the preforms. In this figure, the carriers and the

bobbins holding the bias tows can clearly be seen along with the beginning of the braided

tube. The axial tows can also be seen coming out of the long, slender rod in the center of

the base at which the carriers rotate. This type of preform is required to be braided into

a tube because it is triaxial in nature. The machine works by pulling up the braided tube

through four rollers. This is referred to as the uptake. As it is being pulled up, the axial

tows remain straight and taught while the bias tows are being maneuvered around them. By

changing the speed at which the carriers holding the bias tows are moving along with the

speed of the uptake, the braided angle can be changed.

Once a sufficient length of tube is braided, it can be cut from the machine to be infused

8

Figure 2.3: Braided Q3D tube braided from machine

or laid up.

It can be kept as a tube, but for this study, the tube is cut length-wise to

create a flat preform. Figure 2.3 and 2.4 show this Q3D preform as a tube and flat preform,

respectively. More information about the manufacture of a triaxially braided Q3D composite

can be found in the aforementioned patent [18].

2.1.2 Hand Braiding

When investigating the fracture toughness of a composite, a pre-crack is required in the end of

the specimen. The machine being used to braid the Q3D architecture does not have a method

for adding a pre-crack, thus the fracture toughness samples must be hand braided. Because

the hand braiding of the pre-forms is extremely labor intensive and time consuming, a single

9

Figure 2.4: Flat Q3D preform crease by cutting tube length-wise

[0◦/60◦/-60◦] layer of the investigated architecture is used symmetrically on each side of the

crack surface. If the laminate is too thin, failure can occur at the outsides of the laminate or

the arms in mode I testing or the outer skins will fail before crack propagation during mode

II. To counteract this, the thickness can be built up using a unidirectional or textile material

stacked on each side of the desired preform until the optimal thickness is achieved [17]. To
achieve this higher thickness, five layers of A&P Technologies triaxially braided [0◦/60◦/-
60◦] carbon fiber is used on either side of the studied layers. This thickness is required such

that delamination is the failure mechanism during the experiments as opposed to in-plane

fiber or matrix failure. It has been shown in previous works that a sufficiently thin PTFE
(Teflon R(cid:13)) or polyimide (Kapton R(cid:13)) film is sufficient for creating the necessary pre-crack in
fracture toughness testing [17, 22, 23, 24]. For this study, a 0.001 inch thin (Kapton R(cid:13)) film

is used to create the pre-crack between the two layers of the studied architecture. Because

the Q3D architecture has the through-thickness tows throughout the preform, one end was

braided such that there were not any of these tows for approximately 65mm such that the

film could be inserted. Figure 2.5 shows a schematic of the stacking sequence used to create

10

Figure 2.5: Schematic showing layers of laminates for fracture toughness testing with the tested architectures
in the middle and the commercially available triaxially braided composite on the outer layers.

Figure 2.6: Images showing how the pre-crack is generated in the Q3D composite. The orange, polyimide
film can be seen in the cured laminate.

the fracture toughness laminates. Figure 2.6 shows how one end of the preform does not have

the through-thickness tows, and the film is inserted to create the pre-crack during infusion.

2.2 Infusion

The preforms are infused with API SC-15 toughened epoxy through vacuum-assisted resin

transfer molding (VARTM). The in-plane and fracture toughness specimens are manufac-

tured on a flat, aluminum caul plate, or tool, that is sanded and prepped after every use.

The preform is placed straight onto the caul plate, and a layer of peel ply is placed on top of

the preform to separate the infusion media from the preform such that is does not stick. The

11

Figure 2.7: Schematic of infusion showing the different layers of consumables

infusion media is then added from the resin inlet to the vent with approximately a 20mm

break in it at the end of the preform. This is to force the resin front to slow such that it

diffuses fully through the preform. Another layer of peel ply is added on top of the infusion

media to dull the sharp edges and protect the bag from popping under vacuum. Small pieces

of breather can be added over the resin inlet and vent to ensure that the bag does not create

a seal at these two ports. The bag is then carefully stuck to the caul plate. Great care is

taken to ensure there are no wrinkles in the bag at the sealant tape and no leaks form once

vacuum is applied. Figure 2.7 shows a schematic of the different layers within the bag during

an infusion. Once fully assembled, the resin inlet is clamped and vacuum is pulled. Vacuum

is then pulled at 28inHg, measured in the catch pot. Once this full pressure is reached, the

pump is turned off for 15 minutes, and the pressure is monitored. If there is no discernible

pressure drop in the analog gauge after those 15 minutes, the bag is deemed to be leak free.

The vacuum will then be turned back on for 15 minutes to boil any residual moisture in

the bag and preform and suck it out. If this is not done, this moisture could be trapped

within the resin, causing voids. During this time, the epoxy resin is mixed and degassed

for 10-15 minutes in a vacuum chamber. Once the resin is degassed, the tube at the resin

inlet is placed in the resin pot, and the clamp is released. The infusion process begins, and

this will take typically 30-60 minutes depending on the architecture and size of the preform.

Three steps of the infusion process are shown in figure 2.8. Once the preform has been fully

infused, it is placed in a convection oven to cure. There is a joint in the tubing between the

12

VentResin InletCaul PlateSealant TapeVacuum BagFlow MediaPeel PlyPreformFigure 2.8: Infusion process at three different points

catch pot and caul plate that allows the tool to be placed into the oven and still be under
vacuum during the full cure cycle. The laminate is then vitrified at 60◦C for two hours, and
then it is post-cured at 94◦C for four hours.

2.3 Volume Fraction Determination

The volume fractions are obtained initially by weighing the dry fiber preform before resin

infusion then dividing that mass by the mass measured after the laminate is cured to obtain

a weight fraction. The densities of the resin and the fibers are then used to get a volume

fraction. The weight of the polyamide film is taken to account in this measurement, and

the values are confirmed by sulfuric acid digestion following the ASTM standard D3171-15

[25]. Figure 2.9 shows the setup for this procedure for the acid digestion. Small pieces of

the laminate are weighed to the nearest milligram, and then placed in the heated acid inside

of a fume hood to digest the matrix. Once the matrix has been digested, the mixture is

poured into the funnel to get rid of the acid and matrix mixture. The fibers are sprayed

with distilled water and then acetone to remove all of the acid. Once all of the acid is

13

Figure 2.9: Setup for acid digestion of the matrix to determine fiber volume fraction

Figure 2.10: Carbon fibers after acid digestion and oven drying

removed, the fibers are oven dried and weighed again. Figure 2.10 shows the carbon fibers

after oven drying. The weight of the dry fibers divided by the weight of the original laminate

will give the weight fraction of the fibers. Comparing the weight fraction from acid digestion

to the weight fraction from weighing the preform before infusion and the laminate after

infusion yields the same results within certainty. Table 2.1 shows the volume fractions for

Table 2.1: Volume Fractions [%]

Tests

Fracture Toughness

Hat Section

In-Plane

Vf

60.0±0.5
55.5±0.5
55.5±0.5

14

the laminates in this study. The reason for the fracture toughness samples having a higher

volume fraction is likely due to the tighter weave of the commercial fabric used to build up

the thickness.

15

Chapter 3: Fracture Toughness

3.1 Background

Fracture toughness is an important property of laminated composite materials due to their

relative ease to delaminate. Laminated composites are made of layers, or plies, of unidirec-

tional or woven, textile composites within a matrix. In this study a thermoset epoxy-resin

polymer is used as the matrix. The fibers have a modulus of 240 GPa and a tensile strength

of 4200 MPa [2]. In contrast, the SC-15 toughened epoxy has a modulus of 2.62 GPa and

tensile strength of 56.4 MPa [26]. The fibers have a stiffness and strength nearly two mag-

nitudes above that of the matrix. In a traditional laminate, the layers are only separated by

this relatively weak matrix. The methods for increasing the fracture toughness for a compos-

ite typically consist of either increasing the fracture toughness of the matrix or mechanical

interlocking from stitching, z-pinning, or Q3D braiding. The SC-15 epoxy-resin used in this

study is already rubber toughened, thus improvements seen by the Q3D structure will not

be as prominent as they would in a composite where the resin is not toughened.

There are three modes when looking at fracture toughness as shown in figure 3.1. The

two studied in this thesis are mode I: opening and mode II: in-plane shear. The mode II

in-plane shear is especially relevant for the out-of-plane impact seen in crash scenarios for

automotive vehicles.

In this work, only mode I and mode II fracture toughness is investigated because these

are the two commonly seen in the real world. Mode III is also difficult to test for, and there

is not a currently accepted standard for it as of the time this thesis is written. In mode I

fracture toughness, the toughing mechanism of the interlaminar fiber will have a large effect

16

Figure 3.1: The 3 different fracture modes

due to the requirement for the tow to break. In mode II, the interlaminar fiber tow will

require the crack to propagate around the tow, but not necessarily break it.

3.2 Mode I

3.2.1 Experiment

The mode I experiments follow the testing basis given by ASTM standard D5528 [3]. The

two methods for loading the specimen are piano hinges and loading blocks. Piano hinges

were first implored, but the large deformation required to fully split the sample made it such

that the beam came into contact with the grips. For this reason, loading blocks are used with

a steel wire through the block and into the grips. This allows for sufficient space between

the sample and the grips so there is no interaction. Similar to previous studies, the side of

the specimen is painted white, and markings are made every 5mm [22]. The first 20mm is

marked every 1mm. This is so that the initial crack propagation can be better visualized

because the initial fracture toughness, GIC, is defined by the first 1mm of crack growth. It

is also very difficult to determine where the manufactured crack tip is, and precise marking

17

Figure 3.2: Mode I specimen at beginning of test showing the wires for loading and markings on side on
specimen

near the estimated crack tip allows for accurate adjustment after the test. An example

of a specimen during the test is shown in figure 3.2. Using a camera to record the crack

length during the test has been shown to be an effective method for measuring the crack

propagation during the test [17]. Using the camera to record the experiment allows the test

to be run continuously without the need to unload after crack growth. This is useful for

saving the amount of time required to run a test, accuracy of measuring the location of the

crack tip without the possibility of damaging the specimen further, and it reduced the risk of

loading history causing artifacts within the data. The camera used records the experiment

by acquiring images at a rate of 1 image

second to match the sampling rate of the machine at 1 Hz.

The setup of the experiment with the camera and light source is shown in figure 3.3. The

machine used for this and all fracture toughness tests is the MTS Insight Electromechanical

test frame. In these experiments, the crosshead speed is set at 5 mm

min. The test is run until

the crack propagates completely through the specimen, and the specimen is split into two.

18

Figure 3.3: Test setup for mode I experiment with camera and light source

Figure 3.4: Critical dimensions for DCB test [3]

19

Table 3.1: Mode I Sample Dimensions [mm]

Specimen

UD1

UD2

2DW1

2DW2

Q3D1

Q3D2

b

25.4

25.1

25.9

25.8

25.5

25.5

h

3.528

3.606

3.554

3.564

3.620

3.480

ao

56.1

54.1

56.3

55.0

61.8

59.9

Table 3.1 shows the dimensions of each mode I sample with the corresponding dimensions

showing in figure 3.4.

3.2.2 Results

For mode I, the fracture toughness is calculated as a function of load, extension, crack length,

and specimen geometry. In this study, the modified beam theory (MBT) is used to calculate

the GI value. The initial fracture toughness value, GIC , is calculated at the point when the

crack has propagated by 1mm, and this is determined by using the images acquired during

the test. Equation 3.1 is used for determining the GI value.

GI = F

3P δ

2B (|∆| + a)

(3.1)

In this equation, F is an inclusion factor that takes into account how the specimen geometry

is changing during the test. Equation 3.2 shows how the inclusion factor is calculated,

and figure 3.5 shows a representation of the variables used in calculating the GI value. The

rotation correction factor, ∆, is determined by performing a linear regression on a plot of the

cube root of compliance, C1/3, as a function of the delamination length, a. The compliance,

20

(a) Mode I fracture toughness

(b) Inclusion factor, F

Figure 3.5: Representation for nomenclature in mode I fracture toughness calculations from ASTMD5528
[3].

(cid:104) J
(cid:104) J

m2

(cid:105)
(cid:105)

Table 3.2: Mode I Averages

UD

563.6

732.8

3.345

2DW

564.1

730.4

3.756

Q3D

466.3

884.7

4.994

GIC

GI

m2
Energy [J]

Table 3.3: Q3D Mode I Architecture Advantages [%]

GIC
GI
Energy

2DW

-17.3

21.1

32.9

UD

-17.3

20.7

49.3

21

Figure 3.6: R-Curve comparison for all architectures.

C, is defined as δ

P . ∆ is then found to be the x-intercept of the linear regression.

(cid:32) δ

a

F = 1 − 3
10

(cid:33)2 − 3

(cid:32)δh

(cid:33)

2

a2

(3.2)

Tables 3.2 and 3.3 show the average mode I values for each architecture and the percentage

difference between them. It can be seen that the Q3D structure has a lower GIC value, but

a larger GI value. This is due to difficulties in creating the pre-crack in the Q3D structure

because of the through-thickness tows. The leading edge of the film can wrinkle, causing a

stress concentration at the crack tip and a lower GIC value. It can be seen by the resistance

curves shown in figure 3.6 that the GI value increases quickly beyond this initial value. More

R-curves for each of the fiber architectures are shown in the appendix.

Along with the fracture toughness values, the energy required to fully split the sample

22

Figure 3.7: Comparison of load-displacement curves for all architectures.

is reported. Each sample has the same length from the crack tip to the free end, so these

energy absorption values can be properly compared. This value gives a better comparison of

the delamination resistance of each of the material architectures because the amount of crack

propagation and the lower initial values have less of an effect. Looking at a comparison of

typical load-displacement curves for each of the architectures in figure 3.7, it is seen that it

takes more extension for the 2DW to fully split compared to the UD sample. The 2DW will

also hold a higher load throughout the experiment, so it would have been expected to have

a higher GI value, but this is not the case. This is due to the 2DW having lower fracture

toughness values compared to UD initially, but they increase as the crack propagates. The

crack also propagates in a stick-slip fashion for the 2DW causing it to have fewer GI values

to report, thus lowering its average. The same can be said about the Q3D structure with

the larger amount of low GI values initially, but fewer values as the crack propagates and the

23

Figure 3.8: Fiber bridging crack gap of Q3D specimen.

fracture toughness increases. The energy absorption takes this into account and allows for

a better comparison to be made. The Q3D also has a larger increase in energy absorption

compared to both the UD and 2DW than it does in fracture toughness.

The Q3D has a more pronounced stick-slip condition compared to the 2DW, and this

is due to the interlaminar tows in the Q3D composite bridging the crack interface. The

introduction of these tow requires fiber breakage for the crack to propagate, where the UD

and 2DW simply require matrix failure. These tows support more load and require a larger

crosshead extension before failure. Figure 3.8 shows the interlaminar tow clearly bridging

the gap in the middle of a mode I experiment. After failure of this interlaminar tow, the

crack will propagate until it reaches the next tow, and crack propagation will cease. At this

point, the load will begin to rise again until the next tow breaks. The interlaminar tows are

24

Figure 3.9: Crack interface of Q3D specimen after DCB test. The 26mm spacing of the broken interlaminar
tows can be seen.

spaced out by 26mm as shown in figure 3.9. Looking at the resistance curve for the Q3D

architecture, it is seen that fracture toughness values are given for crack lengths spacing

approximately 26mm as well. The points for the resistance curve correspond to the points

immediately before crack propagation that correspond to the max load.

3.3 Mode II

3.3.1 Background

The current standard for measure the mode II fracture toughness of a material is to perform a

3-point end notch flexural (ENF) test. This method gives consistent results, but it only gives

a fracture toughness value for one crack length. Furthermore, two tests must be conducted

on the same specimen such that a fracture toughness from the blunt crack tip manufactured

by the film and the sharp crack tip generated by shear induced crack propagation can be

compared. If an R-Curve is desired, multiple tests must be conducted with different initial

crack lengths. This can be tedious and also allows opportunity for error and loading history

effects.

If an R-curve is desired, another test is required and is investigated later in this

study.

25

Figure 3.10: ENF specimen during test

3.3.2 3ENF

3.3.2.1 Experiment

Experiments for mode II fracture toughness follow the 3ENF test outlined in ASTM standard

D7905 [4]. As with the mode I samples, the sides are painted white such that the crack and

the compliance calibration markings can be seen more easily. This is shown in figure 3.10.

Crosshead speed during the test is 0.5mm/min.

Compliance calibration is done at 3 initial crack lengths of 20mm, 30mm, and 40mm.

For the 20mm and 40mm crack lengths, the test is stopped at half of Pcrit, where Pcrit is

the load required for crack propagation. For the 30mm crack length test, the test is run

until crack propagation and a significant load drop is achieved. The Pcrit value is determined

experimentally for each crack length. Table 3.4 shows the dimensions of each of the mode II

samples.

3.3.2.2 Results

The mode II tests have two different tests conditions: no pre-crack (NPC) and pre-crack

(PC). For the NPC condition, the initial delamination for the NPC test is simply the manu-

factured pre-crack, where the initial delamination for the PC test includes crack propagation

26

Table 3.4: Mode II Sample Dimensions [mm]

Specimen

UD1

UD2

2DW1

2DW2

2DW3

Q3D1

Q3D2

b

25.2

25.3

25.0

24.8

25.4

24.6

25.4

h

1.790

1.766

1.729

1.778

1.724

1.814

1.837

(a) NPC

(b) PC

Figure 3.11: Specimen pre-crack conditions for NPC and PC

beyond this point as shown in figure 3.11.

The ENF test used for mode II fracture toughness calculation does not give stable crack

growth, and thus, only one GII value can be calculated for a single test. Once crack prop-

agation begins, the delamination rapidly expands until it reaches the loading roller. The

fracture toughness is calculated as a function of the initial crack length, max load, specimen

width, and compliance calibration curve slope as shown in equation 3.3.

GQ =

3mP 2
M axa2
0
2b

(3.3)

The compliance calibration curve slope is obtained by the linear regression of the compliance

vs crack length cubed plot. The compliance for a particular crack length is taken from the

range of 90N to the

Pcrit
2 value that was determined. This is done such that any non-linearities

27

Figure 3.12: Representation of nomenclature for ENF test from ASTMD7905 [4].

from the fixture at the beginning of the test and any non-linearities from the laminate at

higher loads are not influencing the compliance.

For a pictorial representation of the variables used for calculating GQ, figure 3.12 can be

referenced. GQ can be considered to be the GII value for the sample only when equation 3.4

is satisfied. If it is not satisfied, then the test must be redone to obtain a GII value.

15 ≤ 100 (Piai)2

(PMaxa0)2 ≤ 35;

i = 1, 2

(3.4)

Pi and ai in this equation are the Pcrit
2

and crack length values used in each compliance

calibration test.

As with the mode I tests, Q3D has a higher fracture toughness after initial crack propa-

gation, but a lower value when the crack tip is manufactured from the film. The Q3D sees a

higher increase in GII versus that 2DW and UD compared to the increase in GI. The Q3D

architecture shows nearly a 47% increase in mode II fracture toughness compared to the UD

and an 18% over the 2DW architecture. The GII values and percent difference are shown in

tables 3.5 and 3.6, respectively.

28

Table 3.5: Mode II Averages

(cid:105)

(cid:104) J

m2

NPC

PC

UD

644.5

559.5

2DW

502.2

696.1

Q3D

613.8

824.3

Table 3.6: Q3D Mode II Architecture Advantages [%]

NPC

PC

UD

-4.8

47.3

2DW

22.2

18.4

Similar to the mode I DCB test, the load-displacement plots for the PC mode II tests in

figure 3.13 show that the Q3D architecture requires a larger load and extension before crack

propagation. The PC plots are shown because they give a better representation of the effect

that the Q3D architecture has on fracture toughness. The NPC test does not initially engage

the interlaminar tow that increases the fracture toughness for the Q3D composite, and as

with the mode I test, the issue with manufacturing the pre-crack for the Q3D structure also

gives unrepresentative results. The larger load and extension required for Q3D shows that

in mode II testing, the Q3D architecture will have a larger energy absorption than UD or

2DW. More plots for each architecture are shown in the appendix.

3.3.3 4ENF

3.3.3.1 Background

While the ENF test shows that the Q3D architecture has an advantage, it may not show

everything. As it is seen in the mode I tests, the Q3D laminate has the largest advantage

during steady-state crack growth. With the ENF test, steady-state crack growth is not

achievable because of the inherent unsteadiness of the crack.

There have been a few proposals tested to achieve the steady-state crack growth. Some

29

Figure 3.13: ENF test load-displacement curve comparisons for three different architectures during the PC
test

of these include the end-loaded split (ELS) test, stabilized ENF, over-notched flexure (ONF)

test, and 4ENF. Of these tests, the 4ENF test proved to be the most promising because it

produced stable crack growth while also being a relatively simple experiment [24, 27, 28].

Figure 3.14 shows why 3ENF is unstable while 4ENF is stable by using moment diagrams.

As the crack grows in the ENF test, it propagates into regions of higher moment, thus causing

the crack propagation to continue. However, with the 4ENF test, the crack remains in an

area of constant moment and maintains stable crack growth. As the crack grows, the load

will drop slightly, decreasing the moment. With the decrease in moment, the crack will cease

to propagate until the load increase back to where crack propagation occurred initially. In

theory, this will create a load-displacement diagram with a nearly horizontal line after peak

loading. Due to issues in manufacturing, friction, and non-ideal loading conditions, this is

30

(a) 3 point bend

(b) 4 point bend

Figure 3.14: 3 point and 4 point bend moment diagrams

31

Figure 3.15: UD specimen at high deflection during 4ENF test showing asymmetry

not necessarily the case.

3.3.3.2 Experiment

While the 4ENF test is the simplest of the proposed tests for steady-state, mode II crack-

growth, it still comes with issues to be overcome. Without additional measures, the specimen

will begin to slide [29]. The phenomenon is not documented in all cases, so it appears to

be relative to the specimen being tested. The specimen is likely to slide in the fixture at

relatively high deformation, so thinner specimens with a lower flexural modulus and higher

fracture toughness are more likely to slide. This is due to the asymmetry of which an ENF

specimen will deform when loaded, causing the end with the pre-crack to have a higher angle

of contact with the roller relative to the original position. At high deformations, this end

will become closer to parallel with the loading as opposed to perpendicular. Figures 3.15

and 3.16 shows this issue for the tests conducted.

Literature has shown different ways to combat this issue of sliding. Wang et. al. used

rubber bands to hold the specimen in place at the support rollers [24], Schuecker and David-

son used a fixture on the pre-cracked end of the specimen to hold it and keep it from sliding

[29], and Kageyama et. al. used a PTFE film between the rollers and the specimens to

reduce friction, and in turn, reduce the stick slip seen in the load-displacement curve [30].

32

(a) Beginning of test

(b) End of test

Figure 3.16: Images showing relative sliding of specimen on fixture.

33

The method by Wang was trialed first to combat the sliding of the specimen. The reason

that this was chosen first is because it would not completely constrain the specimen from

sliding, and it would put fewer artificial loads into the specimen. As opposed to constraining

the pre-cracked end, the rubber bans did not add any lateral load into the specimen, and

it required less post-processing of the specimen. This method was found to eliminate the

relative sliding in some specimens, but there was still reduced sliding in others. Due to the

inconsistencies in the specimens and the relative sliding that still existed in some, another

method was pursued.

The next method investigated was to reduce the friction in the two loading rollers and

the support roller under the pre-crack. Kageyama et. al. found this to reduce the stick-slip

in the load-deflection curve, and this is what the relative sliding appears as in the data. The

friction was reduced by using a 0.001 in thick polyimide film between the rollers and the

specimen. When this is implemented, there is still some sliding, but this can be expected in

a four-point bend test. Instead of the stick-slip condition seen before, the sliding was gradual

throughout the test, and it was reduced when compared to the original tests.

The stick-slip condition was nearly, if not completely, eliminated by reducing the friction

in the loading rollers and support roller under the pre-crack. There is still some relative

sliding between the specimen and the fixture, and this was sought to be eliminated by

increasing the friction at the fourth roller. This would allow for the beam to be nearly

completely simply supported with no sliding allowed on the end of the specimen without the

pre-crack and reduced friction sliding at the end with the pre-crack. The increase friction is

achieved by using a thin piece of high-grit sandpaper between the roller and the specimen

with the grit side facing the specimen. When this was implored along with the polyimide

film between the loading rollers and the support roller under the pre-crack, no stick-slip or

34

Figure 3.17: 4ENF fixture with orange polyimide film to reduce friction and gray sandpaper to increase it

Table 3.7: 4ENF Sample Dimensions [mm]

Specimen

UD1

UD2

UD3

UD4

Q3D1

Q3D2

Q3D3

Q3D4

b

25.7

25.6

26.2

25.4

25.3

25.2

25.7

25.7

h

1.805

1.809

1.820

1.789

1.796

1.863

1.877

1.866

relative sliding is seen. The final setup of the fixture is seen in figure 3.17. Figure 3.18 shows

how this fixture’s adaptation changes how the specimen is sliding on the fixture.

For the 4ENF tests conducted in this study, the spans are changed from the previous

ENF tests and what is seen in literature [5, 30, 31]. The geometry of the test is show in

table 3.7.

A d/2L value of 0.5 is used to remain consistent with literature [5, 30, 32, 33, 34], and this

is the determining factor for the geometry of this test. The loading span is set to 63.5mm

to ensure that two full interlaminar tows beyond the initial crack tip are within the loading

span. With the standard d = 50mm, the second tow is at/beyond the second loading roller,

35

(a) Beginning of test

(b) End of test

Figure 3.18: Images showing no relative sliding of specimen on fixture.

Figure 3.19: Representation of nomenclature for 4ENF test [5]. Not shown is d, which is the spacing between
the loading rollers.

36

Table 3.8: 4ENF Parameter Dimensions [mm]

Parameter

a

L
SL
SR

Value

35

63.5

31.75

95.25

Figure 3.20: 4ENF setup showing cameras of both sides of the specimen

and the influence of the roller is not discernible from the influence of the interlaminar two.

For this reason, the inner span is increased, the the outer is also increased to maintain the

d/2L ration of 0.5. These testing parameters are summarized in table 3.8

Due to the [0◦/60◦/-60◦]s not having the 0◦ply at the interface and the interlaminar tows

of the Q3D not always being located in the center of the beam, two cameras are used to

record the crack growth. Figure 3.20 shows the cameras on both sides of the specimen. The

sides of the specimen are painted white, and marks are made every millimeter such that an

accurate measure of the crack length can be taken. The cameras are each taking images

at 1 image

second to match the data recording of the testing frame at 1 Hz. The tests are also

conducted at 0.5 mm

min to coincide with the standard for ENF and the paper proposing 4ENF

to measure mode II fracture toughness [5]. An initial crack length, ao, of 35 mm to put the

crack tip just inside the first loading roller.

37

Table 3.9: 4ENF Mode II Averages

Initial

Steady-State

UD

759.5

449.4

(cid:105)

(cid:104) J

m2
Q3D

697.3

653.1

3.3.3.3 Results

In the 4ENF tests, there is no need to perform PC and NPC tests. This is because there is

continuous crack propagation throughout the 4ENF test, so only the initial crack propagation

is considered NPC, and the rest is considered PC. For this study, the initial crack propagation

is ignored because the steady-state crack resistance is desired. As seen in the mode I and

mode II tests, the Q3D architecture shows its advantage beyond initial crack propagation,

and this is where the comparison will be drawn.

The values for fracture toughness through the 4ENF test are found in a similar manner

to the ENF test. Similar to the ENF test, the compliance calibration is also used, and these
measurements are taken at crack length of 25mm and 45mm to keep with the ±10mm from

the ENF test. The main difference, however, comes from the equation used to calculate the

fracture toughness. For ENF, the GII value is dependent upon the crack length, but with

4ENF, it is not. Beam theory shows this relationship, but in reality, the loading rollers have

an effect on the G value [5]. The equation used to calculate GII for the 4ENF test is:

GII =

P 2m
2b

(3.5)

Table 3.9 shows the average values obtained in the initial and steady-state crack growth

regimes of the R-curves.

In the steady state regime, these values are about 20% smaller

than what is obtained through ENF. This is contradictory to what is observed by Martin

[5]. However, Martin mentioned that the specimens were tested a year later, and moisture

38

Figure 3.21: Comparison of representative load-deflection curves from UD and Q3D 4ENF tests

absorption could cause an increase in the perceived toughness of the composite. A different

span is used in the ENF and 4ENF tests in this study, and that could have an effect as well.

Figure 3.21 shows a comparison of representative curves from the UD and Q3D 4ENF

specimens. This shows that the Q3D maintains a higher peak load, and it does not have that

immediate drop. This would require more energy to achieve the same deflection and damage

in the Q3D specimen as is seen in the UD specimen. This observation also agrees with the

work done comparing Q3D and UD in out-of-plane impact tests. Liu, Rosario, Klann, and

Zhou found that a Q3D structure has lower energy absorption and lower deflection than

a UD sample, but it also has a much smaller damage area, leading to a higher specific

energy absorption [2, 19, 20]. Coppens and Liu also found that a larger amount of energy

is required to fully perforate a Q3D laminate compared to a UD laminate [35]. The higher

fracture toughness for steady-state crack growth along with the higher load sustainability

39

Figure 3.22: Comparison of representative R-Curves from UD and Q3D specimens

point towards this higher specific energy absorption and larger required energy for perforation

found in literature.

A similar trend to the mode I R-curves is shown for the mode II R-curves in figure 3.22.

Initially near the pre-crack, the Q3D has a fracture toughness which is lower than that of

the UD. However, beyond 50mm, the effect of the through thickness tows is shown with

a higher fracture toughness. Both the UD and the Q3D show the same downward trend
beyond the large crack growth from ∼50-∼65mm. This shows that crack propagation is

similar for both the UD and the Q3D between the interlaminar tows, but the interlaminar

tows have a significant improvement on the fracture toughness: nearly 45%. More R-curves

for both architectures are shown in the appendix.

In the mode I specimens, the higher fracture toughness comes from the interlaminar

tows physically bridging the gap and breaking at a higher load. However, this is not the

40

(a) UD specimen after 4ENF

(b) Q3D specimen after 4ENF

Figure 3.23: Images showing relative sliding of specimen on fixture.

case for mode II. In mode II tests, the interlaminar tow of the Q3D laminate does not

break. This is seen post-mortem in figure 3.23. In these images, the same crack separation

is given at the location of the support during the 4ENF test. The UD specimen shows less

resistance to being pulled apart along the length of the crack where the Q3D is being held

together by that interlaminar tow. This would mean that for the same crack growth, the

Q3D structure will likely have better in-plane properties. This carries significant meaning

for this study because this architecture is being investigated for a crash-sensitive component

in an automotive vehicle.

41

Chapter 4:

In-Plane Characteristics

4.1 Background

The in-plane properties of the Q3D architecture are also investigated in this study. The

main purpose is to gather data for material characterization such that a material model can

be made. This material model would be applied to the hat section that this material is being

investigated for, and the theoretical results can be compared with the experimental results.

The in-plane results can also be compared experimentally with UD and 2DW architec-

tures if the same fibers are used. This would show the true effect of the 3 dimensional

braiding on the in-plane properties of the laminate as compared to a UD or traditionally

woven laminate.

4.2 Tension

4.2.1 Experiment

The tensile tests for the Q3D material are performed in accordance with ASTM D3039 [36].

Specimens are tested at a rate of 2 mm

min on a MTS 810 hydraulic test frame. A typical gage

length of 33.5 mm is used, and strain measurements are made by utilizing a laser exten-

someter to measure strain before failure. After failure, the reflective tape on the specimen

often falls off or moves, thus making the strain reading inaccurate. The experimental setup

is shown in figure 4.1 and the specimen with the reflective tape is shown in figure 4.2. The

tabs are made from a woven fiberglass composite and adhered by using 24 hour, room-temp

cure epoxy. The tabs are 8.89 mm thick such that the MTS machine could be used for its

high load capacity. The dimensions for the specimens are shown in table 4.1. In this table,

42

Figure 4.1: Setup of tensile tests showing laser extensometer and specimen

43

Figure 4.2: Close-up of specimen showing the silver tape that is reflecting the laser

44

Table 4.1: Tensile Sample Dimensions [mm]

Specimen

Q3D1

Q3D2

Q3D3

b

26.1

25.1

25.5

h

4.380

4.813

4.353

Lg

103.2

94.6

101.7

Lg refers to the length between the reflective tape used by the laser extensometer to measure

the strain. The overall length of 250mm and general width of 25.4mm is recommended by

ASTM D3039 [36].

4.2.2 Results

The stress and strain for the test are calculated using equations 4.1 and 4.2, respectively.

σ =

P
A


 =

δ
Lg

(4.1)

(4.2)

After failure, the laser extensometer can no longer accurately measure the strain, so to obtain

a full stress-strain curve to include after failure characteristics, a relation with the machine

displacement output is made. A correction value is obtained by doing a linear regression

between the displacement given by the laser extensometer and by the test machine. This

correction factor is then applied for all of the machine displacement values to obtain strain

beyond initial failure. These stress-strain curves are shown in figure 4.3. The values for the

ultimate stress and elastic modulus are then shown in table 4.2.

Figure 4.4 shows the failure of each of the specimens. Each failure is slightly different

with Q3D1 and Q3D2 failing at the top of the specimen as opposed to in the middle like

45

Figure 4.3: Stress-strain curves from Q3D tensile tests

Table 4.2: Q3D Tensile Properties

Specimen

Ultimate Stress [MPa]

Elastic Modulus [GPa]

Q3D1

Q3D2

Q3D3

Average

232.3

236.9

240.4

236.5

27.28

27.38

28.39

27.68

46

Figure 4.4: Tensile failures of the Q3D1, Q3D2, and Q3D3 specimens, respectively

Q3D3. The Q3D2 fails across the specimen where the Q3D1 and Q3D3 fail in the direction
of the -60◦tow. This is more evident in the Q3D1 specimen with failure occurring completely
parallel to the -60◦tow.

4.3 Flexural

4.3.1 Experiment

The flexural properties of the Q3D composite are determined by following ASTM D7264 [37].

Instead of using the standard width of 12.7 mm, a width of 40mm is used. This is because

the standard prescribes a width of two unit cells for textile composites. The standard span-

to-thickness ratio of 32:1 is maintained. The typical thickness of a specimen is 4mm, and

thus the support span, L, used is 128mm. Figure 4.5 shows the nomenclature for the test,

47

Table 4.3: Flex Test Dimensions [mm]

Specimen

Q3D1

Q3D2

Q3D3

Q3D4

Q3D5

b

39.45

40.29

39.50

38.85

38.90

h

4.233

3.889

4.250

4.118

4.127

and table 4.3 shows the specimen dimensions.

For the flex tests, h is the full thickness

and not the half thickness. Figures 4.6 and 4.7 show the specimen mid-test. The tests are

conducted with a crosshead displacement rate of 1mm
min.

4.3.2 Results

Equations 4.3 and 4.4 are used to calculate the stress and strain of the specimen in the

flexural test. The load is taken from the load cell in line with the loading roller, and the

displacement is taken from the crosshead displacement of the machine. This is used because

the output was confirmed to be within 0.5% by a laser extensometer for all expected loads.

σ =

3P L
2bh2

(4.3)

Figure 4.5: Diagram showing flexural test nomenclature

48

Figure 4.6: Isometric view of 3 point bend setup

Figure 4.7: Front view of 3 point bend specimen mid-test

49

Figure 4.8: Stress-strain curves from Q3D flexural tests


 =

6δh
L2

(4.4)

From the results of the flexural tests, the average ultimate stress is found to be 265.9

MPa with a flexural modulus of 18.49 GPa. Figure 4.8 shows the stress-strain curves from

the test. Strain values are taken from 0.001 to 0.003 mm
mm.

Initial failure occurs on the compression side of the specimen. Figures 4.9 and 4.10 shows

failures for the Q3D4 specimen. The compression failure occurs first at a stress of just under

300 MPa and a strain of 0.015 mm

mm. Failure occurs solely on the compression side of the

specimen until a strain of about 0.0375 mm

mm where the first failure is seen on the tensile

surface of the specimen. Table 4.4 shows the properties for each specimen tested along with

the average for the five.

50

Figure 4.9: First failure at the compression side of the Q3D4 specimen at the peak load

Figure 4.10: Failure at the tensile side of the Q3D4 specimen

Table 4.4: Q3D Flexural Properties

Specimen

Ultimate Stress [MPa]

Flexural Modulus [GPa]

Q3D1

Q3D2

Q3D3

Q3D4

Q3D5

Average

266.5

255.6

243.3

295.0

269.1

265.9

51

17.14

19.17

16.31

20.93

18.93

18.49

Chapter 5: Hat Section

The purpose of this work is to test the feasibility of using a CFRP composite as the material

for the roof-header in an automotive vehicle. The critical section of this roof-header is

deemed the “hat section,” and figure 5.1 shows a cross section of the studied hat section

with dimensions removed. Bayraktar used a 3D woven composite in a similar hat section

and found that the 3D woven composite will absorb less energy than a comparable steel

hat section, but it will be more efficient in terms of energy absorption per mass [21]. When

compared to the optimized hat section, the composite beam did not perform as well, but the

composite beam was not optimized at all.

Figure 5.1: General cross section of the hat section

52

5.1 Manufacture

A special tool is required to create the hat section, and it is machined out of aluminum. The

tool is aluminum so that it is durable and sand-able for multiple infusions. The aluminum

will also be able to handle the heat of the cure and is light enough to be transferred from the

work area to the oven by one person. It is also manufactured such that the resin inlet and

the vent are integrated into the tool. The tool is tapped with 1/4” NPT threads such that

barbed tube fittings can be placed on either end. An image of the tool is shown in figure

5.2.

Infusion of the hat section is very similar to that of a flat plate. The major difference

between the two is that the hat section needs pleats in the concave sections of the mold to

reduce bridging in the bag.If the bag bridges, there is no pressure in the concave section,

and the preform with not take the shape of the tool. This also gives an area where resin

will pool and is a likely site for voids. The pleats used for this part are shown in figure 5.3.

Figure 5.2: Tool used to manufacture hat section

53

When vacuum is first pulled, the pressure is only brought to ∼20inHg. This is so that there

is partial vacuum holding the bag in place, but it can still be easily manipulated. The crease

that is caused by the pleats is then manipulated to be on top of the concave section. When

full vacuum is pulled, instead of the bag bridging, the bag will come into full contact with

the preform because of the extra bagging from the crease. Once this has been done and

all bridging has been eliminated, that infusion process can continue as normal. Figure 5.4

shows a hat section during the infusion process.

The Q3D material also comes with issues when it comes to manufacturing. Because it is

a single preform and not multiple layers, this can cause poor drapability. There were issues

involving draping the Q3D preform onto the hat section mold which caused bridging in the

concave section. The effect of this bridging can be seen in figure 5.5 where there is an excess

of resin in one part of the component. It is possible to overcome this, but greater care needs

to be taken in manufacturing. The bridging can also be overcome with more gradual corners.

A flat plate of the same material is also made, and it is bonded to the bottom of the hat

section. It is crucial that this flat plate is made of the same or similar material because if

it has a different stiffness, this could cause premature or uncharacteristic failure in the hat

section. This flat plate is bonded to the bottom of the hat section such that the beam bends

along its span as opposed to opening up and becoming wider. Along with the Q3D material,
hat sections and flat plates made from A&P Technologies triaxially braided [0◦/60◦/-60◦]

carbon fiber are also manufactured for a comparison to a textile composite. Previous tests

done by Ford show delamination in the center at the hat section that the Q3D seeks to solve.

Figures 5.6 - 5.8 shows example hat sections that will be tested.

54

(a) Pleat in the sealant tape

(b) Pleat in the vacuum bag

Figure 5.3: Pleats in the vacuum bag of a hat section infusion

55

Figure 5.4: Infusion of a hat section

Figure 5.5: The resin rich area on the left side of the part is a direct effect of preform bridging during
composite manufacture

56

Figure 5.6: Isometric view of a Q3D hat section

Figure 5.7: Comparison of Q3D and 2DW hat sections

Figure 5.8: Q3D hat section with flat plate to be bonded for testing

57

5.2 Experiment

The hat section has not been tested as of the writing of this thesis, but tests are currently

in progress. The test being performed is a dynamic 3 point bend test to simulate an impact

occurring transverse to the length of the hat section. Both the textile hat section and the

Q3D hat section will be tested, and they will have similar thicknesses to draw conclusions

based simply upon the fiber architecture. 3 specimens of each are tested to show repeatability

of the tests such that an accurate conclusion can be drawn about the feasibility of using the

Q3D architecture for structural applications.

58

Chapter 6: Conclusion

6.1 Summary and Conclusions

Braiding a composite structure with the bias tows spanning multiple layers has shown to

improve the interlaminar fracture toughness of laminated composite materials. Under mode

I loading, a fracture toughness increase of over 20% is found when compared with UD and a

traditionally braided laminate of the same fiber orientation. It has also been found that the

Q3D composite requires more energy for the same amount of crack propagation compared to

the UD and 2DW in both mode I and mode II testing. Using the energy absorption is found

to be a good method to compare the fracture performance of composites because there is no

discrimination due to the number of times that the crack propagates during the test. The

Q3D architecture also shows an even greater increase in mode II fracture toughness with

47.3% and 18.4% increase over UD and 2DW composites, respectively when looking at the

PC condition for the 3 point ENF test.

The 4ENF test was also investigated to determine its effectiveness for producing an R-

curve for mode II tests. There are initial issues with the tests due to specimen sliding,

but by reducing the friction of the loading rollers and support roller under the pre-crack

and increasing the friction of the support roller under solves this issue. An R-curve can

be generate for mode II fracture toughness because the 4ENF test will give stable crack

growth. When comparing the fracture toughness values of the 4ENF test to the ENF tests,

the initial fracture toughness is over predicted, but the steady-state fracture toughness is

under predicted.

At initial crack growth, the Q3D architecture does not perform as well as the UD, but

59

this agrees with the mode I results and the mode II results from ENF. However, in the steady

regime of crack growth, Q3D shows an average 41.3% increase in fracture toughness when

compared to a UD laminate with the same fiber orientation. This increase is also maintained

as the crack propagates.

Tensile and flexural tests have been conducted on the Q3D composite to view the stress-

strain behavior. The results from these tests will also be compared to a UD laminate to

compare properties and investigate if this fiber architecture causes any degradation in the

in-plane properties.

The tri-axial, Q3D preform used in this study is braided into a tube, but when it is cut

lengthwise and spread, it is able to be draped over a tool with curvature to make a complex

part.

6.2 Future Work

Beyond the work presented in this paper, a better comparison of the Q3D architecture

with that of UD and 2DW should be determined for in-plane properties. This will give a

better understanding of how the out-of-plane braiding degrades the in-plane properties of

the composite. The Q3D architecture will also be tested in the hat section structure to

determine is the fracture toughness benefits seen in the mode I and mode II testing translate

well when applied to a structure.

Along with the tensile and flexural in-plane results that have been obtained, compressive

tests will be conducted for the material model and for a comparison to a UD laminate.

The experimental results for in-plane properties and fracture toughness will also be used

to create a material model for the Q3D architecture. Once this material model is made

and confirmed for the in-plane and fracture toughness experiments, the hat section will be

60

simulated with the material model to determine how well the simulation model predicts the

actual performance and failure of the structure. abc

61

APPENDIX

62

Figure A.1: UD mode I R-curves

Figure A.2: 2DW mode I R-curves

63

Figure A.3: Q3D mode I R-curves

Figure A.4: UD mode II load-displacement plots

64

Figure A.5: 2DW mode II load-displacement plots

Figure A.6: Q3D mode II load-displacement plots

65

Figure A.7: UD mode II 4ENF R-curves

Figure A.8: Q3D mode II 4ENF R-curves

66

BIBLIOGRAPHY

67

BIBLIOGRAPHY

[1] X. Mao, “Mode I and II interlaminar fracture toughness simulation of unidirectional
and quasi-three-dimensional composites,” Master’s thesis, Michigan State University,
2019.

[2] W. Zhou, Peridynamic modeling and impact testing of dynamic damage, fracture, and
failure process in fiber-reiforced composite materials. PhD thesis, Michigan State Uni-
versity, 2018.

[3] ASTM International, “D5528-13 Standard Test Method for Mode I Interlaminar Frac-
ture Toughness of Unidirectional Fiber-Reinforced Polymer Matrix Composites,” stan-
dard, ASTM International, West Conshohocken, PA, 2013.

[4] ASTM International, “D7905/D7905M-19e1 Standard Test Method for Determination
of the Mode II Interlaminar Fracture Toughness of Unidirectional Fiber-Reinforced Poly-
mer Matrix Composites,” standard, ASTM International, West Conshohocken, PA,
2019.

[5] R. H. Martin and B. D. Davidson, “Mode II fracture toughness evaluation using four
point bend, end notched flexure test,” Plastics, Rubber and Composites, vol. 28, pp. 401–
406, aug 1999.

[6] D. G¨okta¸s, W. R. Kennon, and P. Potluri, “Improvement of mode I interlaminar fracture
toughness of stitched glass/epoxy composites,” Applied Composite Materials, vol. 24,
pp. 351–375, Nov. 2016.

[7] A. P. Mouritz, C. Baini, and I. Herzberg, “Mode I interlaminar fracture toughness prop-
erties of advanced textile fiberglass composites,” Composites Part A Applied Science and
Manufacturing, 1999.

[8] I. Gnaba, X. Legrnd, P. Wang, and D. Soulat, “Through-the-thickness reinforcement

for composite structures: a review,” Journal of Inudstrial Textiles, 2018.

[9] V. Dikshit, S. Bhudolia, and S. Joshi, “Multiscale polymer composites: a review of the

interlaminar fracture toughness improvement,” Fibers, vol. 5, p. 38, Oct. 2017.

[10] B. M’membe, S. Gannon, M. Yasaee, S. R. Hallett, and I. K. Partridge, “Mode II delam-
ination resistance of composites reinforced with inclined z-pins,” Materials & Design,
vol. 94, pp. 565–572, Mar. 2016.

[11] A. P. Mouritz, M. K. Bannister, P. J. Falzon, and K. H. Leong, “Review of applications
for advanced threee-dimensional fibre textile composites,” Composites Part A Applied
Science and Manufacturing, 1999.

68

[12] A. P. Mouritz, K. H. Leong, and I. Herszberg., “A review of the effect of stitching on
the in-plane mechanical properties of fibre-reinforced polymer composites,” Composites
Part A Applied Science and Manufacturing, 1997.

[13] D. D. R. Carti´e, M. Troulis, and I. K. Partridge, “Delamination of z-pinned carbon fibre
reinforced laminates,” Composites Science and Technology, vol. 66, pp. 855–861, May
2006.

[14] B. Zhang, G. Allegri, and S. R. Hallett, “An experimental investigation into multi-
functional z-pinned composite laminates,” Materials & Design, vol. 108, pp. 679–688,
Oct. 2016.

[15] A. P. Mouritz, “Review of z-pinned composite laminates,” Composites Part A: Applied

Science and Manufacturing, vol. 38, pp. 2383–2397, Dec. 2007.

[16] A. P. Mouritz, “Compression properties of z-pinned composite laminates,” Composites

Science and Technology, vol. 67, pp. 3110–3120, Dec. 2007.

[17] A. Kerber, A. Gargano, K. Pingkarawat, and A. P. Mouritz, “Explosive blast damage
resistance of three-dimensional textile composites,” Composites Part A: Applied Science
and Manufacturing, vol. 100, pp. 170–182, Sept. 2017.

[18] D. Liu, C.-Y. Lee, C.-F. Yen, C. Anderson, and G. Li, “Composite structure and method

of making a composite structure,” Mar. 2016.

[19] K. Rosario and D. Liu, “Assessment of quasi-three-dimensional composites - with dis-
cussions on fiber straining and weaving effectiveness,” Journal of Composite Materials,
2010.

[20] D. Liu, K. Rosario, and S. Klann, “Behavior of quasi-three-dimensional woven compos-

ites,” American Society of Composites 2008, 2008.

[21] H. Bayraktar, D. Ehrlich, J. Goering, and M. McClain, “3D woven composites for energy
absorbing applications,” 20th International Conference on Composite Materials, 2015.

[22] Y. Zhao, M. Cao, W. P. Lum, V. B. C. Tan, and T. E. Tay, “Interlaminar fracture
toughness of hybrid woven carbon-Dyneema composites,” Composites Part A: Applied
Science and Manufacturing, vol. 114, pp. 377–387, Nov. 2018.

[23] T. O’Brien, W. Johnston, and G. Toland, “Mode II interlaminar frafture toughness and

fatigue characeterization of a graphite epoxy composite material,” NASA, 2010.

[24] W.-X. Wang, M. Nakata, Y. Takao, and T. Matsubara, “Experimental investigation
on test methods for mode II interlaminar fracture testing of carbon fiber reinforced

69

composites,” Composites Part A: Applied Science and Manufacturing, vol. 40, pp. 1447–
1455, sep 2009.

[25] ASTM International, “D3171-15 Standard Test Methods for Constituent Content of
Composite Materials,” standard, ASTM International, West Conshohocken, PA, 2010.

[26] A. Balakrishnan and M. C. Saha, “Tensile fracture and thermal conductivity character-
ization of toughened epoxy/CNT nanocomposites,” Materials Science and Engineering:
A, vol. 528, pp. 906–913, jan 2011.

[27] B. R. K. Blackman, A. J. Brunner, and J. G. Williams, “Mode II fracture testing of
composites: a new look at an old problem,” Engineering Fracture Mechanics, vol. 73,
pp. 2443–2455, nov 2006.

[28] P. Davies, G. D. Sims, B. R. K. Blackman, A. J. Brunner, K. Kageyama, M. Hojo,
K. Tanaka, G. Murri, C. Rousseau, B. Gieseke, and R. H. Martin, “Comparison of test
configurations for determination of mode II interlaminar fracture toughness results from
international collaborative test programme,” Plastics, Rubber and Composites, vol. 28,
pp. 432–437, Sept. 1999.

[29] C. Schuecker and B. D. Davidson, “Evaluation of the accuracy of the four-point bend
end-notched flexure test for mode II delamination toughness determination,” Compos-
ites Science and Technology, vol. 60, pp. 2137–2146, aug 2000.

[30] K. Kageyama, I. Kimpara, T. Suzuki, I. Ohsawa, M. Kanai, and H. Tsuno, “Effects of
test conditions of mode II interlaminar fracture toughness of four-point ENF specimens,”
Proceedings of the 1999 International Conference on Composite Materials, 1999.

[31] P. Davies, P. Casari, and L. A. Carlsson, “Influence of fibre volume fraction on mode II
interlaminar fracture toughness of glass/epoxy using the 4ENF specimen,” Composites
Science and Technology, vol. 65, pp. 295–300, feb 2005.

[32] G. Marannano, F. Parrinello, and A. Pasta, “Numerical and experimental analysis of
the frictional effects on 4ENF delamination tests performed on unidirectional CFRP,”
Procedia Engineering, vol. 109, pp. 372–380, 2015.

[33] J. R. Reeder, K. Demarco, and K. S. Whitley, “The use of doubler reinforcement in de-
lamination toughness testing,” Composites Part A: Applied Science and Manufacturing,
vol. 35, pp. 1337–1344, nov 2004.

[34] B. D. Davidson and X. Sun, “Effects of friction, geometry, and fixture compliance on the
perceived toughness from three-and four-point bend end-notched flexure tests,” Journal
of Reinforced Plastics and Composites, vol. 24, pp. 1611–1628, jun 2005.

70

[35] G. C. Coppens and D. Liu, “Effects of through-thickness fiber geometry and snall angle

on impact resistance,” Journal of Advanced Materials, 2007.

[36] ASTM International, “D3039/D3039M-17 Standard Test Method for Tensile Proper-
ties of Polymer Matrix Composite Materials,” standard, ASTM International, West
Conshohocken, PA, 2017.

[37] ASTM International, “D7264/D7264M-15 Standard Test Method for Flexural Prop-
erties of Polymer Matrix Composite Materials,” standard, ASTM International, West
Conshohocken, PA, 2015.

71