Experimental Setup For Fibre Reinforced Composite

Experimental Setup and Testing of Fiber Reinforced
Composite Structures
by
Scott Robert John Bumpus

Bachelor of Engineering, University of Victoria, 2002
A Thesis Submitted in Partial Fulfillment of the

Requirements for the Degree of
Master of Applied Science
in the
Department of Mechanical Engineering.
c Scott Robert John Bumpus, 2005

University of Victoria
All rights reserved. This thesis may not be reproduced in whole or in part, by
photocopy or other means, without the permission of the author.
ii
Supervisors: Dr. Afzal Suleman
Abstract
Fiber-reinforced composite structures have seen an increased application in aero-

nautics and in other industries such as automotive, marine transportation, civil en-
gineering, sporting goods, medical equipment and prosthetic devices. With the in-
creased use of composite materials, there is a need to develop methods to predict the
material properties and behavior of composite materials and structures made of these
materials under a variety of loading and environmental conditions.
In this thesis, an experimental test procedure was designed and implemented to
determine the mechanical properties of fiber reinforced composite structures. Resin
transfer molding was used to manufacture the test specimens. Large panels were
molded with different constituent concentrations. The test coupons were cut from
a single plate using a water-jet cutting technique. Tensile and flexural tests were
performed and tables of new material properties have been created. Each of the spec-
imens were tested in a random order and the stress and strain data was calculated
from the load and displacement results. The experimental tests were performed at
two perpendicular orientations to determine the influence of fiber orientation on the
material mechanical properties. Experimental values were obtained for tensile mod-
ulus, maximum stress, the strain at maximum stress, and Poissons ratio in all three
directions.
iii
Table of Contents
Abstract ii
List of Tables v
List of Figures vi
1 Introduction 1
1.1 Background and Motivation . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Testing Guidelines . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3 Scope of the Thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.4 Structure of the Thesis . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2 Testing Environment Design and Setup 9

2.1 Coupon Manufacturing Methods . . . . . . . . . . . . . . . . . . . . . 10
2.1.1 Small Scale Resin Transfer Molding . . . . . . . . . . . . . . . 10
2.1.2 Large Scale Injection Molding . . . . . . . . . . . . . . . . . . 15
2.2 Cutting Pattern . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.3 Test Procedures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.3.1 Tensile . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.3.2 Flexure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.4 Synopsis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3 Data Processing and Analysis 29

3.1 Data Processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.1.1 Tensile Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.1.2 Flexure Data . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.1.3 Database Design . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.2 Design of Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.2.1 Analysis of Variance . . . . . . . . . . . . . . . . . . . . . . . 34
3.2.2 Multiple Factor Analysis of Variance . . . . . . . . . . . . . . 37
TABLE OF CONTENTS iv
3.2.3 F Value . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.2.4 Confidence Interval . . . . . . . . . . . . . . . . . . . . . . . . 38
3.2.5 Box Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
3.3 Methods of Modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
3.3.1 Curve Fitting . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
3.3.2 Rule of Mixtures . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.4 Synopsis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
4 Experimental Results 45
4.1 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
4.1.1 Tensile Results . . . . . . . . . . . . . . . . . . . . . . . . . . 50
4.1.2 Flexural Results . . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.1.3 Effects of Orientation . . . . . . . . . . . . . . . . . . . . . . . 69
4.2 Modeling of Experimental Data . . . . . . . . . . . . . . . . . . . . . 74
4.2.1 Rule of Mixtures . . . . . . . . . . . . . . . . . . . . . . . . . 74
4.2.2 Analysis of Variance . . . . . . . . . . . . . . . . . . . . . . . 76
4.3 Synopsis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
5 Conclusions and Future Work 85

5.1 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
5.2 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
References 89
v
List of Tables
1.1 Factor Ranges . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
4.1 Phase Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

4.2 Panel Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
4.3 Mass Fractions of the Three Phases . . . . . . . . . . . . . . . . . . . 49
4.4 Volume Fractions of the Three Phases . . . . . . . . . . . . . . . . . . 50
4.5 Tensile Panel Properties . . . . . . . . . . . . . . . . . . . . . . . . . 58
4.6 Confidence Interval for Tensile Properties . . . . . . . . . . . . . . . . 59
4.7 Flexural Panel Properties . . . . . . . . . . . . . . . . . . . . . . . . 65
4.8 Confidence Interval for Thickness and Flexural Properties . . . . . . . 65
4.9 ANOVA Results for Individual Panels . . . . . . . . . . . . . . . . . . 69
vi
List of Figures
2.1 Small Scale Molds: Flexure Mold (left) and Tensile Mold (right) . . . 11
2.2 Injection Port Options: Flexure Mold (left) and Tensile Mold (right) . 11
2.3 Steel Rule Dies: Flexure (left) and Tensile (right) . . . . . . . . . . . 12
2.4 Mold Ready For Injection . . . . . . . . . . . . . . . . . . . . . . . . 14
2.5 Molded Test Coupons: Flexure (top) and Tensile (Bottom) . . . . . . 15
2.6 Mold for Creating Large Flat Composite Panels . . . . . . . . . . . . 16
2.7 Inlet and Outlet for Mold Heating System . . . . . . . . . . . . . . . 17
2.8 Resin Mixer and Pump For Hybrid Resin . . . . . . . . . . . . . . . . 18
2.9 Resin Flow Pattern Through the Composite . . . . . . . . . . . . . . 19
2.10 Water Jet Cutter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.11 Preform Container On Scale . . . . . . . . . . . . . . . . . . . . . . . 22
2.12 Water Jet Cutting Pattern for Panels . . . . . . . . . . . . . . . . . . 24
2.13 Specimen Loaded and Ready for Tensile Test . . . . . . . . . . . . . . 26
2.14 Specimen Loaded and Ready for Flexure Test . . . . . . . . . . . . . 27
3.1 Coordinate System . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

3.2 Composites Compared to Isotropic Materials for Beam Theory . . . . 33
4.1 Variation of Weight (Top) and Thickness (Bottom) with Panel Position 46
4.2 Tensile Stress-Strain Plots for Panels 1 and 2 . . . . . . . . . . . . . . 51
4.6 Tensile Stress-Strain Plots for Panels 9 and 10 . . . . . . . . . . . . . 53
4.7 Tensile Strain-Strain Plots for Determination of Poissons Ratio of Pan-
els 1 and 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
els 3 and 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
els 5 and 6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
LIST OF FIGURES vii

els 7 and 8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
els 9 and 10 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
4.12 Confidence Interval of Poissons Ratio . . . . . . . . . . . . . . . . . . 58
4.13 Flexure Stress-Strain Plots for Panels 1 and 2 . . . . . . . . . . . . . 62
4.18 Confidence Interval of Tensile and Flexural Moduli . . . . . . . . . . 67
4.19 Confidence Interval of The Maximum Tensile and Flexural Stresses . 68
4.20 Confidence Interval of The Strain at Maximum Tensile and Flexural
Stress . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
4.21 Box Plot Orientation Comparison for Tensile Modulus . . . . . . . . 70
4.22 Box Plot Orientation Comparison for Maximum Tensile Stress . . . . 70
4.23 Box Plot Orientation Comparison for Strain at Maximum Tensile Stress 71
4.24 Box Plot Orientation Comparison for Flexural Modulus . . . . . . . . 71
4.25 Box Plot Orientation Comparison for Maximum Flexural Stress . . . 72
4.26 Box Plot Orientation Comparison for Strain at Maximum Flexural Stress 72
4.27 Hi Flow Mat Core . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
4.28 Comparison of Rule of Mixtures Estimation With Experimental Data 75
4.29 Scatter Plot Showing ANOVA Points . . . . . . . . . . . . . . . . . . 77
viii
Acknowledgements
I would like to thank my supervisor, Dr. Afzal Suleman, for his support and guidance
throughout the time it took to complete this thesis and my degree. The graduate
research assistance provided by ASI BC and Profile Composites Inc. is acknowledged.
I would also really like to thank the people at Profile Composites Inc; Geoff Wood,
Peter, Quinn, Graham, Ryan, Tamra, and Dan. Thanks to Sean Taylor at Aquashear
for dealing with all the panels cut and the specific requirements that came with that
task. I would also like to thank my graduate student colleagues working under Dr.
Suleman, Sandra Makosinski, and the mechanical engineering graduate secretaries for
their support, suggestions, and friendship throughout this degree. Last but not least
I would like to thank my friends and family for their support.
Chapter 1
Introduction
Historically a predominantly aerospace material, fiber-reinforced composite structures

have seen an increased application in other industries such as automotive, marine
transportation, civil engineering, sporting goods, medical equipment and prosthetic
devices. Fiber-reinforced composite structures provide high strength, high stiffness
mechanical properties, unique flexibility in design capabilities, and ease of fabrication.
Also, they are lightweight, corrosion resistant, impact resistant, and have excellent
fatigue strength. With the increased use of composite materials, there is a need
to develop methods to predict the material properties and behavior of composite
materials and structures under a variety of loading and environmental conditions.
Fiber-reinforced composites are composed of two or more materials which, when

properly combined, form a different material with properties not available from the
ingredients alone. Depending on the ingredients chosen and the method of combining
them, a large spectrum of material properties can be achieved. A brittle material
can be made more ductile (flexible) by adding a softer material; conversely a soft
material can be made stiffer. Wood is a good example of a composite. The cellulose
fibers provide the strength and are held together by the resin. Reinforced concrete is
CHAPTER 1. INTRODUCTION 2
another example. The steel re-bars provide excellent tensile strength and the concrete
provides compressive strength and transfers the load between the steel bars. Modern
composites or FRP (Fiber reinforced polymers, or plastics) are the newest addition
to the structural engineers toolbox. Although the materials have been available for
decades, a reduction in cost, combined with newer understanding of the versatility and
benefits of the material properties, has allowed composites to move into mainstream
construction.
Composite properties are phase dependent, and every phase included in the com-
posite has a significant effect on overall material properties. There are a number of
models that can be used to predict the composite properties, but the more accurate
the model, the more one must know about the different phases. Extensive tests must
be carried out on the individual phases to determine the material properties that are
used for a specific model. Even if the properties of the materials are known, these are
considered an approximation due to the intrinsic nature of composites. To demon-
strate this point, the composites factor of safety in the marine industry is taken as
10 [1].
1.1 Background and Motivation
There are a number of methods for manufacturing composite structures. The pre-
dominant method used in the aerospace and other high cost industries is the autoclave
method. Another common method of composites construction is resin transfer mold-
ing (RTM). This more modern method is a faster and more cost effective way to
create parts. Comparisons of these methods have been reported in the literature such
as [2, 3]. The autoclave has been around for a longer period and is therefore a more
researched and well known method. However, the RTM method is becoming more
wide spread and further research is being conducted due its lower cost production.
The cost savings comes from faster production and lower material costs.
For the autoclave production method, the part is first molded using a prepreg
material. A prepreg consists of fibers in sheet form already impregnated with resin.
The prepreg is placed in a mold and then heated to 120-180o C and at this temperature
the resin cures. The part is placed in the autoclave which is a large temperature
pressure vessel. The temperature is then raised and the pressure is increased in order
to consolidate the part, thus removing air bubbles and squeezing out any extra resin.
Bringing the autoclave up to temperature is a slow process which depends largely on
the size of the autoclave. The pressure chamber can reach up to 5 bars, creating very
high quality and consistent parts.
Resin transfer molding is a process which involves placing dry fibers in a mold
and adding the resin separately. Resin is injected in the mold at the inlet port (or
ports depending on how the mold is designed) until it flows out the outlets and the
fibers become completely wetted out. The mold can be a closed mold or it can be
composed of a single flow surface and a vacuum bag. A vacuum bag is just a thin
film that covers the fiber preform and is connected to the base of the mold, or it may
encompass the entire mold. A vacuum is then pulled on the part in order to impart
a one atmosphere pressure over the part surface. This process provides for a large
reduction in cost because purchasing the fibers and resin separately are cheaper than
prepreg materials, and vacuum bags are much cheaper than an autoclave. Also, cure
times are faster because a catalyst is used in RTM as an alternative to waiting for the
system to get to really high temperatures. As an example, Ramulu et al. [4] reported
cure times of 27 to 32 hours to create panels using the autoclave method. During the
course of this thesis, it was possible to manufacture 5 panels in a single day using the
RTM procedure.
The advantage of using an autoclave over an RTM process is that it produces

higher fiber concentrations. Due to the higher pressures involved, it packs the fibers
tighter and essentially squeezes the resin into small areas between fibers that may
not be possible using the RTM process. McIlhagger et al. [3] did a comparison with
similar input fiber and resin and found about a 25.7-27.2% reduction in flexural
properties and a 15-17.7% reduction in tensile properties. When using the RTM
method, Marsh [5], also concluded that RTM is a technology that can produce a part
in minutes which would take hours or even days with previous technology. This is
one of the reasons that autoclaves are still used in the aerospace industry, however
for parts that do not require such high quality, the RTM can be a very affordable
solution. For example, in the automotive industry there are numerous parts that do
not require such high quality composites.
Hybrid composites have been extensively studied and reported in the open liter-
ature. Continuous fiber oriented hybrid composites have been studied by Junior et
al. [6] and are a very popular composite because of its superior strength character-
istics. Also, Fu et al. [7, 8] have done extensive research on short fiber composites.
These short fibers are on average less than 1mm in length and are mixed with the
resin matrix prior to being injection molded to create the samples. Joseph et al. [9]
and Rana et al. [10] also investigated this composite with a maximum unbroken fiber
length of about 10mm. This method is simple but results in weak reinforcement due
to the short fiber lengths.
The main purpose of this thesis is to create a database of tensile and flexural
properties for chopped fiber composite structures. With this information, models can
be created to predict how different phase combinations effect the composite material
properties. A large number of tests were performed. Also in order to ensure that
the data is statistically accurate, a number of replicate tests were performed and a
statistic method based on the Design of Experiments (DOE) was used.
Design of Experiments (DOE) is a method for selecting the best combinations of

factors in order to determine the optimal response of the system. Given a test matrix,
the DOE determines the necessary experiments and the appropriate order of testing.
The method is used to reduce the number of experiments by testing the extremes
of the test matrix and determining the factors that do not significantly affect the
response of the experiment.
Sutherland [1] has done a number of tests using design of experiments and has
found large deviations in the results when testing composites. He has concluded that
when designing an experiment to test a composite material, it is difficult to perform
fractional factorial designs. Fractional factorial designs allow the experimenter to
further reduce the number of tests. It is a technique used when there are a large
number of input variables. It allows the researcher to find out information on all the
factors without actually running every combination of variables.
1.2 Testing Guidelines
In the current research, two fiber phases are examined: the carbon fiber (TORAY
T700SC -12000 -50C) and the E-glass fiber (from either PPG or Owens Corning).
R
Polymat Hi Flow made by Scott and Fyfe is used as a core for the composites.
This core is a prefabricated material that contains a polypropylene core with random
chopped E-glass mat stitched to both surfaces. The resin phase is a urethane and
polyester hybrid resin system (DION ITP 31638-00). A preform is created from the
Hi Flow core with extra glass or carbon chopped fibers added as needed to both sides
of the core. The method for making the composite used in this research is resin
transfer molding (RTM). Test samples were manufactured from a series of composite
sheets about 30cm x 60cm and the test coupons were cut out of the sheet using a
waterjet cutter.
ASTM D3039 [11] is used for determining tensile properties. The shape of the
tensile coupon is taken from ASTM D368 [12]. ASTM D790 [13] is used as the
guideline for performing the flexure tests. These are the recommended ASTM test
procedures by the Mil Handbook 17 [14] and have been used by other researchers
such as [4, 15].
With plastics in general, tensile and flexural properties are tested independently.
The tests determine the elastic modulus, the maximum stress, and the strain at
maximum stress for the material in tension and flexure, as plastics behave differently
in these two modes. ASTM D790 has been adapted from a strictly plastics test
and is not reliable for determining actual flexural properties of composites. The Mil
Handbook 17 does not recommend using the flexural properties. In the aerospace
industry, flexure testing is used mainly for quality control [14]. These tests have been
reported in the literature by [10, 16], and are used to compare production methods
and composite compositions.
The experiments were performed on an MTS hydraulic test machine. The ma-
chine has a load cell to measure the applied force. For tensile tests, an MTS biaxial
extensometer is used for calculating the strain in the axial direction, in either the
width or the thickness direction of the specimen. This allows the Poissons ratio to
be calculated. The flexure tests are performed using a three-point bending jig and
the displacement is measured using the sensor on the MTS machine.
1.3 Scope of the Thesis
The motivation for this thesis is to test hybrid composites that can be used for re-
placing traditional construction materials in automobiles and similar applications.
Advanced composites can have better strength to weight characteristics than tradi-
tional materials such as steel and aluminum. There is also the potential for cost
reduction.
In this thesis, a table of properties for the destructive tensile and bending tests
was constructed. The factors to be considered are shown in Table 1.1. Factors like
temperature have not been considered.
Factor Value Range

Carbon Percentage 0% - 57%
Specimen Thickness 3mm - 6mm
Fiber Angle Random orientation
Table 1.1: Factor Ranges
The results of the tests have been analyzed using the statistical design of ex-
periments tools. This technique provides the amount each factor contributes to the
strength of the material and the interaction effects between factors.
1.4 Structure of the Thesis
The thesis is composed of 5 chapters. Chapter 1 presents the background and moti-
vation of the thesis. Chapter 2 presents the experimental setup and the fabrication of
the test coupons. Several techniques were attempted to create the test coupons, and
this chapter describes the steps taken to manufacture the test specimens. Chapter 3
discusses the analysis of the raw data and how the design of experiments method is
used to help perform the experiments in a reliable and systematic manner. Chapter 4
states the results and findings of the experiments. The results are then compared to
numerical data obtained using a basic rule of mixtures model and a response surface
model. Finally, the conclusions and future work is outlined in Chapter 5.
9
Chapter 2
Testing Environment Design and

Setup
In this chapter, the design and manufacturing process for the test coupons are de-
scribed. The manufacturing process was carried out at Profile Composites Inc. The
dimensions of the test coupons and method of testing used in this research were de-
termined using the American Society for Testing and Materials (ASTM) standards.
The tensile coupons were designed in accordance with ASTM standard D 638 -
99 [12]. According to the standard, the specimen should be a dogbone shape. If the
material is less than 7mm thick the width should be 130.5mm with a length of the
narrow straight section of 570.5mm. ASTM standard D3039 [11] is the standard
for polymer matrix composites. The dogbone shape was chosen to ensure that the
samples would break in the gauge area. D3039 cautions that samples should be
individually molded or if cut from a panel care must be taken to avoid a rough finish
from the cut or delamination as a result of poor cutting methods.
The flexure coupons were designed based on ASTM standard D790 [13]. The
CHAPTER 2. TESTING ENVIRONMENT DESIGN AND SETUP 10
specifications for the size of the coupons given in the standard are that the specimen
width shall not exceed 1/4 the support span. The support span is 161 times the
specimen thickness. The ends of the specimen should extend at least 10% of the
support span over the support. The dimensions chosen to fulfill these requirements
were 13mm wide by 126mm long.
2.1 Coupon Manufacturing Methods
Resin Transfer Molding is a common composite manufacturing process. The pre-

catalysed resin is injected into a mold containing the dry reinforcing material arranged
in the desired number of plies and orientation. The injection pressure varies between
0.5-4 bars while the curing of the resin is often assisted by heating.
2.1.1 Small Scale Resin Transfer Molding
In the first instance, it was decided that the composite coupons would be molded
individually using a small scale resin transfer molding process. One mold was made
for two tensile test coupons and another mold was made for two flexure test coupons
(see Figure 2.1).
These molds were made from 1.25 inches thick aluminum on the bottom and 1
inch thick on the top. These dimensions were determined so that the mold can be at
least 1 inch larger than the part in all directions. This allowed for high temperature
injection molding of plastics so that the molds could be used for testing plastics
properties as well. They were created for use on a 20 ton Morgan injection molder.
The molds were also created with the intention to use them for resin transfer molding
and thus have injection ports on the edge of the molds as well. The two injection
Figure 2.1: Small Scale Molds: Flexure Mold (left) and Tensile Mold (right)
Figure 2.2: Injection Port Options: Flexure Mold (left) and Tensile Mold (right)
port choices are illustrated in Figure 2.2.
In order for the molds to be designed properly there are a number of features that
must be included [17]. All cuts into the mold must be tapered by at least 2o . That
includes the edges of the test coupons and the central injection port on both molds.
Under the central injection port there is a small cavity to capture the initial shot of
Figure 2.3: Steel Rule Dies: Flexure (left) and Tensile (right)
injected plastic because it has a tendency to not be fully melted and would not allow
proper flow of the plastic. In order to get the injected material to all parts of the
mold without air bubbles, the mold must be properly vented. These vents are only
0.5mm deep to allow the overflow to be cleanly trimmed from the finish part. The
gates where the injection ports meet the specimen cavity must be designed properly
so they are not too large, making it difficult to remove from the finished product, or
too small, making it impossible to get the necessary flow rates in order to get the
injected plastic to the ends of the cavity before it solidifies. The chosen size was to
have the gate as wide as the depth of the cavity.
Before coupon production could commence tools had to be created to cut the
fibers into the shape of the mold cavities, steel rule dies were chosen for this task.
These are essentially like a cookie cutter and are illustrated in Figure 2.3. They
consist of steel bands with a knife edge on one side which are inserted into a plywood
backing in the desired shape of the object to be cut. The fibers are sprayed into a
flat mat preform and then a 20 ton press is used to stamp the shapes of the steel rule
dies out of the mat. These preforms are then inserted into the molds to create the
coupons.
The process of creating the coupons is to first create a preform mat so that the
coupon shapes can be cut out. These mats were constructed using E-glass mat
stitched to a Rovacore center as the base. This mat was used before the Hi Flow
mat was found to be better. 3M contact cement is then sprayed on the E-glass mat
as a binder for the carbon fibers. Carbon fibers are then sprayed onto both sides of
the mat from the chopper gun to the desired thickness of fibers. This preform is then
stamped to the specimen shape using the steel rule die. The mold is preheated in
the oven. It is then removed from the oven and a mold release is sprayed into the
mold cavity for easy sample removal when curing has occurred. The fiber preform is
then placed in the mold and the appropriate plugs are placed in the injection ports
that are not to be used. The top half of the mold is connected to the bottom and
four hand screw clamps are used, one on each corner of the mold for a good seal (see
Figure 2.4). The mold with the clamps in place is then inserted in the oven to bring
it back to a good molding temperature. Once the mold is up to temperature the resin
is mixed and the mold is once again removed from the oven and the resin is injected
into the mold port using a syringe with sealing tape around the nozzle. After the
injection is complete the excess resin is wiped off the mold. After the specimens have
cured they are removed and the mold is scraped clean of excess resin and prepared
for the next samples.
This molding technique had a few problems that made it an unfeasible method for
coupon production. The molds were very ridged in their design making it very difficult
to vary specimen thickness. Inserts would have to be used to decrease specimen
thickness and spacers would have to be used to make the specimen thicker. Inserts
would be difficult to remove after the resin cures. Also, the steel rule dies were very
difficult to use. The dies were very sensitive to the slightest variations in the steel
height. The carbon fibers used have a filament diameter of 7m, making it exceedingly
Figure 2.4: Mold Ready For Injection
difficult to make a perfectly clean cut. When a group of filaments is left uncut, upon
removing the stamped shape it may still be connected to the source material creating
a shift in the fibers before the filaments break away from the shape or the attached
filaments cause a large section of fibers to be removed from the stamped shape. Also
when removing the preform from the dies, the preform would often come apart and
would not maintain its original shape.
In order to ensure the fiber preform would fit easily in the mold the steel rule dies
were created slightly smaller than the mold. This created a problem. When the resin
was injected it would push the fiber to one side of the mold and there would be a
strip of un-reinforced resin as seen in Figure 2.5. This results in a need to machine
out the un-reinforced resin because the inconsistency would give inaccurate results.
The machining of the specimens would be counter productive to the molding process.
The problem is further complicated by the fact that sometimes the fiber would not be
pushed all the way to one side so the un-reinforced resin would have to be removed
from both sides of the specimen. The objective of molding individual samples was to
test the samples straight out of the mold, but the need to machine the samples would
create an unnecessary overhead to the process.
Figure 2.5: Molded Test Coupons: Flexure (top) and Tensile (Bottom)
These factors made the small scale injection molding impractical. It was decided
to manufacture a large panel and cut the individual coupon samples from this panel.
This process is discussed in the next section.
2.1.2 Large Scale Injection Molding
The method used to create the test coupons was to use a resin transfer process
similar to the individual sample creation process described above. The fundamental
difference is that instead of molding just two coupons at once, a panel about 30cm
by 60cm was created and the coupons were cut from the panel. This is accomplished
using the mold shown in Figure 2.6.
Figure 2.6: Mold for Creating Large Flat Composite Panels
The mold consists of a steel frame with eight locking clamps around the edge. The
mold surface is created from a composite that is formed in the shape of the mold and
is coated with a smooth gel coat surface. Embedded in the bottom of the composite
mold is a system of pipes as seen in Figure 2.7. Hot water is pumped into this piping
and is used to heat the mold. This allows for the mold to be brought up to the
optimum cure temperature of the resin matrix which is 45o C. In the top of the mold
is a glass window which allows the operator to view the resin flow front as the part is
injected. At one end of the mold is the inlet port for the resin to be injected. At the
other end are two exit holes to allow the resin to flow through the fiber preform and
escape. The exit holes are connected to a cylinder from which a vacuum is pulled to
enhance flow and collect excess resin.
After the preform is created the mold is preheated to about 45o C. The preform is
Figure 2.7: Inlet and Outlet for Mold Heating System
then placed in the mold and an inflatable o-ring type dynamic seal is placed around
the mold cavity so that the resin is stopped from escaping though the sides of the
mold. Two rubber strips are then laid between the seal and the preform to block the
resin from traveling down the sides of the preform instead of through the preform.
After the seal is in place and the mold is closed up and clamped, the seal is then
inflated to between 20 and 50 psi depending on desired thickness of the part. The
resin used is a three part hybrid resin system. It consists of a polyester resin, a
polyurethane resin, and a catalyst. The resin is mixed to the desired ratios and
pumped using the equipment seen in Figure 2.8. This equipment also keeps the resin
heated to about 45o C. The resin is pumped into a pressure pot and initially the resin
is injected by pulling a vacuum in the mold. After the part is completely wetted
out and resin flows out the outlet ports, or 5.5 minutes after the injection started,
the vacuum is removed and the pressure pot is used to provide positive pressure. At
about 7 minutes after the start of injection the outlet lines are pinched off and the
Figure 2.8: Resin Mixer and Pump For Hybrid Resin
resin is pumped in for about 45 seconds more. The timing is critical because the resin
starts to cure about 8 minutes after it is mixed. A picture of the resin flow front from
the window in the top of the mold can be seen in Figure 2.9.
After the resin has cured the mold can be opened and the completed composite
removed. The composite panel is then shipped to Aquashear in Vancouver so the test
coupon shapes can be cut using a high pressure water jet with the equipment seen
Figure 2.9: Resin Flow Pattern Through the Composite

Figure 2.10: Water Jet Cutter
in Figure 2.10. This water-jet cutting is very useful for composites because to cut
them with a saw requires diamond tipped cutters and produces dust that is harmful
to breathe. The water-jet cutter can start a cut in the middle of the sheet, it is
computer controlled, and the cut is only 0.8mm wide. The piece to cut is placed on
sacrificial steel strips which are suspended above a large tank of water. The high
pressure water cuts the material and then defuses into the water tank below.
Manufacturing Issues
The large mold method is not free of problems. One problem is that the thickness
throughout the panel is not uniform. This is due to flexing in the mold during the
injection and curing process. This results in up to a 20% increase in thickness from
the edge of the panel to the center.
Originally the resin was injected into the preform only by drawing a vacuum. This
caused a significant amount of air bubbles in the panel. By curing a cup of resin in
a vacuum chamber it was discovered that the vacuum was causing small bubbles to
form in the resin. The process was changed to start with drawing a vacuum and then
turning off the vacuum and applying positive pressure to push the resin in instead
of pulling it with the vacuum, thus minimizing the bubbles.
The preform design proved to be a difficult process. Initially the same process
was used as the small scale RTM method. A rectangular section of the Rovacore mat
was cut to the dimensions of the mold cavity. 3M contact cement was applied to each
side and carbon is added by visual inspection to both sides equally. This was not
a very accurate way of adding carbon. When these preforms were used in the mold
problems developed stemming from inability of resin to flow through the preform.
The problem that would occur was that the preform was not fully wetted so there
would be dry fiber in the final part. Also, for thin panels the fiber would be packed
especially close together and the resin, when injected, would push the preform down
the mold cavity and the part would be unusable. The preform process developed into
more careful addition of fibers, a different core, and a different binding agent.
The final method of preform creation is as follows. A container was created the
size of the mold cavity, this can be seen in Figure 2.11. This container is placed on a
scale and the scale is zeroed. The core mat is placed in the container and weighed.
Binder is sprayed on the exposed surface of the mat and the scale is zeroed again.
If glass fiber is to be added, it is added at this stage. The glass fibers are chopped
into a bucket and sprinkled on by hand because the chopper gun cannot chop the
glass fibers as easily as the carbon fibers. The chopper gun produces fibers that are
about 30mm in length. After the glass is added, more binder is sprayed and the scale
is zeroed again. The carbon fiber is then sprayed on using the chopper gun. The
preform is then flipped over and the process is repeated on the other side. With the
scale under the container it is easy to determine how much fiber is being added. Care
must be taken to ensure even distribution of fibers and make sure there are no bare
spots. It is very difficult to ensure even distribution and this is a likely cause for
scatter in material data.
Figure 2.11: Preform Container On Scale
The core mat used is made by Scott and Fyfe and is called Polymat Hi flow.
The Polymat consists of two layers of glass chopped strand mat stich-bonded to a
polypropylene core. The core allows the resin to flow through the middle of the glass
sandwich and then the resin seeps out the the top and bottom surface. The Rovacore
mat is supposed to behave similarly but the Polymat core allows the resin to flow
much more freely, thus allowing the mold to be filled more smoothly. The binder used
is called InfuZene. It is a spacial binder designed for use with polyester resin so that
it does not inhibit curing, but actually redissolves into the resin during molding and
helps the bond.
2.2 Cutting Pattern
Once the panels are cured they are removed from the mold and the edges are trimmed
using a circular tile saw with a diamond blade. The panel is then weighed and
measured so that the percentage of carbon fibers, E-glass fibers, and resin can be
determined. This calculation is made assuming no loss of fiber weight after the
trimming has occurred. The pattern used for samples to be cut from the panel can
be seen in Figure 2.12. The ASTM standards state that at least five samples per
composite should be tested. Six was the chosen number of samples because this
would allow for three measurements of Poissons ratio in each direction (thickness
and width directions). This pattern was chosen because it allows six samples in
the vertical direction and horizontal direction. There are actually nine samples cut
horizontally, this is because there is enough space on the panel so if a sample must be
discarded for some reason there are others to replace it. The problem with this layout
is that all the vertical samples are down the sides of the panel and all the horizontal
samples are down the middle. This is not a good layout because if the properties
in the middle are different to the properties at the edge it may be confused with an
orientation effect.
The pattern is printed out on paper and the sample shapes are cut out of the
paper to leave a master template. This template is placed over the panel and each
sample is numbered with a silver paint marker. Consistent numbering of the samples
allows comparison of results with the knowledge of where it was cut from on the
panel. With this information position influences can potentially be investigated. The
Autocad file of the pattern is sent to Aquashear along with the panels. They line up
the cutter in the same position as the template was lined up and the samples are cut
and shipped ready for testing.
Figure 2.12: Water Jet Cutting Pattern for Panels
2.3 Test Procedures
In both types of tests the specimen must first be measured. The critical measurements
are the thickness and width of the specimen. The dimensions are important in order
to get accurate results due to the fact that some of the material properties rely on
the cross-sectional area. Before each specimen is tested the width and thickness are
recorded in a text file named either tensile or flexure with the date the experiment is
occurring. Also any abnormal feature noticed in the part is recorded.
2.3.1 Tensile
For the tensile tests, due to the smooth surface finish on the specimens, the metal
grips for the MTS machine sometimes slip when tension is applied. To solve this
problem, emery cloth is placed in the grips which provides enough friction to stop
the slipping. Initially, before a specimen in loaded in the MTS machine the load
cell channel is zeroed because it has a tendency to drift. The MTS machine is set
to a home position that is a good distance for specimen testing. A test specimen is
carefully measured and the width and thickness of the neck region are recorded in
a file along with the specimen and panel number and any observations made during
the experiment run. The specimen is then placed in the grips, care is taken to ensure
good alignment of the specimen. The biaxial extensometer is then placed on the
specimen using the conical point grips. The conical points are placed in the center of
the specimen, either on the edge or on the face depending on the data to be obtained.
Alignment pins are used to ensure the axial extension gauge length will be exactly
one inch. A picture of a specimen to be tested is shown in Figure 2.13. The transverse
extension gauge length is either the initial material thickness or width depending on
which is being monitored. The transverse channel must be zeroed using the signal
conditioning software. Since the alignment pins are used for the axial direction the
axial channel is automatically zeroed. The tensile setup for the TestStar IIs software
is then run and the grips separate at a displacement rate of 12.7 mm/min. After
complete fracture of the sample the test is stopped and no more data is collected.
The specimen is then removed and kept for observation purposes.
2.3.2 Flexure
For flexure tests, the metal grips on the MTS machine are replaced by a three point
bend test jig. According to the ASTM standard, the bottom supports of the three
point bend test jig must be spaced apart 161 times the thickness of the specimen.
Each of the pedestals must be set equally from the center of the jig so the load is
applied at the center of the supports. Care must be taken to ensure that each of the
cylindrical loading noses are aligned. The specimen is measured and the thickness and
width are recorded in a file with the specimen and panel number. The support span is
recorded in this file and any observations of the experiment are recorded as well. The
Figure 2.13: Specimen Loaded and Ready for Tensile Test
specimen is centered on the bottom pedestals and the top pedestal is hydraulically
lowered to just above the surface of the specimen. A picture of a specimen ready to
be tested is shown in Figure 2.14. The flexure setup for the TestStar IIs software
is then run and the center loading nose is lowered at a rate of 15.2 mm/min. The
program is stopped after significant breakage in the specimen occurs. The loading
nose is then raised and the specimen is then removed and replaced by the next test.
Figure 2.14: Specimen Loaded and Ready for Flexure Test

2.4 Synopsis
This chapter covers the selection of the size and shape for advanced composite test
coupons using ASTM standards for tensile and flexure testing. In order to create
these composite samples two separate resin transfer molding procedures were tested.
The first involved molding each sample individually with the use of a small aluminum
mold. The mold cavity consisted of two locations for the coupons to be molded to
their final desired shape. The second method of coupon production involved using a
larger steel mold to manufacture a large panel out of which test coupons were cut. The
second method proved to be superior due to more efficient test coupon production.
The tensile and flexure test procedures were outlined at the end of this chapter. The
next chapter will discuss how the collected data was analyzed.
29
Chapter 3
Data Processing and Analysis
This chapter describes the analysis and processing of the large amount of data col-
lected during the research. For tensile tests, the collected data includes the elapsed
time, load, axial extension, transverse extension, and grip displacement. The load
and extension data are critical in the analysis whereas the time and displacement
data are collected in case they can be used in the future. For flexure tests the mon-
itored variables are elapsed time, load, and load-nose displacement. Again load and
displacement are critical for analysis and time data is just collected for completeness.
For both flexure and tensile tests, a Matlab script reads in the data file with the
specimen number and specimen sizes and stores this data into an array. The script
then reads in the necessary data columns from the file and stores the information
in cell structured arrays. Each experiment is then scanned for the maximum load
data point, this is recorded in an array and another subset of arrays is created with
just the data prior to the peak load point. The relevant data can then be calculated
and the results are discussed in the Tensile Data3.1.1 and Flexure Data3.1.2 sections.
After the calculation of data, the results for each panel are inserted into the Microsoft
Access database for convenient storage and further analysis.
CHAPTER 3. DATA PROCESSING AND ANALYSIS 30
3.1 Data Processing
3.1.1 Tensile Data
The properties that must be calculated in the tensile data scripts are Youngs modulus,
maximum tensile strength and Poissons ratio in all directions. Initially, arrays of
stress are created for each test by dividing the applied load by the specimens cross-
sectional area, and arrays of strain by dividing the extensometer displacements by
the initial gauge lengths. The maximum tensile stress data in this thesis is taken as
the maximum from the stress array. Typically Youngs modulus would be found by
plotting a stress verses strain curve and reading the slope of the initial linear portion
of the curve. To simulate this the script takes the stress and strain data and fits a
first order polynomial to the linear portion and the slope that is calculated is Youngs
modulus for that test. The Poissons ratios are calculated in a similar manner except
they use the strain data in two directions to calculate the relationship as in equation
3.1. It can also be seen in equation 3.1 that Poissons ratio in one direction can be
determined from the ratio in the two other directions. In this research the 23 direction
is found from the 12 and 13 directions. The directions are defined in Figure 3.1.
j
ij = where : i 6= j
i
also : (3.1)
kj
ij =
ki
Where:
x is the strain in the x
xy is Poissons ratio in the xy direction
Figure 3.1: Coordinate System
3.1.2 Flexure Data
The material properties that can be determined from the collected flexure data are
the maximum flexural stress, the flexure modulus, and the flexural strain at the
maximum flexural stress. The load and displacement data of each of the experiments
is stored into an array. Once the load and displacement data is recorded the loads are
converted to stress and displacements are converted to strain both at the outer surface
of the specimen where the values are at their highest. The stress is calculated from
equation 3.2 and the strain is calculated from equation 3.3. The maximum flexural
stress can easily be obtained from the array of stresses. The flexural modulus can be
obtained from the slope of the stress verses strain relation as in the tensile properties.
3P L
f = (3.2)
2bd2
Where:
f is the flexural stress of the outer surface at the midpoint of the specimen,
P is the applied load to the load nose,
L is the support span,
b is the width of the specimen,
d is the depth of the specimen.
6Dd
f = (3.3)
L2
Where:
f is the maximum flexural strain at the outer surface of the specimen,
D is the maximum deflection (at the midpoint) of the specimen.
There is a problem with using equations 3.2 and 3.3 for determining stress and
strain. These equations are based on a beam theory that uses the assumption that the
beam is isotropic, homogeneous and of equal tensile and compression strengths [18].
From [19] there are assumptions made on the beams normals, which are imaginary
lines that are perpendicular to the neutral plane (the doted line in Figure 3.2). The
assumption is that the normals remain straight, unstretched, and remain normal. In
bending, one side is in tension and one is in compression, if the material does not
respond the same way in compression as in tension the normals would not remain
straight. Also with a sandwich type structure the elasticity of the material is likely
to be different from one layer to another causing the normals to not be straight as
seen in the top of Figure 3.2.
These assumptions do not necessarily eliminate the usefulness of flexure tests on

composites. If the effects are small enough they can be considered insignificant, in
Figure 3.2: Composites Compared to Isotropic Materials for Beam Theory
fact that is the case with isotropic materials. As the composite moves further away
from isotropy the worse the assumptions become. Flexure tests are still used in the
literature [10, 16, 20, 21] and are useful for comparing properties of composites with
varying input factors.
3.1.3 Database Design
After the experimental results are analyzed and the material properties are deter-
mined the data must be stored in an easily accessible place. To this end, Microsoft
Access was chosen as a database program to store the relevant data. Access records
were created to store all the tensile and flexural data from the test coupons. A record
is also created with the average results to make an estimate for each of the properties
of the original panels. The benefit to using a database program such as Access is
that queries can easily be made up in order to sort through all the data and search
for whatever criteria the user desires. These query results can be used by Matlab or
Excel and ANOVA can be performed on the necessary data.
3.2 Design of Experiments
Design of experiments (DOE) is a combination of statistical techniques used to choose

the values at which the desired experimental factors will be run, give rules on how to
run the experiments, and then how to analyze the results. DOE is often used when
optimization of a problem is needed. One can determine sensitivities to changes in
factors and use DOE for estimating the optimal factor combination from a few tests.
In this project an optimum is not of interest because there is nothing specific to
optimize. One can use DOE to estimate the sensitivities from a small subset of ex-
periments to determine if all factor combinations must be run, or if some combinations
can be eliminated. For good references and examples on DOE see [2226].
DOE also provides a method of determining whether or not a factor has a signif-
icant effect on the response. When looking at raw data sometimes it is difficult to
determine what amount of the changes are random error and what is a result of a
factor changing. Analysis of variance is a tool to evaluate this problem and will be
discussed in the next section.
3.2.1 Analysis of Variance
The core of the DOE analysis is called ANOVA, which is short for analysis of vari-
ance. This can be used for determining the effect of a single factor or the effect of
multiple factors and their interaction effects. The variance of a population is defined
as equation 3.4.
P
2 (X )2
= (3.4)
N
Where:
is the population mean,
N is the size of the population, and
X is the individual population values.
In order to use equation 3.4, the whole population must be tested. As this is
often infeasible, a small sample of the population must be taken and the mean can
then be estimated with the sample average. The standard deviation of the sample is
an unbiased approximation of the variance of the whole population and is given by
equation 3.5. P
2 (X x)2
s = (3.5)
n1
Where:
x is the sample average,
n is the size of the population, and
X is the individual population values.
In equation 3.5 one is subtracted from the sample size n in the denominator in
order to make it an unbiased estimate. This is a well examined problem and is
explained in many statistics books such as [27].
ANOVA uses the variance of a response from changing a factor and compares it to
the variance of the error of the experiment. Using this comparison one can determine
if a change in the response is due to the factor the experimenter is trying to test or
just due to error in the system. ANOVA is based on equation 3.6. The term (i)j is
normally and independently distributed with an average of 0 and a variance 2 . This
is one reason design of experiments states that the experiments should be run in a
random order. If the experiments are not run in a random order the error will not
have a normal distribution.
yij = + Fi + (i)j (3.6)
Where:
yij is the response of the ith factor level and j th experiment repeat of that factor level,
is the overall mean,
Fi is the effect factor that was varied has on the mean,
(i)j is the random error of the j th repeat in the ith factor level. [23]
The variance is used to check the closeness of the data to this assumption. First
the factor sum of squares (SSF) is found using equation 3.7 and then the error sum
of squares (SSE) is found using equation 3.8.
I J
1 XX
CT = ( Xij )2 (3.7)
IJ i=1 j=1
XI J
1 X
SSF = ( Xij )2 CT
i=1
n j=1
I X
X J
SST = Xij2 CT (3.8)
i=1 j=1
SSE = SST SSF
In equation 3.7 the correction term CT is introduced. It allows for the deviation
around the mean and is given a special designation because it is used for calculating
the each sum of squares. In equation 3.8 the SST term is introduced, this is the total
sum of squares. Using this notation the equations can be expanded to a multiple
factor ANOVA which is demonstrated in section 4.2.2 of this thesis.
After the sum of squares are found, the mean squares can be calculated by dividing
the sum of squares by the degrees of freedom minus 1. This is displayed in equation
3.9, where MSF is the factor mean squares and MSE is the error mean squares.
SSF
M SF = (3.9)
I 1
SSE
M SE =
IJ 1
The ratio of the factor mean squares and error mean squares can then be compared
to the appropriate F value which is discussed in the next section. If the ratio is larger
than the F value the factor is significant. This means the variation in data when
the factor level is changed is larger than the variation in error throughout to whole
experiment.
3.2.2 Multiple Factor Analysis of Variance
When dealing with more than one factor in an analysis of variance, instead of an-
alyzing each factor individually the data can be analyzed with multiple factors at
once. This is beneficial because it can bring to light interaction effects. Interaction
effects let the experimenter know how a change in one factor effects the response to a
change in another factor. So if one factor interacts with another then when one factor
is increased the response when the other factor is changed will change at a different
rate.
The principles for the the multiple factor ANOVA is the same as for the single
factor version. The sum of squares is calculated for each factor effect as well as the
sum of squares for the interaction effects. These are then divided by their respective
degrees of freedom to obtain the mean squares. The mean squares of an effect is
divided by the mean squares of error to obtain the F value to determine significance.
An extensive example is given in section 4.2.2 of this thesis.
3.2.3 F Value
The F distribution is a probability distribution that is based on two different degrees

of freedom. This is a good distribution for comparing factor effects with error effects.
The F value is determined by the two separate degrees of freedom and by the level
of certainty . F can be found in tables or can be calculated from the ratio of two
chi-squared variables given in equation 3.10 [27].
1 /1
F = (3.10)
2 /2
Where:
is the number of degrees of freedom,
is the chi-squared distribution.
3.2.4 Confidence Interval
The spread in raw data of any specific material property examined in this thesis
is large relative to most isotropic materials due to the macroscopic differences in
material composition of composites. Therefore one cannot give average of the tests
as the definite value of that property for a given panel. It is much more accurate to
give a range in which the property value will fall. This range is centered around the
average and the size of the range is based on the spread of the data, or the standard
deviation. This is what the confidence interval is used for.
With a confidence interval the experimenter can specify a certainty level and
using the sample average, standard deviation, and the students t distribution can
determine a range in which the actual mean will fall. To compute the confidence
interval equation 3.11 is used.
x t, sx < < x + t, sx (3.11)

s
sx =
n
Where:
t, is the students t distribution with
degrees of freedom and a certainty level of [23]
In this thesis, Poissons ratio in the 23 direction is not measured directly. It is
calculated from the 12 and 13 directions. In order to determine the confidence interval
for the 23 direction since the confidence interval cannot be determined directly, error
propagation tools must be employed. For both addition and subtraction the intervals
can be added directly. For multiplication and division the absolute error must be
added. This involves dividing the interval by the mean as seen in equation 3.12 [28].
s 2 2
x y
z = z + (3.12)
x y
3.2.5 Box Plots
Box plots can offer a quick visual comparison of data sets. Some trends can be easily
determined from the plot. The definition of a box plot is taken from [27,29] and is as
follows: In a box plot, each data set is represented by a box with its upper and lower
edges on the one quarter and three quarters percentile of the data. A line through
the box represents the median of the data. Lines extending from the box represents
the upper and lower limit of the data. Outliers, or data that is outside of the upper
and lower limits are marked with an asterisk. The quarter percentiles are the median
of the largest and smallest n/2 observations, or (n+1)/2 if there is an odd number
of observations. The upper and lower limits are a distance of 1.5 times the distance
between the one quarter and three quarters percentile away from the box ends. In
section 4.1.3 box plots are used to observe the effects of orientation on the composite
properties.
3.3 Methods of Modeling
In the results section of this thesis, two methods of modeling are used in an attempt to
create a way of predicting material properties. This section outlines the theory behind
the models used in the results section. There are many published works on modeling
of composite materials, [7, 9, 3033] many are based on the rule of mixtures method
and are just expansions on the theory using more information on the composite and
its constituents than is known for the materials in this thesis.
3.3.1 Curve Fitting
Curve fitting can be performed on the experimental data to obtain an empirical

equation for the material properties. To do this the results were set up in a matrix
form in Matlab and solved by using the method in equation 3.13. These curve fitting
methods were taken from [3436].
Y = 0 + 1 X1 + 2 X2 + ... + n X2m (3.13)
Where:
n are coefficients to be determined
Y is the response value
X1 and X2 are the factor input levels
Equation 3.13 can be written in matrix form, as in equation 3.14, to be used with
many input and response levels.

y1 x11 . . . x1m 1

.. .. .. ..
. = . ... . . (3.14)

yn xn1 . . . xnm m
The matrix equation 3.14 can also be written in the form of equation 3.16. By
simply multiplying by the inverse of the factor input levels matrix the vector of
coefficients can be found (equation 3.16). Unfortunately this only works for if X is a
square, invertible matrix. If this is not the case Matlab uses a least squares approach
to solving the system of equations [36].
Y = [X] (3.15)
= Y [X]1 (3.16)
3.3.2 Rule of Mixtures
The rule of mixtures uses the separate phases that go into creating a composite in
order to estimate the composite properties. It is a very basic model and is good
for a very simple approximation of some material properties. This method works on
the theory that the composite properties are simply the constituent properties added
together, each constituent contributing in portion to their volume fraction. Without
any modifiers this theory is for continuous unidirectional fibers. The composites used
in this thesis contained chopped and randomly oriented fibers of more than one fiber
type so the equation must be modified slightly.
Short fibers, for the purpose of modifying the rule of mixtures are defined as any
fiber less than the fiber critical length [33]. The critical length is the minimum length
a fiber must be in order for the stress to be fully transferred from the matrix to the
fiber and thus to be considered continuous in the rule of mixtures model. This length
can be estimated using equation 3.17.
lC f
= (3.17)
d 2
Where:
lc is the critical fiber length
d is the fiber diameter
f is the maximum fiber stress
is the interface shear stress
Since the fibers are not all aligned in one direction each of the fiber components
must be multiplied by a factor K in order to compensate for the random orientation
[30]. This K factor must be determined experimentally. Thus the rule of mixtures
model for the composites used in this thesis is 3.18 for elastic modulus and 3.19 for
the maximum tensile stress.
EComposite = Kc Vc Ec + Kg Vg Eg + Vm Em (3.18)
0
cu = Ksc Vc c + Ksg Vg g + Vm m (3.19)
Where:
V is the volume fraction of a phase,
E is the Youngs modulus of a phase,
Ks is a correction factor for random orientation fibers.
is the ultimate strength of a phase, The subscripts are:
c is for the carbon phase,
g is for the E-glass phase,
m is for the matrix Phase,
Composite is the final product of the phases in composite form.
0
m is the matrix strength at the failure strain of the fibers. [32]
3.4 Synopsis
This chapter outlines the theory behind the design and analysis of of the experiments.
After the data was collected it was analyzed using a Matlab script and the relevant
data was stored in an Access database. Significant effects can be determined numeri-
cally using analysis of variance or visually by plotting the data using a box plot. Two
models are presented in this chapter: The rule of mixtures is a simple and well known
model that uses the properties of the composites constituents to predict the final
composite properties. The curve fitting models are soley based on the experiment
results. The curve fitting method is beneficial because it can be used on all the mate-
rial properties. The rule of mixtures is used for only tensile modulus and maximum
tensile stress. The next chapter presents the actual results of the experiments.
45
Chapter 4
Experimental Results
The primary concern with the planned experiment when the panels came back from
the waterjet cutter was that a few of the samples were not labeled or the numbers were
almost completely cut off due to poor panel alignment when the cutting occurred.
Thus, the original position the of sample on the panel was not certain. The only
panels that were not possible to decipher the partial and non-existent numbers were
panels 8 and 10. Some of the numbers were visible on panel 10 but none of the
numbers were visible for panel 8. In order to make an educated guess as to where
the samples came from on the panel, a couple of graphs were created. These graphs
can be seen in Figure 4.1, they are plots of sample positions verses sample weights
and thicknesses. As seen in the graphs there is a clear pattern as the specimens
get toward the middle of the panel they get thicker and more heavy. Using this
information, position numbers were chosen for the samples without numbers.
A few samples were lost somewhere during the cutting and were not shipped with
the other samples. Of the flexure coupons these were samples 8 and 14 from panel
10 and samples 7, 8, and 13 from panel 3. Also sample 22 from panel 6 of the tensile
coupons had been misplaced. The only problem with the missing samples is that
CHAPTER 4. EXPERIMENTAL RESULTS 46
Figure 4.1: Variation of Weight (Top) and Thickness (Bottom) with Panel Position
the panels with missing coupons will not have as many data points to calculate the
average, which makes it slightly less likely to be accurate.
After running the tests, there were two samples that produced nonsensical results
due to a malfunction in the bi-axial extensometer. The off axis extensometer did not
record a displacement change for sample 26 of panel 7 and therefore Poissons ratio
could not be determined for that sample. For sample 6 of panel 7 the pins for the
axial extensometer were left in, producing results in which only the maximum tensile
stress can be used due to the fact the other properties rely on the axial extension
measurement.
4.1 Results
The properties of the constituents used in the composites are listed in table 4.1.
These properties are used throughout the results section. The densities are used for
calculating volume fractions and the modulus and strength properties are used for
the modeling calculations. This table shows that carbon fiber is about half the weight
of E-glass while it is about three times as stiff and 1.4 times stronger in tension, but
E-glass is less brittle and breaks at a strain twice as high as carbon.
Phase Density Tensile Modulus Tensile Strength Yield Elongation

[g/cm3 ] [GPa] [MPa] [%]
E-Glass [37] 2.54 72.4 3,448 4.8
Carbon [38] 1.80 230 4,900 2.2
Matrix [39] 1.25 3.4 0.083 5.8
Table 4.1: Phase Properties
Table 4.2 is a table of measurements made prior to the experiments being per-
formed. The panel volume was found by multiplying the panel length by the panel
width and multiplying by the average specimen thickness for that panel. The density
was then found by dividing the panel weight by the volume. The Core+E-glass weight
column is the weight of the Hi Flow mat and the weight of any chopped E-glass added
to the Hi Flow mat is given after the + sign.
Table 4.2 shows that for the most part the core weight stays relatively constant
while the amount of carbon added changes and panel thickness changes.
Panel Core+E- Side 1 Car- Side 2 Car- Panel Specimen Density

Glass bon Weight bon Weight Weight [g] Thickness [kg/m3 ]
Weight [g] [g] [g] [mm]
1 166+30 70 70 842 3.72 1307
2 156+84 94 94 910 3.81 1366
3 160 0 0 802 3.73 1221
4 163 57 55 1261 5.83 1258
5 159 78 78 1361 6.14 1254
6 158 100 105 1232 5.52 1273
7 170 60 66 1129 5.05 1259
8 168 60 69 868 3.85 1307
9 156+30 70 70 933 4.10 1303
10 142+84 94 94 1182 5.10 1317
Table 4.2: Panel Properties
The mass fractions were calculated by dividing the mass of each phase by the mass
of the finished composite panel. The mass fraction for E-glass was found using the
chopped E-glass and the mass of the Hi Flow mat. This is making the assumption that
the polypropylene in the core of the mat can be approximated by E-glass properties.
The total panel weight was measured after the panel edges were trimmed because
there is excess resin on the panel after molding. There is a chance that some of
the fibers were cut out which would make the mass fractions and volume fractions
slightly inaccurate. Regardless, it is not believed that this would change the values by
a significant amount. The fiber mass fractions are relatively low in these experiments.
The low viscosity of the hybrid resin allows for fiber mass fractions up to 75% (using
glass fiber) [39]. The mass fractions are low mainly because it is difficult to stop the
fiber that is added to the Hi Flow mat from bunching and moving during the RTM
process, even with the low viscosity.
Panel Percent E-Glass Percent Carbon Percent Matrix

By Weight [%] By Weight [%] By Weight [%]
1 23.3 16.6 60.1
2 26.3 20.7 53.0
3 20.0 0.0 80.0
4 12.9 8.9 78.2
5 11.6 11.5 76.9
6 12.9 16.6 70.5
7 15.0 11.2 73.8
8 19.3 14.9 65.8
9 19.9 15.0 65.1
10 19.1 15.9 65.0
Table 4.3: Mass Fractions of the Three Phases
The volume fraction is calculated based on the masses of the phases and their
respective densities. The volume of a phase can be approximated by dividing the
mass by the density. The volume of the composite panel is determined by the dimen-
sions of the panel using an average thickness. The volume fraction is then found by
dividing the volume of the respective phase by the total panel volume. The results
are presented in table 4.4.
Panel Percent E-Glass Percent Carbon Percent Matrix

By Volume [%] By Volume [%] By Volume [%]
1 25.1 12.1 62.8
2 26.5 15.7 57.9
3 21.8 0.00 78.2
4 15.1 6.21 78.7
5 14.9 8.00 77.1
6 16.4 11.8 71.8
7 17.9 7.8 74.3
8 20.4 10.8 68.8
9 21.3 10.9 67.8
10 19.9 11.6 68.5
Table 4.4: Volume Fractions of the Three Phases
The next two sections will present and discuss the experimental results.
4.1.1 Tensile Results
The following Figures 4.2-4.6 are plots of the tensile stress verses strain of each of
the samples. The average tensile modulus is represented by the black line. The black
horizontal line represents the average maximum stress level. The red lines on the plots
are the samples in the horizontal direction and the blue lines in the vertical direction
of the panel. The figures all have the same scale in order to easily see differences
between graphs.
These graphs show the variation in the collected data. The graphs also show how
the stress strain curves are highly linear until breakage occurs. All specimens result
in a brittle fracture. Upon visual inspection it appears that the blue lined data in
general might have a higher tensile modulus. In some panels the maximum stress
appears higher in the red lines, and some higher in the blue. These orientation effects
are further explored in section 4.1.3 of this chapter. In the graphs if the curve makes
a sudden jump backward in strain this indicates the extensometer pins slipping on
the specimen surface. Typically this is due to the vibration from the initial cracking
that dislodges the pins from their original seated location.
The effects of adding carbon fiber can be seen by comparing panel 3 to the rest of
the panels. The tensile modulus for panel 3 is the lowest of all panels. The addition of
carbon fiber increases the panel stiffness. The elongation at break is also the highest
in panel 3, which is a property of glass composites. As for maximum tensile stress
the only panel that was lower than panel 3 was panel 4. This is due to the fact only a
small amount of carbon fiber was added, and the thickness was increased. A thickness
increase without much fiber added causes the matrix volume fraction to increase and
the E-glass volume fraction to decrease. As seen in table 4.4, panel 4 has the largest
matrix volume fraction and lowest E-glass volume fraction. Also, the low amount of
carbon added caused some samples to have sections that contained no carbon on one
side, thus essentially making a local volume fraction of carbon equal to zero and thus
lowering the composite strength.
7 Panel 1 7 Panel 2
x 10 x 10
16 16
14 14
12 12
10 10
Stress [Pa]
Stress [Pa]
8 8
6 6
4 4
2 2
0 0
0 0.005 0.01 0.015 0 0.005 0.01 0.015
Strain [m/m] Strain [m/m]
Figure 4.2: Tensile Stress-Strain Plots for Panels 1 and 2

7 Panel 3 7 Panel 4
x 10 x 10
16 16
14 14
12 12
10 10
Stress [Pa]
Stress [Pa]
8 8
6 6
4 4
2 2
0 0
0 0.005 0.01 0.015 0 0.005 0.01 0.015
7 Panel 5 7 Panel 6
x 10 x 10
16 16
14 14
12 12
10 10
Stress [Pa]
Stress [Pa]
8 8
6 6
4 4
2 2
0 0
0 0.005 0.01 0.015 0 0.005 0.01 0.015

7 Panel 7 7 Panel 8
x 10 x 10
16 16
14 14
12 12
10 10
Stress [Pa]
Stress [Pa]
8 8
6 6
4 4
2 2
0 0
0 0.005 0.01 0.015 0 0.005 0.01 0.015
7 Panel 9 7 Panel 10
x 10 x 10
16 16
14 14
12 12
10 10
Stress [Pa]
Stress [Pa]
8 8
6 6
4 4
2 2
0 0
0 0.005 0.01 0.015 0 0.005 0.01 0.015

The following Figures 4.7-4.11 are the strain-strain graphs used for determining
the Poissons ratio for the samples. In the figures below the blue lines are the strain
in the axial direction plotted against strain in the thickness direction (12 ). The red
lines are the strain in the axial direction plotted against strain in the width direction
(13 ).
As seen in the graphs, Poissons ratio is not heavily influenced by changes in

the composites composition. Also, the differences between 12 and 13 are not too
large. In fact, with panel 5 the two averages are almost coincident. As seen in Figure
4.12 13 is more consistent between panels. In fact given the 95% confidence interval
(table 4.6) the range of 0.373 to 0.355 is common for all ten panels. 12 seems to have
more variability between panels. Although with the exception of panel 5 and panel
7 the 95% confidence interval includes the range of 0.422 to 0.408 for 12 . When
23 is calculated from 12 and 13 the 95% interval widens because uncertainty of
a value that is obtained from two experimental values is going to be greater than
that of the individual experimental values. The common accepted value range for
eight of the panels with regards to 23 is between 0.933 and 0.836. Once again panel
5 and panel 7 are outside of this range. It can be seen from graph 4.12 that the
uncertainty of 13 is less than that of 12 . It is not known what would cause this
effect, but it is possibly to do with the surface finishes which are different for the two
directions. When measuring the thickness change the finish is mostly smooth but
has some surface voids either in the direction of the fibers or between the fibers. As
the experiment proceeds the extensometer pins locate these surface voids and cause
erratic effects as the pins slip into the voids. When measuring in the width direction
the surface is that which is left from the water jet cutter. It is more rough than the
resin surface but also it is more consistent.
3 Panel 1 3 Panel 2
x 10 x 10
4 4
3.5 3.5
3 3
Off Axis Strain [m/m]

2.5 2.5
2 2
1.5 1.5
1 1
0.5 0.5
0 0
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9
Axial Strain [m/m] 3
x 10 Axial Strain [m/m] x 10
3
Figure 4.7: Tensile Strain-Strain Plots for Determination of Poissons Ratio of Panels
1 and 2
3 Panel 3 3 Panel 4
x 10 x 10
4 4
3.5 3.5
3 3
2.5 2.5
2 2
1.5 1.5
1 1
0.5 0.5
0 0
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9
3
3 and 4
3 Panel 5 3 Panel 6
x 10 x 10
4 4
3.5 3.5
3 3

2.5 2.5
2 2
1.5 1.5
1 1
0.5 0.5
0 0
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9
3
5 and 6
3 Panel 7 3 Panel 8
x 10 x 10
4 4
3.5 3.5
3 3
2.5 2.5
2 2
1.5 1.5
1 1
0.5 0.5
0 0
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9
3
7 and 8
x 10 x 10
4 4
3.5 3.5
3 3
2.5 2.5
2 2
1.5 1.5
1 1
0.5 0.5
0 0
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9
3
9 and 10
Table 4.5 is a list of the averaged tensile results.
Panel Tensile Maximum 12 13 23 Tensile Strain at

Modulus Tensile [m/m] [m/m] [m/m] Maximum Stress
[GPa] Stress [%]
[MPa]
1 11.9 87.6 0.404 0.356 -1.14 0.937
2 14.9 120 0.460 0.379 -1.21 1.04
3 5.20 48.5 0.417 0.375 -1.11 1.09
4 8.16 47.7 0.395 0.359 -1.10 0.655
5 8.52 55.5 0.367 0.366 -1.00 0.721
6 11.3 74.1 0.388 0.373 -1.04 0.798
7 9.01 55.3 0.346 0.368 - 0.789
0.942
8 11.0 75.2 0.417 0.380 -1.10 0.821
9 11.7 88.9 0.459 0.355 -1.29 0.934
10 13.7 108 0.430 0.368 -1.17 0.991
Table 4.5: Tensile Panel Properties
Table 4.6 is a table of the 95% confidence levels for their respective properties for
each panel.
Figure 4.12: Confidence Interval of Poissons Ratio

Panel Tensile Maximum 12 13 23 Tensile Strain at

Modulus Tensile [m/m] [m/m] [m/m] Maximum Stress
[GPa] Stress [%]
[MPa]
1 (13.2, 10.6) (97.7, 77.6) (0.467, (0.373, (-1.06, (1.04, 0.830)
0.341) 0.338) -0.704)
2 (16.1, 13.7) (130, 111) (0.519, (0.428, (-1.03, (1.15, 0.928)
0.401) 0.330) -0.617)
3 (5.42, 4.97) (51.7, 45.2) (0.436, (0.389, (- (1.16, 1.02)
0.398) 0.360) 0.959,
-0.836)
4 (8.74, 7.58) (50.6, 44.7) (0.422, (0.375, (- (0.707, 0.603)
0.368) 0.343) 0.992,
-0.827)
5 (9.24, 7.79) (61.4, 49.7) (0.389, (0.374, (-1.06, (0.779, 0.663)
0.345) 0.358) -0.937)
6 (12.3, 10.3) (81.1, 67.0) (0.428, (0.402, (-1.07, (0.905, 0.691)
0.347) 0.344) -0.835)
7 (9.70, 8.31) (60.0, 50.7) (0.376, (0.382, (-1.14, (0.869, 0.698)
0.316) 0.353) -0.980)
8 (12.3, 9.70) (81.4, 69.0) (0.454, (0.405, (-1.03, (0.910, 0.731)
0.380) 0.355) -0.798)
9 (12.7, 10.7) (96.5, 81.3) (0.510, (0.380, (- (1.03, 0.838)
0.408) 0.331) 0.933,
-0.617)
10 (14.5, 13.0) (117, 97.7) (0.453, (0.386, (- (1.09, 0.891)
0.407) 0.351) 0.935,
-0.779)
Table 4.6: Confidence Interval for Tensile Properties

Using the data from each successful test, equations 4.1a through 4.1c were obtained
as an empirical model for the material properties listed. Each equation is a special
cubic with each individual factor and their interactions represented, but no squared
or cubed terms are present. The orientation, O, was set as 1 for vertical samples and
-1 for horizontal samples. C and G are the volume fractions of carbon and E-glass
fibers respectively. The matrix volume fraction is not included as a factor because it
is linearly dependant on C and G.
The equations are useful for calculating material properties within the experi-
mental factor range. The model should not be used to calculate properties much
outside the experimental range of the factors. This is know as extrapolation, and
when extrapolating far from observation limits, the model can be very poor as de-
scribed in [27]. One example of this is that the model can be used to determine the
properties of a composite that has a high fiber volume fraction. This would give high
values for each of the properties, but when it came time to make the composite, it
would be impossible to completely wet out the fibers and there would be large dry
spots, causing properties much lower than predicted.
T ensile M odulus :
y =0.46 + 3.69 O + 84.93 C + 21.19 G 31.61 OxC 16.74 OxG (4.1a)
107.70 CxG + 159.15 OxCxG GP a
M aximum T ensile Stress :
y = 49.51 + 24.26 O + 734.39 C + 432.22 G 323.47 OxC 108.32 OxG (4.1b)
1525.04 CxG + 1546.97 OxCxG M P a

Strain at M aximum T ensile Stress :
y = 4.01 0.09 O + 58.34 C + 66.83 G 17.29 OxC + 2.20 OxG (4.1c)
314.79 CxG + 48.42 OxCxG mm/m
Where:
O is the orientation, given a value of 1 for vertical and -1 for horizontal
C is the volume fraction of carbon
G is the volume fraction of E-glass
4.1.2 Flexural Results
The following Figures 4.13-4.17 are the plots from which the flexure data was ob-
tained. Looking at the graphs it is interesting to note that the panels with the lowest
percentage of carbon have very little residual strength after the coupon reaches its
peak stress. Panels 3, 4, 5, and 7 have the lowest carbon percentage and their fig-
ures show a sharp drop down to a very low stress after the peak. The other figures
show, at least some of the time, significant residual strength after failure. Panel 2
has the highest carbon volume fraction and has some of the most significant residual
strengths of any of the samples.
Panel 3 has the lowest average flexural modulus and maximum stress and the
highest strain at maximum stress. This is not surprising, since the outer surfaces are
solely composted of E-glass fibers (and the stitching of the mat). Due to the nature
of the applied load in flexure testing, the maximum stresses and strains occur at the
surface of the specimen. The top surface is in compression and the bottom surface is
in tension and half way through the specimen is the neutral axis where the axial load
is zero. Thus the material at the surfaces carry most of the load. Since carbon fiber
is added to the specimens on the surface, the addition of carbon drastically increases
the rigidity of the material.
In one of the next stages of testing for this project impact of plates using a drop-
weight tester will be performed. With this type of impact test the plate is supported
around its edges and the weight is dropped onto the middle. This loading scheme
is most similar to the flexure testing. The residual strengths seen in the graphs of
the flexure results may be representative of residual strengths after impact. Thus the
higher the carbon volume fraction the stronger a sample will be after initial damage
occurs.
8 Panel 1 8 Panel 2
x 10 x 10
3.5 3.5
3 3
2.5 2.5
Stress [Pa]
Stress [Pa]
2 2
1.5 1.5
1 1
0.5 0.5
0 0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08
Figure 4.13: Flexure Stress-Strain Plots for Panels 1 and 2

8 Panel 3 8 Panel 4
x 10 x 10
3.5 3.5
3 3
2.5 2.5
Stress [Pa]
Stress [Pa]
2 2
1.5 1.5
1 1
0.5 0.5
0 0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08
8 Panel 5 8 Panel 6
x 10 x 10
3.5 3.5
3 3
2.5 2.5
Stress [Pa]
Stress [Pa]
2 2
1.5 1.5
1 1
0.5 0.5
0 0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08

8 Panel 7 8 Panel 8
x 10 x 10
3.5 3.5
3 3
2.5 2.5
Stress [Pa]
Stress [Pa]
2 2
1.5 1.5
1 1
0.5 0.5
0 0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08
x 10 x 10
3.5 3.5
3 3
2.5 2.5
Stress [Pa]
Stress [Pa]
2 2
1.5 1.5
1 1
0.5 0.5
0 0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08

The experimentally obtained values for flexure modulus, maximum stress, and the
strain at maximum stress are presented in table 4.7.
Panel Flexure Modulus Maximum Flex- Flexural Strain at

[GPa] ural Stress [MPa] Maximum Stress
[%]
1 14.6 222 2.18
2 20.8 274 2.00
3 7.28 144 2.24
4 10.3 154 1.78
5 11.7 161 1.65
6 14.9 212 1.94
7 11.7 173 1.80
8 14.5 214 1.90
9 15.9 231 2.05
10 17.5 227 1.70
Table 4.7: Flexural Panel Properties
Table 4.8 lists the 95% confidence interval for the flexural properties along with
the confidence interval for specimen thickness.
Panel Specimen Flexure Modulus Maximum Flex- Flexural Strain at

Thickness [GPa] ural Stress [MPa] Maximum Stress
[mm] [%]
1 (3.82, 3.62) (15.7, 13.5) (237, 206) (2.37, 1.99)
2 (3.91, 3.70) (22.9, 18.8) (296, 252) (2.23, 1.78)
3 (3.86, 3.61) (7.67, 6.89) (157, 131) (2.36, 2.11)
4 (6.07, 5.60) (11.1, 9.57) (172, 136) (1.91, 1.65)
5 (6.37, 5.90) (12.7, 10.8) (185, 137) (1.86, 1.44)
6 (5.73, 5.32) (15.8, 14.0) (230, 193) (2.20, 1.67)
7 (5.29, 4.81) (12.5, 10.9) (193, 154) (1.98, 1.62)
8 (4.02, 3.68) (15.7, 13.3) (234, 195) (2.06, 1.75)
9 (4.21, 3.99) (16.8, 14.9) (246, 216) (2.26, 1.84)
10 (5.29, 4.90) (19.0, 16.0) (248, 207) (1.79, 1.61)
Table 4.8: Confidence Interval for Thickness and Flexural Properties
Using all the data from each successful test, equations 4.2a through 4.2c were
obtained as an empirical model for the material properties listed. Each equation is
a special cubic each of the individual factors and their interactions are represented,
but no squared or cubed terms are present. The variables have the same definition
as those in the models in section 4.1.1.
F lexural M odulus :
y =2.56 + 0.09 O + 89.68 C + 18.63 G + 22.91 OxC 1.75 OxG (4.2a)
46.35 CxG 102.97 OxCxG GP a
M aximum F lexural Stress :
y = 2.35 + 144.73 O + 1382.16 C + 647.68 G 1075.72 OxC 715.39 OxG (4.2b)
2914.72 CxG + 5057.17 OxCxG M P a
Strain at M aximum F lexural Stress :
y =6.23 + 10.80 O + 51.16 C + 73.56 G 68.31 OxC 53.30 OxG (4.2c)
330.19 CxG + 328.37 OxCxG mm/m
Where:
O is the orientation, given a value of 1 for vertical and -1 for horizontal
C is the volume fraction of carbon
G is the volume fraction of E-glass
Plastics often have higher flexural properties than tensile properties [37]. For this
reason designers often will use the material properties from flexural testing for design
instead of tensile testing when a part is subject to a more flexural type load. This
increase in flexural properties can be seen in Figures 4.18 to 4.20. The following
figures are plots demonstrating graphically the 95% confidence interval. The average
is displayed as the solid dots and the upper and lower confidence levels are displayed
with a dash. The modulus and maximum stress comparisons demonstrate very similar
trends between flexure and tension tests. However the stain at maximum stress
comparison does not demonstrate much similarity between the two types of tests.
This data would be more useful if the material was a fully isotropic plastic. Due
to warnings given earlier in this thesis the flexural results should not necessarily be
taken as accurate and thus should not be used for design.
Composite Moduli
2.50E+10
2.00E+10
1.50E+10
Modulus [Pa]
Tensile Modulus
Flexure Modulus
1.00E+10
5.00E+09
0.00E+00
1 2 3 4 5 6 7 8 9 10
Panel Number
Figure 4.18: Confidence Interval of Tensile and Flexural Moduli

Composite Maximum Stresses
3.50E+08
3.00E+08
2.50E+08
Maximum
2.00E+08
Stress [Pa]
Tensile Stress
Maximum
Flexural Stress
1.50E+08
1.00E+08
5.00E+07
0.00E+00
1 2 3 4 5 6 7 8 9 10
Panel Number
Figure 4.19: Confidence Interval of The Maximum Tensile and Flexural Stresses
Composite Strain
0.025
0.02
Strain at
0.015
Maximum
Strain [m/m]
Tensile Stress
Strain at
Maximum
0.01 Flexural Stress
0.005
0
1 2 3 4 5 6 7 8 9 10
Panel Number
Figure 4.20: Confidence Interval of The Strain at Maximum Tensile and Flexural
Stress
4.1.3 Effects of Orientation
Based on an ANOVA of each panel for each property, the orientation significance can
be compared to what is seen in each box plot. The ANOVA results can be seen in
table 4.9, the box plots are shown in Figure 4.21 through Figure 4.26 (orientation 1 is
vertical and 2 is horizontal). The gap between the first and third quarter percentiles
is quite substantial in most cases. This makes it impossible to resolve an effect if it
is minor. Panel 3 has some of the smallest data scatter. This is likely due to the
fact it is just the Hi Flow mat which is manufactured with consistent fiber density
and random orientation. When the carbon or glass fiber is added by a human, it can
be clumped together and not uniformly distributed. It is likely that the positioning
of the horizontal and vertical samples on the original panel can be a major cause of
orientation effects. The vertical samples were all at the edge of the panel and the
horizontal samples were in the middle. The middle samples are thicker and with even
fiber distribution that means there is more resin, giving a lower fiber volume fraction,
weakening the sample.
Property Panel with Significant ANOVA for Orientation

Tensile Modulus 3, 9
Maximum Tensile Stress 6, 7, and 9
Strain at Maximum Tensile Stress 4
Flexural Modulus 3
Maximum Flexural Stress 3, 6
Strain at Maximum Flexural Stress 3, 5
Table 4.9: ANOVA Results for Individual Panels

Figure 4.21: Box Plot Orientation Comparison for Tensile Modulus
Figure 4.22: Box Plot Orientation Comparison for Maximum Tensile Stress
Figure 4.23: Box Plot Orientation Comparison for Strain at Maximum Tensile Stress
Figure 4.24: Box Plot Orientation Comparison for Flexural Modulus

Figure 4.25: Box Plot Orientation Comparison for Maximum Flexural Stress
Figure 4.26: Box Plot Orientation Comparison for Strain at Maximum Flexural Stress
From the ANOVA and box plot results panel 3 is the most common to show
orientation effects. This is most likely because it consisted of only the Hi Flow mat.
The E-glass on the mat is a chopped and in a random orientation on both sides of
a core. The core can be seen in Figure 4.27. The chopped glass layers are held to
the core by stitching. This stitching is done about half an inch apart and is only in
one direction. This would add some directionality to the material properties of the
composite. Also some directionality can be seen in the core which is a polypropylene
material that is essentially a system of loops. Thus it readily stretches in one direction
and not in the perpendicular direction. Once carbon is added to the composite,
the carbon properties dominate the composite and most of the orientation effects
disappear or become indistinguishable from the noisy data.
Figure 4.27: Hi Flow Mat Core

4.2 Modeling of Experimental Data
4.2.1 Rule of Mixtures
The fibers used are about 30mm in length. Equation 3.17 is used to calculate whether
this length will effect the rule of mixture because it is too short. The term is un-
known for the fiber and matrix combination used in this thesis but Fu et al. [7] per-
formed experiments on glass and carbon fiber pullouts and found the critical lengths
to be 887.92 m and 813.87 m respectively. Thus the average fiber length is about
33 times the critical length, so even though the experiments were not conducted with
the same matrix, it is safe to assume the majority of the fibers are over the critical
length.
Figure 4.28 shows the results of the rule of mixtures comparison with experimental
results. The comparison for tensile modulus is very good. The experimental results
are shown with the confidence interval included. For the tensile results only the
modeled value for panel 10 was not within the confidence limits of the experimental
data. For the tensile modulus fit the value for Kc was 0.24 and Kg was 0.17. The fit
for the maximum stress properties was not as accurate, the modeled value was outside
of the confidence interval for panels 4, 7, and 10. The values used for Ksc and Ksg
were 0.07 and 0.065 respectively. It is likely that the maximum stress approximations
were poor due to the fact that the rule of mixtures only models one failure mode. The
mode of failure it models is fiber breakage. The composite can fail in other modes
such as fiber pullout or delamination. In fact some combination of these modes are
more likely due to the chopped nature of the fibers. This also explains the very low
value of Ksc and Ksg . The effect of the non continuous fibers is confounded with the
random orientation effect.
Figure 4.28: Comparison of Rule of Mixtures Estimation With Experimental Data

4.2.2 Analysis of Variance
ANOVA can be carried out in order to determine what factors have a statistically
significant effect on the response. To accomplish this at least 4 points must be used,
and they should consist of only 2 levels from each factor such that it makes a simple
factorial design. Figure 4.29 is a scatter plot of the carbon and E-glass volume
fractions of the 10 panels. A factorial design would be represented by a box shape in
the plot. The points that are surrounded by the quadrilateral in Figure 4.29 are close
to a factorial design but the E-glass volume fraction rises a little when the carbon
volume fraction rises. This is a problem because it confounds the effect of E-glass
with that of carbon. When the method examines the effect of carbon alone it is not
accurate because the E-glass is contributing to the effect as well. Thus the effect of
carbon will seem higher than it is in reality. The other problem that is confounded in
the results is that the matrix volume fraction does not stay constant between panels,
or even between samples in the panels due to thickness variation.
The raw data in the following table is the tensile modulus data (in psi) for panels
5, 6, 7, and 10. Only 5 measurements are used for the data points because panel 7
only had 5 measurements. Thus position 6 was omitted from each panel to keep as
consistent as possible.
Volume Fraction Scatter Plot

0.28
2
0.26
0.24
EGlass Volume Fraction
0.22 3
9
8
0.2
Vc=+1 10
Vg=+1
0.18 ANOVA points 7
Vg=1 6
0.16
5
4 Vc=1
0.14
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16
Carbon Volume Fraction
Figure 4.29: Scatter Plot Showing ANOVA Points

Vg
Orientation +1 -1
2118372 1564158
2129324 1825315
+1 2047249 1787900
2309260 1592432
1620189 1122713
+1
1967820 1506336
1901825 2541999
-1 1773726 1501849
2480527 1640292
1974019 1396935
Vc
1330293 1542247
1287782 1185736
+1 1040898 1623855
1819631 1180878
1675320 1171896
-1
1321056 1299836
1233305 1569204
-1 1022998 967003
1338746 1084765
1259663 919582
The model used for the ANOVA is listed below. Each of the interactions is ex-
amined including the three way interaction. Three way interactions are often left
out of ANOVA because their physical meaning is difficult to understand. In this
case, leaving it out does not make much difference to the results so it is left in for
completeness.
Model:
yijkl = + Ci + Gj + CGij + Ok + COik + GOjk + CGOijk + (ijk)l
i = 1, 2 j = 1, 2 k = 1, 2 l = 1, 2, 3, 4, 5
The summation of the responses grouped by each factor (or treatment, T) and
combination are listed in the following tables. Dot notation is used in the tables, a
dot in place of an index means that the responses have been summed over that index.
Totals:
C G O
T1... 36802240 T.1.. 33652003 T..1. 31975448 Total
T2... 25874694 T.2.. 29024931 T..2. 30701486 T.... 62676934
P 2 P 2 P 2 2
Ti... 2.024E+15 T.j.. 1.975E+15 T..k. 1.965E+15 T.... 3.928E+15
i j k
C by G C by O G by O
T11.. 20322311 T1.1. 18116912 T.11. 17378318
T12.. 13329692 T1.2. 18685328 T.12. 16273685
t21.. 16479929 T2.1. 13858536 T.21. 14597130
T22.. 12545002 T2.2. 12016158 t.22. 14427801
PP 2 PP 2 PP 2
Tij.. 1.01964E+15 Ti.k. 1.01381E+15 T.jk. 9.88076E+14
i j i k j k
The following are the equations to obtain the sum of squares (SS) for each of the
treatment conditions.
XXXX 2
T....
2
SST otal = Tijkl
i j k l
n
1 X 2 T2
SSCarbon = SSC = Ti... ....
jkl i n
1 X 2 T2
SSGlass = SSG = T.j.. ....
ikl j n
1 XX 2 T2
SSCxG = Tij.. .... SSC SSG
kl i j n
1 X 2 T2
SSOrientation = SSO = T..k. ....
ijl k n
1 XX 2 T2
SSCxO = Ti.k. .... SSC SSO
jl i k n
1 XX 2 T2
SSGxO = T.jk. .... SSG SSO
il j k n
1 XXX 2 T2
SSCxGxO = Tijk. .... SSC SSG SSO SSCxG SSCxO SSGxO
l i j k n
SSError = SST otal SSC SSG SSO SSCxG SSCxO SSGxO SSCxGxO
The sum of squares are displayed in the table below along with the mean squares
(MS) and the F value which is used to determine whether the treatment is significant.
SS
MS =
df
M SF actor
F =
M SError
Factor df SS MS F Fcrit
Carbon (C) 1 2.98528E+12 2.98528E+12 35.5 4.15
Glass (G) 1 5.35245E+11 5.35245E+11 6.37 4.15
CxG 1 2.33737E+11 2.33737E+11 2.78 4.15
Orientation (O) 1 40574479436 40574479436 0.483 4.15
CxO 1 1.45298E+11 1.45298E+11 1.73 4.15
GxO 1 21869839310 21869839310 0.260 4.15
CxGxO 1 12511642352 12511642352 0.149 4.15
Error 32 2.69038E+12 84074273244
Total 39 6.66489E+12
It can be seen from the above results table that the only significant effects are
those of the addition of carbon and E-glass fibers.
The following table is for the maximum tensile stress, the significant effects are
those of carbon, glass, the carbon and glass interaction, and the glass and orientation
interaction.
Carbon (C) 1 283332644.1 283332644.1 53.9 4.15
Glass (G) 1 87272976.4 87272976.4 16.6 4.15
CxG 1 73327224.1 73327224.1 14.0 4.15
Orientation (O) 1 3059196.1 3059196.1 0.582 4.15
CxO 1 15765313.6 15765313.6 3.00 4.15
GxO 1 24258062.5 24258062.5 4.62 4.15
CxGxO 1 11274192.4 11274192.4 2.14 4.15
Error 32 168180158.4 5255629.95
Total 39 666469767.6
The following table is the ANOVA for strain at maximum tensile stress. It shows
the carbon, E-glass, and E-glass and orientation interaction have significant effects on
the response. The E-glass and orientation interaction effect in the stress and strain
ANOVA tables are most likely due to the anisotropy of the Hi Flow mat. In the case
of panel 10 chopped E-glass was added to the mat to increase the volume fraction.
The other panels contained just the mat with different amounts of carbon and resin.
As the carbon and resin amounts vary the glass volume fraction changes but the
amount of material stays the same. This changes the significance of the orientation
effects of the mat.
Carbon (C) 1 2.51943E-05 2.51943E-05 8.04 4.15
Glass (G) 1 2.11249E-05 2.11249E-05 6.74 4.15
CxG 1 1.24139E-06 1.24139E-06 0.396 4.15
Orientation (O) 1 1.16315E-06 1.16315E-06 0.371 4.15
CxO 1 1.12312E-05 1.12312E-05 3.58 4.15
GxO 1 1.33532E-05 1.33532E-05 4.26 4.15
CxGxO 1 1.29741E-09 1.29741E-09 0.000414 4.15
Error 32 0.000100319 3.13497E-06
Total 39 0.000173628
The following table is an ANOVA for flexural modulus. It shows that carbon
content is highly significant and dominates the effects on response. This is likely due
to the sandwich structure of the composite. The highest portion of the load on a
flexed specimen is carried by the bottom surface in tension and the top surface in
compression. Therefore with the structure of this composite, when carbon is present
it carries the largest of the load which is why it has the highest influence on the
response.
Carbon (C) 1 5.23898E+12 5.23898E+12 43.4 4.08
Glass (G) 1 3.21132E+11 3.21132E+11 2.66 4.08
CxG 1 2.34028E+11 2.34028E+11 1.94 4.08
Orientation (O) 1 20263574695 20263574695 0.168 4.08
CxO 1 6902954961 6902954961 0.0572 4.08
GxO 1 22149273913 22149273913 0.184 4.08
CxGxO 1 4.17386E+11 4.17386E+11 3.46 4.08
Error 40 4.8271E+12 1.20677E+11
Total 47 1.10879E+13
This next table is an ANOVA of maximum flexural stress. The only factor that
has an influence on the response in this case is the carbon percentage. This is likely
due to the same reason as described for flexural modulus.
Carbon (C) 1 477938341 477938341 10.7 4.08
Glass (G) 1 128782836 128782836 2.87 4.08
CxG 1 4319400.021 4319400.021 0.0964 4.08
Orientation (O) 1 56882833.52 56882833.52 1.27 4.08
CxO 1 173261000.5 173261000.5 3.87 4.08
GxO 1 1286092.687 1286092.687 0.0287 4.08
CxGxO 1 76046122.69 76046122.69 1.70 4.08
Error 40 1792666295 44816657.37
Total 47 2711182921
The last table is the ANOVA for strain at maximum flexural stress. This table
determines that none of the factors have a significant effect on the strain response.
This is likely due to the large spread of data. The standard deviations for each panel
were large, up to 32% of the strain. Therefore each of the averages were within one
standard deviation of each other.
Carbon (C) 1 3.73113E-07 3.73113E-07 0.0282 4.08
Glass (G) 1 1.6413E-06 1.6413E-06 0.124 4.08
CxG 1 1.04394E-05 1.04394E-05 0.790 4.08
Orientation (O) 1 0.000136983 0.000137 10.4 4.08
CxO 1 4.54899E-05 4.54899E-05 3.44 4.08
GxO 1 8.60582E-06 8.60582E-06 0.651 4.08
CxGxO 1 8.21139E-09 8.21139E-09 0.000621 4.08
Error 40 0.000528897 1.32224E-05
Total 47 0.000732437
4.3 Synopsis
In this chapter the results were presented in graphical and tabular form. The results
were analyzed using analysis of variance and box plots and modeled using the rule
of mixtures technique and a simple curve fitting technique. The amount of carbon
fiber was found to have the highest significance in contributions to the majority of the
material properties. Poissons ratio was not very sensitive to changes in constituent
volume fractions. The orientation effects were difficult to draw conclusions from and
were most likely caused by thickness inconsistency throughout the panels. The next
chapter will draw conclusions on the research performed for this thesis.
85
Chapter 5
Conclusions and Future Work
5.1 Conclusions
In this thesis, mechanical properties of advanced chopped fiber reinforced composite

structures were determined. The empirical models developed in this thesis for elastic
modulus, maximum stresses, and strain at maximum stress provide a design tool for
predicting properties of future chopped fiber composites. The industrial partner can
use equations 4.1 and 4.2 combined with the confidence interval information collected
in this thesis for preliminary design of these novel composite structures.
To produce the data in this thesis, tensile and flexural tests based on ASTM
standards for plastics and composite testing were carried out. To manufacture the
coupons, a resin transfer molding (RTM) process was used. RTM was chosen because
it is an affordable method for manufacturing composite structures. After failure of a
small RTM mold, a large mold was designed to create a large flat composite panel.
After the panel was molded, it was sent to be waterjet cut using a template of 6
vertical tensile and flexural samples and 9 horizontal tensile and flexural samples.
CHAPTER 5. CONCLUSIONS AND FUTURE WORK 86
The experiments were performed using an MTS testing machine and the data was
acquired using the load cell, the displacement sensor, and in the case of the tensile
tests, a bi-axial extensometer.
The tensile tests were conducted successfully. There was a large data spread, but
this is to be expected with chopped fiber composites. Estimates were obtained for
tensile modulus, maximum stress, the strain at maximum stress, and Poissons ratio
in all three directions. From the data collected, it can be seen that tensile modulus
and maximum tensile stress respond very similarly to the amount of carbon or glass
fibers added to the composite. Carbon fiber has the most influence on the modulus
and stress responses. The tensile strain, on the other hand, does not follow the same
trends. The reason for this can be seen from the ANOVA in that the F values for the
effect of carbon and E-glass are very close. In the case of maximum stress and tensile
modulus the volume fraction of carbon dominates the response. Thus if the carbon
volume fraction increases the response will increase essentially independently of the
amount of E-glass. With the F values close, if the carbon volume fraction increases
and the E-glass volume fraction decreases it may have no effect on the response. It was
difficult to conclude on Poissons ratio. The values of Poissons ratio were very close
between panels. All the averages were within one standard deviation of each other.
The tensile modulus and maximum stress response to changes in the constituents
were validated and are in agreement with the rule of mixtures.
The flexural tests were also conducted successfully. However, the results found
for material properties using the flexure testing methods can not be used for design
values. This is due to a flaw in the method of calculating the stresses and strains, and
the composites do not fulfill all the basic assumptions of the beam theory used. The
flexural modulus, maximum flexural stress, and the strain at maximum flexural stress
were estimated using the load and displacement data obtained experimentally. These
material properties are found by graphing the stress verses strain calculated from
the load and displacement data. The flexural modulus and stress averages followed
the same basic trends as tensile modulus and stress from panel to panel, once again
reinforcing the results.
The important results from the flexural testing was the residual strengths experi-
enced with the addition of carbon fiber. This almost plastic behavior of the flexural
specimens is a important finding in terms of future impact testing. These findings
forecast the excellent impact damage tolerance capabilities of carbon fiber reinforced
composite structures. Thus, based on this thesis, more work should be performed on
the failure mechanics and determining the levels of residual strength to fully under-
stand this phenomenon.
Each panel had samples cut from it both horizontally and vertically. This was
so that any orientation effects would be seen, if existent. With the variability in
properties for each sample, it is generally impossible to determine, with certainty, if
there is an orientation effect due to the large data scatter. Generally, the means are
different in the two orientations, but the for most panels it can not be said the the two
orientations are statistically different. The clearest indication of an orientation effect
is related to panel 3. This panel has a low spread of test results likely due to the fact
that it is just the prefabricated mat and no carbon or E-glass is incorporated in the
composite. The mat, though, is stitched it one direction only and this likely is the
reason for the dependence on orientation. Other orientation effects that were observed
are likely due to thickness inconsistencies in the panel. The horizontal samples were
on average thicker than the vertical samples, causing a lower fiber volume fraction
and therefore a weaker specimen.
With the inclusion of confidence intervals it is possible to identify more accurate

factors of safety for designing composite structures. In order to decide on a factor of
safety there should be no outlying specimens. To accomplish this much work needs
to be done on the manufacturing side in order to guarantee consistent samples. An
improvement in factor of safety leads to reduced material usage and cost.
5.2 Future Work
The principal focus of the present thesis was to design and implement the experimental
capabilities to test hybrid composites in a laboratory setting. Work is currently being
undertaken on modeling of resin flow in the RTM process. This will allow for more
consistent molding, leading to more reliable confidence intervals so that a lower factor
of safety can be used in designing composite structures.
This work will be expanded in the future to include more comprehensive tests.
Further testing includes the influence of temperature on the hybrid composite. Also of
importance is the response of the advanced composites to fatigue loading. Typically
composites have very good fatigue characteristics but it is important to see how the
varying carbon and E-glass percentages effects fatigue life. Further down the road,
tests on impact of composite structures created using the production method studied
in this thesis must be performed.
89
References
[1] L.S.Sutherland and C.Guedes Soares. The effects of test parameters on the im-
pact response of glass reinforced plastic using an experimental design approach.
Composites Science and Technology, 63:118, 2003.
[2] A. Mills. Automation of carbon fibre preform manufacture for affordable

aerospace applications. Composites Part A: applied science and manufacturing,
32:955962, 2001.
[3] R. McIlhagger D. Abraham, S. Matthews. A comparison of physical properties

of glass fibre epoxy composites produced by wet lay-up with autoclave consol-
idation and resin transfer moulding. Composites Part A: applied science and
manufacturing, 29A:795801, 1998.
[4] N.S.McDevitt I.P.Datar D.Kim M.G.Jenkins M.Ramulu, P.B.Stickler. Influence

of processing methods on the tensile and flexure properties of high temperature
composites. Composites Science and Technology, 64:17631772, 2004.
[5] George Marsh. Composites on the road in the big time. Reinforced Plastics,
47(2):3347, 2003.
[6] V.M. Fonseca S.N.Monteiro J.R.M.dAlmeida C.Z.Paiva Junior, L.H.de Car-

REFERENCES 90
valho. Analysis of the tensile strength of polyester/hybrid ramio-cotton fabric

composites. Polymer Testing, 23:131135, 2004.
[7] E.Mader C.Y.Yue X.Hu Y.-W.Mai S.Y.Fu, B.Lauke. Hybrid effects on

tensile properties of hybrid short-glass-fiber-and short-carbon-fiber-reinforced
polypropylene composites. Journal of Materials Science, 36:12431251, 2001.
[8] B.Lauke S.Y.Fu. Characterization of tensile behaviour of hybrid short glass

fibre/calcite particle/abs composites. Composites Part A, 29A:575583, 1998.
[9] K.Joseph S.Thomas P.Pradeep P.V.Joseph, G.Mathew. Mechanical properties

of short sisal fiber-reinforced polypropylene composites: comparison of experi-
mental data with theoretical preditions. Journal of Applied Polymer Science,
88:602611, 2003.
[10] B.C. Mitra-R. Jacobson R. Rowell A.N. Banerjee A.K. Rana, A. Mandal. Short
jute fiber-reinforced polypropylene composites: Effect of compatibilzer. Journal
of Applied Polymer Science, 69:329338, 1998.
[11] American Society for Testing and Materials. ASTM D3039 Standard Test Method
for Tensile Properties of Polymer Matrix Composite Materials. ASTM, West
Conshohocken, USA, 2000.
[12] American Society for Testing and Materials. ASTM D638 Standard Test Method
for Tensile Properties of Plastic. ASTM, West Conshohocken, USA, 1999.
[13] American Society for Testing and Materials. ASTM D790 Standard Test Methods
for Flexural Properties of Unreinforced and Reinforced Plastics and Electrical
Insulating Materials. ASTM, West Conshohocken, USA, 1999.
REFERENCES 91
[14] Composite Materials Handbook. Department Of Defense, volume 1. polymer

matrix composites guidelines for characterization of structural materials edition,
2002.
[15] A.C.Caba A.C. Loos S.D.Moechnig, T.A.Bullions. Mechanical properties of

feather fiber glass fiber polypropylene composites. In ANTEC 2003 Plastics:
Annual Technical Conference, Volume 2: Materials, volume 2, pages 21322136.
Society of Plastics Engineers, 2003.
[16] C.B.Lee T.M.Wu. Effect of needle density on the mechanical properties of

fiber-reinforced polypropylene composites. Journal of Applied Polymer Science,
73:21692176, 1999.
[17] Cutting Costs in Short-Run Plastics Injection Molding. Morgan Industries, Inc.,
California, USA, 1989.
[18] T.Kincis Yu.M.Tarnopolskii. Static test methods for composites. Van Nostrand
Reinhold Company, New York, USA, 1985.
[19] eFunda. Euler beam equation: Kinetics. http://www.efunda.com/formulae/solid

mechanics/beams/theory kinematics.cfm, 2004.
[20] D.Bhattacharyya K.Jayaraman. Mechanical performance of woodfibre-waste

plastic composite materials. Science Direct, 41:307319, 2004.
[21] K. Friedrich P. Tsotra. Short carbon fiber reinforced epoxy resin/polyaniline

blends: Their electrical and mechanical properties. Science Direct, 2004.
[22] Davic Bacon. Design of Experiments for Engineers Workshops. Council for
Continuous Improvement, 1992.
REFERENCES 92
[23] R.A.McLean V.L.Anderson. Design of Experiments A Realistic Approach. Marcel

Dekker, Inc., New York, USA, 1974.
[24] Charles R. Hicks. Fundamental Concepts in the Design of Experiments. Holt,

Rinehart and Winston, Inc., New York, USA, 1964.
[25] Peter W.M. John. Statistical Design and Analysis of Experiments. The Macmil-
lan Company, New York, USA, 1971.
[26] M.Rajendran M.Muthukumar, D.Mohan. Optimization of mix proportions of

mineral aggregates using box behnken design of experiments. Cement and Con-
crete Composite, 25:751 758, 2003.
[27] Jay L. Devore. Probability and Statistics for Engineering and the Sciences.
Brooks/Cole Publishing Company, California, USA, fourth edition edition, 1995.
[28] Vern Lindberg. Uncertainties and error propagation part i of a manual on

uncertainties, graphing, and the vernier caliper. http://www.rit.edu/ up-
hysics/uncertainties/Uncertaintiespart2.html, July 2000.
[29] Minitab Inc. Minitab statistical software, release 14 for windows. State College,
Pennsylvania, 2003.
[30] E.Mader C.Y.Yue X.Hu S.Y.Fu, B.Lauke. Tensile properties of short-glass-fiber-

and short-carbon-reinforced polypropylene composites. Composites Part A: ap-
plied science and manufacturing, 31:11171125, 2000.
[31] I.Miskioglu J.P.Konell, J.A.King. Tensile modulus modeling of carbon-filled ny-

lon 6,6 and polycarbonate-based resins. Journal of Applied Polymer Science,
90:17161728, 2003.
REFERENCES 93
[32] G.N.Karam. Effect of fibre volume on tensile properties of real unidirectional

fibre-reinforced composites. Composites, 22(2):8488, 1991.
[33] L.H.Sperling J.A.Manson. Polymer blends and composites. Plenum Press, New
York, USA, 1976.
[34] H. Smith N.R. Draper. Applied Regression Analysis. John Wiley and Sons, Inc.,
New York, USA, second edition, 1981.
[35] Fred S. Wood Cuthbert Daniel. Fitting Equations to Data. John Wiley and Sons,
Inc., New York, USA, second edition, 1980.
[36] The MathWorks Inc. Matlab, the laugage of technical computing, release 12.1.
2001.
[37] George Lubin. Handbook of Composites. Van Nostrand Reinhold, New York,
USA, 1982.
[38] Toray Carbon Fivers America Inc. T700s data sheet.
[39] Reichhold. Atlac itp 31638-00 product bulletin. June 2001.

VITA
Surname: Bumpus Given Names: Scott Robert John
Place of Birth: British Columbia, Canada Date of Birth: July 31st, 1979
Educational Institutions Attended:
University of Victoria 1997 to 2005
Degrees Awarded:
Bachelor in Engineering University of Victoria 2002
Publications:
S. Bumpus, A. Suleman, Experimental Evaluation of Two-Phase Composite

Structural Properties, Proceedings of the 19th Canadian Congress of Applied
Mechanics, Calgary, Alberta, June 1-6, 2003, accepted.
S. Bumpus, Testing of Hybrid Composite Materials, Book of abstracts of the

1st Mechanical Engineering Graduate Research Colloquium, Victoria, British
Columbia, May 18, 2004, accepted.
PARTIAL COPYRIGHT LICENSE
I hereby grant the right to lend my thesis to users of the University of Victoria Library,
and to make single copies only for such users or in response to a request from the
Library of any other university, or similar institution, on its behalf or for one of its
users. I further agree that permission for extensive copying of this thesis for scholarly
purposes may be granted by me or a member of the University designated by me. It
is understood that copying or publication of this thesis for financial gain shall not be
allowed without my written permission.
Title of Thesis:
Experimental Setup and Testing of Fiber Reinforced

Composite Structures
Author:
Scott Robert John Bumpus
Date

Experimental Setup For Fibre Reinforced Composite

Uploaded by

Copyright:

Available Formats

Experimental Setup For Fibre Reinforced Composite

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Experimental Setup For Fibre Reinforced Composite

Uploaded by

Copyright:

Available Formats

Experimental Setup and Testing of Fiber Reinforced

Scott Robert John Bumpus

A Thesis Submitted in Partial Fulfillment of the

c Scott Robert John Bumpus, 2005

Supervisors: Dr. Afzal Suleman

Fiber-reinforced composite structures have seen an increased application in aero-

2 Testing Environment Design and Setup 9

3 Data Processing and Analysis 29

5 Conclusions and Future Work 85

1.1 Factor Ranges . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

4.1 Phase Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

3.1 Coordinate System . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

4.10 Tensile Strain-Strain Plots for Determination of Poissons Ratio of Pan-

Historically a predominantly aerospace material, fiber-reinforced composite structures

Fiber-reinforced composites are composed of two or more materials which, when

1.1 Background and Motivation

The advantage of using an autoclave over an RTM process is that it produces

statistic method based on the Design of Experiments (DOE) was used.

Design of Experiments (DOE) is a method for selecting the best combinations of

1.2 Testing Guidelines

1.3 Scope of the Thesis

Factor Value Range

Table 1.1: Factor Ranges

1.4 Structure of the Thesis

Testing Environment Design and

2.1 Coupon Manufacturing Methods

Resin Transfer Molding is a common composite manufacturing process. The pre-

2.1.1 Small Scale Resin Transfer Molding

port choices are illustrated in Figure 2.2.

Figure 2.4: Mold Ready For Injection

2.1.2 Large Scale Injection Molding

Figure 2.6: Mold for Creating Large Flat Composite Panels

Figure 2.7: Inlet and Outlet for Mold Heating System

Figure 2.8: Resin Mixer and Pump For Hybrid Resin

Figure 2.9: Resin Flow Pattern Through the Composite

Figure 2.10: Water Jet Cutter

Figure 2.11: Preform Container On Scale

2.2 Cutting Pattern

Figure 2.12: Water Jet Cutting Pattern for Panels

2.3 Test Procedures

Figure 2.13: Specimen Loaded and Ready for Tensile Test

Figure 2.14: Specimen Loaded and Ready for Flexure Test

Data Processing and Analysis

3.1 Data Processing

3.1.1 Tensile Data

Figure 3.1: Coordinate System

3.1.2 Flexure Data

These assumptions do not necessarily eliminate the usefulness of flexure tests on

Figure 3.2: Composites Compared to Isotropic Materials for Beam Theory

3.1.3 Database Design

Excel and ANOVA can be performed on the necessary data.

3.2 Design of Experiments

Design of experiments (DOE) is a combination of statistical techniques used to choose

3.2.1 Analysis of Variance

yij = + Fi + (i)j (3.6)

3.2.2 Multiple Factor Analysis of Variance

The F distribution is a probability distribution that is based on two different degrees

3.2.4 Confidence Interval

x t, sx < < x + t, sx (3.11)