Study of Forming of Composite Materials with Abaqus CAE and The Preferred Fiber

Orientation (PFO) Model

Thesis

Presented in Partial Fulfillment of the Requirements for the Degree Master of Science in

the Graduate School of The Ohio State University

By

Yumeng Li, B.S.

Graduate Program in Mechanical Engineering

The Ohio State University

2017

Thesis Committee

Farhang Pourboghrat

Prasad Mokashi

1
Copyrighted by

Yumeng Li

2017

2
 Abstract

The forming of composite materials may lead to fiber angle change during the

deformation. The change of fiber orientations can lead to changes in mechanical properties

of the composite material. Therefore, it is important to know the changes of the fiber

orientations in order to calculate effective material properties of the material. The

constitutive model for obtaining properties of the composite material has been investigated

in this study. Zampaloni originally developed the preferred fiber orientation (PFO) model

in his PhD study, in which he tracked the fiber orientation during the composite forming

process. He showed that PFO model gives more accurate results compared with the

Abaqus/CAE model. The simple tension and shear tests were studied with both Abaqus

simulation software, and analytical calculations. Results from both models were compared,

and it was shown that compared with Abaqus, the PFO model tracks the fiber angle

correctly. The stamping of a hat shape section was also simulated with the PFO model and

simulation results were compared with experiments. A good agreement between simulation

results and experimental results were obtained, which suggests that the PFO model can

predict fiber angle change correctly, and its results are closer to the real forming situation.

ii
 Dedication

I dedicate this work to my parents. Thanks for the love, support and guidance.

iii
 Acknowledgments

Thank you to everyone who has helped me with research in my graduate study.

Thanks to Dr. Hyunchul Ahn for his expertise in composite materials.

Most of all, thanks to my advisor, Prof. Farhang Pourboghrat, for this opportunity and

support throughout my graduate study.

iv
 Vita

December 2015………………. .B.S. Mechanical Engineering, The Ohio State University

2015 to present………………...M.S. Mechanical Engineering, The Ohio State University

Fields of Study

Major Field: Mechanical Engineering

v
 Table of Contents

Abstract ............................................................................................................................... ii
Dedication .......................................................................................................................... iii
Acknowledgments.............................................................................................................. iv
Vita...................................................................................................................................... v
List of Tables .................................................................................................................... vii
List of Figures .................................................................................................................. viii
Chapter 1. Introduction ....................................................................................................... 1
Forming of Composite Materials .................................................................................... 6
Chapter 2. Literature Review .............................................................................................. 9
Modeling Techniques...................................................................................................... 9
Material Characterization Methods............................................................................... 12
Chapter 3. Numerical Methods ......................................................................................... 16
Constitutive Model........................................................................................................ 16
Obtaining Material Properties ....................................................................................... 19
Preferred Orientations ................................................................................................... 21
Implementation into Abaqus ......................................................................................... 25
Analytical & Simulation Discussion ................................................................................. 27
Numerical Validation ........................................................................................................ 32
Numerical Setup............................................................................................................ 32
PFO Validation ................................................................................................................. 41
Validation with experiments ......................................................................................... 45
Conclusions ....................................................................................................................... 55
References ......................................................................................................................... 57

vi
 List of Tables

Table 1. Properties of composite ...................................................................................... 27

vii
 List of Figures

Figure 1. Two types of reinforcement for fibrous composite [3] ....................................... 3

Figure 2. Several types of particulate composite [3] .......................................................... 4
Figure 3. Composition of a laminated composite [4] ......................................................... 5
Figure 4. Match die molding [7] ......................................................................................... 7
Figure 5. Material frame and global coordinate [7] .......................................................... 17
Figure 6.Stiffness summation in PFO model [7] .............................................................. 22
Figure 7.Rotation of PFO to the global coordinate [7] ..................................................... 23
Figure 8.Effects of deformation tensor on elements [7] ................................................... 25
Figure 9.Flow chart of PFO model implemented in Abaqus/CAE [9] ............................. 26
Figure 10. Modulus plot for single layer .......................................................................... 28
Figure 11. Properties of a single layer with 0 degree ....................................................... 29
Figure 12. Young's modulus plot in fiber direction .......................................................... 30
Figure 13. Young's modulus in transverse direction......................................................... 30
Figure 14. Shear modulus ................................................................................................. 31
Figure 15. Dimension of a composite specimen ............................................................... 32
Figure 16. Material properties of the specimen ................................................................ 33
Figure 17. Shell composite section of the specimen ......................................................... 34
Figure 18. Material direction of the specimen .................................................................. 34
Figure 19. Bottom boundary condition of the specimen................................................... 35
Figure 20. Top boundary condition of the specimen ........................................................ 35
Figure 21. Result of the tensile test ................................................................................... 36
Figure 22. Stress vs. strain curve for [45/-45/45/-45/45/-45/45]s..................................... 37
Figure 23. Stress vs. strain curve for [90/0/90/0/90/0/90]s in its fiber direction .............. 37
Figure 24. Stress vs. strain curve for [90/0/90/0/90/0/90]s in its transverse direction ..... 38
Figure 25. Simple shear test .............................................................................................. 39
Figure 26. Calculation of shear modulus .......................................................................... 39
Figure 27. Fiber orientation after deformation of the tension test .................................... 41
Figure 28. Tension test result of fiber angle (degrees) after deformation with PFO model
........................................................................................................................................... 42
Figure 29. Coordinates of deformed element with PFO model ........................................ 43
Figure 30. Result of shear test with PFO model ............................................................... 44
Figure 31. Calculation of fiber orientation for shear test .................................................. 44
Figure 32. Undeformed shape of uniaxial extension test .................................................. 46
Figure 33. Deformed shape of uniaxial extension test ...................................................... 46
Figure 34. Deformed preform at elongation of 25 %........................................................ 47
viii
Figure 35. Three different modes of deformation in experiment...................................... 48
Figure 36. Geometry for upper die ................................................................................... 49
Figure 37. Assembly of the stamping simulation ............................................................. 50
Figure 38. Stress distribution of the sheet (0/90 degrees) ................................................. 50
Figure 39. Result of stamping process of the sheet (0/90 degrees), SDV5 is the updated
fiber angle after deformation in degrees ........................................................................... 51
Figure 40. Stamped hat section from experiment (0/90 degrees) ..................................... 52
Figure 41. Stamping simulation with 45/-45 PFO. SDV5 represents fiber angle
orientations after the deformation in degree ..................................................................... 53

ix
 Chapter 1. Introduction

The interest in using composite materials has been on the rise for the last couple of

decades, since composites offer considerable high strength to weight ratio. The modern use

of composite materials, fiber-reinforced polymer composites especially, could be traced

back to the World War II era, with most applications being rocket motor cases, and high-

performance car bodies such as those for 1950’s Chevrolet Corvettes. Recently, in the

aerospace industry, carbon fiber composite is widely used for its high strength, which could

exceed that of steel, as well as its lower specific density. At first, the costs of the composite

material were higher than traditional metal materials. Because of the higher costs of such

new materials, engineering designers often had to make tough choices, since by using

composite materials, the properties of such materials had to overcome the additional costs

compared with conventional materials. As the industry grew, the cost of composite material

decreased, and it is now commonly seen in consumer goods such as automotive

components, sporting equipment, and even some high-end gaming laptops. Products like

these are the first to utilize composite materials as it is becoming more affordable.

However, the price of those products is still out of reach for average consumers.

Nevertheless, the advantages of using composites are too attractive to ignore, especially

for automotive and aerospace industries, where a material with strength as strong as steel

and weight only a fraction of that of steel is needed. Many automotive components made
1
from composite materials can be seen in the market, as the automotive companies are

striving for higher fuel efficiency. Since the 1960s, the composite market has been growing

steadily at the rate of approximately 15% per year [1].

Composites can be categorized based on the type of reinforcement used in the material.

Fibrous and particulate are two common classes. Each type of reinforcement has its own

unique properties and application purposes, and can be subdivided into specific categories

as following:

Fibrous. A fibrous composite consists of either continuous or chopped (whiskers) fibers

suspended in a matrix material. They can be identified from a geometric viewpoint:

Continuous Fibers. A continuous fiber is characterized as having a very high length-

to-diameter ratio based on its geometry. Usually they are stronger and stiffer than bulk

material. Fiber diameters generally range between 0.00012 and 0.0074 μin (3-200 μm),

depending upon the fiber [2].

Whiskers. A whisker is generally considered to be a short, stubby fiber. It is defined to

have a length-to-diameter ratio of 5< l/d <1000 and beyond. Whisker generally have a

diameter range between 0.787 and 3937 μin (0.02-100 μm)

Composites in which the reinforcements are discontinuous or whiskers can be

produced to have reinforcements that have either random or biased orientations. Material

systems having discontinuous reinforcements are considered single layer composites.

Usually the discontinuities will produce anisotropic material response property, but the

random orientation reinforcements have nearly isotropic material property in most

situations. On the other hand, continuous fiber composites can have either single layer or

2
multiple layers. A single layer composite with continuous fiber reinforcement can be either

unidirectional or woven, and multilayered composites with continuous fiber are commonly

referred to as laminates. Generally, a continuous fiber composite is assumed to have

orthotropic material property. A picture of both types of fibrous composites is shown in

Figure 1.

Figure 1. Two types of reinforcement for fibrous composite [3]

Particulate. A particulate composite has particles suspended in a matrix. Particles can have

random shape, size or configuration. The most common particulate composites in modern

days are concrete and particle board. Two subclasses are included in this category: flake

and filled/skeletal:

Flake. A flake composite usually has large ratios of platform area to thickness

within the flakes, which is suspended in a resin system. (Similar to a particle board).

Filled/Skeletal. A filled/skeletal composite is composed of a continuous skeletal

matrix filled by a second material: such as a honeycomb core filled with an insulating

material.

3
 The material response property of such composite can be either anisotropic or

orthotropic. They can be used in the applications that do not require too much strength for

design purposes. A picture of several types of particulate composite is shown in Figure 2

[3].

Figure 2. Several types of particulate composite [3]

In some applications, the fibers are aligned in certain directions to meet different

strength requirements. The most common way is to use a material called “prepreg”.

“Prepreg” is a general term for a reinforcing fabric composite, which has been pre-

impregnated with a matrix system. The resin system already has proper curing agent, which

means the “prepreg” could be directly laid into the mold with no additional resin. The

“prepreg” resin is active, so it is better to keep the material in a freezer to suppress cure

until it is needed for producing parts. When it is needed, the “prepreg” is taken out of the

freezer and stored in room temperature, then cut into needed sizes, and laid up to a laminate,

in which each lamina is oriented in the appropriate direction.

4
 Figure 3. Composition of a laminated composite [4]

For the uncured laminate to be cured, elevated temperature and pressure are applied.

Raising temperature could cause a chemical reaction to progress at a certain speed, and

entrapped air within the laminate could be driven out by high pressure. The curing process

is commonly done by an autoclave.

There are a lot of composites employing matrix systems such as polyester,

vinylester and epoxy. And those kinds of plastics are known as thermoset, which are

applied to the reinforcing fibers in liquid form. And heat or a chemical hardener is used to

catalyze cross-linking of the polymer chains to harden the resin permanently. Low curing

temperature, a low pre-cured viscosity, high stiffness and stability over a wide range of

temperature are several advantages of such resin. However, it also has numerous

disadvantages. The curing cycle of the resin system can take hours, which hinders the

ability for mass production. Moreover, since thermosets are substantially infusible and

insoluble, damaged components are not going to be fixed easily, and often they need to be

replaced completely. Also, recycling resins is also difficult.

Some of the disadvantages can be overcome by using thermoplastic resin systems.

This type of resin system softens and melts at high temperature and solidifies when cooled.

5
This will reduce the manufacturing cycle to a fraction of time when compared with those

from thermosets. Also, thermoplastics are capable of being repeatedly softened and

reshaped with temperature increase and hardened by temperature decrease. This is

convenient for the repair and joining of components by fusion bonding and thermoplastic

welding techniques [5]. In contrast to thermoset resins, thermoplastics do not have cross-

linking; instead they form a solid structure by an entangled network of amorphous or semi-

crystalline polymer chains. Such a structure gives thermoplastic polymer the ability to

reach large deformations before failure, which leads to larger energy absorption than that

from thermoset polymers. Also, thermoplastics are easily recycled, which benefits waste

reduction [6].

However, thermoplastics also have several drawbacks that are worth mentioning.

Often, thermoplastics need a higher processing temperature than thermosets. Take PEEK

for an example, it requires approximately 340℃ to melt and about 385℃ to process.

Moreover, poor interface adhesion between fiber and resin for thermoplastics lead to

reduced mechanical properties. By adding a binding agent between the interface of fiber

and matrix for consolidation can overcome those problems. Still, the benefits outweigh the

cost for using thermoplastic matrix systems [7].

Forming of Composite Materials

The ability to be heated and reformed several times before the final shape has been a

key feature for the thermoplastic composite. There are a lot of methods for forming

composite materials with such matrix system. Match die molding is one of the most

common methods of processing thermoplastic composites. It is similar to sheet metal

6
forming; instead it is composite material that is acted on by a die of the desired geometry.

The composite is inserted between two dies, and the upper die closes on the composite

material and pushes it into the bottom die, which is a mirror image of the upper die but

slightly larger. When the two dies come together, the composite is pushed into the cavity

that is used to determine the shape and thickness of the composite sheet. The process of

match die molding is shown Figure 4.

Figure 4. Match die molding [7]

In this method, the dies need to be heated to the forming temperature for the

composite material to deform, and then cooled to harden the material to its final shape. In

this method, the heating and cooling of the tools are necessary for forming the required

shape, however, it takes relatively long time for the die to be heated to the forming

temperature and cooled for material to solidify. Therefore, this method is relatively time-
7
consuming. Also, the dies only exert downward force onto each other. For deep drawn part

application, the normal directions may be significantly different from the force application

vector along the die surface. Those differences will make it difficult to control the thickness

of the part and maintain even pressure along the surface. And the final part might not have

uniform thickness or be well consolidated.

Thermo hydro-forming can be utilized to form composite materials. This method

uses heated pressurized fluid to conform the composite material to a punch with the desired

part shape. This method is similar to sheet metal hydro-forming, which also uses

pressurized fluid to conform a metal material to a punch of its final shape. The thermo

hydro-forming for composites was developed and patented at Michigan State University’s

Advanced Materials Manufacturing Laboratory.

The thermo hydro-forming has several advantages compared with match die

molding. The heated and pressurized fluid can be used to keep the composite material at

the forming temperature, which eliminates the need for heating the tool required in match

die molding. Also, the use of fluid is similar to using a bottom die, which serves as a female

die. So only the male die is required and there is a reduced cost for designing new tools

and dies. Moreover, since fluid is used to form the part, the force vector that pushes the

material to the punch is always normal to the material surface. The evenly distributed force

along the material surface can keep uniform thickness of the material and reduce out-of-

plane warping significantly. This hydro-forming method allows forming of deep drawn

parts without the problem of developing non-uniform thickness along the material surface

[7].

8
 Chapter 2. Literature Review

Modeling Techniques

ABD matrix is the most common way to characterize laminated composite material in

classical lamination theory (CLT) [8]. The ABD matrix has an assumption of plane stress,

which makes the 9*9 stiffness matrix to be reduced to a 3*3 matrix. Then it is extended to

a 6*6 matrix to account for the effects of extension-bending, and bending stiffness. The

ABD matrix can be used to make good predictions for laminated composite material after

having been formed. But it has limited accuracy for characterizing forming process of

composite material, since it does not take changes of fiber orientation into account. Strains

produced during the forming procedure can cause the fiber reinforcement to change its

direction, which will change the effective properties of the overall composite material.

Also, the model assumes a perfect bonding between laminates.

The fiber reinforced material can have different mechanical behavior, based on the

type of reinforcement, orientation of the fibers, etc. In an ideal situation where fibers are

truly randomly orientated, the composite material can be assumed to exhibit an isotropic

material property. Zampaloni modeled a simulation of forming hemispherical cups from

random continuous fiber reinforced polypropylene matrix composite material, with no

prescribed directionality during the manufacturing process, using a commercial code

MARC [9]. Only half of the sheet was analyzed by assuming symmetric behavior of the
9
overall part, and a rigid-plastic incremental analysis that used large displacements and an

updated Lagrangian procedure was used for modeling the deformation process. It is found

that although the material could theoretically show isotropic property, it is the actual

manufacturing process that does impart some directionality to the material as it is being

deformed. Also, by Environmental Scanning Electron Microscopy (ESEM) image taken of

an undeformed glass mat fiber reinforced thermoplastic with a random orientation, it is

seen that there seem to be some directionality, especially in x-direction. This makes the

isotropy assumption invalid. Furthermore, fiber orientation after deformation in this model

is not available.

Most of woven composite modeling assume homogeneous material properties that

have been developed based on well-developed mathematical theory, which can reduce

simulation time [10]. However, those models assume orthotropy of woven FRT throughout

the whole deformation process. This means that the fiber directions in woven composite

material stay perpendicular to each other, even after forming procedure. However, this

assumption is not valid for analyzing forming of woven FRT, since fiber orientations

change significantly during deformation. Then some efforts have been made to take fiber

angle changes into account to the constitutive model, which leads to a non-orthogonal

continuum constitutive equation.

For textile composites, Peng and Cao proposed a dual homogenization and finite

element approach to characterize this kind of composite material [11]. The model uses a

combined approach of homogenization method and finite element formulation to predict

the effective nonlinear elastic moduli of textile composites. Since there are various kinds

10
of composite materials on the market, it is very time-consuming to obtain material

properties by experiments, they use the homogenization method to study the material

behavior of a single fiber yarn on a meso-microscopic level, and it is based on the properties

of the constituent phases. Then a unit cell is built to enclose the characteristic periodic

pattern in the textile composites, and the effective nonlinear mechanical stiffness tensor

can be obtained numerically as functions of element strains.

Later, Xue, Peng and Cao propose a non-orthogonal constitutive model for

characterizing woven composites [12]. During thermoforming of woven fabric reinforced

composites, there usually will be large in-plane shear deformation, which induces

additional anisotropy into the composite material. The model makes assumptions of

incompressiblity, the tensile and shear responses in the non-orthogonal material

coordinates are decoupled, and the compressive stiffness is ignored. Based on stress and

strain analysis in both the orthogonal and non-orthogonal coordinates, as well as the rigid

body rotation matrices, the model can make a good prediction with respect to the

experimental data. However, this model still needs to consider effects of temperature and

viscous behavior of the resin.

Then, Cao et al. proposed a modeling approach to include the temperature effect in

thermo-stamping of woven composite materials [13]. Since thermo-stamping process has

relatively higher efficiency than autoclave stamping forming process, it is important to

develop a model to predict mechanical behavior of the composite material during

deformation. The model proposed in this paper is based on the non-orthogonal material

model mentioned as above, but now it takes temperature effect into account. The model

11
uses two states of material properties for the simulation, i.e., a high temperature state and

a low temperature state to approximate the contact status between the tooling and the

composite blank. By incorporating temperature effect into the non-orthogonal material

model, the equivalent material properties will be updated, which will affect the stamping

simulation, consequently affecting the maximum draw depth of a composite blank.

However, even this model has drawback. It purely relies on the assumption that all tooling

is represented by analytical surfaces, and this is not true for complex geometries.

Autoclaves are often used for curing process of composite materials, and provides

a controlled cycle of temperature and pressure in this process. However, upon removal

from the tool, residual stresses will rise, and often it causes the deformed composite part to

distort from its desired shape. Yuan et al. developed an analytical model on through-

thickness stresses and warpage of composite laminates due to tool-part interaction [14]. By

making the assumption that slip occurs between tooling and the composite material, and

the knowledge of the stress distribution through thickness, the analytical model can predict

through-thickness residual stresses and warpage for the composite material during the

curing process in the autoclave, without extensive resin characterization.

Material Characterization Methods

Tests such as the three-point bending can be used to evaluate the properties of the

entire specimen as a structure, so it is important to understand the mechanical response

when there is an applied load to the composite structure. It will be a lot easier to understand

homogeneous materials, since they respond isotropically when loaded in different

orientations. However, when dealing with composite materials, the anisotropy created by

12
the orientations of reinforcing fibers for each ply has to be taken into account. Such kind

of material response and structures needs an orthotropic material model coupled with an

assumption of plane stress. The stiffness along the fiber axis and perpendicular to the fiber

axis as well as the Poisson’s ratio and shear modulus will be taken into account in such a

model. With those terms, a reduced compliance matrix can be obtained, which can be used

to predict the mechanical response when a stress is present.

Zampaloni developed a material model which tracks preferred fiber orientations of

randomly oriented, unidirectional or woven composite sheets [9]. At first, the model was

developed to predict the response of randomly orientated fibers. For woven or

unidirectional composites, it will be easy to determine which orientation will give the

stiffest response. However, when dealing with composites with random fiber orientations,

this cannot be done easily. By using a squeeze flow test, Zampaloni was able to determine

which directions have the greatest amount of reinforcing fibers, and which orientations

have the least. He was able to determine the orientations worth tracking in his material

model just by assuming that orientations with the largest amount of reinforcing fibers

would be the stiffest. Then, the material properties can be extracted when these orientations

are determined.

Young’s modulus is typically the first property to be determined. This is standard

characterization for composite materials based on ASTM D3039 [15]. In this standard, a

thin strip of composite specimen with a constant rectangular cross section is mounted in a

mechanical testing machine with a specialized grip to load the specimen in tension

monotonically while recording the force. The stress-strain relationship of the material can

13
be determined through this process. And properties like ultimate tensile strain, tensile

modulus of elasticity, Poisson’s ratio can be obtained.

Measurement of shear modulus is also necessary. Two methods are reviewed for

obtaining shear modulus: the Iosipescu shear test [16] and V-notch rail shear test [17]. Both

of these tests have a rectangular specimen with notches cut in order to concentrate shear

stress at the neck. It is how the specimen is loaded that differentiates these two tests. The

Iosipescu method utilizes a special equipment which loads the specimen along its edges by

exerting compressive force. The V-Notched rail shear method uses another specialized

fixture to grip the faces of the specimens and shears the specimen in tension. These two

methods apply different forces in different locations of the composite specimen, therefore,

it is reasonable to expect that each of the two approaches can give slightly different shear

modulus. Yan-lei et al. evaluated these two methods in his study [18]. Although the

Iosipescu shear specimens can give good results, he claims that edge crushing can be an

issue due to the way the specimen is loaded into the fixture. And V-notched rail shear

method can get rid of such unacceptable failure and it uses a larger gauge section, so this

method has more advantages, and therefore it is preferred.

Thermo-stamping of composite sheets has been an efficient way to impart complex

shapes into composites, since it only takes seconds to form the composite sheet into the

desired shape. According to Liu et al., during the thermo-stamping process, shear is the

main deformation mechanism, and the main parameter that is influencing woven fabric

formability is the shear resistance [19]. A picture frame model was developed, using the

kinematic and elastic analysis of the picture frame experiments. Then, the stress tensor can

14
be obtained from the proposed shear model, and by using Hooke’s law, shear properties of

the woven fabrics can be calculated. An analytical solid mechanics model was also

proposed to predict the shear properties, hoping to eliminate extensive experimental

characterization. However, the analytical model had significant differences with

experimental data. Because frictional resistance factor was not considered in the study.

Recently, Alshahrani and Hojjati proposed a new test method for characterizing the

bending behavior of textile prepregs during forming simulation with both rate and

temperature dependencies [20]. This new method is based on the vertical cantilever test,

with metallic custom grips to clamp the sample vertically, while a linear actuator controls

the sample deflection and applied rate. And a radiant heater is used to provide the sample

with testing temperature, which is monitored by an infrared camera. By adjusting the

testing speed from the actuator’s controller, the rate dependent effect can be measured. The

testing results show that bending stiffness is about 20% higher in the warp direction than

in the weft direction for satin woven carbon/epoxy prepregs. The results can be used for

aiding future design. Moreover, the experimental results show that the bending behavior is

rate-dependent, because of the viscoelastic behavior of the prepreg. However, the new

testing method does not include the viscoelastic behavior at different temperatures. And

the bending friction coupling effects might still have some influences during the forming

simulation of laminated composites.

15
 Chapter 3. Numerical Methods

Numerical simulations require a solid background on properties of laminated

composites, since fiber directions, number of layers, lay up sequence, etc. can affect the

laminate properties during deformation. The goal of developing numerical model is to

produce a good agreement between simulations and experiment results, so that the results

could be predicted before experiments.

Constitutive Model

Composite materials generally have elastic stress-strain behavior until failure. The

stress-strain relation can be given as follows:

𝜎𝑖𝑗 = 𝐶𝑖𝑗𝑘𝑙 𝜖𝑘𝑙 (1)

The stiffness tensor C has 81 independent terms, which are too many to use as a constitutive

equation. Some assumptions have to be made. Tsai et al. showed that the stress, strain,

compliance and stiffness matrices are symmetric, and the stress and strain are 2 nd order

tensors, the number of independent terms is reduced from 81 to 36 [21]. Also, since the

material stiffness is assumed to be symmetric, the number of independent terms in C is

reduced to 21. Then another assumption is made that the material is exhibiting orthotropic

behavior, which contains 3 planes of symmetry. Then equation (1) becomes

16
 𝜎1 𝐶11 𝐶12 𝐶13 0 0 0 𝜀1
𝜎2 𝐶12 𝐶22 𝐶23 0 0 0 𝜀2
𝜎3 𝐶13 𝐶23 𝐶33 0 0 0 𝜀3
𝜎4 = 0 (2)
0 0 𝐶44 0 0 𝜀4
𝜎5 0 0 0 0 𝐶55 0 𝜀5
[𝜎6 ] [ 0 0 0 0 0 𝐶66 ] [𝜀6 ]

Furthermore, individual layer will be modeled, which means that an assumption of plane

stress can be made. Then equation (2) can be reduced to equation (3).

𝜎1 𝑄11 𝑄12 0 𝜀1
[𝜎2 ] = [𝑄12 𝑄22 0 ] [𝜀2 ] (3)
𝜎6 0 0 𝑄66 𝜀6

The reduced stiffness Q assumes that the material frame (1,2,3 direction) is aligned with

its global coordinate (x, y, z axes). However, these coordinates do not have to align with

each other, so a distinct notation is used to avoid confusion. The 1-direction from the

material frame is always aligned with the fiber orientation, while the 2-direction is fiber’s

transverse direction within the plane. The global coordinate is fixed in space, while material

frame can move under loading. The notation can be shown in figure 5.

Figure 5. Material frame and global coordinate [7]

17
Stresses will be experienced in more than one direction within a plate, so Poisson’s ratio

becomes important. The definition of Poisson’s ratio is the ratio of the strain perpendicular

to the loading direction, to the strain along with the loading direction.
𝜀𝑇 𝜀2
𝜈12 = = 𝐹𝑜𝑟 𝑙𝑜𝑎𝑑𝑖𝑛𝑔 𝑎𝑙𝑜𝑛𝑔 𝑡ℎ𝑒 𝑓𝑖𝑏𝑒𝑟𝑠, 𝑜𝑟 (4)
𝜀𝐿 𝜀1
𝜀𝐿 𝜀1
𝜈21 = = 𝐹𝑜𝑟 𝑙𝑜𝑎𝑑𝑖𝑛𝑔 𝑎𝑙𝑜𝑛𝑔 𝑡ℎ𝑒 𝑡𝑟𝑎𝑛𝑠𝑣𝑒𝑟𝑠𝑒 𝑑𝑖𝑟𝑒𝑐𝑡𝑖𝑜𝑛 (5)
𝜀𝑇 𝜀2

Then the strain component will decrease due to the contraction of Poisson’s effect

caused by the force perpendicular to the applied force:

𝜎1 𝜎2
𝜀1 = − 𝜈21 𝜖2 𝑎𝑛𝑑 𝜀2 = − 𝜈12 𝜖1 (6)
𝐸1 𝐸2

The shear force can be defined as:

τ = 𝛾12 𝐺12 (7)

And the above equations can be written in the matrix form as:

𝜀1 𝑆11 𝑆12 0 𝜎1
𝜀
[ 2 ] = [𝑆12 𝑆22 𝜎
0 ][ 2] (8)
𝛾12 0 0 𝑆66 𝜏12

Where,

1 −𝜈12 −𝜈21 1 1
𝑆11 = 𝑆12 = = 𝑆22 = 𝑆66 = (9)
𝐸1 𝐸1 𝐸2 𝐸2 𝐺12

The S matrix is called compliance matrix, and by inverting the matrix, a reduced

stiffness matrix, Q, can be shown as:

𝜎1 𝑄11 𝑄12 0 𝜀1
[ 𝜎2 ] = [𝑄12 𝑄22 0 ] [ 𝜀2 ] (10)
𝜏12 0 0 𝑄66 𝛾12

18
 Where

𝐸1 𝜈12 𝐸2 −𝜈21 𝐸1 𝐸2
𝑄11 = , 𝑄12 = = , 𝑄22 = 𝑄 = 𝐺12 (11)
1−𝜈12 𝜈21 1−𝜈12 𝜈21 1−𝜈12 𝜈21 1−𝜈12 𝜈21 66

Often, the loading direction does not align with material direction, then the stresses and

strains need to be transformed into the coordinates that do coincide with fiber directions.

Then a transformation matrix, [T], is defined as follow:

𝜎1 𝑐𝑜𝑠 2 𝜃 𝑠𝑖𝑛2 𝜃 −2𝑠𝑖𝑛𝜃𝑐𝑜𝑠𝜃 𝜎𝑥

[ 𝜎2 ] = [ 𝑠𝑖𝑛2 𝜃 𝑐𝑜𝑠 2 𝜃 2𝑠𝑖𝑛𝜃𝑐𝑜𝑠𝜃 ] [ 𝜎𝑦 ] (12)
𝜏12 𝑠𝑖𝑛𝜃𝑐𝑜𝑠𝜃 −𝑠𝑖𝑛𝜃𝑐𝑜𝑠𝜃 (𝑐𝑜𝑠 2 𝜃 − 𝑠𝑖𝑛2 𝜃) 𝜏𝑥𝑦

Also, the stiffness and compliance matrix after transformation can be defined as follow:

Stiffness matrix: [𝐶̅ ] = [𝑇][𝐶][𝑇]𝑇 (13)

𝐶𝑜𝑚𝑝𝑙𝑖𝑎𝑛𝑐𝑒 𝑚𝑎𝑡𝑟𝑖𝑥: [𝑆̅] = [𝐶̅ ]−1 (14)

Obtaining Material Properties

With the constitutive model developed, the stiffness matrix, Q, can now be

calculated. Four constants, 𝐸11 , 𝐸22 , 𝐸12 , 𝐺12 , need to be obtained to construct the stiffness

matrix.

𝐸11 is the Young’s Modulus of unidirectional composite in its fiber direction. The

most common way to obtain this constant is to use the rule of mixture as follow:

𝐸11 = 𝐸𝐹1 𝑉𝐹 + 𝐸𝑀1 (1 − 𝑉𝑓 ) (15)

Where 𝐸𝑓 and 𝐸𝑚 are the Young’s Modulus of fiber and matrix, respectively. And 𝑉𝑓

represents the fiber volume fraction of the composite. It is reasonable to use (1 − 𝑉𝑓 ) to

represent the volume fraction of the matrix, since the composite is mainly composed of

fiber and matrix, and the assumption that zero void exists, results in:

19
 𝑉𝑓 + 𝑉𝑚 = 1 (16)

However, even among the highest quality of composite material, voids are inevitable.

Therefore, the zero-void assumption is highly idealized. But usually, the percentage of

voids are small enough to be ignored. Moreover, equation (16) assumes that there is a

perfect bond between fiber and matrix components. And the following equation takes

continuity of mass into account, and it is represented by Mohammed et al [22].

𝑣𝑓𝑜
𝑉𝑓 = (17)
sin 𝛼

Where 𝑣𝑓𝑜 is fiber volume fraction before deformation and 𝛼 is the angle between two

orthogonal fibers, a-fiber and b-fiber, before deformation starts. In general, the assumption

of zero void and perfect bond between fiber and matrix will introduce errors, but they are

usually to be ignored because the error is too small.

𝐸22 is the Young’s Modulus of unidirectional composite in its transverse direction.

And the equation to determine this quantity is to invert the rule of mixture:

𝐸𝐹2 𝐸𝑀2
𝐸22 = (18)
𝐸𝐹2 𝑉𝑀 + 𝐸𝑀2 𝑉𝐹

Poisson’s ratio is also obtained by rule of mixture as below:

𝜐12 = 𝜐𝑀 𝑉𝑀 + 𝜐𝐹 𝑉𝐹 (19)

And
𝐸
𝜐21 = 𝐸22 𝜐12 (20)
11

Also the shear modulus can be defined in the rule of mixture:

𝐺𝑀 𝐺𝐹
𝐺12 = (21)
𝑉𝑀 𝐺𝐹 + 𝑉𝐹 𝐺𝑀

20
 Therefore, once the properties of fibers and matrix and volume fraction of fiber are

obtained, all the constants mentioned above could be calculated accordingly. The stiffness

matrix and compliance matrix of the laminated composite could be calculated,

transformation matrix needs to be considered when the loading direction is not aligned with

the principal material direction, so that the new stiffness and compliance matrix could be

obtained for the laminated composite after rotation.

Preferred Orientations

The constitutive model mentioned above can be used to describe properties of

unidirectional layup laminated composite. However, in real situation, fiber angle will

change during deformation. Therefore, PFO model, which was developed by a former

student, is used to trace the fiber angle change during deformation. Figure 6 represents a

laminated composite composed of two layers of lamina. The stiffness of the overall

material can be assumed to be the summation of the stiffness tensors of the two individual

fiber orientations. Assume that one lamina, A-fiber, has 0-degree fiber orientation, and the

other one, B-fiber has 90-degree fiber orientation. Before deformation, these fibers are

orthogonal to each other. But during the process of deformation, these fibers will not

always remain orthogonal to each other. The summation process can be illustrated in the

following figure.

21
 Figure 6.Stiffness summation in PFO model [7]

If the material cannot be fully characterized by two preferred orientations, additional fiber

orientations could be added.

Global coordinate is needed for using the summation of each PFO. By using

equation (10), the stiffness tensor can be obtained for the composite material. Then by

using transformation matrix from equation (12), the stiffness tensor can be rotated to the

global coordinate system. Next, the sum of stiffness for each PFO can be utilized when the

stiffness tensor is rotated to the global coordinate, in order to get the overall properties of

the material. It can be demonstrated as follows:

22
 Laminated composite with two layers of fiber

The fiber orientation and its transverse direction for a-fiber and b-fiber and a global
coordinate (red)

The stiffness is calculated for each PFO and then rotated to the global coordinate

The overall stiffness is calculated for a-fiber and b-fiber after rotation

Figure 7.Rotation of PFO to the global coordinate [7]

23
 The process mentioned above is used to find the overall stiffness tensor of a

composite material. And typically, the fiber orientations are specified by the user when

creating simulation models, and it will make stiffness easy to calculate for the laminate.

However, as deformation proceeds, the angles of α, and β will change, and it will

result in a change of transformation matrix. The overall properties of the laminated

composite material will change due to the change of fiber angles. These changes are

important, since they will determine how the composite material behave at each

deformation increment. Take a small differential element with a-fiber and b-fiber with its

global coordinate for an example. A deformation gradient tensor acts on the material.

F = RU (22)

Where F is the deformation gradient tensor, and it can be decomposed into a product of

two second-order tensors, which are R the rotation tensor, and U the right stretch tensor.

The differential element is acted by the deformation gradient tensor by stretch and rotation

tensors. The structure coordinate system is acted by the rotation tensor, R, so that the

coordinate remains orthogonal after deformation. The fiber orientations are affected by the

rotation and stretch tensor, which affect the angle of each fiber to the material coordinate

system.

24
 Figure 8.Effects of deformation tensor on elements [7]

From the example, in the undeformed shape, the fibers are aligned with the SCS, and in

the deformed shape, the SCS is changed by the rotation tensor R, and fiber angles are

changed by the deformation tensor, which affects the stiffness matrix of the differential

element, and the element will behave according to the stiffness tensor for the next step of

deformation.

Implementation into Abaqus

The PFO model is implemented in Abaqus explicit via a user subroutine. The

explicit solver allows the strain increment, and deformation gradient tensor to be fed into

the subroutine for each progressive time step. Then the change of SCS and the fiber

orientations during the current time step are determined by the user subroutine, using the

deformation gradient tensor, as shown in figure (8). Next, the stiffness of the PFO can be

obtained once those changes are determined. And the stiffness needs to be rotated to their

material coordinate. These steps are repeated for every PFO. When all the PFO stiffness

are rotated to their material frames, they are then summed up to get overall stiffness.

25
Multiply the strain increment with stiffness, the stress increment can be obtained for the

time step. A flow chart of this process can be seen Figure 9.

Figure 9.Flow chart of PFO model implemented in Abaqus/CAE [9]

26
 Analytical & Simulation Discussion

Some typical values of composite can be found in the table below.

Table 1. Properties of composite

Fiber 𝐸𝑓11 230 Gpa

𝐸𝑓22 230 Gpa

𝜈12 0.2

𝐺12 95.83 Gpa

Matrix 𝐸𝑚 3.5 Gpa

𝜈𝑚 0.4

𝐺12 1.25 Gpa

𝑉𝑓 0.6

Where 𝐸𝑓11 and 𝐸𝑓22 are the Young’s modulus of fiber in its fiber direction, and transverse

direction. 𝜈12 and 𝐺12 are the Poisson’s ratio and shear modulus of fiber. 𝐸𝑚 𝜈𝑚 and 𝐺12

are the Young’s modulus, Poisson’s ratio, and shear modulus of the matrix, respectively.

27
 From the constitutive equations in the constitutive model section for calculating

compliance and stiffness matrix for a single layer of a laminate after transformation, the

new young’s modulus for fiber direction and its transverse direction can be obtained simply

from 𝑆11 , and 𝑆22 elements of the new compliance matrix. Also, shear modulus can be

calculated from 𝑆33 element from the new compliance matrix. Equations are shown below:

1
𝐸11 = (23)
𝑆11

1
𝐸22 = (24)
𝑆22

1
𝐺12 = (25)
𝑆33

By implementing the logic into Matlab, graphs representing Young’s modulus and

shear modulus can be shown in Figure 10. The fiber angle on x-axis means the angle

between fiber and loading directions.

Figure 10. Modulus plot for single layer

28
 By putting 0 degree into the Matlab code, the Young’s modulus along the fiber

direction, transverse direction and the shear modulus are shown in Figure 11 (unit: Gpa):

Figure 11. Properties of a single layer with 0 degree

The results from Matlab are the basic material properties for Abaqus simulation later.

For multi-layer composite material, the new stiffness after transformation is

calculated for each layer. Then all the stiffness matrices are added, and divided by the

number of layers to get an average stiffness matrix. Since a lamina has the strongest

strength along its fiber direction and least strength in its transverse direction, different

stacking sequences can also affect the overall properties of the laminated composite

material. Therefore, different stacking sequences can have different response during certain

loading. For example, consider a laminated composite with 14 unidirectional layers. Three

different stacking sequences are used to plot Young’s modulus and shear modulus. The

different stacking sequences are [90/0/90/0/90/0/90]s, [90/45/0/90/45/0/90]s, [45/-45/45/-

45/45/-45/45]s

29
Figure 12. Young's modulus plot in fiber direction

Figure 13. Young's modulus in transverse direction

30
 Figure 14. Shear modulus

The angle in the plots is the loading angle with respect to the global coordinate

system. The plots give an understanding of how the material behaves under loading in

different directions. And this helps to select appropriate stacking sequence for specific

design requirements.

31
 Numerical Validation

It is important to understand that during numerical simulation, the composite

material will response as the properties calculated from Matlab. The numerical simulation

software used is Abaqus/CAE package.

Numerical Setup

The test used is a simple tension test. A composite material with a 500mm x

100mm dimensions is shown in Figure 15.

Figure 15. Dimension of a composite specimen

32
 The part is created in shell planer base feature. Material properties are the results

calculated from Matlab. And a shell composite section is assigned to the part. The fiber

direction is along the x axis and its transverse direction is along the y axis, which are shown

below.

Figure 16. Material properties of the specimen

33
 Figure 17. Shell composite section of the specimen

Figure 18. Material direction of the specimen

A static general time step with a time period of 1 is created in the model. A standard four-

node, reduced integration, S4R shell element is used with a total of 500 elements in the

simulation. For boundary conditions, the bottom edge is set to YSYMM and 50 mm

displacement is applied to the top edge to stretch the specimen.

34
Figure 19. Bottom boundary condition of the specimen

Figure 20. Top boundary condition of the specimen

35
 Figure 21. Result of the tensile test

From the result, by summing all the reaction forces along the bottom edge nodes,

and dividing the force by the bottom cross-sectional area, a stress vs. time curve can be

obtained from the software. Also, by picking a node on the top edge of the specimen, a

strain vs. time curve can be obtained. Combining these curves, a stress vs. strain curve can

be created for the composite specimen. Young’s modulus can be demonstrated by

calculating the slope from the stress vs. strain curve. The 2 different stacking sequences

shown before are used for validation purposes.

For the stacking sequence [45/-45/45/-45/45/-45/45]s, the Young’s modulus

calculated from Matlab is 11.3396 Gpa for both fiber and its transverse direction, and the

stress vs. strain curve obtained from Abaqus is shown in Figure 22.

36
 Figure 22. Stress vs. strain curve for [45/-45/45/-45/45/-45/45]s

For the stacking sequence [90/0/90/0/90/0/90]s, the Young’s modulus calculated

from Matlab is 64.8745Gpa in the fiber direction, 83.6371Gpa in its transverse direction,

and the stress vs. strain curve obtained from Abaqus is shown below.

Figure 23. Stress vs. strain curve for [90/0/90/0/90/0/90]s in its fiber direction

37
 Figure 24. Stress vs. strain curve for [90/0/90/0/90/0/90]s in its transverse direction

Comparing Matlab and Abaqus results, it can be concluded that the analytical

approach and the numerical approach are in great agreement in regard to the values of

Young’s modulus. This means that the behavior of the composite material in the numerical

simulation follows the properties that are calculated from Matlab. Therefore, the numerical

simulation results are reliable and can be used for more complicated simulation

applications.

A simple shear test is performed in Abaqus/CAE to obtain shear modulus and

compare it with that from Matlab code for the same stacking sequences from tension test.

As it is shown below, the blue area is the deformed shape, and the grey area is the

undeformed shape. The element has a 100mm x 100mm dimension, and the method used

to calculate the shear modulus is also shown in Figure 26.

38
 Figure 25. Simple shear test

Figure 26. Calculation of shear modulus

𝐹
𝜏𝑥𝑦 𝐹𝑙
𝐺= = 𝐴 = (26)
𝛾𝑥𝑦 ∆𝑥 𝐴∆𝑥
𝑙
𝐹
Where 𝜏𝑥𝑦 = 𝐴 = 𝑠ℎ𝑒𝑎𝑟 𝑠𝑡𝑟𝑒𝑠𝑠, F is the force, and A is the area on which the force

∆𝑥
acts, 𝛾𝑥𝑦 = 𝑠ℎ𝑒𝑎𝑟 𝑠𝑡𝑟𝑎𝑖𝑛, 𝜃 = 𝑡𝑎𝑛𝜃 = , since it is a small angle approximation. ∆𝑥 is
𝑙

39
the transverse displacement, and l is the initial length. In equation (26), l and A are known

from geometry of the element, and ∆𝑥 can be set in the displacement boundary condition

of the simulation. Then F can be obtained from summing up the reaction forces in the force

direction.

Both stacking sequences have a ∆𝑥 = 1𝑚𝑚 displacement to the right on top edge

of the element. The bottom edge is fixed in all directions to prevent it from moving, and

both vertical edges are fixed in y direction so that there will not be any changes in the

length. The cross-sectional area is 𝐴 = 280 𝑚𝑚2, and the initial length is l=100mm. For

the stacking sequence [45/-45/45/-45/45/-45/45]s, the shear modulus calculated from

Matlab is 35.3894 Gpa. The force from Abaqus simulation result is 100.702 KN, and by
100.702 𝐾𝑁∗100𝑚𝑚
applying the shear modulus calculation formula, 𝐺12 = = 35.965 𝐺𝑝𝑎.
280 𝑚𝑚2 ∗1𝑚𝑚

For the stacking sequence [90/0/90/0/90/0/90]s, the shear modulus obtained from

Matlab is 3.065 Gpa. The force from Abaqus result is 8.58473 KN, and from the shear
8.58473 𝐾𝑁∗100𝑚𝑚
modulus equation, 𝐺12 = = 3.065975 𝐺𝑝𝑎. From the results, it can be
280 𝑚𝑚2 ∗1𝑚𝑚

seen that both shear modulus results are almost the same, the reason that there are slightly

difference in numbers is that for [45/-45/45/-45/45/-45/45]s, the structure is undergoing

simple shear deformation in Abaqus simulation, and Matlab calculations are based on pure

shear deformation. Although, the shear strain is relatively small, still little difference could

affect the reaction forces in numerical simulation, thus the shear modulus is affected. And

Abaqus takes that into account in the simulation.

40
 PFO Validation

Abaqus/CAE package assumes that the structural coordinate system does not

change during deformation. This may be true with metals in some cases, however, the fiber

orientation always changes under loading, which implies that the material axis also

changes. The PFO model tracks alignment of the fiber during each time step, and then

calculates the overall stiffness and updates material properties during the simulation.

To illustrate this point, a simple tension test model is developed in Abaqus/CAE

package, and its result is compared with the PFO model for validation purposes. Figure 27

shows a 30mm x 30mm single square element subjected to a displacement of 15mm in the

positive direction of the x-axis. The initial fiber orientation is set at 45 degrees (0.7854

radian) from the x-axis. The results are shown in Figure 27.

Figure 27. Fiber orientation after deformation of the tension test

41
 The 1-axis, which is in blue color, is the fiber orientation after deformation, and 2-

axis, which is in yellow color, is the transverse direction to the fiber. From the result, the

fiber orientation remains at 45 degrees from the x-axis after the deformation. This is not a

real situation for composite materials under loading, since fibers are free to move during

the deformation. That means the fiber angle will not stay at 45 degrees after the

deformation.

The same test with PFO subroutine is performed to demonstrate the fiber direction

change after the deformation. The result is shown in Figure 28. SDV5 shows the updated

fiber angle after deformation, unit is in degrees.

Figure 28. Tension test result of fiber angle (degrees) after deformation with PFO model

The PFO tracks the fiber orientation after deformation, which is shown as SDV5 in

Figure 28. From the contour of SDV5, it can be seen that the fiber orientation changes from
42
45 degrees to 31.33 degrees. To validate this angle change, an analytical calculation is

performed based on the geometry of the deformed shape. From Abaqus, the coordinates of

the deformed element are shown in Figure 29.

Figure 29. Coordinates of deformed element with PFO model

The angle θ is the new fiber orientation that needs to be solved. Simply taking

27.39714
𝑡𝑎𝑛−1 ( ), the new fiber angle comes out to be 31.33 degree, which is exactly the
45

same as the one obtained from the PFO simulation. The reason that fiber is always

diagonally connected is that the PFO model always assumes a perfect bonding in the

lamina.

A shear test was also performed to validate the PFO model. The same 30mm x

30mm single element was subjected to a displacement of 5mm on top of the square, and

the initial fiber angle was assumed to be 45 degrees from the x-axis. The result is shown in

Figure 30.

43
 Figure 30. Result of shear test with PFO model

From the result, the fiber orientation that the PFO model tracks after the

deformation is 44.59 degree in SDV5. An analytical calculation is also performed to prove

that this angle is correct. A schematic of the calculation is shown in Figure 31.

Figure 31. Calculation of fiber orientation for shear test

From the PFO model’s subroutine, the rotation of the structural coordinate system

can be obtained, which is 4.834 degree as shown in Figure 31. The red coordinate is the

44
structural coordinate system after the deformation, and the blue dashed line is the global

coordinate. From the Abaqus software, the coordinate of the top right point can be obtained

as (35, 29.1153), as shown. By taking the arctangent of this coordinate, a 39.756 degree

can be calculated, which is the angle of the black dash line along the diagonal. Then by

adding these two angles together, a new fiber orientation can be calculated after the

deformation as: 39.756+4.834=44.59, which is exactly the same as the angle that the PFO

simulation predicted in Figure 30. The shear test also proves that the PFO model uses

deformation tensor to update the overall properties of the composite material, which is

subjected to both rotation and stretch tensors.

Validation with experiments

A uniaxial extension test is simulated to compare with experimental data obtained

from references for the PFO model. This time the tension test is loaded in the vertical

direction, and the positive y-axis is the loading direction. And the composite material is

assumed to be a woven fabric preform with 45/-45 for its preferred fiber orientations. The

composite material is a rectangular shape with 203.2mm x 406.4mm in dimension, as

shown in Figure 32. For boundary conditions, the bottom edge is constrained such that all

nodes are fixed and cannot move in any direction, and the top edge is subjected to a 101.6

mm displacement in the positive y-axis direction, to create a 25% elongation. The

simulated result is shown in Figure 33. The SDV5 parameter shows the updated fiber

angles (in unit of degrees) after the deformation.

45
 Figure 32. Undeformed shape of uniaxial extension test

Figure 33. Deformed shape of uniaxial extension test

From the simulation result, three distinct deformation areas can be identified with

different fiber angle changes. It can be seen that there is no significant fiber angle change
46
in the blue areas on top and bottom parts, as the fiber angle after deformation is around

47.42 degrees compared to 45 degrees for its initial orientation. Also, there are relatively

large fiber angle changes in the middle area of the specimen, and a little less angle change

on the side areas. It is interesting to note that there are different modes of deformations

taking place in the stretched woven sheet. Sidhu et al. also identified three different modes

of deformation in their experiment, as shown in Figures 34 and 35 [23].

Figure 34. Deformed preform at elongation of 25 %

47
 Figure 35. Three different modes of deformation in experiment

From the experimental results, it can be seen that there are no significant fiber angle

changes in Zone I. There are both shearing and sliding between yarns and some fiber angle

change in Zone II. And there is mainly shearing deformation in Zone III. And the

simulation results using PFO model is in good agreement with these experimental results.

That implies that the PFO model’s prediction of the fiber angle rotation is in close

agreement with the real deformation of a composite material subjected to a uniaxial tension

deformation. Furthermore, this will also imply that the PFO model will be able to correctly

predict the updated properties of the composite material after each incremental

deformation. Given that the Abaqus/CAE model does not take into account fiber rotation

after deformation, its prediction of material properties will not be correct either.

Next, simulation of stamping of a woven composite material with initial 0/90

degree preferred fiber orientation to make a hat section was also performed with the PFO

model. The shape and dimensions of the tooling used in this stamping simulation is shown

in Figure 36.

48
 Figure 36. Geometry for upper die

The shape of the lower die is similar to the upper die, with a slightly bigger dimension to

compensate for the thickness of the composite sheet. The dimension of the sheet is 180mm

x 280mm, with general contact property for interaction between the sheet and the dies. The

assembled tooling and the composite sheet are shown in Figure 37. Figures 38 through 41

show the simulation results for woven composite sheets with 0/90 and 45/-45 initial fiber

orientations, as well as experimental results. The SDV5 parameter shows the updated fiber

angles (in unit of degrees) after the deformation.

49
 Figure 37. Assembly of the stamping simulation

Figure 38. Stress distribution of the sheet (0/90 degrees)

50
Figure 39. Result of stamping process of the sheet (0/90 degrees), SDV5 is the updated

fiber angle after deformation in degrees

51
A

B continued

Figure 40. Stamped hat section from experiment (0/90 degrees)

52
Figure 40 continued

Figure 41. Stamping simulation with 45/-45 PFO. SDV5 represents fiber angle
orientations after the deformation in degree

53
 From the finite element simulation results shown in Figure 39, it can be seen that

for 0/90 PFO, there is no fiber angle change in most of the specimen during the

deformation. There is very little fiber angle change in corner areas, where bending stresses

are the highest. When compared with the actual samples, shown in Figure 40, it can be seen

that there were no obvious fiber angle changes after stamping. This is probably due to the

simple shape of the tooling, and that the stamping process does not stretch the sheet off-

diagonally. In other words, there are not much shear deformation imposed in the composite

sheet to cause fiber rotations. However, when the PFO of the composite sheet was changed

from 0/90 to 45/-45 degrees, more fiber angles changed direction during the stamping

process. As can be seen from Figure 41, although the simulation results show no significant

angle changes after stamping, still the change in the fiber orientation around corner areas

is slightly larger than that from 0/90 PFO simulation. Again, this shows the importance of

tracking fiber alignment in the PFO model, which is neglected in the Abaqus/CAE model.

54
 Conclusions

Throughout the course of this study, composite material has been studied for its

types of reinforcements, its unique directionality properties, etc. When forming a

composite material, it is necessary to understand the constitutive relations between stress

and strain, so that the material properties such as Young’s modulus and shear modulus can

be obtained from compliance and stiffness matrices. A former student, Zampaloni,

developed a PFO model which tracks the fiber alignment of the reinforcing fibers within

the material during the deformation. During the FEA simulation, each element is subjected

to a unique deformation gradient tensor, which not only deforms the element but also

changes the angle of PFO. Then, the material properties are updated according to the

change in PFO. The PFO model has been used to simulate the simple tension test and shear

test with Abaqus software. The results from these simulations indicate that the PFO model

works well for predicting fiber orientation change during the deformation, which

Abaqus/CAE model does not account for. Neglecting to update fiber orientations results in

inaccuracies in the shape of deformed part as well as its material properties. Furthermore,

the PFO model was used to simulate the uniaxial tension test and stamping of a hat section,

and numerical results were compared with those obtained from similar experiments. The

comparisons showed that the PFO model can track the fiber orientation correctly, and

provide more precise results, which were also very close to the real forming situation.
55
Although, Abaqus/CAE model is computationally faster in obtaining preliminary results

for composite simulation, the PFO model implemented into Abaqus as a user material

subroutine (VUMAT), provides more reliable results. A shortcoming of the PFO model is

the assumption of perfect bonding between layers, which can lead to wrong predictions

when out of plane warping occurs. In the future, the PFO model needs to be updated in

order to relax this assumption for forming simulations.

56
 References

[1] [Online]. Available:

http://web.mit.edu/course/3/3.064/www/slides/composites_overview.pdf.
[2] W. a. B. P. Watt, Handbook of Composites, Vol. 1: Strong fibers, Amsterdam:
North-Holland, 1985.
[3] G. H. Staab, Laminar Composites, Boston: Butterworth-Heinemann, 1999.
[4] R. M. Koide, G. v. Z. d. Franca and M. A. Luersen, "An ant colony algorithm
applied to lay-up optimization of laminated composite plates," Latin American
Journal of Solids and Structures, vol. 10, 2013.
[5] A. Yousefpour, "Fusion Bonding/Welding of Thermoplastic Composites," Journal
of Thermoplastic Composite Materials, vol. 17, pp. 303-341, 2004.
[6] G. Schinner, "Recycling Carbon-Fiber-Reinforced Thermoplastic Composites,"
Journal of Thermoplastic Composite Materials, vol. 9, pp. 239-245, 1996.
[7] N. E. Kuuttila, "Composite Thermo-Hydroforming of Millitary Ballistic," East
Lansing, 2014.
[8] R. Jones, "Chpater 4: Macromechanical Behavior of a Laminate," in Mechanics of
Composite Materials, Washington D.C., Scripta Book Company, 1975, pp. 147-
173.
[9] M. Zampaloni, "A multipreferred fiber orientation constitutive model for fiber mat-
reinforced thermoplastics with a random orientation applied to the stamp
thermohydroforming process," East Lansing, 2003.
[10] P. L. A. P. P.de Luca, "Numerical and experimental investigation of some press
forming parameters of two fibre reinforced thermoplastics: APC2-AS4 and PEI-
CETEX," Composite Part A: Applied Science and Manufacturing, vol. 29, no. 1-2,
pp. 101-110, 1998.
[11] J. C. Xiongqi Peng, "A dual homogenization and finite element approach for
material characterization of textile composites," Composites Part B: Engineering,
vol. 33, no. 1, pp. 45-56, 2002.
[12] X. P. J. C. Pu Xue, "A non-orthogonal constitutive model for characterizing woven
composites," Composites Part A: Applied Science and Manufacturing, vol. 34, no.
2, pp. 183-193, 2003.

57
[13] P. X. X. P. N. K. Jian Cao, "An approach in modleing the temperature effect in
thermo-stamping of woven composites," Composite Structures, vol. 61, no. 4, pp.
413-420, 2003.
[14] Y. W. X. P. J. W. S. W. zhenyi Yuan, "An analytical model on through-thickness
stresses and warpage of composite laminates due to tool–part interaction,"
Composites Part B: Engineering, vol. 91, pp. 408-413, 2016.
[15] A. International, "Standard Test Method for Tensile Properties of Polymer Matrix
Composite Materials," ASTM International , 2008.
[16] "Standard Test Method for Shear Properties of Composite Materials by the V-
Notched Beam Method," ASTM International, 2012.
[17] A. International, "Standard Test Method for Shear Properties of Composite
Materials by V-Notched Rail Shear Method," ASTM International, 2012.
[18] W. Yan-lei, "Evaluation of in plane shear test methods for composite material
laminates," Journal of Chongqing University, vol. 6, pp. 221-226, 2007.
[19] J. C. X. L. J. S. Lu Liu, "Two-dimensional macro-mechanics shear models of
woven fabrics," Composites Part A: Applied Science and Manufacturing, vol. 36,
no. 1, pp. 105-114, 2005.
[20] M. H. Hassan Alshahrani, "A new test method for the characterization of the
bending behavior of textile prepregs," Composites Part A: Applied Science and
Manufacturing, vol. 97, pp. 128-140, 2017.
[21] S. W. Tsai, "Mechanics of Composite Materials," Part II, Technical Report AFML-
TR-66-149.
[22] C. L. a. M. B. U. Mohammed, "Experimental studies and analysis of draping of
woven fabrics," Composites: Part A, vol. 31, pp. 1409-1420, 2000.
[23] R. A. M. R. F. P. R.M.J.S Sidhu, "Finite element analysis of textile composite
preform stamping," Composite Structures, vol. 52, no. 3-4, pp. 483-497, 2001.

58