COMPOSITE THERMO-HYDROFORMING OF MILITARY BALLISTIC
HELMETS
By
Nicholas Eric Kuuttila

A THESIS
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
Mechanical Engineering—Master of Science
2014

ABSTRACT
COMPOSITE THERMO-HYDROFORMING OF MILITARY BALLISTIC HELMETS
By
Nicholas Eric Kuuttila
Composite thermo-hydroforming is an MSU patented process similar to sheet metal
hydroforming. This process uses heated and pressurized fluid to form composite blanks to a punch of
desired geometry. Prior to this study, a small 40 ton press served as a proof of concept by forming 4”
diameter hemispheres, allowing the lab to acquire a larger 300 ton press. This study focuses on the
modification and design of the 300 ton press and using it to form advanced combat helmets using the
ballistic composite Spectra Shield SR-3136. Good results are achieved in the forming process showing
that composite hydroforming is a viable means for manufacturing thermoplastic composite materials. The
process showed good results in the ability to form these deep drawn parts by reducing wrinkling of the
final product.
Concurrent to these forming experiments, the forming process is numerically modeled using
Abaqus/CAE. The material is modeled using a Preferred Fiber Orientation model developed by a past
student of the hydroforming lab. The model is adapted to work with a thick composite laminate consisting
of many layers. Model parameters are also updated to work with a significantly thicker laminate than has
been used in the past. Results of the numerical modeling show good correlation with the forming
experiments. The model still shows need for improvement due to the premature onset and severity of the
out of plane warping observed.

Copyright by
NICHOLAS ERIC KUUTTILA
2014

I dedicate this work to my mother and father.
Thanks for the love and encouragement, but mostly thanks for supporting my continued education.

iv

ACKNOWLEDGMENTS

Thank you to everyone who helped guide me along the path or research and discovery. Thanks to
Dr. Loos and Dr. Hossain for their expertise and support in material characterization. Thanks to all that
helped at Mott Community College for supplying the lab with pre-cut blanks. Thanks also to BAE
Systems and TARDEC for supporting this work through funding, materials and tooling.
Thank you to my hardworking undergraduate assistants; Scott Belonge, Luke Ferguson, Matt
Ayers, Ryan Volkman and Tim Maul.
Most of all, thanks to my adviser, Dr. Farhang Pourboghrat, who has not only advised me but
mentored me as well throughout the majority of my higher education. Thanks for the opportunities and
for the continued support.

v

TABLE OF CONTENTS

LIST OF TABLES ..................................................................................................................................... viii
LIST OF FIGURES ..................................................................................................................................... ix
Introduction ................................................................................................................................................... 1
Composite Thermo Hydroforming............................................................................................................ 2
Topics Investigated During This Study .................................................................................................... 4
Literature Review.......................................................................................................................................... 7
Consolidation of CFRTP Composites ....................................................................................................... 7
Material Characterization Methods ......................................................................................................... 12
Thermoplastic Composite Forming Methods ......................................................................................... 14
Numerical Methods for Simulating the Forming of Composites ............................................................ 17
Consolidation of Flat Panels and Material Characterization....................................................................... 20
SR-3136 .................................................................................................................................................. 20
Blank Consolidation and Shape .......................................................................................................... 20
Material Characterization.................................................................................................................... 23
Shear Modulus ................................................................................................................................ 23
Young’s Modulus............................................................................................................................ 27
Polypropylene Reinforced Carbon Fiber ................................................................................................ 31
Blank Consolidation............................................................................................................................ 31
Consolidation Properties Tool ............................................................................................................ 34
Future Recommendations ....................................................................................................................... 36
Experimental Methods ................................................................................................................................ 37
Experimental Setup ................................................................................................................................. 38
High Pressure System (HPS) .............................................................................................................. 38
Pressure Producing Skid (PPS) ........................................................................................................... 40
Forming Dies ...................................................................................................................................... 44
Punches ............................................................................................................................................... 50
Oil Heater ............................................................................................................................................ 52
Infrared Heater .................................................................................................................................... 53
Computer Control System................................................................................................................... 55
Forming Process...................................................................................................................................... 57
Results and Discussion ........................................................................................................................... 63
Phase 1 ................................................................................................................................................ 64
Phase 2 ................................................................................................................................................ 65
Phase 3 ................................................................................................................................................ 67
Phase 4 ................................................................................................................................................ 73
Future Recommendations ....................................................................................................................... 83
Numerical Methods..................................................................................................................................... 85
Preferred Fiber Orientation (PFO) Material Model ................................................................................ 85
Constitutive Model.............................................................................................................................. 86
Obtaining Material Properties ............................................................................................................. 88
vi

Preferred Orientations ......................................................................................................................... 91
Implementation in Abaqus .................................................................................................................. 95
Simulation Changes and Adaptions ........................................................................................................ 96
Multiple Layer Simulation .................................................................................................................. 97
Transverse Shear Stiffness ................................................................................................................ 100
Transverse Young’s Modulus ........................................................................................................... 102
Use of Material Characterization Data.............................................................................................. 102
Results and Discussion ......................................................................................................................... 103
Impact of Transverse Shear Stiffness................................................................................................ 104
Square Blank Simulation Results ...................................................................................................... 106
Over-prediction of Shear Stiffness .................................................................................................... 110
Comparison to Experimental Results ................................................................................................ 112
Future Recommendations ..................................................................................................................... 116
Conclusions ............................................................................................................................................... 118
REFERENCES ......................................................................................................................................... 120

vii

LIST OF TABLES

Table 1 – Shear Modulus ............................................................................................................................ 26
Table 2 – Tensile Test Data ........................................................................................................................ 30
Table 4 – Consolidation Conditions............................................................................................................ 34
Table 3 – CF-PP Layup Tool ...................................................................................................................... 35
Table 5 - Traction Separation Properties .................................................................................................. 100
Table 6 – Original Transverse Shear Properties ....................................................................................... 101
Table 7 – Simulation Properties ................................................................................................................ 106

viii

LIST OF FIGURES

Figure 1 - Matched Die Molding .................................................................................................................. 3
Figure 2 – CMA Blank Shape Prediction Capabilities ............................................................................... 22
Figure 3 – Shear Modulus Test Specimens ................................................................................................. 24
Figure 4 – Shear Fixtures ............................................................................................................................ 25
Figure 5 – Shear Stress Strain Curves ......................................................................................................... 26
Figure 6 – Difficulties with Tensile Specimens .......................................................................................... 28
Figure 7 – Dog Bone Specimens ................................................................................................................ 28
Figure 8 – Tensile Test Results ................................................................................................................... 29
Figure 9 – Tensile Specimen from Russell Study ....................................................................................... 31
Figure 10 – CF-PP Consolidation Mold ..................................................................................................... 32
Figure 11 – HPS Components..................................................................................................................... 39
Figure 12 – Simplified Hydraulic Schematic of PPS and Forming Die...................................................... 41
Figure 13 – The Blank Sandwich................................................................................................................ 42
Figure 14 – PPS Cooling Loop ................................................................................................................... 43
Figure 15 – Regulator Operating Configurations........................................................................................ 44
Figure 16 – Original Hydroforming Dies ................................................................................................... 45
Figure 17 – Redesigned Hydroforming Die Set.......................................................................................... 47
Figure 18 – Reduction in Punch Cavity Gap .............................................................................................. 48
Figure 19 – Hydrostatic Clamping .............................................................................................................. 49
Figure 20 – Tongue and Groove Feature .................................................................................................... 50
Figure 21 – Section View of Punch-Ram Adapter Interface ...................................................................... 51
Figure 22 – ACH Punch .............................................................................................................................. 51
Figure 23 – Hemispherical Punch ............................................................................................................... 52

ix

Figure 24 – IR Heating Array ..................................................................................................................... 54
Figure 25 – Forming - Step 1 ...................................................................................................................... 58
Figure 26 – Forming - Step 2 ...................................................................................................................... 59
Figure 27 – Forming - Step 3 ...................................................................................................................... 60
Figure 28 – Forming - Step 4 ...................................................................................................................... 61
Figure 29 – Forming - Step 5 ...................................................................................................................... 62
Figure 30 – Forming - Step 6 ...................................................................................................................... 63
Figure 31 – First Hydroforming Results ..................................................................................................... 65
Figure 32 – Double Sided Forming Pressure on Dyneema Blank .............................................................. 66
Figure 33 - Rapid Decompression of the Fluid Cavity ............................................................................... 68
Figure 34 – Blank Formed with Single-Side Forming Pressure ................................................................. 69
Figure 35 - Fiber Orientations of Dyneema Blank...................................................................................... 70
Figure 36 - Reducing Blank Anisotropy ..................................................................................................... 71
Figure 37 – Specialized Stacking Sequence ............................................................................................... 72
Figure 38 – CMA Predicted Blank Shape................................................................................................... 74
Figure 39 – Limited Forming Pressure ....................................................................................................... 75
Figure 40 - Forming Pressure and Punch/Clamp Displacement ................................................................. 76
Figure 41 – Improvement in Results ........................................................................................................... 78
Figure 42 - Blank Lifting Problem.............................................................................................................. 79
Figure 43 - Change in Clamping Scheme ................................................................................................... 81
Figure 44 – Clamping Scheme Difficulties................................................................................................. 82
Figure 45 – Material and Structural Coordinate Systems ........................................................................... 87
Figure 46 – Stiffness Summing Used in PFO Model.................................................................................. 91
Figure 47 – Rotation of PFO’s to the Structural Frame .............................................................................. 93
Figure 48 – Modification of the Structural Coordinate System and Fiber Orientations ............................. 94
Figure 49 – User Subroutines Within Abaqus/CAE ................................................................................... 95
x

Figure 50 - Model Setup ............................................................................................................................. 96
Figure 51 – Fracture Modes ........................................................................................................................ 98
Figure 52 - Traction-Separation Curves...................................................................................................... 99
Figure 53 - Tansverse Shear ..................................................................................................................... 101
Figure 54 – Transverse Shear Stiffness Effects ........................................................................................ 105
Figure 55 – Deformed Blank Shape .......................................................................................................... 107
Figure 56 - Previous Simulation Results................................................................................................... 108
Figure 57 – Fiber Shear Contour Plot ....................................................................................................... 109
Figure 58 -

Contour Plot .................................................................................................................... 110

Figure 59 - Shear Stiffness vs Volume Fraction ....................................................................................... 111
Figure 60 - Frame Rotation Effect on Shear Stiffness .............................................................................. 111
Figure 61 - Pressure Profiles ..................................................................................................................... 113
Figure 62 - Force-Displacment Comparison ............................................................................................. 114
Figure 63 – Blank Shape and Wrinkling Region Comparison .................................................................. 115

xi

Introduction

The use of composite materials in high strength, light weight structures has been occurring for
decades. Fiberglass reinforced plastics have been employed since the 1950’s in the building of high
performance cars and boat hulls. More recently, carbon fiber composites have gained acceptance in the
aerospace industry as a high strength structural material, capable of exceeding the strength of steel. It has
only been a couple of decades that carbon fiber and other high performing composites has become
affordable enough to appear in consumer goods such as automobile components and sporting goods.
These products are the first among many consumer products that will employ carbon fiber composites as
the material becomes more affordable. Currently these products are still cost prohibitive to the average
consumer. Nevertheless, costs continue to decrease and the strength and weight reduction benefits are
becoming more attractive, especially to the automotive industry as CAFE standards continue to push for
higher fuel efficiency (50 MPG by 2025).
The costs of raw materials alone do not shed light on the reason for the high cost of composites.
A vast majority of composites employ matrix systems such as polyester, vinylester, and epoxy. This class
of plastics, known as thermosets, is applied to the reinforcing fibers in liquid form and either heat or a
chemical hardener is used to catalyze cross linking of the polymer chains which permanently hardens the
resin. This type of resin has several advantages such as relatively low curing temperature, a low pre-cured
viscosity, high stiffness and stability over a wide range of temperatures. However the drawbacks
associated with this resin system are numerous. The curing cycle of the resin system can take hours,
severely reducing the ability to mass produce components. Additionally, components that have been
damaged are not able to be repaired easily and more often than not need to be completely replaced.
Recycling these resins is also difficult.

1

Many of these short comings can be addressed by employing thermoplastic resin systems.
Thermoplastics soften and melt at elevated temperatures and solidify when cooled. This allows
manufacturing cycles that are a fraction of the time when comparing them to thermosets. This also allows
the repair and joining of components through fusion bonding and thermoplastic welding techniques [1].
As opposed to thermoset resins, thermoplastics do not cross link but instead form a solid structure by an
entangled network of amorphous or semi-crystalline polymer chains. This gives the thermoplastic
polymer the ability to reach large deformations before failure, leading to increased energy absorption over
thermoset polymers. Thermoplastics are also easily recycled which leads to waste reduction benefits [2].
A few drawbacks exist when working with thermoplastic resins. Thermoplastics often require a
higher processing temperature than thermosets. PEEK, for example, melts at approximately 340° C and
requires processing around 385° C. Many thermoplastics do not bond well to fiber constituents resulting
in poor interface adhesion and reduced mechanical properties. This problem can be avoided with good
consolidation techniques and the use of a binding agent at the fiber-matrix interface. Benefits in using
thermoplastics still outweigh the costs. This study highlights the design and manufacturing processes used
in the manufacture of a thermoplastic military helmet.

Composite Thermo Hydroforming
As discussed above, a key advantage of thermoplastic matrix composites is the ability to be
heated and reformed several times before the final net shape is created. A wide number of forming
processes can be used to form composites utilizing this type of matrix. Perhaps the most common of these
methods is matched die molding. This method is similar to sheet metal forming in that the composite is
acted on by a die of the desired geometry. This die pushes the composite into a cavity which is a mirror
image of the other die but slightly larger. As the two matched dies come together the composite is
compressed into a cavity that dictates its final shape and thickness. The basic tooling and process can be
seen in Figure 1.

2

Flat composite panel is
preheated and placed
within forming dies

Dies close over panel,
forming it to the desired
shape and thickness

Figure 1 - Matched Die Molding
Typically the tooling used in match die forming is heated to keep the composite at the proper forming
temperature as it is deformed to its final shape. Heating and cooling of the tooling is necessary to keep the
composite at its forming temperature and to accelerate the solidification process. This heating and cooling
can significantly increase process time. The dies are only able to exert force downward onto each other.
In the case of deep drawn parts, the normal direction of the die surface may differ significantly from the
force application vector. This makes controlling thickness and applying even forming pressure difficult.
This can lead to parts that are not well consolidated or have significant variations in thickness.
Thermo hydro-forming is a composite forming process that utilizes a heated pressurized fluid to
conform the composite material to a punch of the desired part geometry. This process is similar to sheet
metal hydro-forming in that it utilizes pressurized fluid to form the blank to a male plug (punch). This
process was developed and patented in the Michigan State University Advanced Materials Manufacturing
Laboratory. The hydroforming press and process will be described in more detail in the Experimental
Methods section.
This method of forming presents several advantages over match die forming. Since fluid is used
to form the material, only a single punch is needed where as in matched die forming a male and female
3

die is required to form the part, leading to reductions in tooling cost and complexity. The heated
pressurized fluid can be used to keep the composite at the appropriate forming temperature, eliminating
the need for tool heating. Since the forming force utilized is applied by pressurized fluid, the force vector
that conforms the material to the punch is always normal to the material surface. This evenly distributed
forming force has been shown to reduce out of plane warping significantly; especially in high shear areas
of the formed part. This allows the hydroforming process to form deep drawn parts and other challenging
part geometries. This is the greatest benefit that composite hydroforming offers over other thermoplastic
composite forming methods.

Topics Investigated During This Study
The focus of this study was to develop a thermoplastic composite processing methodology and
system. This goes far beyond just the forming of the composite blank. Rather, it involves developing a
deep understanding of the microstructure, mechanical response and failure mechanisms of the composite
system being formed. Understanding these properties is critical in developing a process and setting
forming conditions that will produce a high quality part.
The first portion of the study involves understanding the consolidation of the composite material
to be formed. Two separate composite systems were investigated during this study; carbon fiber
reinforced polypropylene and the ballistic composite Spectra Shield SR-3136. Spectra Shield is a
thermoplastic composite consisting of highly crystalline ultra-high molecular weight polyethylene fibers
reinforced by a sparse matrix consisting of a proprietary mixture of lower density polyethylene. A single
layer of this film is made up of four unidirectional layers of UHMWPE fibers oriented [0/90/0/90]. This
film is cut and laid up to the desired areal density where it is then consolidated using a heated platen
press. Honeywell provides a recommended pressing temperature time and pressure in order to achieve the
best ballistic performance possible.
The carbon fiber polypropylene composite was selected to display the versatility of the thermo
hydro-forming system. This composite is formed from 5x5 satin woven 6k carbon fiber and block
4

copolymer polypropylene film. The constituents are laid up into a special mold in alternating order
forming a sandwich structure. The mold is then loaded into a heated platen press where heat and pressure
are applied to the mold forming the plate. For this material, only a consolidation study is performed.
The next portion of this study is obtaining the mechanical properties of the consolidated
composites. Mechanical properties are obtained to provide material models with the needed engineering
constants in order to make the simulation produce realistic results. A preferred fiber orientation (PFO)
material model is utilized in this study to predict the results of the composite forming process. Several
common engineering constants are required for the characterization of the laminated composite in this
material model. The methods used to obtain these constants will be discussed later in the Material
Characterization section.
The main focus of this study is the forming of composite blanks into useful parts. A 300 ton
thermo hydroforming press is used to form the flat blanks into military ballistic helmets. This press is
based off a small 40 ton press that served as a proof of concept for the process of composite
hydroforming. The 40 ton press was successful in forming 4” diameter hemispheres out of polypropylene
reinforced fiberglass. The 300 ton press is capable of forming pieces in excess of 10” in diameter and up
to a depth of 8”. Forming pressures of up to 2400 psi can be applied, allowing the forming of thick
laminated plates. Many changes were made to the press in order to produce high quality parts. Changes
included developing a new bladder system to protect the composite, implementing new die sets and
changing the blank clamping scheme. These methods and process changes will be described in more
detail in the Experimental Methods section.
Finite element analysis was performed in order to develop a deeper understanding of the blank
material and of the hydroforming press. Simulations were performed in Abaqus CAE using a user defined
material model. As stated before, a PFO model that was developed by Mike Zampaloni during his PhD
studies was utilized [3]. The main advantage of this model is its ability to track the alignment of the
reinforcing fibers as the material is deformed during the forming process. Several updates were made to
the material model during this study, including the implementation of cohesive surfaces, updating the
5

transverse shear stiffness and using material properties obtained during material characterization. The
results of these simulations were compared to experimental forming results in order to assess the validity
of the material model. The material model can also be used to predict the effect that changing the forming
conditions will have on the final part. A simulation with sufficient accuracy will ideally be used to
determine the optimal forming conditions without having to physically form a single part. The numerical
simulations and material model will be further discussed in the Numerical Methods section.

6

Literature Review

Consolidation of CFRTP Composites
Consolidation quality of composites has a profound effect on their mechanical properties.
Excessive void content and poor fiber matrix adhesion can greatly limit the mechanical properties of the
consolidated laminate. In the 1980’s, thermoplastic composites began to gain traction because of the
potential to significantly reduce manufacturing costs. Up to this point, composite materials had been
employing thermoset matrix systems such as polyester, vinylester and epoxy.
In order to engineer optimal matrix fiber adhesion for thermoset or thermoplastic composites, the
microstructure of the interface must be well understood. Bascom and Drzal studied the interface of carbon
fibers as the bonding mechanisms used in carbon fiber composites which was not well understood at this
time [4]. Techniques used to construct carbon lattices from polyacrylonitrile (PAN) are presented. The
ways in which these lattice structures pack as they are drawn into fibers has great effect on the crystal
alignment of the fiber and consequently the mechanical response. The surface of a carbon fiber is
composed of the same material structure as can be found at the core of the fiber which is graphitic crystals
(hexagonally packed carbon atoms). Oxidation is used to bond functional groups to the surface of the
fiber which can be used in bonding mechanisms for epoxy. In order for this to happen the resin must
make intimate contact with the inclusions. This is true for both thermoplastic and thermoset composites.
Recent studies have been conducted by Semoto et al. showing the formation of bonds between
epoxy resins and functionalized oxygen based molecules on the surface of graphite lattices [5]. This study
involved molecular dynamic simulation of hydrogen bonds between a fragment of epoxy resin and OH
and COOH functionalized molecules on the surface of carbon fibers that are the product of oxidation
surface treatments. The simulation shows strong atomic bonds reinforcing the interface adhesion between
fiber and matrix constituents but is highly idealized and beyond the scope of this study. Nevertheless, this

7

study shows the crucial importance of functionalized bonds in the adhesion in carbon fiber-epoxy
composites.
Thermoplastic polymers do not typically form these strong bonds with fibers but instead rely on
an entangled polymer chain network to transfer load to the fibers. Upon melting these molecular chains
become less entangled and can slide and flow past one another. However the long chains do not break
down resulting in relatively high molten viscosities. Ye performed a study which investigates how the
high viscosities of these polymers can affect consolidation of composites [6]. The composite system used
in this study is woven CF pre-impregnated with polyether ether ketone (PEEK). The consolidation of a
laminate from individual prepreg sheets is performed using a heated platen press. Ye found that heat,
pressure, and time are the three main variables that affect the consolidation quality of the composite. In
order to achieve the best properties, the fiber yarns (tows) must be wet out with molten polymer. The key
factors shown to vary the wet out of the fiber tows is their permeability, their radius, the matrix viscosity
and the pressure of the molten matrix. For laminates consolidated under varying conditions, the void
volume fraction, ultimate tensile stress and transverse shear modulus are measured. It is shown that
increases in void volume fraction lead to decreases in the measured mechanical properties.
By analyzing the matrix flow equation of the previous paper, it is clear that reductions in matrix
viscosity will lead to faster consolidation with fewer voids. Decreasing a thermoplastic polymers
viscosity can be accomplished by raising its temperature above the melt point. For many high performing
polymers such as PEEK, the temperature cannot be raised significantly above the melting point because
of the onset of thermal degradation. Thermal degradation and its effect on the microstructure of polymers
is described by Beyler [7]. During thermal degradation, thermoplastic polymers are attempting to
volatilize but cannot due to their high molecular weight (polymer chain length). As the degradation
temperature is passed, polymer chains begin to break apart. This reduction in molecular weight can
significantly reduce the mechanical properties of the material and further heating will bring on the
formation of char as broken polymer chains begin to cross link forming chemical species that are more
carbon rich. Therefore, it is ideal to thermoform with polymers that possess a large gap between the
8

melting temperature and thermal gradient temperature to take advantage of reductions in viscosity.
Polypropylene is ideal in this application as the melting temperature and degradation temperature differ
by ~88° C.
Polypropylene has become a material of interest for fiber reinforced composites recently due to
its low price. Mass production industries such as automotive, packaging and consumer goods have taken
an increased interest in PP composites due to their potential for high performance, low cost and excellent
formability. Zampaloni et al. investigated the consolidation and forming of kenaf natural fiber reinforced
polypropylene [8]. Several consolidation techniques and PP blends were evaluated during this study. A
method was developed to consolidate PP-Kenaff composites into sheets. A small amount of maleated
polypropylene (MaPP) was added to the PP blend to improve bond quality between the kenaf fibers and
PP matrix. Zamplaoni also showed that these pre-consolidated sheets can be effectively formed by the
hydroforming process similar to the one used in this study.
MaPP is commonly used to strengthen the bond between PP and reinforcing fibers. In his study of
PP composites, Shubhra states the MaPP can be used to increase the matrix-inclusion bond strength and
stiffness of several types of thermoplastic composites including fiber types of hemp, cotton, carbon, glass,
and several others [9]. On the molecular level, most reinforcing fibers used in composite applications are
polar in nature or have polar molecular groups on the surface. Since the molecular structure of PP is nonpolar, only mechanical forces such as friction and polymer chain entanglement resist separation at the
fiber-matrix interface. The introduction of polar maleic anhydride (MA) to the polymer chain, allows a
physical bond to attach between the fiber and matrix. A small amount of MaPP is required to improve the
properties of the PP composite (only about 2-4% by weight of the matrix). Excess MaPP has been shown
to decrease tensile and bending strength.
Several studies have been focused on adding MaPP to CF-PP composites as well. Karsli et.al.
show increases in the strength and stiffness of composites prepared by twin screw extrusion of chopped
fiber CF-PP composites [10]. The composites in this study were prepared with a 2-20% fiber weight
fraction. Maleic anhydride graphed polypropylene (MAH-g-PP) was added to the matrix constituent in
9

amounts 0-5% by weight. Since preparation of the samples was carried out via plastic injection molding,
the MAH-g-PP was evenly distributed throughout the matrix. Increases in both tensile strength at break
and tensile modulus were obtained by increasing both the fiber and binder content. Micrographs of the
failure sites showed an increase in fiber-matrix adhesion in the MAH-g-PP blended composite over those
without the matrix compatibilizer.
Giruad et. al. investigated the effects of other sizing agents for CF reinforced PEEK [11]. The
composite is consolidated using a resin film infusion technique similar to the one in this study. Prior to
consolidation, a sizing of polyetherimide (PEI) is applied to continuous CF tows by spraying a prepared
emulsion of the polymer sizing directly onto the fiber surface. The emulsion consists of polymer
suspended in water and other solvents which is sprayed directly onto the fiber surface. Micrographs show
an enhancement in bond quality for the sized fibers when compared to the un-sized fibers.
Ultra high molecular weight polyethylene (UHMWPE) composites are a more recent
development in the world of composites. These composites have a high fiber volume fraction (typically
80% or above) and are employed primarily in blast or ballistic protection devices (body and vehicle
armor). Since these materials are in an early stage of development, literature on particular types of
UHMWPE composites is sparse. Most of the literature presented for these types of materials focuses on
products that are similar but not exactly the same as SR-3136.
One of the only pieces of literature existing specifically for SR-3136 comes from the
manufacturer and provides consolidation and forming recommendations [12]. This document
recommends that the processing temperature should be between 240°F and 260°F when using the
compression molding process. A pressure between 500 and 5000 psi should be applied to fully
consolidate the sheets. The laminate is then cooled under pressure until the center of the material is below
125°F. This document also states that improved ballistic performance may be achieve through the use of
higher consolidation pressure. It should be noted that these recommendations are made for the entire
Spectra Shield line and not just for SR-3136. Optimal forming conditions for SR-3136 may lie
somewhere within these relatively broad guidelines.
10

UHMWPE fibers have been used to reinforce high demand structures over the past couple of
decades. Since the fibers have a density less than water, they have been utilized in marine applications to
take advantage of their buoyancy. Heavy duty ropes, sails and fishing lines as well as body and vehicle
armor have been offered by both Honeywell and DSM, the two main manufacturers of UHMWPE fiber
composites. These fibers have been offered in various forms as well, ranging from matrix free braids and
woven fabrics to unidirectional prepreg which is held together with a sparse matrix constituent that is
typically held proprietary.
Xu et. al studied UHMWPE composites in their matrix free form called Spectra Cloth [13]. A
method for forming this cloth was developed called high-temperature high-pressure sintering. This
method involves bringing the polymer cloth close to its melting temperature allowing just the surface of
the fibers melt. Pressure is applied allowing the molten fiber surfaces to unite and become bonded to one
another. Care must be taken to avoid over-melting the fibers during processing however. Excessive
melting destroys the highly crystalline structure of the fibers leading to drastically reduced strength and
stiffness.
In a later study, Xu and Farris compare the ballistic performance and formability of Spectra cloth
with Dyneema Fraglight and Spectra Plus PCR [14]. Dyneema is a non-woven felt-like material with
random fiber orientations, and the Spectra plus PCR consists of fibers impregnated with a proprietary
matrix and stacked in a [0°/90°] tape. Spectra cloth performed the best in impact energy absorption
but only over a very narrow temperature window (2° C). Both the Dyneema and Spectra Plus PCR
absorbed an impact energy of at least 60 KJ/m over a 10° C spread in processing temperature. Even
though these values are slightly lower than the max energy absorbed by Spectra cloth (78.1 KJ/m @
150°C) the wider band of acceptable processing temperature allows for better quality assurance in
the end product. Comparisons in material formability showed that the Spectra cloth and Dyneema
were easily formed with Dyneema requiring slightly less punch force. Problems with forming
Spectra Plus PCR are attributed to matrix squeeze-out and other matrix flow issues. Since all of

11

these materials are processed using the same conditions and same method (match die molding), it
is impossible to say whether or not optimizing the process for each material could result in better
performance or ease of forming.
Greenhalgh et. al performed a more comprehensive study on the effects that processing
parameters have on the response of Dyneema UHMWPE composites [15]. Micrographs revealed that the
matrix constituent of this composite is very sparse and void ridden. Unlike in structural composites, this
matrix structure assists in energy absorption and is a key feature in ballistic composites. Further analysis
revealed that increases in processing pressure lead to changes in the inter-laminar matrix thickness.
Differences in matrix thickness affect the toughness of different modes of fracture and delamination.
Even though processing pressures show a difference in the failure mechanism, the overall target response
between panels processed under low and high pressures is the same.

Material Characterization Methods
When analyzing the forming of materials of any type, it is important to understand their
mechanical response to applied deformations. This is easily done with homogeneous materials that do not
change their response when loading is applied along different orientations. Obtaining the response of
composites is not as trivial due to the anisotropy created by the orientation of the reinforcing fibers. This
type of material response and structure calls for the use of an orthotropic material model coupled with a
plane stress assumption. This model will take into account the stiffness along the fiber axis and
perpendicular to it as well as the Poison’s ratio and shear modulus. With these engineering terms, a
reduced compliance matrix can be assembled that will predict the response for a given stress.
Zampaloni developed a material model which tracks preferred fiber orientations of random
orientation, woven or unidirectional composite sheets [3]. The method was first developed to predict the
response of randomly oriented fibers. With woven or unidirectional composites, it is easy to visually
determine which orientations will give the stiffest response but this cannot be done with composite
12

possessing random fiber orientations. By using a squeeze flow test, he was able to determine which
directions possessed the greatest amount of reinforcing fibers and which orientations had the least. By
assuming that orientations with the greatest fiber content would be the stiffest, he was able to determine
the orientations worth tracking in his material model. After identifying these orientations, material
properties were extracted.
The first property to be determined is typically the Young’s modulus. This is a routine
characterization for composite materials and is typically performed in accordance with ASTM D 3039
[16]. This standard utilizes a rectangular specimen with specialized grip tabs to determine the axial
response of the composite being evaluated. For this standard, both the Young’s modulus and Poisson’s
ratio of a composite material can be determined.
This standard does not work well with all composites however. The use of this standard with
UHMWPE composites often causes inter-laminar shear failure at the gripped area. Russell et al.
developed a specialized gripping mechanism in order to resolve this issue [17]. The gauge section of the
specimen is small in comparison to the gripped area. This ensures that failure does not occur in the
gripped region.
Direct measurement of the shear modulus is also necessary. Two methods for evaluating the shear
modulus are reviewed; the Iosipescu shear test [18] and V-notch rail shear test [19]. Both of these tests
rely on a rectangular specimen with notches cut in order to concentrate shear stress at the neck. The main
difference in these methods is how the specimen is loaded. The Iosipescu method uses a special fixture
which loads the specimen along its edges by exerting compressive force. The V-Notched rail shear
method grips the faces of the specimens using a special fixture and shears the specimen in tension. Since
these methods apply different loads in different locations of the laminated coupons (face or edge) it is
reasonable to expect that they will give slightly different values for the shear modulus. Yan-lei et al.
performed a study where these two methods were evaluated [20]. Although the Iosipescu shear specimens
can produce good results, they claim that edge crushing can be an issue due to the way that the specimen

13

is loaded. The V-notched rail shear method is preferred because of the elimination of this unacceptable
failure mode and the use of a larger gauge section.
The determination of the two Young’s moduli, the Poisson’s ratio and the shear modulus are all
that is needed to characterize the constitutive relation. For large deformations, however, inter-laminar
separation plays a role. This effect becomes pronounced in materials where delamination commonly
occurs, such as in UHMWPE fiber composites. This makes determining the inter-laminar traction
separation response necessary in order to fully characterize the material. Liu et al. utilized a shear test
method to determine the traction-separation response of Dyneema HB26 undergoing mode II interlaminar fracture [21]. A special double notched beam is loaded in tension in order to initiate delamination.
The shear stress can be calculated using the area of the planes in which fracture occurred and
displacement is measured using a clip gauge. The exact traction separation quantities are then determined
via the calibration of a similar FEA model.

Thermoplastic Composite Forming Methods
The manufacturing of composite materials requires many considerations in order to produce a
high quality end part that is visually appealing and performs well. Many methods have been developed
exclusively for thermoset materials. These methods include wet layup, vacuum molding, resin transfer
molding, vacuum assisted resin transfer molding, autoclave, and a host of others. These methods take
advantage of the low temperature and viscosity of the pre-cured resin and often rely on skilled laborers to
mix and distribute the resin properly. The use of prepregs eliminates the need for the manual mixing of
resin and ensures a uniform distribution. Once laid up, these composites are typically cured in an autoclave oven. It is by this method that the highest quality aerospace composite parts are made today. The
need for skilled labor and massive autoclaves drives price up significantly as does the cost of the prepreg
itself.

Thermoplastic composites have the ability to surpass thermosets in the ease and cost of

14

manufacturing. Fully automated methods can be utilized for more consistent parts and shorter cycle times.
These composites also feature a nearly infinite shelf life further increasing their economic value.
By the late 1980’s, thermoplastic composites had risen in popularity and forming methods were
developed with the ability to produce high quality composites. Richard Okine was among the first to
evaluate these methods [22]. Several thermoplastic composite forming techniques were evaluated
including match die forming, vacuum forming, diaphragm forming and hydroforming. Of these the most
commonly used today is match die forming. Match die forming allows for excellent temperature control
and the creation of parts with high quality surfaces. This technique has several drawbacks though. Deep
drawn parts are difficult to form effectively since forming pressure is only applied axially. This leads to
wrinkling and non-uniform part thickness. Resin can also be squeezed away from the fiber structure
leading to resin poor areas.
Yin et.al. performed a more recent study which highlights some of the shortcomings of match die
molding [23]. A layer of pellets is placed between two layers of plain weave carbon fiber and is placed in
a furnace to melt the polymer. Once molten, the composite is transferred to the match die mold where it is
quickly formed into the shape of a wingtip. Most of the formed parts exhibited a resin rich tip indicating
molten polymer squeeze out. This is a common occurrence for deep drawn parts being formed using the
match die technique due to the way the molds apply force.
McCool et al. performed a comprehensive study of match die molding [24]. During this study,
effect of the composite formability and final part quality were investigated. Part temperature was closely
monitored and the concept of a blank forming temperature “window” was investigated. The cooling rate
of the mold was also changed in order to optimize the crystallinity of the polymer (PPS). A greater degree
of crystallinity was shown to improve flexural strength, making the cooling rate, mold tool temperature
and blank forming temperature all parameters of interest.
Another method of composite fabrication is injection molding, provided the fibers are relatively
short. Karmaker and Youngquist explored methods to incorporate polypropylene injection molded parts
reinforced with short natural fibers [25]. The reinforcing fiber constituent in this study was Jute at a
15

volume fraction of 50%. Uniform size fibers dispersed with random orientations give significant gains in
stiffness and strength over the unfilled polymer. The addition of MAPP acting as a binder further
increases the mechanical properties. Further analysis on the injection molding of short fiber composites
was performed by Gupta and Wang [26]. This study concentrated on the effects that plastic injection
molding had on the orientations of the dispersed fibers. It was found that both the injection location and
distance from the boundary layer of the mold had significant effect on the fiber orientation as the fibers
align themselves with the flow of the molten resin. This preferential material orientation must be taken
into account in the design of molds utilizing this process.
Pourboghrat et.al. patented a composite hydroforming process which demonstrated the ability to
form composite materials using pressurized fluid [27]. This process uses pressurized and heated hydraulic
fluid to form thermoplastic composites against a hard punch of desired geometry. The composite blank
can be physically clamped using a blank holder mechanism or hydrostatic fluid pressure can be used to
apply a more uniform clamping force. The main advantage this process presents over other forming
techniques is that fluid pressure acts over the entire composite surface, reducing out of plane material
warping. Since pressure is used as a “soft die”, part thickness remains uniform, unlike in match die
forming. Resin squeeze out is also reduced due to uniform force being applied over the surface of the
blank. Composite Hydroforming has been shown to work with a variety of material types.
Zampaloni showed the ability of this technique to form kenaf reinforced polypropylene natural
composites [8], randomly oriented glass fiber composites [28], and woven glass fiber composites [3].
Since pressurized fluid is used to heat and cool the composite blank directly, process time can be
drastically reduced making composite hydroforming an ideal manufacturing method for the automotive
and aerospace industries.

16

Numerical Methods for Simulating the Forming of Composites
Simulating the forming of composite structures is an important and often necessary practice in
order to determine optimal processing conditions and produce the highest quality parts. Through
numerical modeling, a plethora of boundary conditions can be applied and their effects on the composite
blank evaluated. Additionally, the modification of the laminate stacking sequence can also be varied,
allowing the engineer to investigate how these changes will affect the formed part. In order to achieve the
accurate simulation results, the composite laminate must be accurately modeled. This however is no easy
task as composite materials are inherently anisotropic. Other considerations such as inter-laminar ply
shearing, delamination and updates to the fiber orientations compound the complexity of the material
models that are used.
The most classic way in which laminated composite materials can be characterized is with the
ABD matrix [29]. The ABD matrix relies on a plane stress assumption allowing the 9x9 anisotropic
material stiffness matrix to be reduces to a 3x3 matrix which characterizes that material’s orientation
dependent tensile stiffness qualities as well as shear stiffness. This matrix is extend to 6x6 in order to take
into account the effects of bending stiffness and coupling that may exist between bending and in plane
extension.
This model is able to make good predictions for composite structures after having been formed
but it is not able to characterize the forming of composite materials with sufficient accuracy. Perhaps
most importantly, the model is not able to take into account changes in fiber orientation. Strains produced
by the forming operations change the orientation of the fibrous reinforcements which in turn change the
properties of the material. Also, the ABD matrix assumes perfect boding and load transfer between the
layers of the laminate.
Several material models have been developed which take into account changes in fiber
orientations. Yu et. al developed a material model which exclusively deals with plain weave composite
fabrics [30]. This material model directly takes into account the structure of the fabric. Warp and weft

17

yarns are characterized by the number of fibers per tow and the apparent material properties of an
individual fiber. A unit cell is built based on the constituents of the structure and fabric. Numerical
simulations show that this model is able to accurately predict changes in the angle between the warp and
weft fiber as the fabric is elongated when loaded 45° offset from the warp and weft orientation. Good
results were also obtained when forming a hemisphere by a match die molding.
This material model was improved upon by including additional parameters that took into
account the more complex phenomena when a woven fabric deforms. A significant improvement was
made when taking into account the undulation geometry between the warp and weft yarns [31]. This
study characterized the change in shear stress due to warp and weft yarn geometric considerations. The
model predicts that shear stress increases in a non-linear fashion as shear angle increases. This is due to
shear locking that occurs as the fibers interact with one another. This non-linear response in shear allowed
for accurate predictions in wrinkling severity and location when simulating the deformation of a fabric as
it experienced hydroforming type boundary conditions. An additional study took into account the
asymmetric bending stiffness of the reinforcing fibers [32]. The asymmetric bending stiffness assumes
that the reinforcing fibers do not possess the same stiffness in tension as they do in compression. This
difference leads to a shift in the neutral axis of the fibers when equilibrating stresses during bending. This
property is incorporated into the material model and shows good results when predicting the shape and
fiber shear when a fabric is draped over complex geometry.
Boisse et al. developed a similar material model that takes into account geometric factors such as
undulation geometry and shear locking [33]. This material model takes into account non-linear uniaxial
loading results. As the fibers yarns are loaded they first need to straighten themselves due to the
undulation inherent in biaxial weaving. As the fibers straighten out, the slope of the load increases and the
fibers are directly strained leading to increased stiffness. This leads to an overall non-linear response. The
behavior is evaluated biaxially to establish a weave geometry dependent Poisson’s ratio. This model has
been shown to accurately predict fiber shearing.

18

Another material model was developed during the PhD study of Mike Zampaloni [3]. A
continuum model was developed that focused on tracking orientations of fiber features. These orientations
of interest or preferred fiber orientations (PFO) refer to a particular orientation at which a composite sheet
material is stiffest. Once determined, the orientations are characterized by obtaining the axial and
transverse Young’s modulus, Poisson’s ratio and shear modulus. The model takes into account these
PFOs and their related material properties along their initial orientations. During FEM simulations, a
unique deformation gradient tensor acts upon each element which not only deforms the element but also
changes the orientation of the PFO’s. At the end of each step the stiffness of the PFOs are updated and
used to calculate the material stiffness which changes as the PFOs change. This material model is used in
this study and will be further described in the numerical methods section. The model has been used to
determine the best forming conditions during the hydroforming process and has provided excellent
predictions of the final formed shape using methods developed by Abedrabbo et al. [34].
The aforementioned models all possess a common flaw; they either assume perfect lamination
between the composite layers or only model a single layer. The assumption of perfect bonding is
insufficient due to the finite strength of the inter-laminar adhesive. During forming, and especially when
deep drawing, the assumption of perfect bonding between layers can lead to complex out of plane
warping that does not manifest itself in experiments. Typically this behavior follows a traction-separation
response and can be incorporated into FEM models through cohesive elements. Harper et al. demonstrates
how this behavior is implemented in the study of the mechanical response of panels delaminating [35].
Specialized cohesive zone elements are implemented at the predicted crack boundary, in this case the thin
adhesive film between two laminated composite strips. These elements are given special fracture
mechanics material properties. In its simplest form, fracture mechanics can be characterized by a bi-linear
traction separation curve. An initial elastic stiffness gives the slope for the undamaged elongation on the
adhesive until the damage initiation point is met. Displacements beyond this point undergo linear
softening until failure.

19

Consolidation of Flat Panels and Material Characterization

The materials in this study are selected to exhibit the capabilities of the thermo-hydroforming
process. Materials are laid up to their desired thickness or areal density an consolidated into flat plates.
This allows the forming process to occur without having to consolidate simultaneously. Consolidation of
the sheets prior to the final forming step allows for a reduction in processing time due to the extended
heating, pressing and cooling time required to reduce void fraction and increase inter-laminar bond
strength. Theoretically, a loose ply stack could be loaded into the press, formed and consolidated all in
one step but increases to the cycle time would be expected. A far higher throughput can be achieved by
using pre-consolidated laminates to avoid a bottleneck occurring at the final forming stage.
The laminates are consolidated and formed using a large heated platen press capable of housing
circular blanks up to 36” in diameter. This press is used to consolidate both the spectra shield and carbon
fiber blanks investigated during this study. Heating and cooling rate is not able to be precisely controlled
with this press; both occur at a very slow pace. Also, the press makes use of four independently controlled
heating elements which cool at drastically different rates, possibly leading to residual stresses and
warping post cooling. If these flaws exist they are miniscule and have been ignored.

SR-3136
Blank Consolidation and Shape
Since Spectra Shield comes as a prepreg, a recommended consolidation procedure is provided by
Honeywell [12]. The procedure for consolidating the flat sheets is based off the recommendations of this
document. Other researchers claim that temperatures of up to 300°F can be used to sinter the fibers in
UHMWPE composites that do not contain a matrix constituent [13]. The consolidation technique in this
study required that a small portion of the highly crystalline fibers be melted and fused to its neighboring
fibers requiring a higher processing temperature. Temperatures this high become dangerous due to the

20

onset of excessive fiber melting which ruins the highly crystalline structure of the polymer chains, greatly
reducing their stiffness and strength. SR-3136 contains a sparse matrix constituent of low density
polyethylene (LDPE) which melts at a lower temperature than the UHMWPE fiber, allowing the
composite to be consolidated without the risk of damaging the load carrying fibers.
Taking into account the recommendations from Honeywell, the following procedure for
consolidation of UHMWPE sheets was utilized:
1. The SR-3136 is cut to the desired shape and stacked to the desired thickness or areal density.
2. The ply stack is loaded into the heated platen press. Teflon release film is used as a barrier
between the heated platens and the SR-3136 to prevent the composite from sticking to the press
after cooling.
3. The press is closed such that the upper and lower platens make contact with the composite in
order to ensure effective heat transfer. The temperature of the platens is set to the desired forming
temperature of 250°F.
4. Once the platens reach 250°F an additional ten minute waiting period is utilized to allow the
thermal gradient within the core of the composite to dissipate, ensuring the composite is evenly
heated
5. After waiting 10 minutes, a pressure of 500 psi is applied to the composite (this pressure refers to
the force exerted over the surface area of the blank, not hydraulic pressure). This pressure and
temperature is held for 10 minutes.
6. After 10 minutes under heat and pressure, the heater elements are disabled and cooling measures
are employed. The holding pressure is maintained throughout cooling.
7. Once the temperatures of all zones are below 150°F, pressure is released, the platens are opened
and the composite can be removed.

Blank shape is determined through the use of an Abaqus/CAE add-on called Composite Modeler
for Abaqus (CMA). This add-on is used primarily to construct models for complex composite
21

components that have irregular geometry or strange stacking sequences. The most useful tool in this addon generates a flat pattern. The flat pattern represents the shape of the composite before it is draped over
and formed to the punch. The flat pattern is generated for the desired final shape and includes the extra
material that will be trimmed off after the parts have been formed. It is important to include this extra
material because the material acts as a barrier which contains the pressurized forming fluid in the thermohydroforming process.
In order to generate the flat pattern, the shape of the formed part plus the excess material must be
modeled as a sheet. This sheet model is then loaded in CMA and meshed. Once meshed, the composite
type (unidirectional, bidirectional, woven, etc.), maximum shear angle and region are assigned. The flat
pattern is then generated and shear angles are plotted over the surface of the formed part. This process is
highlighted in Figure 2.
Part geometry is fed
into CMA where flat
pattern and fiber
orientations are
predicted

20.9”

Desired final shape
plus excess clamping
material

Fiber
Orientations

Figure 2 – CMA Blank Shape Prediction Capabilities
Flat pattern prediction is computed by calculating the Gaussian curvature over the entire surface
and unfolding it onto a flat plane. This prediction only takes into account one layer and does not take into
account stretching of the material but it provides a quick prediction for ply shape.

22

Material Characterization
Material characterizations for spectra shield SR-3136 and the CF-PP are aimed at populating the
PFO material model with the appropriate parameters. The fiber properties for this model are linear elastic
so the material parameters required are the Young’s Modulus (E), Poisson’s ratio (ν), and the shear
modulus (G). It is preferred to characterize the traction separation response as well but machining the
double notch shear specimen required to obtain these properties has presented many challenges.
Machining SR-3136 has led to many issues. Unlike epoxy reinforced composites, these materials
are very difficult to machine. Standard mills and drill bits do not pull chips away from the material very
well and typically lead to the edges of specimen being frayed. Delamination and back face deformation
are other common flaws that occur during the machining of Spectra Shield. Water jetting has recently
been attempted and has shown very good results with minimal delamination. This method alone should be
used in the future to prepare test specimens.
Shear Modulus
Perhaps the simplest specimens to prepare are the ones used to determine the shear modulus. Two
standards were selected and utilized for the evaluation of the shear modulus; the v-notch rail shear test
[19] and the Iosipescu shear test [18]. The specimens for these two tests are similar in that they are
rectangular in shape and feature aligned notches to localize shear stress at a predictable and easily
measurable cross section. See Figure 3 for the recommended geometry of the different shear specimens.

23

Iosipescu Specimen

Rail Shear Specimen

Figure 3 – Shear Modulus Test Specimens
The Iosipescu specimen is loaded by a special fixture that applies compressive forces that initiate
shearing in the gauge section of the specimen. Force is applied on the edges of the specimen. The V-notch
rail shear specimen is loaded on the faces of the gripping tabs which initiate shearing in the notched gauge
section. The fixtures used to load these two specimens can be seen in Figure 4. Difficulties with the Vnotch rail shear specimen were observed during initial tests. The fixture itself is very heavy and
cumbersome resulting in the damage of specimens during mounting. For this reason the Isosipescu shear
results will be reported and used.

24

Compressive force
applied to shear
specimen

Tensile force
applied to
shear
specimen

V-Notch Rail
Shear Fixture

Iosipescu Fixture
Figure 4 – Shear Fixtures

The Iosipescu specimens are evaluated using a uniaxial tensile testing machine. A load cell is
used to measure the amount of force required to deform the specimen. The force is then normalized by the
area to calculate the shear stress. Strain is obtained by a strain gauge mounted to the specimen in the
direct vicinity of the gauge section. Strain is measured in the +45° and -45° directions. The absolute
values of these strains are summed in order to obtain the shear strain. The shear modulus (G) is then
calculated by taking the change in shear stress over the change in shear strain over a preselected chord.
These methods are given in more detail in the ASTM recommendations ( [19], [18]). Four specimens are
prepared and tested to obtain the shear modulus. Figure 5 shows the obtained stress-strain curves.

25

Shear Stress-Strain Curves

3

Stress (Mpa)

2.5
2

Specimen 1
Specimen 2

1.5

Specimen3
1

Specimen 4

0.5
0
0

0.005

0.01

0.015

0.02

0.025

0.03

Strain
Figure 5 – Shear Stress Strain Curves
Specimen 1 and specimen 2 clearly give irregular results. The reason for these issues is difficulty
in bonding the strain gauge to the specimen. The trends for these specimens indicate a high stress for a
relatively low strain which is readily explained by ineffective strain gauge adhesion. Specimens 3 and 4
indicate the expected linear trend up to a point where once again, a decrease in strain is observed due
again to failure of strain gauge adhesive. The chord is selected a safe distance away from this adhesive
failure in order to obtain the most accurate data. The shear modulus values and their average can be seen
in Table 1.

Specimen

Cross-Section Area (mm^2)

Δτ (MPa) Δε

3

38.76

0.50567

0.004337 116.59

4

39.56

0.55361

0.004066 136.16

Shear Modulus (MPa)

Average 126.37
Table 1 – Shear Modulus
Ideally, at least three successful test specimens are preferred but difficulty in obtaining good
results prevented additional tests. Also, room temperature tests do not provide the necessary data to

26

characterize the composite as it is formed at elevated temperatures. Since the material model relies on a
linear elastic-transversely isotropic assumption, no temperature dependent terms are used in the
calculation of the materials stiffness. The most appropriate way to characterize the material would be at
the temperature at which it is formed.
Young’s Modulus
SR-3136 is difficult to fit empirical relations to due to poor fiber-matrix adhesion and void ridden
matrix constituent. This makes using empirical relations to estimate the Young’s modulus wildly
inaccurate. Unfortunately, these issues also make material characterization difficult.
Characterization via ASTM standard D3039 was attempted first [16]. This standard recommends
the use of a rectangular specimen that is gripped directly by the jaws of the machine. The first few
attempts resulted in the gripped plies shearing away from the bulk of the laminate. This damages the
material resulting in low force measurements. Emery cloth was utilized at the gripping interface per the
recommendation of the standard. This resulted in the gripped plies shearing away from the bulk as well as
the gripped layer being torn at the gripping boundary. Examples of this issue can be seen in Figure 6.

27

Direct gripping causes
load to localize in the
gripped layer

Gripping with emery
cloth results in
delamination of the
gripped surface

Figure 6 – Difficulties with Tensile Specimens
It was determined that this delamination was due to the low quality bond created by the void
ridden LDPE matrix constituent. Based on this assumption, specimens were machined with a larger
gripping face based on the geometry of the specimens in ASTM D638 [36]. These specimens showed
better results but still seemed to be localizing most of the load on the gripped layers.
Specimens show
evidence of loading in
the gauge section but
still fail at the grips

Figure 7 – Dog Bone Specimens

28

Data were analyzed from this round of tests regardless of the failure mode. Force was measured by a load
cell and normalized by the cross sectional area of the specimen in order to calculate stress. Axial strain
was directly measured using a laser extensometer. The calculated stress strain curves for the dog bone
specimens can be seen in Figure 8.

Figure 8 – Tensile Test Results
The stress strain curves exhibit a relatively tight grouping in the linear elastic region of loading (below
1% strain). When examining the failure region of the specimens, it was clear that two main modes of
failure occurred. Sample 2 seemed to exhibit delamination of the outer layers, resulting in little to no
additional load for increasing strain. The remaining specimens looked as though they experienced a
decrease in strain. This can be attributed to the gripped layers ripping as was seen with the previous
specimens. Since these failures were within the grip area, concern was raised as to the validity of the
results.
The stiffness for each of these specimens was calculated by extracting the change in stress and
strain for a small chord between 0% and 0.5% strain. This data can be seen in Table 2.

29

Specimen

Cross-Section Area (mm^2)

1
2
3
4
5

Δτ (MPa)
65.847
75.810
71.653
73.519
79.137

45.085

Δε

Modulus (GPa)
0.004
0.004
0.004
0.004
0.004
Average
STD

16.408
20.326
17.943
18.366
19.976
18.604
1.302

CV

0.070

Table 2 – Tensile Test Data
The stiffness results are fairly similar to one another with the exception of specimen 1 which seems to
experience slipping though the grips at relatively low strains (see Figure 8). The standard deviation is 1.3
GPa which indicates a relatively tight grouping. This indicates that the loading is repeatable and
uniformly distributed at least at low strains. However the gripping interface is also experiencing loading
hence the failure occurring in this region for all specimens. Undoubtedly the effect of inter-laminar
stresses is affecting the stress-strain response leading to decreased accuracy. Additionally, the ultimate
stress is ~33% of the ultimate stress obtained during the characterization of a similar UHMWPE
composite (Dyneema) by Russell et al. [17].
An additional tensile specimen was fabricated to ensure that the response being obtained during
the characterization is purely a function of the material and not of the gripping scheme. The gauge section
was very small when compared to the gripping area in order to ensure that the stress is localized though
the gauge section and not through the bond line of the gripped layer. The specimen geometry was used
based upon the recommendation from the Russell study mentioned above. The geometry of the specimen
can be seen in Figure 9.

30

Specialized grips are used which apply force over the
specimen face and prevent slipping via though bolts

Specimen features a thin gauge
section with large gripping faces
Figure 9 – Tensile Specimen from Russell Study
Results from initial tests have shown an increase in the ultimate stress to ~410 MPa. This
indicates that the results seen in Figure 8 are experiencing reductions in load carrying capacity due to
failure initiating at the gripped area. Unfortunately, further difficulties with strain gauge mounting
prevented strain measurements. The specimen in Figure 9 is currently being redesigned to allow easier
mounting of the strain gauge, in order to determine the stiffness more accurately.

Polypropylene Reinforced Carbon Fiber
An additional material was also selected to show the versatility of the hydroforming process. A
consolidation technique was devised for this material for characterization and forming of the material.
The carbon fiber used in this study is a 5x5 6k satin woven fiber using HTS40 E-13 carbon fibers from
Toho Tenax. The matrix constituent is block copolymer polypropylene film from Bloomer Plastic
Incorporated. Film of thickness 0.008” is used to target a fiber volume fraction of 50% in the final
composite. The process used to consolidate the composite is referred to as resin film infusion (RFI) and
will be described below.
Blank Consolidation
The consolidation technique of the carbon fiber-polypropylene composite (CF-PP) differs from
SR-3136 due to the lack of concern regarding fiber degradation at elevated temperatures. The
31

consolidation temperature exceeds the melting temperature of the matrix, allowing intimate fiber-matrix
contact and the elimination of voids. Ensuring these two criteria lead to a composite with predictable
mechanical properties. The main difficulty in consolidation above the matrix melting point is preventing
matrix squeeze out. This occurs as the platens exert force on the composite during consolidation and the
molten composite is forced out of the woven CF fabric. The best way to prevent this is to fabricate a
special mold which will allow high pressure to be exerted on the composite but prevent the liquid matrix
from escaping. Such a device was fabricated and can be seen in Figure 10.

Alternating layers
of CF and PP are
laid into the mold

1)
Upper Plate

2)
Center Piece

Lower Plate

The upper plate is placed
on top of the laid up
composite constituents

3)
Compressive force and heat
are applied to the mold to
consolidate the composite

Figure 10 – CF-PP Consolidation Mold
This device consists of three components; a lower plate, center piece and upper plate. The composite is
laid up within the cavity created by the lower plate and the center piece. The upper piece is then placed on
top of the composite. As heat and pressure are applied to the stencil, the matrix melts but is prevented
from escaping from the stencil. This device has been shown to significantly reduce matrix squeeze out
and produce high quality consolidated plates with excellent surface finish.
The following procedure is used to consolidate the CF-PP:

32

1. Composite is laid up inside of the cavity created as the lower plate and center piece are attached.
Care must be taken to ensure that the right amount of material is utilized as to avoid over packing
or under packing the cavity.
2. The top plate is placed on the stencil, allowing the compression of the composite constituents.
3. The stencil is placed in the heated platen press and the press is closed in order to initiate contact
between the heated platens and the stencil but also to ensure that very little pressure is applied.
4. The heating elements of the heated platen press are activated and the platens are raised to the
desired consolidation temperature.
5. Once the platens reach the desired temperature, an additional heating period is utilized in order to
ensure that the composite within the stencil has reached the specified temperature.
6. Once heating is complete, the desired consolidation pressure is applied via the heated platens.
This pressure is applied for an extended period of time to allow the molten polymer to fill the
gaps in between the woven fabric structure and to “wet out” the fiber tows.
7. A second stage of higher pressure can be applied if desired.
8. Once pressure and heat have been applied for a sufficient time period as to eliminate voids, the
heating elements are deactivated and cooling measures can be implemented (active platen
cooling, forced convection via fan, etc.)
9. The stencil is held under pressure during the cooling phase. Cooling continues until the composite
within the stencil reaches a temperature sufficiently lower than the melting temperature of the
polymer.
10. Release pressure from stencil and remove from the press. Allow additional cooling time once the
stencil is removed from the press if necessary.
11. Disassemble the stencil and recover the consolidated laminate.
The above procedure is very general and can be used with any thermoplastic matrix constituent. It
is very important to ensure that the cavity of the consolidation mold is not grossly over packed or under
packed. Over packing involves putting more material in the cavity than can be housed in its fully closed
33

position. This will undoubtedly lead to matrix squeeze-out. Under packing the mold involves placing too
little material in the cavity. Under packing can lead to excess void content in the consolidated composite
and poor mechanical properties. It is preferable to slightly over pack the stencil (less than 5% over the
volume capacity of the cavity) in order to generate pressure within the matrix. This matrix pressure forces
the polymer to fill any voids in the fabric structure leading to a lower void fraction. It is not always
possible to slightly over pack the cavity because the number of layers used in both the RFI and prepreg
CF must be an integer. The best way to ensure that the cavity is sufficiently packed is to layup the
composite constituents so that the total volume of constituents is within ±5% of the closed cavity volume.
The consolidation mold features through-passes on the centerpiece which allow thermocouples to
be imbedded within the composite during trial consolidations. These sensors allow the actual temperature
at the core of the composite to be taken in order to establish the additional heating and cooling times
mentioned in the procedure above. Once these times are determined, composites without thermocouples
can be consolidated while still giving confidence that the composite has reached the desired consolidation
temperature.
Pressure is applied in two stages per the recommendation of Zampaloni [8]. The consolidation
baseline conditions and other experimental conditions can be seen in Table 3.
Specimen

PP plys

CF plys

Temp (f)

P1 (psi)

Time at P1

P2 (psi)

Time at P2

RFI-1

6

5

335

150

20 minutes

350

5 minutes

RFI-2

6

5

375

150

20 minutes

350

5 minutes

Table 3 – Consolidation Conditions

Consolidation Properties Tool
A tool was developed to determine the appropriate number of layers of woven carbon fiber and
polypropylene film in the layup of the RFI composite. The main goal of the tool is to determine the
number of alternating layers of PP and CF that will fit into the cavity of the stencil when it is fully closed.
34

The tool operates on the assumptions that the final composite will have no voids once consolidated. The
volume of the mold when it is fully closed is easily computed since it is perfectly rectangular. Calculating
the volume of the film is also easy as it has constant thickness and is cut to fit tightly in the 10.5”x10.5”
cavity. Calculating the volume of the woven carbon fiber is not necessarily as trivial. The areal density is
used to compute the mass of each ply. This quantity is commonly used in the composite industry and is
given as
(1)
The mass is then multiplied by the number of CF plies used, then the volumetric density is applied to
calculate the total volume of CF to be used in the composite. Knowing the volume of both constituents is
necessary in obtaining the volume fraction. Several other useful parameters are calculated using this tool
as well. An example of the output can be seen in Table 4.
RFI specimen
area of pp ply
area of cf ply
n carbon plies
n polymer plies
mass carbon
mass carbon ply
Vol. carbon
Vol. polymer
Vol. total
V.f.
Vol. mold
Vol. voids
specimen t

110.25
0.071129032
110.25
0.071129032
5
6
0.134647258
0.026929452

in^2
m^2
in^2
m^2
plys
plys
kg
kg

7.65041E-05
4.668564706
5.292
8.67204E-05
0.000163225
9.96
0.469

m^3
in^3
in^3
m^3
m^3
in^3

10.3359375
0.000169376
0.38
0.09375

in^3
m^3
in^3
in

Table 4 – CF-PP Layup Tool
35

The tool allows the user to manually input the number of CF and PP plies as well as the area of the plies.
Using this information, the tool also outputs the specimen’s volume fraction, void volume, thickness, and
other useful quantities.

Future Recommendations
Material characterizations are nearly complete for the SR-3136 samples. The specimen from the
Russell study shows better results than the previously used specimens but the thin gauge section leads to
difficulties mounting strain gauges. A similar specimen with a slightly thicker gauge section is being
designed in order to mitigate this issue. Measuring the out of plane shear modulus is also desired but very
difficult to achieve due to the thick laminate needed for the specimen. Machining a specimen of this
thickness was not possible, but the recent discovery of excellent results via water jetting may make
characterizing this property a possibly in the future. Additionally traction separation properties need to be
determined. A double notch shear beam specimen can be used to determine these properties and was
devised by Liu et al. [21].
Mechanical properties during this study have all been measured at room temperature. In order to
obtain the most realistic forming response, the material must be elevated to its forming temperature when
the properties are being measured. Additionally, the CF-PP material properties need to be obtained. This
phase was abandoned to pursue other endeavors within the study. Characterization of this material will
likely be much easier due to the increase in bond strength between the fiber and matrix constituents,
allowing the easier machining of specimens and fewer issues with gripping.

36

Experimental Methods

The main focus of this project was to show that scaling up the thermo-hydroforming
manufacturing process produces good results and can be used to make functional pieces. The punch used
in this study was designed for forming Advanced Combat Helmets (ACH). The helmet punch was
obtained through a partnership with BAE systems and is the first functional piece to be made with the
composite hydroforming process. The geometry of the helmet requires deep drawing of the composite
blank and present many difficulties.
In the past, methods used to form deep drawn parts using composite materials typically rely on
hand layup techniques coupled with matched die tooling. This process has difficulty forming the parts
without excessive wrinkling. One method used to mitigate the wrinkling issue is darting or cutting away
blank material in high shear regions [37]. This method has been shown to significantly reduce wrinkling
in deep drawn parts but also has many drawbacks. By darting the blank, continuous fibers are cut which
reduces the overall strength of the composite. Complex darting also requires extended blank preparation
time, increasing the part cycle time and decreasing manufacturing efficiency. Ideally simple blank shapes
should be used with minimal trimming required before and after the part is formed.
Thermo hydroforming has shown the ability to reduce wrinkling in deep drawn parts and allows
significant draw depths to be obtained [28]. The promise of high quality formed parts coupled with
significantly reduced cycle times has attracted the attention of industry and prompted the acquisition of
the 300 ton press, which is capable of forming full scale parts. Many modifications were made to the
press to bring it to full operational capacity. This chapter will explain the modifications made to the press
as well as the process used to form composite parts. The formed parts will be discussed as well in
addition to future recommendations which will both increase quality and decrease process time.

37

Experimental Setup
The 300 ton thermo hydroforming press was delivered to MSU in the summer of 2011 in a
partially completed state. Almost every aspect of the press was either only partially operational or not
operational at all. The die set was too small to accommodate the blank of sufficient size to form the ACH
and seal housings were designed in such a way that the seals were impossible to install. A single punch
was delivered but was quickly broken due to a poor ram attachment mechanism. The pressure producing
skid (PPS) was only able to operate in one of the three fluid zones designed into the tooling due to a lack
of pressure regulation hardware. The high pressure pumps used in the PPS were also in danger of
becoming damaged due to a lack of filtration and adequate heat removal from the fluid reservoir.
These issues and several others were discovered shortly after delivery of the press. This resulted
in several key systems having to be either modified or completely replaced. This chapter will not only
describe the modifications made to the press but also explain their functions in order to give the reader a
better understanding of the operational capabilities of the 300 ton thermo-hydroforming (THF) press.
High Pressure System (HPS)
The high pressure system consists of the hydraulics and hardware that actuate the dies and punch
of the press. This system was not heavily modified but contains several design flaws that limit the
operational capacity of the press significantly. These flaws should be taken into account during the design
of future THF presses. The main features of the HPS are a pressure producing and regulating system, the
hydraulic rams that actuate the punch and clamp independently, and the mounting plate for the dies and
hydraulic rams. The main components of this system can be seen in Figure 11.

38

Clamp
Actuation Rams

Mounting
Plate

Punch
Actuation Ram

Clamp
Upper
Zone Die
Support
Columns

Lower
Zone Die

Tooling Bed

Base

Front View

Isometric View
Figure 11 – HPS Components

Not pictured in Figure 11 is the pump and pressure regulators that control the rams. The pressurized fluid
from these unseen devices is controlled and provided to the five rams seen mounted to the top of the
mounting plate. The four outer rams actuate the clamp which houses the upper zone die. These rams can
apply up to 135 tons of holding force to the composite blank as it is being formed. Even if the blank is not
being actively clamped, the fluid pressure generated within the closed dies attempts to open the dies in
order to alleviate the pressure within them. The rams apply a reaction force which holds the dies together
under the high pressure forming conditions. A single ram in the center of mounting plate actuates the
punch. This ram is actuated independently of the clamping rams making the press double acting.
The base is simply a heavy steel weldment with t-slots milled on the surface, allowing the lower
zone die to be firmly attached to its surface. The support columns constrain the movement of the clamp
and support the mounting plate to which the rams are attached.

39

The clamping rams are currently operating at 1300 psi, well short of the standard 3000 psi that
most hydraulic equipment is designed for. By increasing the hydraulic pressure in the clamping rams,
clamping force can be increased 230%. The lack of documentation and knowledge on these rams makes
increasing their operating pressure risky as bursting a ram can lead to serious physical injury and
extensive damage to the press. The problem is further explained in the Results and Discussion section of
this chapter.
Pressure Producing Skid (PPS)
The primary function of the PPS is to provide high pressure hydraulic fluid to the lower zone die.
This hydraulic fluid acts upon the composite blank forcing it to take the shape of the punch. A high
pressure fixed displacement pump is used on the skid in order to pressurize the fluid up to 5000 psi. The
fluid is fed to an array of air piloted pressure regulators. Three pressure regulators are used to
independently control the fluid pressure in the top zone, side zone and bottom zone of the forming dies.
The forming fluid flows from the regulator array into the fluid cavity of the die through one port
and out of the fluid cavity through a different port to the downstream pressure regulator. This regulator
controls the pressure in the forming chambers of the dies. Excess fluid is constantly being allowed to
enter the die from the upstream pressure regulator which is fed by the fixed displacement pump. This
excess fluid is vented through the downstream regulator and returned to the fluid reservoir where it is
cooled to 65°F and circulated through several filtration units. In the event that the oil in the forming
chamber is heated and rapid cooling is desired, the operating point of the downstream pressure regulator
can be reduced which allows the hot fluid to escape to the reservoir while cold oil is injected into the die.
This operating condition is used to cool composite parts immediately after forming in order to drive
process time down. A simple hydraulic schematic highlighting the components and fluid path through the
PPS can be seen in Figure 12. It is important to note that there is only one path leading from the PPS to
the forming dies. In reality there are three paths leading from independently controlled pressure regulators
to separate forming zones of the system. These features were omitted in order to simplify the diagram.

40

Pressurized
Air Supply

Oil Heater and
Reservoir

Lower Zone
Die Fluid
Cavity

Cool Oil
Reservoir

Legend
Pressure reducing valve with
overflow to cool reservoir

PPS

Air piloted pressure regulator
Note:
Regulators and valves do not have to be
computer controlled or air piloted but it is
preferred for accuracy and repeatability

Air piloted open-closed valve.
Normally closed
Hydraulic line
Air pressure line for pilot control
Drainage or exhaust line
System grouping.
Flow direction. Hydraulic fluid
and air respectively
Fixed displacement hydraulic
pump

Figure 12 – Simplified Hydraulic Schematic of PPS and Forming Die
Filtration of the forming fluid is critical as high pressure pumps and regulators are very sensitive
to debris and can be damaged by even the smallest particulate suspended in the oil. During early forming
simulations many difficulties were experienced with the bladder system used to prevent the composite
from coming into contact with the oil. Original bladders consisted of 1/8” rubber sheets and were prone to
tearing and bursting under high forming pressures and clamping loads. Once the bladder was
41

compromised, the composite made contact with the oil, causing two major problems. The composite is
ruined when it contacts the oil and the oil accumulates fibrous debris after contacting the composite. This
debris can restrict flow in high pressure lines, prevent valves from sealing and damage high pressure
pumps and regulators. A new bladder system was devised which consists of nylon vacuum bagging and
Teflon release film which is sealed using vacuum bagging tacky tape. A schematic of this bladder system
can be seen in Figure 13.

Teflon film or other release agent

Vacuum pulled to
produce flat blank
sandwich surface
Sealant Tape

Blank

Vacuum bagging film
Figure 13 – The Blank Sandwich
This bagging system proved to be far more reliable than the previous rubber bladder and eliminated the
threat of oil coming into contact with the composite. This drastically reduced the rate at which debris
accumulated in the oil.
An additional heat exchanger was added to the system in order to reduce the temperature of the
oil in the reservoir. The heat exchanger was implemented through a cooling loop which draws oil from
the reservoir, passes it through the heat exchanger, then through a filter before it is returned to the
reservoir. An example of this cooling/filtration loop can be seen in Figure 14.

42

Cool Oil
Reservoir

Heat
Exchanger

Oil Filter

Figure 14 – PPS Cooling Loop
An additional feature was added to the cooling loop. By incorporating additional valves and fluid lines,
the fixed displacement pump used in the cooling loop can be temporarily repurposed to pump fluid out of
the forming dies and return it to the cool oil reservoir.
The last and perhaps most important modification to the PPS was the implementation of the
downstream regulator. The addition of this regulator allows the pressure chamber of the forming dies to
operate under several different conditions. Originally the forming dies did not feature a downstream
regulator. The only means of pressure control was an open\closed pneumatically actuated valve which did
not provide accurate control of the forming fluid pressure.
By implementing the downstream pressure regulator, pressure in the chamber can be raised or
lowered to any desired set point. Examples of several operating conditions can be seen in Figure 15.

43

Lower Zone
Die Fluid
Cavity

Fluid from High
Pressure Pump

Upstream
Regulator
Condition

Fluid Returning
to Reservoir

Downstream
Regulator
Deadheading

Pressurized Circulation

Oil Flood

Pressure Set Points
Die Cavity Pressure

ambient

Result

Pressure supplied by
US reg. cannot
overcome DS reg.
Figure 15 – Regulator Operating Configurations

Fluid is circulated
through the die cavity
under the pressure

Fluid is circulated
through the die cavity
at ambient pressure

The ability to operate at several different conditions make the hydroforming press versatile and help drive
down process time. The ways in which these different operating conditions are used will be discussed in
the Forming Process section.
Forming Dies
The hydroforming dies are the most critical components in the hydroforming process. The other
systems of the press are made of generic components where the dies are the first tooling of their type.
Since there is no bench-mark for these types of dies, design has gone through several iterations. The
initial tooling was designed to allow for rapid reconfiguration, allowing different forming conditions to be
implemented.
The first iteration of the hydroforming dies was the prime example of this. These dies were
designed with three separate pressurized fluid zones which ideally could be controlled independently of
one another. A cut section of these dies can be seen in Figure 16.

44

Punch-Ram
Adapter

Mounting Plate
Upper Zone Die
Lower Zone Die

Composite Blank
Punch
Upper Zone
Clamping Surface

Side Zone

Lower Zone/
Fluid Cavity

Upper Zone/
Punch Cavity

18.0”
12.0”

Cut Section View
Figure 16 – Original Hydroforming Dies
The lower zone/fluid cavity is designed to allow fluid pressure to act against the composite blank and
conform it to the punch. The composite blank is clamped between the upper zone die and the lower zone
45

die clamping surfaces. This effectively separates the bottom zone from the side zone as is seen in Figure
16. The side zone is designed to apply pressurized fluid to the sides of the blank in order to apply force
which promotes the blank drawing in if high clamping loads are used. The side zone was never used and
was phased out in future designs.
The upper zone/ punch cavity houses the punch and the punch adapter. Pressurized fluid can be
used in this chamber to balance the pressure exerted on the composite blank from the bottom zone. In the
event that pressurized fluid is utilized from the lower zone alone, the blank acts as a diaphragm which
contains the pressurized fluid. Before the punch is traversed downward through the material a gap will
exist between the portion of the blank that is supported by the punch and the portion of the blank that is
supported by the clamping surfaces. This area of the blank will bulge into the upper zone punch cavity
and possibly rupture if the forming pressure becomes too high. Pressurized fluid can be added to the
upper zone to help support this gap in blank support area and reduce the severity of the bulge.
Unfortunately, applying pressure in the upper zone directly resists and cancels out the forming pressure in
the lower zone. Additionally, the side zone severely reduces the outer diameter of the blank holding area,
limiting the size of blanks that can be used to a circle with a diameter of 18”.
This die set was re-machined once in order to fill the side zone cavities in on both the upper zone
and lower zone dies. This increased the clamping area, allowing circular blanks as large as 21.0” to be
utilized. This would still prove to be slightly too small to form deep drawn components. For these reasons
and several others, new dies were designed and fabricated. In the next design iteration, the side zones
were removed and blank holding area was increased all the way to the edge of the tongue and groove. The
redesigned upper and lower zone dies can be seen in Figure 17.

46

Seal
Retainer

Seal

Air Bleeding Port
Punch
Cavity

Forming Fluid
Outlet

Groove

Forming Fluid
Inlet

10.4”

Sealing
Surfaces

Clamping
Surface

Redesigned Upper Zone Section View

Tongue

Clamping
Surface

Seals

31.0”
Fluid
Cavity
Forming Fluid
Outlet

Forming Fluid
Inlet

14.0”

Redesigned Lower Zone Section View
Figure 17 – Redesigned Hydroforming Die Set
The most obvious of these changes is the significant increase in the outer diameter of the clamping
surfaces. The largest blank that can be clamped with these dies is a circle with 31” diameter. A new
feature called the seal retaining ring was implemented in the redesigned upper zone die. This feature
allows a third seal (colored orange in Figure 17) to be easily replaced and inspected if necessary. The flaw
47

with the original design is that the seal could not be inserted or removed without damaging the seal. With
the implementation of the retaining ring, this seal can be inserted and removed several times without
damage occurring.
Not pictured but also implemented in this design was the use of quick disconnects between the
fluid cavity and the punch cavity. When connected, these two cavities are linked through flexible tubing
which allows equal fluid pressure. Several experiments were performed with equal pressure in the upper
and lower zone but, as stated above, this canceled out the forming pressure that forms the composite
blank to the punch. With the quick disconnects removed, the fluid cavity is isolated and single side
pressurization is engaged.
Since equal pressurization is abandoned as a viable option to reduce the blank bulging into the
upper zone punch cavity, reduction in the gap between the unsupported portions of the blank is utilized.
This gap is significantly reduced and can be seen in Figure 18.
Original Design

Revised Design

Gap between punch and
cavity wall is 0.2”

Gap between punch and
cavity wall is 1.68”
Figure 18 – Reduction in Punch Cavity Gap

There are several other important design changes to be noticed in the second revision of the
hydroforming dies. The fluid cavity has been expanded to a diameter of 14.0” while the punch cavity has
been reduced to a diameter of 10.4”. The cavities of the original die set had the same diameter at 12.0”.

48

This change was made in order to take advantage of a specific type of blank clamping called hydrostatic
clamping. Hydrostatic clamping relies on pressurized fluid to clamp the composite blank to the upper
zone die clamping surface. Instead of applying the clamping force by pressing the two dies against one
another, the pressurized fluid forces the blank against the upper zone blank holding surface. This style of
clamping is implemented to reduce the probability of the bladder system tearing. Figure 19 shows a free
body diagram that highlights the forces that are used in hydrostatic clamping.

Unsupported
blank will
bulge into the
punch cavity

Force applied by pressurized fluid
FR= FN*f

Legend
Blank will draw in
as punch is traversed

Forces
Displacements

FN=P*A

Figure 19 – Hydrostatic Clamping
The tongue and groove features of the dies are also modified to allow for a greater gap between
the upper zone and lower zone die clamping surfaces. The tongue and groove features were adopted from
the original die design. Seals are contained within grooves of the tongue which contact the sealing surface
49

of the groove, forming a seal. The tongue and groove are aligned in such a way that the gap between the
upper and lower zone die clamping surfaces is variable as long as the seal contacts the grooves sealing
surfaces. A schematic of this interface can be seen in Figure 20.

Tongue fits
into groove
Seals contact sealing surface
allowing pressurized fluid
containment

Variable clamping surface gap
as long as seals engage sealing
surface
Figure 20 – Tongue and Groove Feature
The design of this feature was changed from the first die set. The seal grooves are raised to a higher
position on the tongue, allowing the gap between the clamping surfaces to be larger while still
maintaining a sealed fluid cavity.
Punches
The function of the punch is to provide the geometry to which the composite blank will be
formed, similar to the function of the punch in sheet metal stamping. Since the materials formed in this
process are much softer than metals, there is little concern over the punch becoming fatigued. The punch
is connected to the punch-ram adapter by an opening that receives a cylindrical boss from the ram
adapter. Several screws pass through the cylindrical opening into holes on the cylindrical boss that locks
the punch in place. In order to prevent these screws from being sheared off, a ring is used to transfer load

50

from punch directly to the punch-ram adapter. The screws are used only to prevent the punch from
rotating about the adapter. A schematic of the punch and the punch ram adapter can be seen in Figure 21.

Tool Coolant/
Heating Channels
(optional)

Punch-Ram
Adapter

Load Carrying
Ring
Set Screws Prevent
Punch Rotation

Cylindrical Boss
Constrains Punch

Punch

Figure 21 – Section View of Punch-Ram Adapter Interface
Two separate punches were designed and machined for the hydroforming press. Both of the
punches are male plugs that require the material to be wrapped around them in order to form the desired
shape. The ACH punch was provided by BAE Systems through a partnership that provided the
hydroforming lab with the punch and Dyneema composite blanks.

Figure 22 – ACH Punch
51

The other punch is a simple hemisphere. It was sized to create as small of a gap as possible
between the blank holding surface and the punch. It can be seen in Figure 23.

Figure 23 – Hemispherical Punch
The isometric view in Figure 23 provides a good view of the cavity that accepts the cylindrical boss from
the punch-ram adapter in order to constrain the punch. Both punches feature positive draft angles when
viewed from the side. This prevents the formed parts from locking onto the punches and becoming
difficult to remove.
Oil Heater
Heating of the forming oil is necessary for the thermo-hydroforming process. The use of heated
oil reduces the cooling rate of the composite, allowing sufficient time for the press to close, raise the
forming fluid pressure and form the part before it solidifies. A Mokon oil heating unit is utilized to
provide forming dies with heated oil. This unit heats the oil within an internal heat exchange loop and
uses a low pressure pump to pump hot oil into the dies. This heater has the ability to raise the forming oil
to temperatures in excess of 500°F. The oil is pumped directly into the room temperature dies and begins
to cool rapidly. For this reason the hot oil should be flooded into the dies immediately before the
preheated blank is loaded. Typically the commanded temperature is at least 50°F hotter than what is
desired in the die in order to compensate for this cooling.
52

The oil pump used on the heater unit is not able to build fluid pressure higher than 50 psi. In order
to raise the pressure past this point cool oil from the PPS must be utilized. The addition of this cool oil
lowers the temperature of the fluid within the fluid cavity, reducing the amount of time that the composite
will remain in its formable state. Typically this is not an issue as fluid pressure is built immediately prior
to the forming of the blank to its final shape. In the case of thermoset composites the ability to flow hot
oil at high pressures is needed to catalyze the curing process. Since the thermo-hydroforming process
currently only deals with thermoplastic materials the inability for the heater to pressurized hot oil before
send it to the forming dies is of little concern.
Infrared Heater
An infrared heater array was acquired by the lab in order to preheat the composite samples before
forming them in the hydroforming press. This is a common industrial practice as heating within the
tooling requires complex heating elements and increases the time that the part spends within the die.
Experiments were performed where heated oil was circulated through the forming dies for over an hour in
an attempt to bring the composite to its forming temperature. Even this extended period of time was not
sufficient to bring the composite to the appropriate forming temperature. Direct temperature
measurements of the composite when it is within the closed die are also quite difficult further prompting
the abandonment of this preheating technique.
A 18.8 kW IR heater array was purchased from Radiant Energy Systems as an alternate to in-die
composite heating. A simple gantry was constructed that allowed the IR elements to be raised and
lowered over the composite to allow efficient heating of the specimen as well as easy recovery. This IR
heating array is only able to apply heat to a single side of the composite. This resulted in a significant
thermal gradient developing during the preheating of the composite. In order to mitigate this issue, a
ceramic stone is utilized as a platform for the heated composite. Prior to the heating of the composite
blank, the IR heaters directly heat the stone to approximately 20°F above the desired forming temperature
of the composite. Once the stone is preheated, the composite is placed on top of the stone. The stone

53

remains hot for some time and help heat the composite surface that is not exposed to the IR heaters. This
practice has been shown to significantly reduce and even eliminate the thermal gradient throughout the
composite during preheating. A picture of the IR heating device can be seen in Figure 24.

IR Element
Control Box

Gantry for
Raising and
Lowering
Elements

Ceramic
Heating
Surface

IR Heating
Elements

Figure 24 – IR Heating Array
The IR heating elements have been shown to localize heat at the exposed surface of the
composite. The first heating experiments resulted in the top layers of the composite becoming overheated
and charred while the interior layers barley increased temperature at all. In order to prevent this from
happening, a 1/8” rubber pad is placed over the composite to protect the surface layers during preheating.
This method prevents the top layers from becoming overcooked and allows sufficient protection for the
top layer while heat is being conducted though the thickness of the blank.
Initially, a blank was prepared with embedded thermocouples in order to investigate the severity
of the temperature gradient within the composite during the preheating process. Results obtained with this
test setup gave exact times and temperatures for heating a blank of a certain thickness. Comparable results
were easily obtained by a far simpler method.

54

1. First the stone is preheated for 30 minutes until a thermocouple at the surface of the stone reads
~20°F above the desired forming temperature for the composite. Place 2 1/8” rubber sheets on
top of the heating stone and the thermocouple during this process.
2. Place the composite within its bladder on top of 1 rubber sheet on the stone. Place one
thermocouple between the rubber pad and the composite. Place another thermocouple on top of
the composite and the other rubber pad on top of this.
3. Lower the IR heaters to within 0.5” of the rubber pad covering the composite.
4. Set the IR heater temperature ~100°F above the desired forming temperature of the composite.
5. Monitor both thermocouples. Make adjustments to the IR element temperature until both
thermocouples read ±10°F of the desired forming temperature.
6. Maintain this thermocouple temperature for 30 minutes. Adjust the element temperature if
necessary.
7. Once 30 minutes has elapsed, raise the heater elements and remove the rubber sheet. Using heat
shielded gloves, move the composite from the heating platform to the forming dies.
This process has been shown to uniformly heat the composite to the appropriate forming
temperature. Monitoring the through thickness temperatures using the blank with embedded
thermocouples showed that the thermal gradient using this procedure is negligible.

Computer Control System
The entire press including the PPS, HPS, oil heater and die set is controlled using a computer
system. The program used to control the system is UniTest and is programmed by Interlaken Technology
Incorporated. This system has the ability to control each aspect of the press manually including the
displacement of the punch and clamping mechanisms as well as their applied loads, the pressure applied
to the separate pressurization zones and the valves which allows hot and cold oil to enter and leave the

55

die. This system can be programmed to automatically implement the desired forming conditions but, since
forming is still in a highly experimental phase, commands are delivered manually.
The UniTest program interprets the data from numerous sensors around the press and uses this
data to seek the inputted set-points through PID control. The data from these sensors can be output to a
.csv file in order to analyze the data acquired during an experiment. Data from nearly any system of the
press can be outputted including forming fluid pressure and temperature, punch and die displacement,
pressure provided to the clamping and punch rams, and the forces applied by the clamping and punch
rams just to name a few. These data provide a large amount of information but one piece of data was
missing.
Since the punch and the ram can act independent of one another it is not useful to know the
displacement of these two systems independently. The composite blank always remains in contact with
clamping surface of the upper zone die. Therefore, the distance that the punch has been drawn through the
composite material has to be related to the clamp displacement since the upper zone die is mounted to the
clamp. By simply subtracting the displacement of the punch from the displacement for the clamp and
adding an offset factor is can be determined how far the punch has penetrated the original plane that
comprises the blank surface. This information is important to know for two reasons. First when the die is
closed there is no visual confirmation of how far the punch is above the blank surface. By being able to
calculate and display the displacement between the punch in the ram allows the operator to know how for
the punch is from the composite once the upper and lower zone initiate clamping of the composite blank.
Secondly, it is important to know when the punch has been traversed far enough to completely form the
part. Calculating the punch to clamp (PtoC) displacement allows the operator actuate the press
appropriately in order to ensure that the part is completely formed.

56

Forming Process
Even though several pieces have been formed successfully with the thermo-hydroforming process
over the past decade, the process has not been well documented. Even if the process was well documented
for the 40 ton press, it is likely that this process would not work well with the 300 ton press due to
changes in tooling design and increases in thickness of the composite blank. The forming process has
undergone numerous changes from the first attempt at forming the helmet. Many of these changes were
made by simply altering the forming pressure or temperature. Others were more complex in nature such
as application of single or double side forming pressure and either physically clamping the blank or
hydrostatically clamping it. The more complex changes often required changes to the forming dies or the
PPS.
The forming procedure presented below is the latest iteration on a forming process that is
constantly changing. With each additional blank formed the process is further refined with the hope of
increasing the part quality or decreasing process time. What is presented has given the best results so far.
Prior to forming, the composite is laid up inside of a bagging system referred to as the blank
sandwich (see Figure 13). A vacuum is pulled within the blank sandwich, smoothing any wrinkles on the
outer surfaces on the flat surfaces of the blank. The blank is then heated to its forming temperature while
inside of the bagging system by any number of methods. As mentioned above, an IR heating system has
been shown to work well. Unconsolidated composites can be laid up and consolidated while within the
blank sandwich if the consolidation temperature does not exceed the service temperature of the bagging
film. This method was not investigated during this study but is presented as an additional way to reduce
processing time.
As the blank is heating, the lower zone fluid cavity is flooded with hot oil from the oil heater.
This can be seen in Figure 25.

57

Heated oil
flooded into the
fluid cavity

Figure 25 – Forming - Step 1
The temperature of this oil is typically 50°F above the forming temperature to reduce the effect of the die
cooling the oil as described above. It is preferable to heat the upper clamping surface, lower clamping
surface and punch as well in order to maintain the temperature of the blank sandwich. This requires
heating elements built into the blank or some sort of external heating system. Tool heating was not
utilized during this study.
Once the blank sandwich has been raised to the appropriate forming temperature, it is laid flat on
the lower zone die clamping surface. In this step, it is important to properly align the blank sandwich with
the punch so that the part forms properly. This can be seen in Figure 26.

58

Blank sandwich must
be carefully indexed to
the punch in order to
ensure even blank draw
in (blank width = w)

w/2

w/2

Tool heating and
cooling can be
employed on these
surfaces to decrease
process time and
reduce thermal
gradient in blank

Figure 26 – Forming - Step 2
Next, the upper zone die is lowered onto the blank until a light clamping force is generated to
firmly grip the blank sandwich between the upper zone and lower zone clamping surfaces. The punch is
also brought down simultaneously until it is very close (approximately 0.040”) from the blank surface.
This step can be seen in Figure 27.

59

Clamping force
applied by hydraulic
clamping rams

Punch
approximately
0.04” above
blank sandwich
Pressurized
heated
forming
fluid

Figure 27 – Forming - Step 3
Depending on the thickness of the composite sheet, a relatively low pressure (e.g. about 10-100 psi) is
built in the fluid cavity. Pressure is built using cool oil from the pressure producing skid (PPS). It is
preferable to build pressure using hot fluid from the heater. However, that is not necessary as the volume
of the fluid required to raise the pressure of the fluid cavity is small compared to the total volume of fluid
in the fluid cavity. Once the pressure is raised, the upper zone is slowly raised up about 0.040” to keep the
blank hydrostatically clamped to the upper zone clamping surface. When hydrostatically clamped, the
blank will no longer physically contact the lower zone die. Hydrostatic clamping can be seen in Figure
28.

60

Upper zone is
raised about
0.04”

Fluid pressure
keeps blank
clamped to the
upper zone
clamping surface
(Hydro-Static
Clamping)

Figure 28 – Forming - Step 4
When hydrostatic clamping is being initiated, a portion of the blank is not supported by either the
clamping surface or the punch. This section of the blank will bulge into the gap between the side wall of
the punch cavity and the punch. As mentioned in the forming die section, care is taken to reduce the size
of this gap in order to reduce the severity of the material bulge into the upper zone punch cavity. If an
excessively large gap exists between the punch and the upper zone die clamping surface, the blank may
be drawn too far into the upper zone leading to excessive wrinkling and the inability for the blank to form
a seal and contain the pressurized fluid in the fluid cavity.
After the hydrostatic clamping has been accomplished, the punch is traversed down into the
blank sandwich and eventually into the lower zone fluid cavity. As the punch is drawn into the lower
zone, the blank conforms to the punch and is drawn across the upper zone clamping surface. This can be
seen in Figure 29.

61

Blank is drawn in
as punch traverses
downward

Volume
increase of
fluid cavity
leads to
pressure
increase

After punch is
traversed warm or
cool fluid is
circulated through
the forming cavity
to solidify the part

Figure 29 – Forming - Step 5
During this process, the volume of the fluid cavity is reduced leading to an increase in pressure. As the
pressure increases and the blank draws across the clamping surface, the initial positioning of the blank
becomes increasingly important. If the blank is off center before it is clamped then a discrepancy in
contact area will exist when hydrostatic clamping is initiated. The side of the blank that engages more
clamping surface will draw-in less due to a higher frictional force holding this section of the blank in
place. This is why proper indexing of the blank to the punch and clamping surface before clamping is
crucial. Having a clean smooth surface on the blank sandwich also helps ensure even draw-in of the
blank.
As the punch is drawn through the composite material and into the fluid cavity, the rate of
pressure increase can follow a predetermined profile which is controlled by the downstream pressure
regulator. Once the punch has reached its fully traversed position, cool hydraulic oil is circulated through
the forming cavity to solidify the thermoplastic resin. In the event that a composite uses a thermoset resin
system, hot oil can be circulated through the die to accelerate the curing process. Once the part has

62

solidified, pressure from the lower zone can be relieved and the upper zone die and the punch can be
traversed upward so that the finished part can be recovered and removed from the bladder.

Punch is raised to
forcibly remove formed
part if necessary

Fluid
pressure is
lowered to
ambient

Figure 30 – Forming - Step 6
A passage for oil recovery is added to the lower zone which allows the hydraulic fluid to be
returned to the oil reservoir. By integrating the plumbing with a cooling loop pump on the PPS, valves
can be actuated that allow oil to be pumped at a high rate from the fluid cavity of the lower zone to the oil
reservoir. An optional overflow passage from the reservoir can return excess oil back to the heater
reservoir. This option has been implemented to eliminate the need to manually transfer oil from the PPS
reservoir to the heater reservoir.

Results and Discussion
Many trials have been attempted to successfully form a part using the 300 ton hydroforming
press. These experiments have helped shape the process described above as well as the tooling and

63

pressurization mechanisms of the press itself. This section will highlight the major discoveries that were
made over the last two years as the process and the press itself has been refined. Pictures of the formed
blanks will be displayed with the description of the discovery as well as the direction that this finding
would have on the continuing development of the process and equipment.
Phase 1
The first hydroforming attempts were made with single layer polypropylene reinforced fiber
glass. This material used is commercially available under the name TwinTex. It is a plain woven
fiberglass that is comingled with strands of polypropylene. During consolidation these polypropylene
strands melt and impregnate the woven glass structure. The fiber glass was sandwiched in between two
rubber mats in order to protect the composite from the forming fluid. The upper zone was not able to
contain pressurized fluid at this time so single sided pressurization was utilized. The blank and bladder
system was clamped between the clamping surfaces of the upper zone die and the lower zone die. The
method used to heat the composite to its forming temperature was circulation of hot oil in the fluid cavity
using the oil heater. This method proved to be very ineffective and required over an hour to melt the
polypropylene strands.
Single sided pressurization of the thin blank sandwich led to additional problems. At low
pressures (~50psi) the blank sandwich would bulge excessively into the upper zone and burst,
contaminating the composite and introducing glass fibers to the forming oil. It was determined that the
main reason the blank sandwich burst was due to single sided pressurization. After this realization was
made the upped zone die was redesigned. The design revisions allowed the die to contain pressurized
fluid. Quick disconnects were used to connect the fluid cavity of the lower zone die to the punch cavity of
the upper zone die, ensuring that the pressure on both sides of the die was equal. In reality, excess gap
between the punch and the clamping surface played a large role in bursting the bladder as well. The punch
used for this experiment is a double tapered cylinder. It can be seen in Figure 31.

64

Double taper
punch utilized

Un-melted PP
strands lead to poor
consolidation

Figure 31 – First Hydroforming Results
Also in Figure 31 is a picture of a partially formed fiber glass blank. As mentioned above, the heating
process used was very inefficient and did not completely consolidate the TwinTex material.
Several important lessons were learned during these initial experiments. Preheating the blank was
required. Also, rubber bladders are prone to tearing and bursting. They would be abandoned in favor of
the vacuum bag bladder system after the completion of this set of experiments. Finally, pressure
regulation is critical. During these trials a downstream pressure regulator was not utilized which greatly
hindered the operator’s ability to control the forming pressure.
Phase 2
The second phase of forming trials saw several changes. The double taper punch was replaced by
the ACH punch. The material used in this phase was an UHMWPE composite called Dyneema HB-80.
This material is very similar to SR-3136 in laminated structure (4 unidirectional cross plys), fiber material
and fiber volume fraction (~80%). The material is stacked 66 layers thick and consolidated to form a plate
0.5” thick. The blank shape used is a 19” circle.

65

The press was extensively modified as well. Filtration units were added in order to remove the
fibrous material that contaminated the forming fluid in the previous phase of experiments. The upper zone
die was also redesigned to allow the application of fluid pressure on both sides of the blank. A
downstream pressure regulator was used to provide extra control over the pressure of the forming fluid.
An IR heater was utilized to preheat the blank, significantly decreasing the time required to perform a
single forming experiment. The vacuum bag bladder was also utilized instead of the rubber pad bladder.
The first experiments using this set up were performed using double sided pressurization. No
compressive force was utilized to fix the blank during these experiments. A low forming pressure of 50
psi was utilized and the punch was drawn to a depth of 5.0”. The results at these conditions were poor.
The blank did not conform well to the punch and excess wrinkling covered nearly the entire surface of the
blank. A diagram of the forming parameters and images of the blanks can be seen in Figure 32.

5”

50 psi

Blanks show little
evidence of conforming
to the punch
Figure 32 – Double Sided Forming Pressure on Dyneema Blank

66

The primary discovery made from this phase of experiments is that double sided pressurization of
the blank essentially leads to the forming forces cancelling one another out and eliminating the forces that
form the blank to the punch. Equal fluid pressure and no clamping load exerted by the clamping surfaces
means that the blank is free to wrinkle, further reducing part quality. Finally, a draw depth of 5.0” is not
nearly enough to form all of the features of the helmet. According to the geometry of the helmet and
assuming that the punch starts by touching the upper surface of the helmet blank, a displacement of 7.5”
minimum is required to fully form the helmet.
Phase 3
During the next phase of experiments, the choice was made to abandon double sided
pressurization in order to form the composite to the punch as tightly as possible. Hydrostatic clamping is
utilized to clamp the blank in order to reduce the likelihood of bladder tearing.
The first few experiments proved that the 19” blank was too small to form the helmet. As the
punch is drawn downward the volume of the fluid cavity is decreased leading to an increase in forming
pressure. This increase in pressure indicates that the blank has been hydrostatically clamped successfully
and that the fluid is being contained within the fluid chamber. Every blank formed under these conditions
failed to have the punch drawn to the desired depth of 7.5” before the fluid chamber experienced rapid
decompression along with a loud pop. Data acquisition was used to acquire the forming parameters as this
phenomenon occurred and can be seen in Figure 33.

67

Figure 33 - Rapid Decompression of the Fluid Cavity
Upon recovery of the blank, it became clear what has happened. As the punch was drawn
downward, the blank conformed to the punch and was drawn downward as well. A large portion of the
blank was not directly acted on by the punch and was drawn along the clamping surface of the upper zone
die. The increasing fluid pressure kept this portion of the blank tightly pressed against the clamping
surface and maintained the seal that kept the pressurized fluid within the fluid cavity. The material
continued to be drawn across the clamping surface until the edge of the blank reached the edge of the
clamping surface. Since a large gap existed between the clamping surface and the punch, the blank was
no longer able to contain the pressurized fluid. The blank was rapidly pushed into the upper zone die
(punch cavity) which released the pressurized fluid, creating an audible pop. Once this has occurred,
additional forming was not possible and the part had to be removed. This phenomenon was termed
“popping through”. An example of a blank that popped through can be seen in Figure 34.

68

Single sided
pressure reduces
wrinkling on dome
of helmet
significantly

Ear line and other
markings used to
align blank to punch
(ear line is aligned
with 0° fiber)
Blank does not
draw in evenly
leading to a cloverlike shape

2.5”

Uneven draw-in pulls
the edge of the blank
past the clamping
surface and causes this
area to “pop through”

450 psi
Figure 34 – Blank Formed with Single-Side Forming Pressure

It can clearly be seen in Figure 34 that even though the blank only popped through in one location, the
blank was traversing a large gap around the entire circumference of the punch cavity. This gap can be
seen on the blank as the trench that forms between the formed part and the clamped region. In these
experiments, the maximum pressure obtained before decompression varied from ~150 psi (seen in Figure
33) all the way to above 450 psi. The depth the punch is drawn also varies between 2.5” and 3.5”. These
relatively wide variances were the cause of changes in blank size and poor alignment of the punch to the
blank.
It was observed that the blank appeared to draw-in the most along the orientations of the fiber
reinforcements. This is expected because the fibers are stiffest in their axial direction and most compliant
in the transverse direction. Since the blank is a laminate consisting of a layers oriented [0/90/0/90], we

69

expect that the in plane stiffness along the 0° and 90° orientations will be the greatest. This is shown in
Figure 35.
Blank consists of 66 layers
consolidated so that all fibers are
aligned either 0° or 90° (punch
orientation shown as well)
These regions will
be stiffest since
they are aligned
with the fibers

90°

0°
Individual layer of Dyneema
consists of 4 laminated
unidirectional plies with
[0/90/0/90] fiber orientation
Figure 35 - Fiber Orientations of Dyneema Blank
Since the 0° and 90° orientations of the blank are the stiffest, we expect them to resist deformation
during clamping the most, leading to greater draw-in around these areas. We would expect the
material to be far more compliant in the 45° directions for two reasons. First, this direction is not
aligned with the axis of the fiber. Second, the material has a tendency for the fibers to scissor over
one another, leading to further increases in deformations. For these reasons we expect that the 45°
orientations will draw-in very little.
This is exactly what we see in Figure 34. The ear line on the underside of the partially
formed blank is aligned with the 0° fiber orientation leading to the blank popping through in this
region. This also explains the clover shape that the circular blank takes on in the region that
remains clamped. This continued to be an issue, even when larger 21. ” blanks were utilized.
Having identified the main issue as the blank drawing-in unevenly, an attempt was made to
design a blank that would draw in evenly in order to maintain sufficient clamping area around the
70

circumference of the blank. A single blank was laid up using the stacking sequence [0/45/0/45]16.
Figure 36 shows the fiber orientations of this special blank.

Dyneema layers are oriented
[0/45/0/45]16. This leads to an
actual stacking sequence of
[0/90/0/90/45/135/45/135] 16

Fibers now
aligned along 0°,
45°, 90°, and 135°
orientations

90°

0°
Increased dispersion
of fiber orientation
should reduce
uneven draw-in

Figure 36 - Reducing Blank Anisotropy
The idea behind using this stacking sequence is to reduce the anisotropy of the blank in order to
make it draw in more evenly. By reducing the unevenness of the blank draw-in, the hope was to
increase depth of draw as well as forming pressure. The results of forming this blank can be seen in
Figure 37.

71

Relatively even
draw-in observed
3.5”

930 psi

Small but numerous wrinkles
on dome of helmet
Figure 37 – Specialized Stacking Sequence
This blank was only drawn 3.5” to avoid popping the clamping are of the blank into the upper zone.
However, it appears that the punch could have been drawn further due to the large amount of blank
clamping material remaining. It is because of this, that the forming pressure was able to be raised to its
highest yet obtained value of ~930 psi. This exceptional forming pressure allowed the blank to be formed
tight to the punch. This increased forming pressure is indeed necessary for forming parts of high quality
and within tight tolerance.
72

This blank exhibited strange wrinkling patterns over the dome of the helmet that had not yet been
previously observed. It was determined that these wrinkles were products of the abnormal stacking
sequence used to form the blank. As the blank is formed to the shape of the punch, the adjacent material
layers experience a similar deformation. However, due to their differing stiffness, stress levels are not the
same. These differing stress levels do not satisfy force equilibrium and bend out of plane in order to do
so. This out of plane bending is what we see all over the surface of the formed helmet. We do not see
warping on the very top of the helmet because very little curvature is applied to the material in this area.
Down the sides of the helmet we see increases in warping due to increases in both in plane strains and
curvatures that are required to form the helmet to its final shape.
Several important discoveries were made during the experiments performed in phase 3. The most
important discovery was that the gap between the punch and the clamping surface needs to be reduced in
order to prevent the blank from popping through. It also appeared to be necessary to utilize larger blanks
in order to prevent the loss of the pressurized fluid. This would lead to a complete redesign of the forming
dies for the next phase of experiments. It was also discovered that the blanks will not draw in evenly due
to their anisotropic material behavior. Additionally, the blank must be designed in a way that it allows the
containment of the pressurized forming fluid as it is being drawn across the blank clamping surface.
Phase 4
This phase of experiments was performed on a completely new set of dies. These dies were
designed to house circular blanks up to 30.0” in diameter. The gap between the punch and the clamping
surface was drastically reduced and the lower zone die fluid cavity size was increased. These changes
were made to the die design to reduce the amount of blank material that is bulged into the upper zone by
the forming fluid and to increase the ease with which hydrostatic clamping can be initiated. The new die
designs can be seen in Figure 17 and Figure 18.
The material used in this set of experiments is SR-3136 instead of Dyneema. The blank shape is
also altered during this phase. CMA is used to predict the blank shape required for the helmet. The

73

stacking sequence was not altered due to fear of wrinkling behavior seen in Figure 37. The entire blank
stack was rotated in order to see what orientation allowed for the smallest flat pattern. It was found that a
45° rotation minimized the overall blank shape. The blank shape predicted can be seen in Figure 38.

Fiber Orientations

CMA
Prediction
Idealized Blank Shape
Blank Shape Cut by Mott
Figure 38 – CMA Predicted Blank Shape
The blank’s shape predicted by CMA is the inner rounded square seen in the figure above. The outer
square represents the simplified blank shape that is easy to cut. The blank shape on the right is used due to
difficulties with the cutting system utilized.
The leading issue experienced in this phase of forming is the difficulty to build pressure in the
forming cavity when hydrostatic clamping is used. This has proven to be the leading issue in the thermohydroforming process. Pressurized fluid in the dies exerts force over every interior surface of the forming
dies. The surface of concern is the upper zone die clamping surface. The pressure acting on this surface
produces a force in the upward direction. This force must be matched by the 4 clamping rams on top of
the press. If these rams cannot match the force applied to the upper zone die by the forming fluid pressure
then they will be back driven causing the dies to open and relieving the forming pressure. If the applied
74

forces are not in equilibrium with one another, the upper zone die will be moved away from its desired
location. This comparison can be seen in Figure 39.
4 hydraulic rams use fluid pressure acting
on a piston to produce downward force

Force must balance one
another to prevent dies
opening or closing
Hydraulic pressure (1300
psi max) acts over four
8.0” diameter pistons
Fram

F rams

F die

Forming fluid pressure
produces an upward acting
force on the upper zone die
upper zone upwards

Forming pressure acts over a ring
with OD 31.0” and ID 10.4”

Assuming
and
, the
maximum forming
pressure is ~390 psi

Figure 39 – Limited Forming Pressure
There are a few methods that can be used to increase the forming pressure. The easiest one is to increase
the maximum operating hydraulic pressure that is supplied to the forming rams. As mentioned previously,
the clamping rams used are not marked in any way and the maximum operating pressure is not known. If
standard hydraulic equipment from a reputable company would have been used, the operating pressure of
the rams would be around 3000 psi, increasing the maximum forming pressure to ~900 psi. Rams with
larger pistons can be utilized, which allow the hydraulic fluid to act over a larger area. Similarly, a
smaller upper zone die can be utilized in order to decrease the area that the forming pressure is acting
over.

75

The forming experiment in this phase once again relies on the punch traversing downward to
increase the forming pressure. As the calculation in Figure 39 indicates, this pressure produces an
upwards force on the die that is far too great for the clamping rams to resist.

Figure 40 - Forming Pressure and Punch/Clamp Displacement
As we can see in Figure 40, the forming pressure peaks out at approximately 500 psi. After this, the clamp
begins to displace since the rams cannot contain this high pressure. Around the 80 second mark, the
pressure reduces to around 390 and the clamp does not displace any further. This is exactly what was
predicted by the calculation performed in Figure 39, indicating a decent level of understanding of the
systems hydraulics and limitations. Around the 100-second mark, the operator adjusted the downstream
pressure regulator to not allow any fluid to escape, leading to a steady state pressure increase to around
430 psi. The punch displacement was then continued to around the 6.5” mark. At his point the pressure is
held as well as the punch displacement. In can be seen that the clamp rams cannot resist this forming
pressure and are slowly back driven as the experiment progresses (negative displacement is upwards).

76

Throughout the remainder of the experiment, the upstream pressure regulator is deadheading into the fluid
cavity leading and driving the upper zone die upwards. The punch draw parameter appears to be creeping
as well. The punch draw parameter (P to C) is calculated as the difference in displacement between the
punch and clamp. Since the clamp is being driven backwards, the apparent displacement between the
separate components appears to be increasing even though the punch is not moving within its own
reference frame. Also important to note is that all displacement plots are normalized to 0 at the beginning
of the forming experiment. So, Zero displacement for the clamp indicates the clamps displacement after
hydrostatic clamping is initiated just prior to drawing the punch. Zero punch displacement indicates the
punch’s position just above the blank within the punch cavity before it is drawn through the blank.
The data acquisition indicates there are still issues with the operation of the press. However, the
result is that the formed part has increased in quality. An example of a part formed in this phase is shown
in Figure 41.

77

Material is forming to
punch without wrinkling
on top of helmet

Front

Ear feature
becoming
apparent

Back

Punch orientation for
top view

Left

Excess blank material
as one would expect

Left
Back

Wrinkling occurring around
base of helmet (allows forming
fluid to leak into punch cavity)
Figure 41 – Improvement in Results

The change in blank shape and orientation as well as the reduction in the gap between the clamping
surface and the punch has allowed blanks to be formed at nearly the entire punch draw depth required
(7.2” vs the needed 7.5”). This allows a more complete forming of the helmet where the finer features and
details can be observed. Wrinkling around the base of the helmet is occurring where the blank has almost
completely drawn to the base of the helmet. The more severe wrinkles lift the blank from the clamping
surface and allow fluid to leak into the upper zone fluid cavity. Although the leak is not severe enough to
cause a decrease in forming pressure, it does eventually starve the PPS reservoir. It is initially thought that
these wrinkles could be ironed out with the use of greater forming pressure. However, these forming
pressures cannot be obtained due to the issue with the clamping rams described above.
Additional experiments are performed with this clamping scheme. For the next trial the punch is
drawn the full 7.5” to fully form the helmet. The results of this experiment can be seen in Figure 42
78

Rear

Front
Top of helmet is
once again well
formed with most
wrinkling occurring
at the base

Upper zone die clamping
surface edge is pressed
into excess blank material

Wrinkling is severe
leading to leaking of
pressurized forming fluid

Figure 42 - Blank Lifting Problem
This blank was the first to be aligned using a laser alignment system. This helps ensure that the blank is
perfectly aligned with the punch before forming it. This experiment reached a maximum pressure of ~430
psi. The material is conformed well to the top of the helmet, but once again, severe wrinkling occurs near
the base of the helmet leading to the loss of fluid pressure. During this trial the circular edge of the upper
zone die clamping surface can be seen. It is important to note that once the blank has been pulled past the
edge of this circle, the pressurized fluid will not be contained within the fluid chamber any longer. This is
exactly what we see in Figure 42, indicating that the blank is slightly undersized as was predicted earlier.
It is also clear that hydrostatic clamping alone will not be able to prevent the blank from wrinkling
especially at the base of the helmet where the critical seal that contains the pressurized fluid is formed.

79

Physical clamping of the composite blank between the upper zone die and lower zone die was
attempted in order to see if wrinkling at the base of the helmet could be eliminated. A slightly different
forming procedure was used for this experiment and it is highlighted below.
1. The fluid cavity is filled with hot forming oil.
2. The pre-heated blank is placed in die and indexed to the punch using the alignment laser method.
3. The upper zone is lowered onto the punch and a 50 kip clamping load is applied. The punch is
lowered until it is just above the blank.
4. The pressure in the fluid cavity is built to 150 psi. The downstream regulator is set to 100 psi to
allow pressurized circulation.
5. The punch is drawn 6” through the material. Since the downstream regulator is set up to allow
circulation, the decrease in the fluid cavity volume will not result in a pressure increase.
6. The forming pressure is increased to 1000 psi. The downstream regulator is set to 900 psi to
continue pressurized circulation.
7. The punch is drawn the final 1.5” under this condition. This also stretches the blank helping to
eliminate any wrinkles.
8. Pressurized circulation is continued until the part cools and solidifies.
9. Pressure is released and the part is recovered.
This process amendment is in the very early stages of development but it has already shown promising
results as can be seen in Figure 43.

80

Grid allows fiber orientation to
be visually tracked

Blank is formed to upper helmet
region with no signs of wrinkling

Circle indicates boundary
of fluid cavity and edge of
clamped region

Blank barley drew in past
lower zone fluid cavity

Figure 43 - Change in Clamping Scheme
Also, implemented during this stage of forming experiments is the grid on the top side of the blank. This
grid is aligned with the fiber orientations of the blank. In the future, it will be used to track the amount of
shear seen at different locations of the blank and compared to the results of the numerical simulations. As
can be seen in the figure, the amount of wrinkling of the blank is significantly reduced. Only a few small
areas of the blank show any wrinkling whatsoever. For this experiment, the punch was only drawn 4.5” to
prevent the blank from drawing in past the punch cavity which would result in a loss of forming fluid.
The maximum forming pressure was ~440 psi, well below the desired 1000 psi desired in the revised
procedure. However, this pressure was able to conform the blank to the punch surprisingly well.

81

Ideally, the blank should not draw past the edge of the lower zone fluid cavity. If this can be
accomplished, then the area being acted on by the pressurized fluid will be reduced to a circle of 14” in
diameter. This significant reduction in area will allow increases in forming pressure with the rams being
able to hold the upper zone die at a fixed position. As can be seen in Figure 43, a few small areas of the
blank were pulled past this boundary allowing fluid to act over the full upper die clamping surface. This
caused the upper die to be lifted which in turn initiated hydrostatic clamping. The operator was able to
reduce the forming pressure, allowing upper zone to be traversed downward and reinitiate physical
clamping.

Figure 44 – Clamping Scheme Difficulties
It is hard to say whether physical clamping or coupled physical and hydrostatic clamping is the cause of
the high quality results seen in the experiment. Further tests are required to determine the true reason for
the success of this forming experiment.
Many important discoveries were made in this phase. Perhaps the most important discovery is
that the use of physical clamping of the blank will not lead to tearing of the bladder system. It can also be

82

used to reduce the wrinkling of the blank over the clamping surface, eliminating the ability for the fluid
pressure to leak into the punch cavity. It was also discovered that low forming pressures (around 450 psi)
can produce high quality formed parts as can be seen in Figure 43. It was also discovered that the blank
used is slightly too small for the helmet, at least in the press’s current configuration. The ability to
generate higher clamping loads can allow higher forming pressures to be used when hydrostatically
clamping the blank. Increasing clamping load will more than likely require the replacement of the
clamping rams.

Future Recommendations
The final sets of thermo-hydroforming experiments show that the process has the ability to form
high quality parts. The helmet possesses complex geometry and is a very deep drawn structure. The
abilities that the process has shown on such an advanced part indicate that this process can be used as a
viable manufacturing method for parts with complex geometry. Forming parts of a simpler geometry can
no doubt be accomplished seeing the results that have been produced initially.
The next set of forming experiments should investigate the effect of physical clamping as these
were showing very impressive results after the first few attempts. Large blanks should be implemented for
the initial forming experiments. Trying to form with an optimized blank before the process was well
understood leading to several failed attempts that could have been surpassed. The hemispherical punch
was not able to be used in this study but should be used in the future. This punch has a very simple
geometry and allows for a simplified analysis of fiber shearing when comparing to numerical results.
The next round of modifications to the press should mostly be centered on the implementation of
high pressure clamping rams. Additionally, a better computer controller needs to be implemented. During
several experiments the operating pressure would differ drastically from the desired pressure producing a
large error. During several experiments it would take up to 20 seconds before the regulators input voltage
was adjusted, leading to long lag times and pressure oscillations. This issue appears to happen with the

83

downstream pressure regulator and could not be fixed by altering the PID parameter of the controller.
This issue should be fixed in the future for better pressure control.

84

Numerical Methods

Numerical simulations require a deep understanding of both the material properties of the
laminated blank and the boundary conditions. The purpose of these simulations is to be able to develop a
model which produces results that compare well with the experimental results. The goal of modeling the
process is to be able to determine forming conditions and optimal blank shape so that the highest quality
parts can be formed with minimal waste. First, a model of sufficient accuracy must be developed.
This chapter will summarize the material model and the changes made to it in order to improve its
accuracy and increase practicality. The results of the simulations will be discussed and compared to
experimental results that were obtained in the previous section.

Preferred Fiber Orientation (PFO) Material Model
The material model utilized in this study comes from the PhD dissertation of Mike Zampaloni [3].
This material model operates on the premise that a majority of a fiber reinforced composites stiffness
comes from the fiber itself. A complex laminate or woven structure will possess many fiber orientations
which all have an orientation in which they are stiffest and an orientation in which they are most
compliant. The model tracks the orientation of these separate fibers and applies their stiffness to a global
tensor in order to obtain the overall stiffness of the material. The model sets itself apart is its ability to
update the fiber orientations as the body is deformed.
The Preferred Fiber Orientation (PFO) model uses a time marching explicit solver that calculates
the change in fiber orientation for each time increment and updates the stiffness tensor accordingly. This
model tracks the preferred fiber orientations and recalculates the stiffness tensor, allowing the materials
response to be updated as the material deforms and fiber orientations change. Multiple fiber orientations
can be tracked, allowing this material model to accurately characterize several types of composite
materials, including biaxial weaves, triaxial weaves, unidirectional tapes, and even randomly oriented
fiber mats.
85

Constitutive Model
Typically composite materials follow linear elastic deformation up to failure. UHMWPE
composites are no different in this respect as is shown by Russell et al. [17]. This allows the individual
fibers to be characterized as a linear elastic material. The stress-strain relation is given by Hooke’s Law.
(2)
When Hooke’s Law is fully expanded, the 4th order stiffness tensor, C, contains 81 independent terms.
Since the stress and strain are 2nd order tensors, the number of independent C terms is reduced to 36. The
material stiffness is assumed to be symmetric, so the number of C terms is further reduced to 21 and can
be displayed as such.

(3)
[

]

][ ]

[

[

]

[

]

[ ]

[

]

The stiffness tensor as it exists in this form can be further reduced by the assumption that the
material is orthotropic. An orthotropic material contains 3 planes of elastic symmetry. This means that the
material will have a unique stiffness in each of its three principal directions. This simplification dictates
that when a normal strain is applied, only normal stresses will be generated (no shear-extension coupling).
The application of shear strains within the planes of symmetry will only produces shear stresses within
this plane and no normal stress (no shear-shear coupling). This allows the stiffness tensor C to be reduced
even further.

(4)
[

]

][ ]

[

Individual layers of composite materials are modeled, allowing a plane stress is assumption. Their
thickness prevents out of plane stresses from becoming prominent and allows for their removal from the

86

stiffness tensor. Assuming the thickness of the composite is in the z-direction, this allows the removal of
any z terms which correspond to the 3rd, 4th and 5th rows and columns of the C tensor in equation 4.
[

]

[

][ ]

(5)

The formulation of the reduced stiffness tensor, Q, has, up to this point, assumed that the material
reference frame (1, 2, 3) has been aligned with the structural reference frame (x, y, z). These reference
frames do not always align which requires the use of distinct notation in order to avoid confusion. The 1direction of the material reference frame is always aligned with the fiber orientation, while the 2-diection
is transverse to the fiber direction but still within the plane of the material. The structural frame rotates
with the body of interest while the material frame rotates with the fiber orientation. This can be seen in
Figure 45.
3
2

1

Material reference frame is
aligned with fiber orientation

Z

1

Y
θ

X
The two coordinate
systems differ by
the angle θ

X
Structural reference
frame is fixed in space
Figure 45 – Material and Structural Coordinate Systems

The best way to define stiffness components is to align them with the material coordinate system. By
doing this, the material properties have a deeper physical meaning due to their alignment with the
reinforcing fiber structures giving the following stiffness components.
(6)
is the elastic modulus of the material measured along the fiber orientation.
material measured transverse to the fiber orientation.
87

is the stiffness of the

is Poisson’s Ratio based on the 1, 2, 3

coordinate system.

is the shear modulus based on this coordinate system as well. Based on this

adjustment, equation 5 needs to be reformed in order to give the stresses and strains in terms of the
material coordinate system and not the structural coordinate system.
[

]

[

][

The stiffness matrix can be condensed and written simply as

]

(7)

. Now that the constitutive relation has

been related to the material frame, the stiffness needs to be able to be related back to the structural frame
to provide a common reference point. A transformation matrix based on the stiffness tensor is used.
( )

[

]

(8)

The stiffness in terms of the structural frame can then be solved for by performing a change of basis
operation.
̅̅̅̅̅
[̅̅̅̅̅
̅̅̅̅̅

̅̅̅̅

̅̅̅̅̅ ̅̅̅̅̅
̅̅̅̅̅ ̅̅̅̅̅]
̅̅̅̅̅ ̅̅̅̅̅

̅̅̅̅ is the stiffness of the material rotated to the global frame.

(9)

is the transpose of T. Now that the

stiffness in the material frame is known, it is possible to calculate the stress in the global frame.
̅

( 10 )

Similarly, the strain in the material frame can be found if the stress is known using the compliance tensor.
̅

̅

̅

( 11 )

Obtaining Material Properties
Now that the constitutive model has been developed the stiffness matrix,
The constants

,

,

and

, must be populated.

are required in order to fully characterize the stiffness matrix in

equation 8. All of these constants can either be measured directly (as was performed in the Material
Characterization section) or can be calculated using accepted relations such as the various rules of
88

mixtures. This section will review the various ways in which the engineering constants can be calculated
and discuss their shortcomings.
is the Young’s Modulus of the unidirectional composite measured in the direction of the
fibers. The classical approach in determining this constant is to use the rule of mixtures.
(

)

are the Young’s Moduli of the fiber and matrix components specifically.

and

( 12 )
is the fiber volume

fraction of the composite it represents the percent volume of the composite that is fiber. The calculation
(1

) is used to calculate the matrix volume fraction

. This leads to a zero void assumption.
( 13 )

In reality, even the highest quality composites contain voids so equation 14 is highly idealized. This leads
to inaccuracies when using equation 12 but generally they are small. Furthermore, equation 13 assumes
perfect bonding between the fiber and matrix constituent. Equation 14 takes into account continuity of
mass and is presented by Mohammed et al. [38].
( 14 )

is fiber volume fraction prior to deformation and α is the angle between the a-fiber and the b-fiber
assuming that they are orthogonal to one another prior to deformation.
Generally, these void free and perfect bonding assumptions introduce error but it is usually small
enough to be ignored. In UHMWPE composites, void fractions have been reported in excess of 10% [39].
Additionally, the bond between UHMWPE fiber and a PE matrix is shown to be weak at best, leading to
issues with the perfect bonding assumption. Nevertheless, the rule of mixtures gives a ballpark estimation
as the longitudinal stiffness that should be expected.
is the transverse Young’s Modulus and represents the stiffness of the material when loaded
transverse to the fiber axis. The simplest calculation to determine this constant is with the inverse rule of
mixtures.
89

(

)

( 15 )

This equation makes the same assumptions as the rule of mixtures of equation 12, leading to errors. Okoli
developed an empirical relation based on the Halpin-Tsai equation for

[40]. This equation shows

reductions in error when compared with experimental results.
(

)

( 16 )

This equation contains several nested relations and values which can be seen in equation 17.
.

.

( 17 )

These equations are configured to work with random fiber orientations reducing their accuracy when used
with unidirectional tapes.
Poisson’s ratio is found using a simple rule of mixtures as is presented by Mohammed [38].
(

)

( 18 )

is found through a relation taking into account the longitudinal and transverse Young’s Modulus.
( 19 )
The shear modulus can be found via several methods. The simplest one involves a simple rule of
mixtures approach.
(

)

( 20 )

Halpin and Tsai propose another method to empirically determine the shear modulus [41].
(

)

( 21 )

Much like equation 16, equation 21 contains many nested variables and factors which can be seen below.
( 22 )

90

The

term in equation 23 is presented separately by Okoli and Smith [40]. This term is once again based

on a randomly oriented fiber structure and leads to errors when used with unidirectional tapes. Many of
these empirical relations lead to poor predications of the engineering constants needed for the stiffness
tensor

. They would be modified to fit the expected trends and would eventually be replaced by direct

material characterization. This process will be further described in the Use of Material Characterization
Data section.

Preferred Orientations
The constitutive relation developed in the previous sections is able to accurately characterize a
unidirectional material. As seen in Figure 35, the UHMWPE composites utilized in this study (both
Dyneema and Spectra Shield) consist of numerous layers of orthogonally stacked unidirectional
composite layers. The stiffness tensor of the laminated composite can be assumed to be the sum of the
stiffness tensors of the two separate fiber orientations. In this case the fiber orientations are oriented in the
0° and 90° directions (classified as the a-fiber direction and the b-fiber direction respectively). These
orientations are orthogonal to one another in the undeformed setting. It is not necessary that these
orientations remain orthogonal to one another as the laminate is deformed. Figure 46 shows the
summation of the a-fiber and the b-fiber to yield the laminated material stiffness.

Two non-orthogonal
preferred fiber orientations

A-fiber orientation

Figure 46 – Stiffness Summing Used in PFO Model

91

B-fiber orientation

Additional fiber orientations may be utilized if the material cannot be adequately characterized by two
preferred orientations. Randomly oriented fiber mats, tri-axial weaves and unidirectional layups where the
fibers have three or more orientations in the stacking sequence are examples of materials that can benefit
from the use of additional PFOs. Determining the fiber orientations is an easy task when the materials
used have highly regular and repeated orientations such as with woven fabrics and unidirectional layups
with a repeated stacking sequence. With other materials, the PFO’s may not be obvious such as with
random mat orientations. A squeeze flow test was developed by Zampaloni that can be used to determine
the preferred orientations [3].
The materials in order to sum the stiffness of the individual PFO’s, a common structural frame
must be utilized. The stiffness tensors of the individual fiber orientations in the structural frame are
obtained using equation 7. The stiffness tensors are then rotated to the common structural frame with
equation 8. Once the individual PFO’s have been rotated to the common structural frame they can be
summed in order to determine the materials stiffness. This method is highlighted in Figure 47.

92

⊥a
b

⊥b

a

α
β
Composite material with two
obvious fiber orientations

̅

̅

Each fiber orientation is assigned its own
material coordinate system and a common
structural coordinate system is set (red)

̅

̅

( )
̅

Laminate stiffness is found by
summing rotated stiffness tensors

( )
(

)

(

)

Stiffness is calculated for each PFO
and rotated to the material frame

Figure 47 – Rotation of PFO’s to the Structural Frame
The method highlighted in Figure 47 is used to find the stiffness of a composite laminate. The fiber
orientations are typically known at the beginning of a numerical simulation, allowing the stiffness to be
updated throughout the simulation.
As one would expect, changes to the angles α and β will change the transformation matrix T.
Alteration of the transformation matrix results in the alteration of the laminate stiffness ̅ leading to a
changes in behavior of the laminate such as increased stiffness in a particular direction or the introduction
of shear-extension coupling. These changes have a profound effect on how the material will deform and
therefore must be taken into account. Consider a small differential element of laminated composite
material with a-fiber and b-fiber directions as well as its own structural coordinate system. This material

93

is acted on by an arbitrary deformation gradient tensor. Polar decomposition of the deformation yields a
stretch tensor and a rigid body rotation tensor.
( 23 )
F is the deformation gradient tensor, R is an orthogonal tensor representing a rigid body rotation and U
right stretch tensor. This deformation gradient acts on the differential element by stretching and rotating
it. The structural coordinate system is acted on by the rotation tensor (R) in order to remain aligned with
the element and still provide an orthogonal coordinate system. The a-fiber and b-fiber directions are acted
on by the stretch tensor and the rotation tensor (RU), which alters the angles that these fiber directions
make with respect to the material coordinate system. This can be seen in Figure 48.
Differential element
deformed according to F

Undeformed
differential element

F=RU

SCS is only rotated by R
Fibers are subject to the whole
deformation gradient tensor F

Structural Coordinate System (SCS)
A-fiber Direction
B-Fiber Direction

Figure 48 – Modification of the Structural Coordinate System and Fiber Orientations
In this example the fibers are aligned with the SCS in their undeformed configuration. The deformation
gradient tensor alters their alignment with the SCS, which in turn alters the stiffness of the differential
element.

94

Implementation in Abaqus
This scheme is implemented in Abaqus explicit through a user subroutine. The explicit solver
uses a time marching scheme allowing the strain increment, deformation gradient tensor to be fed into the
subroutine during each step. The user subroutine uses the deformation gradient tensor to determine how
the SCS and the fiber orientations are changed during the current time step as was shown in Figure 48.
Once this is determined, the stiffness (

) of the PFO is determined. Next, the stiffness is rotated to the

structural frame, giving ̅ . This is repeated for each PFO. Once all the PFO stiffness are rotated to the
material frame, they are summed to obtain ̅ . The strain increment is then multiplied by the stiffness
giving the stress increment for the time step. This information is fed back into the Abaqus solver. A flow
chart of this process can be seen below.

FEA Code
Pre Processing

Load Increment and Time
Stepping

Internal Force Calculation
(Explicit Code)

Time Information
Incremental Strain
Old Stress
Deformation Gradient Tensor
(Stretch Tensor)
Spin Tensor

User Subroutine
Stiffness Matrix
Calculation
(Implicit Code)

New Stress
Stiffness Matrix (Implicit)
New Energy (Explicit)

Solution Solver

Output Print

Figure 49 – User Subroutines Within Abaqus/CAE
The elements of the hydroforming press are implemented as rigid bodies. The blank is comprised
of 6 layers of shell elements tied together by cohesive interactions.

95

Upper Zone Die
Blank Clamping
Surface

Helmet
Punch

Lower Zone Die
Blank Clamping
Surface

Blank is represented by
laminated shell elements

Figure 50 - Model Setup

Simulation Changes and Adaptions
The model displayed several short comings at the beginning of the study, the most prominent of
which was its inability to simulate a blank of thickness required for the helmet forming project. The
transverse shear stiffness was incorrectly defined and it was also found that the transverse Young’s

96

modulus and the shear modulus were significantly over-predicted. This led to the numerical simulation
giving poor results.
Originally, the composite materials simulated using this model were one or two layers thick and
could be idealized as a single layer with minimal voids. This allowed the use of many of the rules of
mixtures equations to predict the response of the composite relatively accurately. In this study increased
blank thickness as well as a different composite system do not allow the use of these simplifying
assumptions.
Multiple Layer Simulation
The Spectra Shield blanks used in the forming experiments consist of 45 layers of SR-3136 which
consist of 4 plies a piece. In actuality, the composite consists of 180 layers of unidirectional composite
material with a layup sequence [0/90]90. To simulate each one of these layers individually would be
computationally expensive. Instead, the simulations utilize 6 layers of shell elements (plane stress
elements) that are tied together using cohesive surface interactions.
One of the biggest problems when attempting to model this thick blank as a single layer is a very
coarse mesh. In order to use the thin shell theory, the shell elements must have a length to thickness ratio
of 10:1 or greater. By using a single layer of shell elements, a very coarse mesh would need to be utilized.
Even if these large elements used quadratic shape functions, they would give very poor predictions for
stress, strain and every other field variable. Therefore, multiple layers must be used, not just to simulate
the effect inter-laminar separation but also to refine the mesh in order to obtain accurate results.
The cohesive surfaces interaction uses the fracture mechanics principal of traction-separation in
order to take into the adhesive forces that hold the separate layers of the blank together. Similar methods
have been used to simulate adhesive bonds in composites [42], [43]. Principals of elasticity and plasticity
are utilized in order to determine crack growth along a predetermined path. An applied force can initiate
crack growth in three ways as can be seen in Figure 51.

97

Mode III:
Twisting in Plane
of Crack Boundary

Mode II:
Shearing Along Plane
of Crack Boundary

Mode I:
Opening Normal to
Crack Boundary
Figure 51 – Fracture Modes

As cracks open, a measurable separation between the two sides of the crack form. As the crack grows,
traction attempts to return the crack to its initial closed position. This traction is a linear function of the
separation and is represented in matrix form.
[ ]

[

][

]

The subscripts n, s and t refer to fracture modes I, II and III respectively. The

( 24 )

,

and

terms refer

to the terms that directly relate a traction with a separation of the same mode. Mixed mode stiffness (
and

,

) can be defined to implement coupling between the fracture modes or may be left as zero

indicating that there is no coupling between the fracture modes.
This linear traction can be viewed as an elastic deformation (fully recoverable) between the two
elements and represents the stretching of the adhesive bond between two elements prior to the initiation of
damage. Eventually the adhesive bond will become damaged when a certain damage initiation criterion is
met. This damage criterion can be based either on a critical traction or separation value. Additionally, the
damage criterion can occur as a function of a single mode (equation 25) or as a quadratic function
(equation 26) of all three modes. These criterions are expressed mathematically below.
〈

{

〉

,

98

, }

( 25 )

〈

{

〉

}

{ }

{ }

( 26 )

The angle brackets around the normal separation term mean that only tensile separations are taken into
account since compression of the interface does not result in fracture.

is the critical separation for

mode i that will lead to damage initiation.
After the damage criterion has been met, the interfaces and their load carrying capacity
deteriorate. With increased separation, the traction will gradually reduce to zero. Damage evolution is
typically represented as linear softening of the interface but can be represented by exponential decay as
well. This study will utilize linear damage evolution as it is the most commonly used. Linear damage
evolution can be characterized in two ways; defining a failure separation (

) or specifying the total

fracture energy (area under the traction-separation curve). These specifications will yield a traction

𝑡𝑛𝑜

𝑡𝑛𝑜

Traction

Traction

separation curves similar to the ones seen in Figure 52.

𝛿𝑛𝑜

𝑓

𝛿𝑛

Separation

Linear Damage Evolution

𝛿𝑛𝑜

𝑓

𝛿𝑛

Separation

Exponential Damage Evolution

Figure 52 - Traction-Separation Curves
This traction separation behavior is implemented through the cohesive surfaces interaction. The
interaction is implemented by attaching adjacent elements of the two bonded constituents. Methods for
empirically determining the cohesive properties are not well documented or widely accepted. The best
way to determine these properties is through a specialized material characterization. Liu et al. developed a
method using a double notch shear beam to directly measure the mode II traction separation curve [21].
99

Time did not permit for these tests to be performed. Luckily, Liu characterized and published the traction
separation behavior for Dyneema HB-26 UHMWPE composite. Since, this composite is similar in
structure and materials to SR-3136 the properties were adapted for the cohesive surface interaction. Table
5 shows the properties utilized for this interaction.
8 GPa/m
0.2 mm
0.4mm
Table 5 - Traction Separation Properties

Transverse Shear Stiffness
The transverse shear stiffness has shown to be a critical property in forming simulations involving
blanks of substantial thickness. The simulations performed by Zampaloni [3] were mostly concerned with
thin composites mostly one or two layers thick. When working with materials this thin, transverse shear is
negligible.
During this study, however, the blanks used did have substantial thickness (up to 0.5”) making
the correct assignment of the transverse shear stiffness critical to obtain good results. When a user
subroutine is used to evaluate the material properties of shell elements in Abaqus, the transvers shear
stiffness must be manually defined according to the following equation.
( 27 )
t is he thickness of the shell elements to which the transverse shear stiffness is being applied. The Scaling
factor of

is a shear correction term to account for the non-uniform shear stress distribution through the

thickness of the element. The shear moduli directions are given assuming that the 1-2 plane represents the
plane of the element. Figure 53 shows the orientation of transverse shears.

100

3

/=τ

– in plane
/ = τ23 – transverse
/ = τ13 – transverse

2

1

12

Figure 53 - Tansverse Shear
Prior to this study, hydroforming simulations were performed using properties similar to what can
be seen in Table 6.
Element thickness

0.0002 m
2.84*106 Pa
1000.0 Pa*m

Table 6 – Original Transverse Shear Properties
At the time these simulations were performed, direct measurement of the out of plane shear modulus was
not performed. Davalos et al. reported that materials similar to the ones used in these simulations showed
a transverse shear modulus that was typically between 60% and 80% of the value of the in-plane shear
modulus [44]. If we calculate the transverse stiffness under the assumption that the transverse shear
modulus is 70% of the in plane shear modulus, then the value used should be 331 Pa*m as opposed to the
1000 Pa*m used in the simulation. Thanks to very small thickness of the composites being evaluated, this
overestimation did not have a significant impact on the results of the simulation.
The transverse shear stiffness has proven to have an enormous effect on the deformation of the
blank since the thickness has been increased for the forming simulations. The most prevalent effect that
this property has on the deformation of the blank is in the amount of in plane shearing exhibited and the
prevalence and realism of out of plane warping (wrinkling). The out of plane (OOP) shear modulus for
SR-3136 was not measured and is estimated as 70% of the in-plane shear modulus. The results of varying
this parameter will be discussed further in the Results and Discussion section.

101

Transverse Young’s Modulus
Measuring the Young’s Modulus for SR-3136 is relatively straightforward, but applying the data
in a meaningful way is less obvious. Tensile specimens are loaded along the 0° direction meaning that
approximately 50% of the fibers are aligned with applied load while the other 50% of the fibers are
transverse to the applied load. Several publications have reported low bond strength between fiber and
matrix as well as a lack of continuity when observing the microstructure of the composite [15], [17]. For
these reasons it is assumed that the plies with fibers aligned with applied tensile load carry a vast majority
of the load while the transverse fibers carry almost none of it. This assumption in terms of stiffness can be
expressed as

. Since the material has the stacking sequence [0/90/0/90], there is no way to

directly measure the transverse Young’s Modulus (

).

The only way to obtain this quantity therefore is through an empirical relation. The rule of
mixtures is chosen as the method to compute the properties. A fiber volume fraction of 80% and stiffness
value for the matrix constituent (LDPE) is used (0.75 GPa). Using equation 12 and assuming that tensile
test results give
constants,
compute

, the apparent Young’s modulus of the fiber is found (

). Using this same set of

is found using the inverse rule of mixtures (equation 15). The Halpin and Tsai method to
can also be used, but the use of the inverse rule of mixtures yields a lower stiffness and

complies better with the original assumption of

.

Use of Material Characterization Data
SR-3136 was characterized in order to obtain the Young’s Modulus and the shear modulus as
was. As mentioned above in the Transverse Young’s Modulus section, the value obtained from these tests
was assumed to be primarily from the fibers aligned with the load direction. Therefore, the modulus
obtained from this test is assumed to be

and is applied to equation 7 in order to populate the

portions of the stiffness tensor.

102

and

The shear modulus is directly obtained from the Iosipescu test highlighted in the Consolidation of
Flat Panels and Material Characterization section. Since the fiber alignment of the test specimen was
aligned with the load applied by the test fixture, the modulus obtained by this test is taken as
implementation of this parameter requires additional insight. The
orientation is

. The

parameter of each individual fiber

of a unidirectional composite. The specimen tested is a laminate of two unidirectional

composites where the fiber orientations of the layers are 0° and 90°. Since two fiber orientations are
used in the material model, applying
the value 2
̅

̅

value will result in ̅

as the

for the laminate having

.
°

̅

°

[

]

[

]

[

] ( 28 )

Seeing that the laminate value for the shear modulus has taken on a value of 2
made by changing the individual fiber orientation to ̅

̅

, an easy correction is

2. These properties are

implemented in the user subroutine by adapting the FORTRAN input file to use them directly instead of
using them as inputs to the various rules of mixtures.

Results and Discussion
Many simulations have been performed which utilize the modified PFO model described above.
In order for the modeled laminate to take on a realistic thickness, many plies were needed which
drastically increased the amount of time required to complete the simulation. For this reason, most
simulations were conducted with 6 plies, giving the laminate a total thickness of ~0.3”. Several blanks
were laid up to this thickness and experimentally formed so that good comparisons could be made with
the numerical simulations.
The impact of the transverse shear stiffness of the results of the simulation, especially pertaining
to wrinkling, was huge. This was determined to be the material property that had the largest effect on the
outcome of the simulations. The effects of varying this property will be investigated as well.

103

Due to difficulties with the forming procedure and equipment, very few parts have been
experimentally formed successfully at this time. This makes it difficult to compare the numerical results
directly with the experimental results. Most of the evaluation of the numerical simulations has been based
on the trends that are seen in the experiments for this reason. Several aspects of the model show good
correlation with the experiments and others need to be improved.
Impact of Transverse Shear Stiffness
The impact of the transverse shear stiffness was noted early on. During this stage of the
simulations, a circular blank was utilized. Additionally, material properties were inferred from the
literature and ended up not being very accurate due to the issues with the empirical estimations of the
properties, as was described above. Nevertheless these simulations show the important effect that
correctly defining transverse shear has on wrinkling prediction and on the amount of in plane shear
allowed.
As multiple layer simulations were began, interesting patterns in wrinkling began to be observed.
The wrinkling observed was very tight and showed very little in plane shear occurring. Upon modifying
the transverse shear stiffness, the wrinkling pattern changed entirely. The wrinkling observed was far less
prevalent and took on realistic patterns. The in plane shearing also increased significantly and reflected
more accurately what was seen in the forming experiments. Figure 54 shows a side by side comparison of
the effect of under defining and correctly defining the transverse shear stiffness.

104

TSS is too low

TSS correctly defined

Angle between
Fiber Orientations
(degrees)

1

1 ,

,

Figure 54 – Transverse Shear Stiffness Effects
The initial results on the left were obtained using the default value of the transverse shear stiffness (TSS)
left over from the Zampaloni simulations. The simulations on the right are performed using the
empirically obtained value from equation 27. The undeformed configuration of the fiber orientations is
90°, so the simulations with the correct TSS show more widespread values of fiber shearing which is
closer to what is observed in Figure 43. The wrinkling severity is reduced and is more gentle that the
sharp sporadic wrinkling seen in the simulation with the low TSS. This is similar to what we see in the
forming experiments. Another benefit of correctly defining the TSS is the reduction in element warping
which causes the simulation to fail. Prior to updating this parameter, many simulations ended prematurely
due to excessive element warping in the areas of severe wrinkling. Updating the TSS has significantly
reduced the occurrence of these problems.

105

The wrinkling in these forming simulations is occurring too soon and is occurring too far up the
dome of the helmet. One reason for this is that these simulations utilize properties from the empirical
relations which compare poorly to the measured values from the material characterizations. Upon the
implementation of the measured material properties, simulation quality increases.
Square Blank Simulation Results
Most of the more recent simulations have been performed using boundary conditions and blank
shapes that directly mimic the forming experiments being performed. The only substantial difference
between the simulations and the experiments is in the application of forming pressure. In the simulations
pressure increase is a linear function of time. In the forming experiments, pressure is largely a function of
the punch displacement. Pressure also reaches a limiting value due to the limited holding capacity of the
clamping rams described in Figure 39. Pressure is modeled for the ideal forming situation. Other than this
discrepancy, the simulation inputs are set to reflect the exact properties of the blank and can be viewed in
Table 7.
18.6 GPa

Characterization Results

3.32 GPa

Equation 15

0.46

Equation 18

0.1264 GPa

Characterization Results

88,480 Pa*m

Equation 27

Laminate Thickness

0.3”

-

Number of layers

6

-

Cohesive Zone Properties

Table 5

[21]

0.1

-

1000 psi

-

Table 7 – Simulation Properties

106

The blank shape utilized is shown in Figure 38. The latest results of the simulations can be seen
below in Figure 55.

Overhead View

Punch Orientation

Undeformed
Fiber
Orientations

Isometric View

Figure 55 – Deformed Blank Shape
It can clearly be seen that the simulation results exhibits out of plane warping along the 0° and 90°
orientations of the blank. Even though the extent of the warping is more than is seen in the experiments, it
is occurring in the same regions seen in Figure 41 and Figure 42. In previous numerical simulations,
warping occurs along the ±45° orientations which are directly conflicting with what was seen in the
forming experiments. These results can be seen in Figure 56.

107

Overhead View

Punch Orientation

Isometric View
Figure 56 - Previous Simulation Results
It can be seen that the wrinkling in these simulations is along the ±45° direction of the blank. The main
reason for out of plane warping in these areas is the over prediction of shear stiffness in these regions.
This phenomenon will be further discussed in the next section.
The advantage of utilizing a user subroutine is that contour plots of any variable can be created.
One of the most useful plots is the angle between the a-fiber orientation and the b-fiber orientation. This
provides a physical quantity which can be measured and compared on the formed helmet blanks in order
to compare the numerical results to experimental ones. The fiber shear plot can be seen in Figure 57.

108

b

a

b

a

Angle between
Fiber Orientations
(degrees)

b

High Fiber Shear

a

No Fiber Shear
Figure 57 – Fiber Shear Contour Plot
The contour plot image is taken immediately prior to the wrinkling of the blank in order to show the
distribution of the value of the shear stiffness. It can be seen that the composite experiences very little
fiber shearing along the directions that the fibers are oriented. ±45° degree offset from these orientations
the maximum value of fiber shear is seen.
The fiber shear angle as well as the angle that each PFO makes with respect to the structural
coordinate system is plotted. The distribution of the fiber orientations is important due to their impact on
the material properties. Another important material property is the shear stiffness, especially when
evaluating parts to be deep drawn. The property is so important because the composite is required to
undergo large shear strains in order to conform to the punch. The shear stiffness is updated during each
step based on the fiber orientation of the PFO’s. As the PFO’s deviate from their initial orientation, the
shear stiffness ( ̅ ) increases due to the changes manifested in the transformation matrix (T). This effect
can be seen in Figure 58.

109

Value of Shear
Stiffness - ̅
term (Pa)

Figure 58 - ̅

Greatest increases in
shear stiffness seen
areas of greatest fiber
shear

Contour Plot

If the shear stiffness becomes too high, the blank may warp out of plane instead of shearing in plane. This
is one of the main reasons that abnormal wrinkling patterns were developing in the simulations depicted
in Figure 56. Lower shear stiffness will allow more in plane shearing as opposed to out of plane warping
as is seen in the most recent simulations (Figure 55).
Over-prediction of Shear Stiffness
As mentioned above the over prediction of the shear stiffness led to poor results in the initial
numerical simulations. This over prediction in the stiffness was not attributed to the empirical relation
used to calculate the shear stiffness (equations 21 and 22). In fact, the value for the shear modulus is
lower using the empirical formulas than it was found to be in the material characterization (42.1 MPa vs
126.4 MPa). The over prediction of shear stiffness comes from the use of equation 14. This equation is
used to predict the increase in volume fraction as a biaxial fiber shears due to matrix squeeze out. When
the fibers shear during deformation, the volume fraction predicted by equation 14 increases, which in turn
increases the shear stiffness (predicted by equations 21 and 22). This trend can be seen in Figure 59.
110

Figure 59 - Shear Stiffness vs Volume Fraction
The shear stiffness is yet again increased as the individual PFO’s are rotated to the structural frame using
equation 9. The effect of the rotating a single PFO from 0° to 90° on the shear stiffness ( ̅ ) can be
seen in Figure 60.

Figure 60 - Frame Rotation Effect on Shear Stiffness

111

Figure 60 shows that the value of the shear stiffness is increased due to rotation of the material frame to
the structural frame. The use of equation 14 increases the shear stiffness twice leading to an over
prediction of the parameter. This is evidenced by looking at similar time steps in two simulations; a
simulation that utilized equation 14 and a simulation that does not utilize equation 14. In the simulation
that utilizes equation 14 the minimum and maximum shear stiffness’s are 42.1 MPa and 883.2 MPa. In
the simulation that does not use equation 14 the minimum and maximum shear stiffness’s are 126.4 MPa
and 769.0 MPa. The minimum value indicates the shear modulus in the undeformed configuration of both
materials ( ̅

when undeformed). It is clear that the simulation that used equation 14 experiences

a much more rapid increase in shear stiffness as well as a much larger gap between the minimum value
and the maximum value. The simulation seen in Figure 56 uses equation 14 to update the volume fraction
and exhibits wrinkling in unexpected regions. The results in Figure 55 utilize a static value for the volume
fraction leading to a more realistic shear stiffness prediction and therefore a more realistic result.
The use of equation 14 (or an equation similar to it) has its uses, however. The artificially high
shear modulus prediction can be used to predict the onset of shear locking. Shear locking occurs in woven
fabrics that are experiencing high shear. As shear strain increases, eventually the warp and weft fibers will
jam over one another leading to increases in the shear stiffness. Equation 14 can be used in this regard to
simulate the effects of shear locking. Zampaloni showed good results using this equation when modeling
the forming of simple weave fiber glass composites due to this artificial shear locking [3]. The fiber
structure in SR-3136 is not woven and therefore will not experience shear locking. Equation 14 has been
omitted for this reason.
Comparison to Experimental Results
Making comparisons between numerical simulations and forming experiments is a good way to
validate the accuracy of the material model and the boundary conditions. One of the easiest ways in which
to do this is to compare the force-displacement curves of the punch for the simulation and experiment.
Upon initial comparison the force-displacement plot showed very poor correlation. After realizing that the

112

forming pressure profile was completely different in the simulation and the experiment, an additional
simulation was run with a similar pressure profile similar to the ones observed in the experiments. The
comparison showed good results after this change was made. Examples of the pressure profiles used can
be seen in Figure 61.

Figure 61 - Pressure Profiles

113

Figure 62 - Force-Displacment Comparison
The pressure profile exhibited in the experiments follow an unusual path due to the limited clamping
capability of the clamping rams described in Figure 39. Tailoring a pressure profile to fit the experimental
results allows the pressure applied in the simulation to be similar to the pressure applied in the
experiment. The displacement of the comparison is restricted to 4 inches to stay within the displacements
where pressure is building during the forming experiments.
Figure 62 shows good correlation between the load-displacement curves up to a value of 2.5
inches. After this point the numerical results see a large increase in load while the experimental values
taper off. The reason for this is that the numerical simulation begins to exhibit severe wrinkling around
this point. The out of plane warping leads to the increased punch load needed to deform the blank. If we
restrict the comparison to punch displacements below 2.5 meters then the curves show excellent
correlation, showing that the model is able to predict the punch force needed to form the blank accurately.
Another method to compare the simulation results to the experimental results is to compare the
shape of the blank clamping areas. This comparison can be seen in Figure 63. The blank shape

114

comparison shows the same blank shape after forming. This shape is produced primarily by the
orientations of the reinforcing fibers. They tend to draw in most along their axial alignment while 45°
from these orientations the blank draws in very little, producing the tabs features seen in Figure 63.
The biggest difference seen between the two results is the wrinkling on the numerical result is more
prevalent than on the experimental results.

Model shows ability to
correctly predict the areas in
which wrinkling will occur

Similar tab features remain after punch
is drawn through the blank

Fiber orientations
for the blanks
shown

Blank shapes nearly perfectly
overlay one another
Figure 63 – Blank Shape and Wrinkling Region Comparison
115

Figure 63 also shows that the model has the ability to predict the location that wrinkling will
initiate in effectively. A vast majority of the wrinkling occurs along the blank holding areas in the regions
that extend away from the center of the helmet along the fiber orientations. The model shows wrinkling
extending a little too far up the dome of the helmet giving room for improvement.

Future Recommendations
The accuracy of the numerical mode has been greatly improved with the implementation of the
material properties from the material characterizations. Further increases in accuracy could be obtained
from obtaining the last few properties need to completely characterize the material and forgo empirical
formulas altogether. These last few properties include the transverse shear stiffness and the traction
separation properties. The tests required to obtain these properties were not performed due to time
constraints and difficulty in machining the specimens required to perform the tests.
The next improvement that should be made is to evaluate the material properties at the
temperature at which the composite is formed. All of the properties were evaluated at room temperature.
Undoubtedly the properties would deviate significantly if evaluated at the forming temperature. The
material is currently treated as an elastic material. This should raise concerns instantly, being that the
deformation being simulated is a forming procedure which implies permanent deformation. Therefore, a
model which incorporates plasticity is needed order to increase simulation accuracy. More specifically,
the model should incorporate a visco-plastic response. Thermoplastic polymers are highly viscous
materials when being formed close to their melting temperature so the need for viscous effects in the
material response is crucial. Implementing an elasto-viscoplastic material model coupled with the effects
of the PFO’s was outside the scope of this work and satisfactory results have been obtained without it.
Perhaps a more obtainable improvement could be made by simply increasing the versatility of the
transverse shear stiffness. The transverse shear stiffness is implemented as a static value for set of shell
elements in the Abaqus input file. The stiffness does not take into account changes in fiber orientation,
thickness or any other changes that occur during the simulation. An additional user subroutine should be
116

written which allows more control over this parameter which has shown to be critical throughout this
study.

117

Conclusions

Throughout the course of this study, composite thermo-hydroforming has been taken from
process that only has the ability to for thin composite hemispheres, to a process that has the ability to
form thick composite laminates. Better pressure control and temperature control are needed to further
increase the quality of the final part. Different clamping mechanisms also need to be investigated to
determine which ones give the best final results. Initial results have indicated that composite
hydroforming is a viable manufacturing method that has the ability to form deep drawn parts. With
additional investigation it is highly likely that this process could be implemented on an industrial scale in
the automotive and aerospace industries in order to drastically decrease process time. Processing time for
the numerical experiments was on average approximately 5 minutes. This time can be further reduced
with better pressure and temperature control in addition to programming the forming steps into the
controller software. Additionally, there is a dire need for more clamping force. Future designs should take
into account the die shape and desired forming pressure when sizing and selecting the clamping rams.
Numerical results were also obtained that showed good correlation with the experimental results
on the basis of wrinkling location, blank draw-in shape and load displacement curves. Out of plane
warping prediction is an issue that needs to be investigated further. The general wrinkling area can be
predicted but the wrinkling shape is not very accurate. Mesh refine may lead to better results but also
leads to decreases in thickness that shell elements can accurately represent. The best way to increase the
accuracy of the simulations at this point is to re-measure the material properties at elevated temperatures.
This will give the most accurate response of the material as it is being formed.
The ACH is a difficult piece to form, especially since it is the first functional part manufactured
with the composite thermo-hydroforming process. It should be noted that ACH’s are currently formed
using Kevlar that is excessively darted in order to reduce wrinkling and form a part with a high quality
surface finish. CTH is nearly able to perform this task using a far simpler blank and using a process that

118

has the ability to be completely automated. These good results indicate the CTH should be able to form
parts with simple curvature as it currently exists. With minimal additional investigation, this process can
be implemented on a massive scale.

119

REFERENCES

120

REFERENCES

[1] A. Yousefpour, "Fusion Bonding/Welding of Thermoplastic Composites," Journal of Thermoplastic
Composite Materials, vol. 17, pp. 303-341, July 2004.
[2] G. Schinner, "Recycling Carbon-Fiber-Reinforced Thermoplastic Composites," Journal of
Thermoplastic Composite Materials, vol. 9, no. 3, pp. 239-245, July 1996.
[3] 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.
[4] W. D. Bascom and L. T. Drzal, "The Surface Properties of Carbon Fibers and Their Adhesion to
Organic Polymers," 1987.
[5] T. Semoto, Y. Tsuji, H. Tanaka and K. Yosizawa, "Role of Edge Oxygen Atoms on the Adhesive
Interaction between Carbon Fiber and Epoxy Resin," The Jounral of Physical Chemistry, vol. 117,
no. 47, p. 24830–24835, 2013.
[6] L. Ye, "Manufacture of CF/PEEK composites from powder/sheath fibre preforms," Composites
Manufacturing, vol. 5, no. 1, pp. 41-50, 1994.
[7] C. Beyler, "Thermal Decomposition of Polymers," in SFPE Handbook of Fire Protection
Engineering, 2002, pp. 110-131.
[8] M. Zampaloni, "Kenaf natural fiber reinforced polypropylene composites: A discussion on,"
Composites: Part A, pp. 1569-1580, 2007.
[9] Q. Shubhra, "Mechanical properties of polypropylene composites: A review," Journal of
Thermoplastic Composite Materials, vol. 26, no. 3, pp. 362-391, April 2013.
[10] N. G. Karsli, "Effects of maleated polypropylene on the morphology, thermal and mechanical
properties of short carbon fiber reinforced polypropylene composites," Materials & Design, vol. 32,
no. 7, pp. 4069-4073, August 2011.
[11] I. Giraud, "Preperation of aqueous dispersion of thermoplastic sizing agent for carbon fiber by
emulsion/solvent evaporation," Applied Surface Science, vol. 266, pp. 94-99, 2013.
[12] Honeywell, Processing Honeywell Spectra Shield, 2011.
[13] T. Xu, "Matrix Free Ultra-High Molecular Weight Polyethylene Fiber-Reinforced Composites:
Process, Structure, Properties and Applications," New Polymertic Materials, vol. 916, pp. 391-405,
September 2005.
121

[14] T. Xu and R. J. Farris, "Comparative Studies of Ultra High Molecular Weight Polyethylene Fiber
Reinforced Composites," Polymer Engineering and Science, vol. 47, no. 10, pp. 1544-1553, 2007.
[15] E. Greenhalgh, "Fractographic observations on Dyneema composites under ballistic impact,"
Composites: Part A, vol. 44, pp. 51-62, May 2012.
[16] ASTM International, Standard Test Method for Tensile Properties of Polymer Matrix Composite
Materials, ASTM International, 2008.
[17] B. Russell, "The high strain rate response of Ultra High Molecular-weight Polyethylene: From fibre
to laminate," International Journal of Impact Engineering, vol. 60, pp. 1-9, 2013.
[18] ASTM International, Standard Test Method for Shear Properties of Composite Materials by the VNotched Beam Method, ASTM International, 2012.
[19] ASTM International, Standard Test Method for Shear Properties of Composite Materials by VNotched Rail Shear Method, ASTM International, 2012.
[20] W. Yan-lei, "Evaluation of in plane shear test methods for composite material laminates," Journal of
Chongqing University, vol. 6, no. 3, pp. 221-226, September 2007.
[21] G. Liu, "Collapse of a composite beam made from ultra high molecular-weight polyethylene fibres,"
Journal of the Mechanics and Phiysics of Solids, vol. 63, pp. 320-335, 2014.
[22] R. Okine, "Analysis of Frming Parts from Advanced Thermoplastic Composite Sheet Materials,"
Journal of Thermoplastic Composite Materials, vol. 2, pp. 50-76, 1989.
[23] H. Yin, X. Peng and T. Du, "Forming of thermoplastic plain woven carbon composites: An
experimental investigation," Journal of Thermoplastic Composite Materials, vol. 27, no. 5, pp. 1-13,
2013.
[24] R. McCool, "Thermoforming carbon fibre-reinforced thermoplastic composites," Journal of
Materials: Design and Applications, vol. 102, no. L, pp. 91-102, 2011.
[25] A. C. Karmaker, "Injection Molding of Polypropylene Reinforced with Short Jute Fibers," Journal of
Applied Polymer Science, vol. 62, pp. 1147-1151, 1996.
[26] M. Gupta and K. K. Wang, "Fiber Orientation and Mechanical Properties of Short-Fiber-Reinforced
Injection-Molded Composites: Simulated and Experimental Results," Polymer Composites, vol. 14,
no. 5, pp. 367-382, 1993.
[27] F. Pourboghrat, M. Zampaloni and A. Benard, "Hydroforming of Composite Materials". United
States of America Patent 6631630 B1, 14 October 2003.
[28] M. A. Zampaloni, "Stamp Thermo-Hydroforming: A new Method for Processing Fiber-Reinforced
Thermoplastic Composite Sheets," Journal of Thermoplastic Composite Materials, vol. 17, no. 1, pp.
31-50, Janurary 2004.
122

[29] R. Jones, "Chapter4: Macromechanical Behovior of a Laminate," in Mechanics of Composite
Materials, Washington D.C., Scripta Book Company, 1975, pp. 147-173.
[30] W. R. Yu and F. Pourboghrat, "Non-orthogonal constitutive equation for woven fabric reinforced
thermoplastic composites," Composites: Part A, vol. 33, pp. 1095-1105, 2001.
[31] W. R. Yu, M. Zampaloni and F. Pourboghrat, "Sheet hydroforming of woven FRT composites: nonorthogonal constitutive equation considering shear stiffness and undulation of woven structure,"
Composite Structures, vol. 61, pp. 353-362, 2003.
[32] W. R. Yu, M. Zampaloni and F. Pourboghrat, "Analysis of flexible bending behavior of woven
preform using non-orthogonal constitutive equation," Composites: PArt A, vol. 36, pp. 839-850,
2005.
[33] P. Boisse, "Finitie element simulation fo textile composite forming including the biaxial fabric
behavior," Composites: Part B, vol. 28, no. B, pp. 453-464, 1997.
[34] N. Abedrabbo, M. Zampaloni and F. Pourboghrat, "Wrinking control in aluminum sheet
hydroforming," International Journal of Mechanical Sciences, vol. 47, pp. 333-358, 2005.
[35] P. Harper, "Cohesive zone length in numerical simulations of composite delamination," Engineetring
Fracture Mechanics, vol. 75, pp. 4774-4792, 2008.
[36] ASTM International, Standard Test Method for Tensile Properties of Plastics, ASTM International,
2010.
[37] D. T. Campbell and D. Cramer, "Hybrid Thermoplastic Composite Ballisitic Helmet Fabrication
Study," Advanced Materials and Process Engineering, 2008.
[38] U. Mohammed, C. Lekakou and M. Bader, "Experimental studies and analysis of the draping of
woven fabrics," Composites: Part A, vol. 31, pp. 1409-1420, 2000.
[39] N. Chang, A. Bellare, R. Cohen and M. Spector, "Wear behavior of bulk oriented and fiber
reinforced UHMWPE," Wear, vol. 241, no. 1, pp. 109-117, 2000.
[40] O. Okoli, "Development of a Semi-Empirical Method for Obtaining the Dynamic Young’s Modulus
in Random Continuous Reinforced Glass/Epoxy Composites," ournal of Reinforced Plastics and
Composites, vol. 19, no. 4, pp. 292-300, 2000.
[41] J. Halpin and J. Kardos, "The Halpin-Tsai Equations: A Review," Polymer Engineering and Science,
vol. 16, no. 5, pp. 344-352, 1976.
[42] C. G. Davila, "Effective Simulation of Delamination in Aeronautical Structures Using Shells and
Cohesive Elements," Journal of Aircraft, vol. 45, no. 2, pp. 1-10, 2007.

123

[43] V. Kushch, "Explicit Modeling the Progressive Interface Damage in Fibrous Composite: Analytical
vs. Numerical Approach," Composites Science and Technology, vol. 71, pp. 989-997, 2011.
[44] J. Davalos, P. Qiao, H. Salim and J. Schlussel, "Shear Moduli of Structural Composites from Torsion
Tests," Jounral of Composite Materials, vol. 36, no. 10, pp. 1151-1173, 2002.
[45] A. R. Offringa, "Thermoplastic composites - rapid processing applications," Composites Part A:
Applied Science and Manufacturing, vol. 27, no. 4, pp. 329-336, July 1996.
[46] J. Moore, "Manufacture and Mechanical Testing of Kenaf Fiber Reinforced Polypropylene
Composites with Aluminum Faces," 2007.
[47] T. J. Keener, "Maleated coupling agents for natural fibre composites," Composites Part A, vol. 35,
no. 3, pp. 357-362, March 2004.
[48] K. Karthikeyan, "The effect of shear strength on the ballistic response of laminated composite
plates," European Journal of Mechanics and Solids, vol. 42, pp. 35-53, April 2013.
[49] I. Chang, "Recent Developement in Thermoplastic Composites: A Review of Matrix Systems and
Processing Methods," Journal of Thermoplastic Composites, vol. 1, no. 3, pp. 277-296, July 1988.
[50] W. J. Cantwell, "The impact resistance of composite materials — a review," Composites, vol. 22, no.
5, p. 347–362, September 1991.
[51] H. Toftegaard, "Determination of In-Plane and Out-of-Plane Elastic Constants for Medium
Thickness Sandwich Composite Skins," in Sandwich Structures 7: Advancing with Sandwich
Structures and Materials, Aalborg, Denmark, 2003.

124