.l}4.1?.!§~rn 1.. .

I.

 

 

. u. o. . l -6. i1!:.!‘ 1‘“,

 

. , . . C. ..
u?....r..|l........

 

 

llll‘lllllllllilllllll

31293 0

This is to certify that the
thesis entitled

POWDER PROCESSING OF COMPOSITE PREPREG TAPE:
PARTICLE SIZE EFFECTS

presented by

SANJAY PADAKI

has been accepted towards fulfillment
of the requirements for

MS degree inLHEMlLAL ENGINEERING

 

Date ._.W

0-7639 MS U is an Affirmative Action/Equal Opportunity Institution

 

lehlgan State

LIBRARY ‘
M
University +

 

 

PLACE IN RETURN BOX to remove thin checkout from your record.
TO AVOID FINES return on or before date due.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

MSU In An Affirmative ActIoNEquel Opportunlty Institution
C.

W31“

 

POWDER PROCESSING OF COMPOSITE PREPREG
TAPE : PARTICLE SIZE EFFECTS

By

Sanjay Padaki

A THESIS

Submitted to
Michigan State University
in partial fitlfillment of the requirements
for the degree of

MASTER OF SCIENCE

Department of Chemical Engineering
Composite Materials and Structures Center

1992

ABSTRACT
POWDER PROCESSING OF COMPOSITE PREPREG TAPE :
PARTICLE SIZE EFFECTS
By

Sanj ay Padaki

The effect of polymer powder particle size on the final volume fraction of the prepreg
tape produced by the powder prepreg process and on the differences in its subsequent
consolidation was the major focus of this investigation. As part of the study, process
enhancements in the areas of aerosolizer wall calcing and particle agglomeration were made. A

semi—empirical model for the process of spreading the fiber tow was also developed.

The results indicate that an optimum polymer powder particle size range exists for the
fabrication of prepreg tape with this fiber-matrix combination in this process. However, the

effect of particle size on consolidation efficiency was not conclusive.

During this investigation, two new applications for the powder prepreg process were
developed. A possible method for creating porous, formable RTM preforms from prepreg tape
made from large particle sizes and low matrix volume fractions was developed. A second
application demonstrated that hollow glass sphere filler material could be fluidized and deposited

in a discontinuous, random fiber mat.

ACKNOWLEDGENIENTS

I would like to thank Dr. L. T. Drzal for his constant guidance and support

throughout the course of this work.

This study is based on the investigation conducted by Dr. S. R. Iyer and I would
like to acknowledge his work. I am also grateful to MMPI who financially supported
this investigation.

I have taken the assistance of my colleagues on numerous occasions and I am
indebted to them. In particular, I wish to thank Murty for his suggestions and assistance

at various times.

Finally, I would like to thank my family who have been a constant source of

support and encouragement.

iii

TABLE OF CONTENTS

List of Tables
List of Figures
1. Introduction
2. The Process
2.1 Process Description
2.1.1 Fiber spool
2.1.2 Spreader
2.1.3 Aerosolizer
2. 1. 4 Heater
2.2 Process Automation
2.2.1 Fiber motion
2. 2. 2 Oven temperature
2.2.3 Data acquisition
2.3 Supplementary Process Description
2. 3.1 Volume fraction determination
2.3.2 Consolidation
2.3.3 Polishing and analysis
3 . Material Properties
3.1 Fiber Reinforcements

3.1.1 Unidirectional carbon fiber
3.1.2 Glass fiber mat

3.2 Matrix
3.3 Filler material

iv

III 10
10
. 14
. 14
16
l6
. 17
. 17
. 17
"I 18
. 21
. 21

21
. 21

.24
.24

4. Process Enhancements
4.1 Line Speed Control
4.2 Caking of Aerosolizer Walls
4.3 Agglomeration °
4.4 Particle Concentration Measurement
5 . Process Modelling
5.1 Choice of Modelling Technique
5 .2 Model Development
5. Particle Size Effects
5.1 Prepregging Experiments
5.2 Consolidation Experiments
6. New Applications
6.1 RTM Preform Manufacture
6.2 Impregnation of Fillers
8. Conclusions
Appendix
Bibliography

.30
.30
.32
.36
.43
.45
.45
.46
.54
.55

82

. 70

.80
.82

Table 1.1
Table 1.2
Table 2.1
Table 3.1
Table 3.2
Table 4.1
Table 5.1
Table 5.2

Table 5.3

Table 6.1
Table 7.1

LIST OF TABLES

‘ Comparision of matrices

Popular thermoplastic matrices

’ Sample calculation of volume fraction

Properties of polyarnide and carbon fiber

Properties of SCOTCHLITB " glass bubbles

Governing equations of particle interaction forces

Parameters involved in spreader modelling

Spreading experiments on A84 carbon fiber at constant amplitude, varying
frequency

Spreading experiments on A84 carbon fiber at constant frequency, varying
amplitude

Results of prepregging experiments

Parameters and weights of typical impregnation run

Figure 2.1 "

Figure 2.2
Figure 2.3
Figure 2.4
Figure 2.5
Figure 3.1
Figure 3.2

Figure 3.3

Figure 3.4

Figure 3.5
Figure 3.6
Figure 4.1
Figure 4.2
Figure 4.3

Figure 4.4
Figure 4.5
Figure 4.6
Figure 5.1
Figure 5.2
Figure 6.1

Figure 6.2
Figure 6.3

Figure 7.1
Figure 7.2

Figure 7.3

LIST OF FIGURES

" Schematic of powder prepregging process

Schematic of the spreader

Schematic of the aerosolizer

Schematic of the control loop

Typical consolidation cycle

SEM micrograph of Orgasol powder

Particle size distributions of Orgasol powders

DSC therrnograms of as received Orgasol powder (A) 5 micrdns (B) 10
microns (C) 20 microns (D) 65 microns

Variation of sintering time with temperature and particle size for
polyamide

Particle size distribution of SCOTCHLI'I'E " glass bubbles

SEM micrograph of SCOTCHLITE " glass bubbles

Schematic of modified control strategy

Schematic of buzzer mechanism

Variation of sound pressure level with time

Characterization of fluidized powders by Geldart

Trend of charge to mass ratio with specific surface area
Schematic of Real-time Aerosol Monitor

Comparison of model and data of spread width versus frequency
Comparison of model and data of spread width versus amplitude
Variation of Va with amplitude

Schematic of consolidation mold

Plot of consolidation pressure versus Vm

Photograph of mold with preform

SEM photograph of RTM preform

Schematic of aerosolizer for filler impregnation

vii

Figure 7.4 Frequency versus bed height at various amplitudes for 18" diameter
aerosolizer (Amplitude values are relative)
Figure 7.5 SEM photograph of hollow glass sphere on glass fiber

Chapter 1

 

INTRODUCTION

Composite materials, as the name suggests, are a combination of two or more
materials. These individual materials are physically and mechanically separate
components, each having its own set of properties. In a composite they are combined
to produce a composite material that has significantly better properties than either of its

constituents alone.

Composite materials can be categorized into three groups :
a) Fibrous Composites : consisting of either short, long or continuous fibers
in a matrix.
b) laminated Composites : consisting of layers of different materials.
c) Particulate Composites : consisting of particles of one or more materials
enveloped in a matrix.
Fibrous composites will be the main focus of this dissertation. Popular fibers in use
currently are carbon, glass and aramid. The matrix material used with these fibers varies
with the application. They ean be ceramics, metals or polymers. Polymer matrix
composites are used in the majority of today’s aircraft, automotive, sporting goods and
other related industries. Polymer matrices are of two types - thermoset and thermoplas-
tic. Thermoset matrices such as epoxy resins are used in the majority of composite

applications because of their ease in processing, low shrinkage, good adhesion and

2
excellent chemical resistance. The drawbacks of using thermoset resins are their limited
shelf life (coupled with expensive storage costs), low resin strengths and brittleness.
Thermoplastic matrices offer potential for lower cost manufacturing due to their
indefinite prepreg stability, fast processing cycles, ease in storage, ability to be
reprocessed, high fracture toughness, good damage tolerance and impact resistance, good
resistance to micro—cracking and ease in quality control. They also have their own
drawbacks such as difficulty in processing because of inherently high viscosity, solvent
sensitivity and poor quality of the prepreg tape. Table 1.1 compares the properties of
thermoset and thermoplastic matrices for composites. Table 1.2 lists the popular

thermoplastic matrices in use today along with their T,’s and Tm’s.

The T, or glass transition temperature of a polymer is the temperature below
which the polymer molecule is frozen in position and exists in a glassy state. At this
stage all molecular motion of the polymer chains ceases. The T, is affected by‘a number
of factors including polydispersity and rate of cooling. The melting temperature, T,, of
a polymer is the temperature at which there is an abrupt change in volume and coefficient
of thermal expansion of the material. Both T, and T,, are practically important. T, sets
an upper temperature limit for use of amorphous thermoplastics like polystyrene and a
lower temperature limit for rubbery behavior of an elastomer like SBR rubber. With
semicrystalline thermoplastics, T,, or the onset of the melting range determines the upper
service temperature. Between T,,, and T,, semicrystalline polymers tend to be rough and
leathery. Brittleness of the amorphous regions begins to set in below T,. As a general
rule the working temperature range for semicrystalline plastics is between T, and a
practical softening temperature which lies above T, and below T. [1]. However, the
viscosity of thermoplastics in this working temperature range is still extremely high and

is one of the main problems facing the processing of thermoplastic composites.

Table 1.1 Comparison of matrices 121

 

Property Thermoset Thermoplastic
Weight + +
Material Cost + +

Processing cost-reduction
potential

Simplicity of chemistry
Melt flow

Prepreg tack and drape
Long prepreg shelf life +
Low processing temperature

Low processing pressure

Low processing cycle time

Low cure shrinkage

Quality control data base
Mechanical property data base
Ability to translate fiber properties
Solvent resistance

Corrosion resistance

Resilience

Toughness +
Lack of time-dependance
Interfacial adhesion

Low thermal expansion

++
++

++
++

++++++
+

+++

 

Table 1.2 Popular thermoplastic matrices

 

Name T, °C T. “C
Victrex PEEK 143 343
poly ether ether ketone

BASF Ultrapek 175 375
poly ether ketone ether ketone ketone

DuPont PEKK 156 33,8
poly ether ketone ketone

Ryton PPS 90 285
poly phenylene sulfide

LaRC-TPI 217-229 275-305
thermoplastic imide

Radel C 260 none
poly ketone

Torlon C 275 none
poly amide imide

 

Many solutions have been presented to the problem of processing thermoplastic

composites. These include:

* Co-mingling - weaving reinforcing fibers with thermoplastic matrix fibers
together to get a shapeable fabric.

* Filament bundle coating - sheathing small filament bundles with the
thermoplastic polymer.

* Film stacking - compacting alternate layers of fiber with thin sheets of
resin.

* In-situ polymerization - preimpregnating the fiber tow with a low
molecular weight prepolymer and subsequently polymerizing a linear chain
in situ.

* Melt impregnation - forcing molten polymer into a fiber tow to form a

prepreg tape.

5
* Powder prepregging - impregnating the fiber tow with a dry powder form
of the polymer.
* Slurry processing - Drawing the fiber tow through a suspension of solid,
small polymer particles suspended in a liquid carrier.
~11 Solution processing - dissolving the resin in a suitable solvent and
impregnating the tow with dilute solution which is later processed to

remove solvent.

The final product of all the above processes (i.e. a material containing the reinforcing
fibers and polymer in an unconsolidated form) is called a prepreg. Prepreg can be in the
form of tape, laminates or fabric. These can be converted into useful parts and
components by a variety of processes usually involving the application of heat and

pressure.

In the case of thermoset polymer composites the prepreg is in a partially reacted
state called a B-staged part. This is because the part is not completely cured. The B-
staged part needs to be kept under refrigerated conditions in order to prevent unwanted
curing. In order to get an ”useful" part, the polymer chemical reaction needs to occur
in the composite which completes the cross-linking of all of the polymer chains. This
reaction often takes several hours of staged heat treatment. Once the reaction is

completed, the part cannot be processed again.

Thermoplastic polymer composite materials do not need to undergo a reaction.
They are molded by simply applying heat and pressure to a geometrical arrangement of
prepreg. This results in faster processing times. Due to the absence of chemical

reactions, thermoplastics have an indefinite shelf life thus reducing handling costs. In

6
order to get a final part the layers or plies of prepreg tape are laid up in the desired
configuration. Any one of the processing methods of stamp molding, laminating,
diaphragm forming, thermoforming, pultrusion, or filament winding can then be used to

get the final composite part.

Thus thermoplastic composites offer a number of potential advantages over the
traditional thermoset matrix composites but the inherent difficulty in their processing has
required newer processing technologies to be developed. One of them based on a
polymer powder prepregging strategy has been developed here at Michigan State
University [3] . The emphasis of the present work was to investigate particle size effects

on the performance of the process and explore new applications for it.

Chapter 2

 

THEPROCESS

The powder prepregging process has many advantages over the other processes

listed in Chapter 1. The following are the most important features of the process :

a)

b)

The average particle size used is approximately in the same order as the
dimensions of the fiber. This enables a precise control over volume fraction.
This also removes matrix viscosity as a variable in the process since the polymer
melt has to flow only over small distances.

Spreading of the fibers enables almost every fiber to be completely wetted by the
matrix.

The deposition of the matrix onto the fiber can be controlled precisely and over
a wide range of volume fractions.

The prepreg tape obtained from the process has good tack and drape qualities.
The tape can be easily laid up and transformed into a composite part by using

simple consolidation cycles.

This chapter describes the various components of the powder prepreg process [4], the

subsequent steps to get a final part and the analysis of the part.

2.1

Process Description

Figure 2.1 shows a schematic of the entire system. A tow of continuous fiber is

unwound from the fiber spool (E) by the nip rollers through a guide slot (1). The nip

8608a wane—noun Season we owed—05m Ha Baum..—

 

 

eae e2
88:— 03:0
.25 so
833
8:886 gee.—

E 9.03.
58mm

 

 

38am
38m 3E

{ddduinifiliu-t

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

9
rollers (I) are all motor driven. The tow passes over the spreader (B) where it is
separated into individual filaments. The section of the tow after passing through the
spreader remains separated though this may not be obvious from the diagram. The
spread tow enters a particle deposition chamber called the aerosolizer (C) where it picks
up the polymer powder. The particle deposition is via an aerosol. The generation of the
aerosol is achieved through an acoustic vibration process which will be explained later
in this chapter. Dry nitrogen is supplied into the aerosolizer through the inlet (H). The
particles deposited on the fibers in the aerosolizer are sintered in the heater (D). The
frequency generator (F) and amplifier (G) provide the acoustic energy to the spreader and

aerosolizer. The prepreg tape produced by the process is wound onto the take-up drum
(15).

2.1.1 Fiber spool .-

The fiber spools are basically fiber unwinding and winding mechanisms.
Continuous fiber is commercially produced in rolls which are either internally or
externally wound and can be directly fitted to the unwinding spool. There is no motor
drive on this spool and the unwinding of the fiber tow is achieved by the tension supplied
to the tow through the nip rollers.

The take-up spool serves to wind the prepreg tow produced by the process and
is driven by a variable speed motor. The motor speed is matched to the speed of the
process by adjusting the voltage supplied to it. The entire take-up drum setup is mounted
on a movable stage. The stage is driven by a linear motor and thus moves the spool in
lateral directions. A combination of the rotary and transverse motion causes the prepreg

tape to raster across the entire spool to maintain an uniform layer. Limit switches at the

10
ends of the motion reverse the direction of lateral travel when the tape reaches the end

of the spool.

2.1.2 Spreader :

The spreader is a patented device which operates using acoustic energy [5]. The
basic function of the device is to spread the incoming fiber tow into a plane so that
individual fiber filaments are exposed. This is necessary for the downstream section of
the process when each individual filament nwds to be coated with matrix. A diagram

of the spreader is shown in Figure 2.2.

The spreader is made up of a speaker over which a number of highly polished
stainless steel rods are mounted. The fiber tow is threaded through these rods. These
rods are mounted on frictionless bearings and rotate when there is tension in the tow.
The underlying principle is that when fibers are vibrated the individual filaments are
shaken apart from the bundles of fiber present in the tow. The rods serve as nodes for
the vibrating fiber filaments. The speaker placed below the rods provides the acoustic
energy. A function generator coupled with an amplifier provide the energy source for
the speaker. Normal operation of the system requires acoustic energy at around 35 hz
and 12 V which gives the maximum spreading widths. In order for efficient spreading
the tension in the spreader should be zero. This is achieved by placing nip rollers at the
ends of the spreader. The two rollers are coupled mechanically and so exert no tension

on the tow between them.

2. 1.3 Aerosolizer :
The aerosolizer is also a patented device and is the key component in the system

[6]. The purpose of the aerosolizer is to generate a controlled concentration of an

 

 

 

 

 

 

s ., //////////// / /
// /

 

Figure 2.3 Schematic of the aerosolizer

 

13
aerosol of fine polymer particles that then adhere to the individual filaments in the spread
fiber tow. Figure 2.3 shows the basic parts of the aerosolizer.

The spread tow that emerges from the spreader enters and exits the aerosolizer
through two coplanar slots made in the acrylic tubing which forms the body of the
aerosolizer. The aerosolizer operates in a batch mode with the polymer powder being
charged into the chamber at the beginning of the run. The top and bottom of the
chamber are closed off by rubber diaphragms which are held in place by O-rings. A dry
nitrogen gas inlet serves to provide an inert atmosphere and a constant buoyant force
within the chamber. The speaker which vibrates the air column inside the chamber

draws its power from a 100 watt amplifier which is coupled to a sine wave generator.

The main process behind the operation of the aerosolizer is the generation of the
aerosol. The bed at the bottom of the aerosolization chamber serves as a reservoir for
powder particles. The energy supplied to the speaker is transferred to the bed and to the
air column within the aerosolization chamber through the rubber diaphragm. The
oscillation of the diaphragm causes the bed of powder to fluidize. The fluidization
results in the collision of clusters of powder particles which releases a fine stream of
powder termed the aerosol. The oscillation of the air column coupled with the buoyant
force of the nitrogen stream carries the aerosol to the top of the chamber where it

interacts with the spread fiber tow. This entire process is termed aerosolization.

Normal operating conditions require resonance in the column which is gaged by
visual observation of the diaphragms. At resonance the diaphragms oscillate with

maximum amplitude. The voltage supplied to the system determines the maximum

l4
arnplitude and the average volume fraction of the outgoing tow. This can range from 6
to 12 volts (ms) with corresponding V, (fiber) ranging from approximately 90% to 50% .
A sound level meter placed at the height of the fiber tow monitors the sound pressure

level in decibels within the system.

2.1.4 Heater :

The purpose of the heater is to sinter the polymer particles to the fiber tow. The
tow from the aerosolizer has particles adhering to it only by molecular and electrostatic
forces. To prevent loss of matrix during subsequent handling, these particle must sinter,
coalesce and adhere to the tow. The heater is a normal convection type oven (VWR
1300, 750 Watts) and is capable of achieving temperatures up to 250 °C. The oven has
been modified for the process by machining slits in the front and back. The top half of
the oven has been sealed off with insulation in order to provide a more concentrated heat

source.

2.2 Process Automation

The different components of the system described in Section 2.1 are integrated
into a control system. A schematic of the control system is shown in Figure 2.4. The
control software is run by a Hewlett-Packard personal computer QS/ 168 with an interface
card (Data Translation DT2811-PGI-I), a controller box (Composite Line Controller or
CLC) and various sensors monitoring the process. Three main tasks are performed by
the control system :

a) Fiber motion control

b) Oven temperature control

c) Data logging of process variables

15

 

 

 

 

{JD
00

 

 

 

 

 

tto

 

 

 

 

 

 

M010”

 

 

 

 

 

QMTW

 

 

 

Interface Board

 

Personal Computer

 

 

 

Figure 2.4 Schematic of the control loop

”2

TDD

§§O

 

16
2.2.1 Fiber motion :

The fiber motion system is a critical part of the powder prepreg process.
Variations in the tow speed result in uneven particle pick-up leading to variations in the
volume fraction of the prepreg tape. Fluctuations in the tow spwd could also lead to tow
breakage or tow scatter. A feedback controller is used in order to maintain a constant
line speed. The controller operates the two nip roller motors as shown in Figure 2.4.
The fwdback control is a C language software program that is run on a personal

computer. The equation governing the PID control loop is [7]

_ At 8 _ T
P.=p+K. [e.+§;<e.+-’;J>+Z%<e.- .-)] (1)

Both motors operate on a 10 V d.c. power supply and are controlled by similar PID
loops. The values of the constants K” 1,, and r, for the two motors are input to the
control program. The take-up drum is driven by its own d.c. motor and runs at a
constant base speed. The direction of travel of the lateral motion motor is switched

through the program by sensing the output from the limit switches.

2. 2. 2 Oven temperature .-

The oven temperature needs to be controlled in order to get the proper quality of
prepreg tape. High temperatures lead to stiff prepreg tapes while low temperatures do
not sinter the particles to the fiber. The temperature is controlled by the control program
through an on-off control loop. The temperature in the oven is sensed by a K-type
thermocouple. The voltage signal is converted into a digital signal by an interface card.
The output of the feedback loop is sent to a switching unit that switches the oven on or
off depending on the deviation from the set point. The set point itself is an input to the

control program.

 

17
2. 2.3 Data acquisition :

The control program logs data from the process during the run. The variables
that are logged are time, motor speeds along with standard deviations, temperature and
sound level. In addition the particle concentration signal from the Real Time Aerosol
(RAM) monitor is also logged.

2.3 Supplementary Process Description
In order to fully study the effects of a particular variable on the process, a number
of subsequent experiments must be carried out to characterize the prepreg tape that is

manufactured and are described below.

2. 3. 1 Volume fi'action determination :

Prepreg tape made from the process is cut into known lengths. A number of them
(typically three out of every ten lengths) are weighed in order to get a statistical
distribution of weights over the run. Knowing the average weight of a similar length of
bare fiber, an approximate average value for the fiber weight fraction (W,) is
determined. Using the appropriate densities for fiber and matrix a fiber volume fraction
(V,) is calculated. The matrix volume fraction (V..) can then be calculated by subtracting
from 1.0. A sample calculation is shown in Table 2.1.

2. 3.2 Consolidation .-

In order that a final part be made, it is necessary to consolidate the prepreg tape
obtained from running the prepregging process. The prepreg tape is cut up and laid up
in a matched die chrome steel mold. Non-porous Teflon' film is placed between the
mold surfaces and the lay-up to facilitate easy removal of the part after consolidation.

A requisite amount of pressure and temperature is applied in a simple consolidation cycle

18

Table 2.1 Sample calculation of volume fraction

 

Sample length, inch = 32
Weight of 32" of prepreg tape, gm = W32
Weight/cm of prepreg tape, gm/cm = W3, / (32 x 2.54)
= Wl
Weight/cm of bare fiber, gm/cm = W,
Weight/cm of matrix, gm/cm = W,-W2
= w3
Weight fraction fiber = W2/W,
= Wf
Density of fiber, gmlcc = 6,
Density of matrix, gmlcc = 6.,
Volume/cm fiber, cc/cm = W2/5,
= V1
Volume/cm matrix, cc/cm = W3/6,
= V2
Volume fraction (fiber) = V,/ (V , +V2)
_—. vf

 

as shown in Figure 2.5. The equipment used for the process is a plate and frame press
(Carver press) with both heating and cooling capabilities. The temperature in the mold
is detected by a K-type thermocouple placed within the mold face. Consolidation
pressure is calculated based on the load applied on the area of the lay-up rather than on

the area of the mold.

2. 3.3 Polishing and analysis .°

In order to view the sample under the microscope to determine V. and V,,, it is
necessary to polish the surface of the composite to below 0.05 micron grit. A section
from the consolidated part is cut using a diamond saw and mounted in an acrylic holder.
The direction of mounting ensures a cross-sectional view of the sample. The sample is
polished in successive steps using a 80 grit belt sander, Struers Abramin Automated
Polisher (240,320,1000,2400 and 4000 grit), LECO GP20 Grinder/Polisher (5 micron

and 1 micron) and a Buelher Vibromet (0.05 micron). The polished samples are cleaned

19

 

 

 

TEMPERATURE

 

 

 

 

 

 

 

3

. ,1... i

— u 3

/ \
l’x/
I"/
30 45
TIME (MINUTES)

—— Supply temperature profile
........... Pressure profile ........... Actual temperature profile

Figure 2.5 Typical consolidation cycle

20
in an ultrasonic bath and then with ethanol. An Olympus microscope with a video
camera attachment is used along with a Amiga personal computer in order to digitize 15
random images. The digitized images are then analyzed for volume fraction and void
fraction using the ONVFA (Optical Numeric Volume Fraction Analysis) program [8].
The program counts the number of fibers in a given screen and calculates the fiaction
occupied by them. Void fraction analysis is a little subjective since the grey levels from

a histogram are used in the determination.

In summary, various parameters in the process need to be analyzed and the
sequence of prepregging, consolidation, mounting, polishing and volume fraction analysis
needs to be performed for each set of these parameters. This would give an understand-
ing of the effects of operating variables on the powder prepreg process and help in scale-

up and commercial application of the process.

Chapter 3

 

MATERIAL PROPERTIES

The materials used in the investigation of particle size effects were carbon fiber
reinforcement and polyamide matrix. A new application based on hollow glass spheres
and random glass fiber mat was also investigated in the present study. This chapter

discusses the properties of these components.

3.1 Fiber Reinforcements
3.1.1 Unidirectional carbon fiber .-

The reinforcing fiber used was the PAN (polyacrylonitrile) based unsizcd
Hercules AS4-3K carbon fiber consisting of 3000 fibers per tow. The mean diameter of

the fibers is 8.0 microns.

3.1.2 Glass fiber mat :

Two types of random glass fiber mats manufactured by Certainde Corporation
and designated as 816 and 850 were used. Each type has different weight percentages
of a proprietary thermoplastic binder. The fiber mat is commercially supplied in 50"
wide rolls. For the purposes of experimentation 12" x 12" sections were cutout from

these rolls.

21

22

3.2 Matrix

The polyamide (N ylon- 12) matrix material for this investigation was obtained
from Atochem Inc. in the form of powders (commercial name Orgasol). Four different
size ranges (mean sizes 5, 10 ,20 and 65 microns) of the powder were obtained. The
properties of all four powders are identical. Figure 3.1 is an SEM micrograph at 720x
of the 10 micron Orgasol powder. Figure 3.2 shows the particle size distributions for
the four powders. It can be noticed that the size distributions for the 5 and 10 micron
powders are extremely tight while those for the 20 and 65 micron are much broader.
Figures 3.3 (A,B,C,D) show DSC (Differential Scanning Calorimetry) thermograms for
the four powders [9]. The thermograms indicate that the powders are very similar and
their melting range is between 160 and 180°C. Figure 3.4 shows the variation of
sintering time with temperature as the particle size is varied. This was estimated using
Frenkel’s theory [10] and the viscosity model for polyamide that was developed by Iyer
[l l]. A listing of the computer program and the program logic is attached in the
Appendix.

3.3 Filler Material

A hollow glass sphere filler material was investigated. The filler is manufactured
by 3M and is commercially called SCOTCHLITE "' Glass Bubbles. The glass bubbles
combine low density with a high level of durability and chemical resistance. The size
distribution of the filler is shown in Figure 3.5. Figure 3.6 is an SEM micrograph of
the spheres at 200x.

The properties of the above materials are summarized in Tables 3.1 and 3.23.

 

" From manufacturer’s literature

 

Figure 3.1 SEM micrograph of Orgasol powder

24

 

 

5

§

 

conga 3&5 a 883%.. one. ocean an 83E

@5825 Saw—“Ev £095—
A

 

ld‘uilfi' A Q -

‘

4\

no»? ems—26 03 3355.8.— E 33:52

Ana—90¢: mg 329%. meow
3:328 05 32D Noam
Amp—83E o c Ban—Km Sou
3:838 3 32GB Sow

O

I>IO

tom
tom
.9»
row
[cc
rob
-ow
-oa

 

 

co—

25

2.23:. no AB 283:. an AB 2238
2 A5 283:. n A3 8258 8390 3382 3 mo «8.23505 0.05 mtm 0.5»...—

81. 0.5.2.0!— .uo_ 9.39.0.3

9&-on.-om.lo~. om.-8¢lorn-wo-fl-ouu.7 .- Hal 1 R. - R. I mnl l

 

 

9.92. o PIE: 0

 

 

 

 

 

9.3.3. 1
33.2.
9.8.2.

 

 

 

("It ”I! 30.4

 

 

 

8.. cites...»

9‘ 8m 1 emu cm. on- v . 1 gm“ 1 cm I a“ 1 m. t m

 

9.2.2. o 9.82. <

 

 

,L
f

 

 

 

L( ll) 1... he . . i,

I\1~.I:
Dean. :-

 

 

 

I I
that can seer-1

 

 

 

27

 

 

 

 

30 a' 4
32 20 «-
0
E .
2
O
>
10 -
o/Iau’asz’)‘ 2145;-

2.8 3.9 5.5 7.6 11 16 22 31 44 62 88 125176200

MICRONS

Figure 3.5 Particle size distribution of SCOTCHLITE " glass bubbles

 

 

28

. ..:

SKU X288 1884 188.8U CEO91

 

Figure 3.6 SEM micrograph of SCOTCHLITE " glass bubbles

29

Table 3.1 Properties of polyamide and carbon fiber “’1

 

ORGASOL POLYAMIDE
Specific gravity 1.02
. Melt Temperature, °C 175

Specific heat, cal/gm/°C

@ 40 °C 0.58

@ 140 °C 0.81
Thermal Conductivity, W/m/°C 0.22-0.31
Viscosity, Pa.s (0.1 sec", 197.3 °C) 52.6
Surface tension, mJ/m2 30.0

HERCULES AS-4 CARBON FIBER

Diameter, microns 8.0
Specific gravity 1.80
Specific heat, cal/gm/°C
@ 75 °C 0.22
@ 175 °C 0.27
Thermal Conductivity, W/m/ °C 5.73

 

Table 3.2 Properties of SCOTCHLITE "' glass bubbles

 

Average particle density, g/cc

Nominal 0.22
Range 0.19-0.25
Bulk density (range), g/cc 0.09-0.17
Floaters
(% by bulk volume) 90
Thermal conductivity,
BTU/hr/ft2/°F/in @room temp. 0.2-1.2

Softening point, °C 715

 

Chapter 4

 

PROCESS ENHANCEMENTS

The process that has been described in the previous chapter has many shortcom-
ings. One of the aims of this investigation was to refine certain sections of the process
primarily to improve the control over volume fraction of the prepreg tape and also to aid
in the development of a scaled-up version of the process. The chief problems with the
present prepreg process are described below. Methods employed to overcome these

difficulties are also described.

4.1 Line Speed Control

One of the drawbacks of the prepreg system is the line speed control strategy.
The control method employed is the matching of speeds of the two separate fiber spool
motors that are placed at the ends of the process. Exact speed monitoring is not possible
by this technique due to the physical limitation of the control system. The speed of
motor 1 is always fractionally different from that of motor 2. This difference in the
speeds causes the fiber tow tends to scatter (when there is little tension) or break (when
there is high tension). The uneven tension in the tow also results in variable tow spread
width.

The possible solution to this problem is the employment of a new control strategy.
A schematic of this strategy is shown in Figure 4.1. This strategy would involve

30

 

.3039. 35:8 wagon. mo ouaoaow fie 85mm

30% 0::
define as
wagon” Honey—
uo=8 3:8
Homoom commend. ESQ EEO 08m
econ com
8% can

 

 

 

\H/

 

 

 

 

 

i

 

 

 

 

 

uozobnoo

32
adjusting Speed according to the tension in the tow. The tow speed itself would be set
at only one point thus eliminating the problem of matching speeds of two independent

motor drives.

4.2 Caking of Aerosolizer Walls

The caking of the walls of the aerosolizer by the polymer powder causes a drop
in the volume fraction of the prepreg tape. The caking affects the system in two ways.
First, it depletes the amount of powder available for aerosolization. Second, it changes
the sound characteristics of the aerosolization chamber which in turn results in poor
aerosolization. Both result in uneven powder deposition and variation in the volume

fraction of prepreg tape.

Caking is a result of the high surface charges present on polymer particles. In

order to quantify this surface charge, a simple equation listed below can be used.

Q

47R2
where (2)
a = surface charge density
Q = total charge
R = radius of the particle

 

This equation indicates that as the radius of the particle decreases the charge density
increases. It also indicates that as the total charge increases the surface charge increases.
However, for a particle in air the surface charge density at the interface with air cannot
grow to an infinitely high level due to the dielectric nature of air. Under normal
conditions, the maximum value of the surface charge density is limited to 2.7 x 10"
C/m’. For a single 5 micron particle this translates into a total charge of 2.12 x 10'" C

and is a fairly high value.

33

In addition to the dielectric properties of air and the high charge density on the
surface of the polymer particle, the particles themselves being insulators have very low
dielectric constants. This causes the charges to concentrate on the surface without being
dissipated into the surrounding medium. When a surface having dielectric constants
lower than air is brought in the vicinity of the polymer powder, the particles tend to
preferentially attach themselves to it and thus eauses caking. In order to solve this
problem, the charge must either be neutralized or a force greater than that exerted by the
charge must be used on the powder particles to dislodge them.

Many techniques employing the above approaches were tried. The wall was
coated with an antistick coating (fluro glide ' CP, manufactured by Chemplast Inc.) but
did not reduce the calcing. An aluminum foil layer inside the chamber reduced the
calcing to a great extent. However, points of higher surface energy (e. g. fingerprints,
surface roughness, etc.) acted as nucleation sites for wall caking. A silver coating on
the inside of the chamber was also tried in order to reduce surface irregularities but did
not show any additional improvement. A combination of surface charge dissipation and

wall tapping was however extremely successful.

The wall tapping mechanism was a simple electromagnetic doorbell (buzzer). A
schematic of the set-up is shown in Figure 4.2. The bell itself has been removed so that
the arm of the buzzer raps against the side of the aerosolizer chamber. The frequency
and time of rapping is controlled through a separate relay. The repeated rapping
dislodges the accumulated particles from the wall. The effectiveness of the buzzer can
be seen in Figure 4.3. The two runs have been performed at different amplitudes
accounting for the fact that the average sound pressure levels (SPL) are different.

However, the trend shows that there is a decrease in SPL when the buzzer is off. There

 

 

 

34

 

AEROSOUZER

 

 

 

ooooooooooooooo
sssssssssssssss
nnnnnnnnnnnnnn

S FROM
W R SUPPLY

It
ooooo

 

 

 

Figure 4.2 Schematic of buzzer mechanism

35

cm

 

on

o

o

2:: £5 _o>o_ 2385 258 .«o 553:3 mi manur—

AmoSEEV 95...
cm av cm on or

 

r8

 

 

_ _ q _ _ _ _ _ . _

 

 

to $35
so 555

 

 

 

 

m?—

our

rm?

NNF

mm?

va

mm?

our

(SP)1dS

36
is a scattering of SPL when the buzzer is switched on but the average SPL is almost

constant throughout the run.

4.3 Agglomeration

Agglomeration of powder particles is a serious problem affecting the process. It
results in poor particle pick-up and large fluctuations in the prepreg tape volume fraction
during the run. In order to solve this problem a thorough understanding of the factors

involved in the agglomeration process is necessary.

The high charge to mass ratio of smaller size polymer particles causes
agglomeration when they are vibrated in the aerosolizer. Physically, agglomeration can
be described as the association of a number of smaller particles forming a much larger
particle. The degree of agglomeration depends on the charge to mass ratio of the
individual particles. Reist [13] has proposed that particle agglomeration can occur in
ordered flow such as that which can be established in a sonic field. Here agglomeration
takes place by the different velocities imparted to particles of differing inertia, by
aerodynamic attractive forces between the particles and by radiation pressure which
moves the particle towards the vibration antinodes. In the aerosolizer a similar situation

arises which causes the polymer particles to agglomerate.

However, agglomeration is not an universal phenomena. Only certain categories
of powders tend to be cohesive in nature. Geldart [14] has classified all powders into
four categories - Types A, B, C and D. Figure 4.4 shows the classification graphically.
Type A powders are characterized by the fact that, when fluidized at a gas rate not too
far above U_, (the minimum fluidization velocity), the expansion is homogeneously

distributed over the powder bed. When the gas rate is further increased above another

37

 

.2. tau—co .3 moo—:58 332:: he 558180830 v.9 9.53..—

.e2 an.
000m 000. 00m 00. Om
1 . - - .1-

 

 

1 d 1 1 1 _ 11 1 1 d 1 1 1

      

35200.

U

00m

1

33323:

<

A

1 000.

e. . 0 v men.» v.35
23.0 v 8a. 235

 

 

b b D b P pIPPP P b 5 I -PPDP1DII 5

 

38
critical gas rate, ”bubbles“ of the gas form and move up through the powder bed. Type
B powders are generally coarser than Type A powders. They begin "bubbling” at the
minimum fluidization velocity and do not exhibit a homogeneous expansion. There are
negligible interparticle forces in this category. Powders in the Type D category are
generally the coarsest of solids and are dense and/or large. Due to their large size and
high particle momentum, these powders do not exhibit any tendency to agglomerate.
All cohesive powders fall into the Type C category and are very hard to fluidize. They

also exhibit extensive agglomeration and large interparticle forces.

There is a general consensus of opinion that cohesion of dry powders is mainly
due to Van der Waals forces [15]. This force operates both in vacuum and in a liquid
environment, although in the latter case the Van der Waals forces is substantially reduced
[16]. The Van der Waals forces are only noticeable when the particles can come
sufficiently close together, on the order of the size of a molecule (0.2 to 1 run). As the
particle size increases, the gravitational forces begin to dominate. Therefore
cohesiveness of powders is usually apparent when the particle sizes possess a major

dimension of less than 10 pm.

In addition to the Van der Waals force, other attraction forces may also operate
between the particles in a powder, such as capillary, electrostatic and magnetic forces.
Table 4.1 lists the governing equations for each of these forces. Grace [17] has also
pointed out that the boundary behaveen A and C type powders is strongly influenced by
the above forces as well as variables such as surface asperities, hardness of the solid
material, humidity, melting or softening point, electrical conductivity, and magnetic
susceptibility. In general, these forces are smaller than the omnipresent Van der Waals

force.

39

Table 4.1 Governing equations of particle interaction forces

 

 

 

Van der Waals force between two spheres, F“ a 1.2%
H
2 1______
. Q ( (R2+H2).5)
Electrostatic force between two spheres, Fc =
16115de
Capillary force between two spheres, F, = 217R
2
Magnetic force between two spheres, F_ = p
611111 2
Force of gravity on a sphere, F = 1.33pg1rR3

where A =Hamaker coefiicient (dependent on the material);

H =separation distance; so = permittivity of vacuum;

Q = particle charge; 7 =surface tension; 11 = permeability;

p = degree of magnetization; p = density; g =acceleration due to gravity

 

 

 

Electrostatic forces can contribute significantly to the cohesiveness of powders and
thereby their agglomeration as long as a gaseous environment is considered. Tribo-
electric charging or the formation of a potential difference when particles of different
work functions are brought into contact are the two major ways of development of
electrostatic charge [13]. The former is due to the transfer of charge between similar
particles due to friction. The latter occurs from a transfer of charge between dissimilar
surfaces such as between the wall of the chamber and the powder particle. The
agglomeration tendency of powder particles increases drastically with increasing surface
area. This behavior has been noticed with the 5 micron powder that has been used in this
investigation. The size of agglomerates varies from a few microns to 3 to 5 centimeters

in diameter.

Figure 4.5 is a graph of the variation of the charge to mass ratio with specific
surface area [18]. The results have been obtained by Glor, et.al. with a lab scale

 

40

 

to“ 111‘ 0' 10'
I T

7‘ ' 7 /
<
O
:1" o
a
V!
O
3' -
é

 

 

 

L l L ‘a-O
to" 10' re‘ 10’

SPECIFIC SURFACE AREA[m2/ltg]

 

Figure 4.5 Trend of charge to mass ratio with specific surface areaml

 

41
pneumatic transport device from many different organic powders of various substances
of different particle size distributions. The overall dependence and trend of the charge
to mass ratio is indicated by the two solid lines. This dependence is almost linear. The
dashed line corresponds to the theoretical limit (based on the assumption of density =
103 kg/m’; maximal surface charge density = 2.7 x 10" Coulombs/m’). For 5 micron
powders, the charge to mass ratio would be significantly higher than for 10 or 20
micron powders because decreasing particle size causes an increase in specific surface
area. Thus this powder would tend to be more cohesive. This is confirmed in the
eXperiments where the 5 micron powder agglomerated extensively compared to either the

10 or 20 micron powders.

A number of methods were tried in order to reduce agglomeration. Initially,
stainless steel ball bearings (5/16' diameter) were introduced into the aerosolization
chamber along with the polymer powder. The choice of ball diameter was 5/16" because
it was approximately the average size of the agglomerates in the chamber. The concept
was to allow the balls to bounce around the aerosolizer and upon impact, break up the
agglomerates. This technique was only partially successful. It reduced the size of the
agglomerates but did not eliminate them. It was noticed that a number of fine
agglomerates still existed in the chamber. Additionally, the spheres preferentially moved
to the edges of the chamber due to the oscillation of the aerosolizer diaphragm and due
to the presence of a vibration free, low energy node. Thus at the end of a given run, the
balls were all found at the periphery of the chamber.

A second similar technique, reported by Sheehan and Carillo [19], involved using
”hair-balls”. These are metal spheres coated with some sort of fiber (e.g. cotton). They

were used to disperse micrometer and submicrometer size particles (dust) in vacuum.

42
The hairballs serve several roles. They grind and mill the powder or dust particles. The
bouncing balls break the Van der Waals and electrostatic bonds that bind the particles
together. The fiber coating on the ball serves as a macroscopic solvent into which the
dust dissolves before it is freed by impact. The fiber also serves as a cushioning which
prevents damage to the diaphragm. The hairballs did not reduce agglomeration to a great
extent. Observations revealed that the hairballs served as accumulation sites for powder.
The fibers around the ball entrapped the powder and prevented their release into the
aerosol. Thus the hairballs continually increased in size forming large agglomerates
themselves. These agglomerates again accumulated at the low energy nodes around the

periphery of the aerosolization chamber.

A third technique utilizing a 1/2' opening wire mesh was also tried. The
rationale behind this method was to prevent agglomerates by the same mechanism as in
the previous two methods but at the same time avoid the migration of the crushing
medium to the chamber walls. The wire mesh was cut square to a size slightly smaller
than the aerosolizer chamber. This allowed the mesh to move and rotate freely inside
the chamber while at the same time vibrate with the powder. The mesh was successful
in breaking some of the agglomerates. However, smaller size agglomerates still existed
in the chamber. In addition, due to the differences in conductivity of the mesh and
powder, the mesh was coated with the powder during the course of a run. Over long

periods of time the mesh openings would close.

Since the technique of physically breaking the agglomerates was not entirely
successful, a different approach was tried. This involved changing the conditions of the
medium surrounding the powder particles. Initially dry nitrogen gas was used. The gas
was allowed into the chamber at the bottom of the chamber. No significant change was

43
noticed in the agglomeration behavior. A humidified nitrogen supply was then
substituted. This was achieved by bubbling dry nitrogen through an flask of distilled

water. No noticeable difference was observed in the size of the agglomerates.

A fifth technique of drying the powder before aerosolization was also tried. The
powder was heated in a pan for 10 minutes at 125 °C. The dry powder was immediately
put into the aerosolization chamber and a positive nitrogen gas supply was started. This
was done in order to prevent any back diffusion of moisture into the chamber. The
aerosolizer was then run for about 30 minutes. This technique did reduce the size of
agglomerates to around 0.5 mm diameter. Moisture is thus an important factor in
agglomeration. It tends to accumulate and form hydrogen bonds with the organic
molecules on the surface of a polymer particle. The free ends of the water molecules
on adjacent particles can also form hydrogen bonds. Thus a number of particles can
associate together and form agglomerates. The last technique described reduces the
moisture content on the surface of the particle and thus reduces agglomeration. It
therefore seems to be the most feasible solution to the problem of agglomeration. This
modification should be incorporated into future systems where the powder would be

continuously dried before feeding into the aerosolization chamber.

4.4 Particle Concentration Measurement

Aerosol concentration is an important design variable needed for scale-up of the
prepreg process. In order to measure this a Realtime Aerosol Monitor (RAM-S) is fitted
online with the system. The operating principle [20] of the RAM-S is based on the
detection of the near-forward scattered electromagnetic radiation in the near-infrared.
Figure 4.6 is a schematic of the system. Samples of the powder in the aerosol chamber
are conveyed to the RAM-S through a copper tube. The optical detection unit in the

 

44
RAM-S measures the scattered radiation and coupled with the flowrate of the aerosol
stream calculates the concentration. However, due to the inherent cohesive nature of
polymer powder the measuring system gets fouled. The instrument zero drifts after a

short sampling period. This causes large errors in the readings.

Since powder charging is the one of the primary reasons for fouling, a possible
solution would be the introduction of an online deionizer unit that would neutralize the
charge on the polymer particles. This must be considered in any future scale-up systems

that are planned.

INLET FROM
AEROSOUZER

 

OPTICAL

IF: DETECTION

ASSEMBLY
FILTERED

AIR MANIFOLD : :

FILTER

 

 

 

 

 

 

 

 

 

 

mi...
6

 

 

 

 

 

 

Figure 4.6 Schematic of Real-time Aerosol Monitor

Chapter 5

 

PROCESS MODELLING

A number of variables affect the powder prepregging operation. A model that
incorporates these variables would aid in understanding their effect on the process.
Three main areas of the powder prepreg process can be identified for the purposes of
modelling : the spreader, the aerosolizer and the sintering oven. The last two have been
characterized physically [4]. The aim of this chapter is to develop a model for the

spreading operation.

5.1 Choice of Modelling Technique

Three approaches to the modelling are possible. A purely theoretical method
could be followed which involves determining the individual forces and parameters and
then using fundamental principles to develop a mathematical representation for the unit
Operation. This requires a thorough knowledge of the fundamental quantities and forces
that are involved. The operation of spreading is mathematically not well understood and
hence this method would be quite difficult.

Another method which would be the other extreme of modelling involves

generating data from the experimental setup and then using a graphic software package

to fit the data to a curve. This would give a mathematical model that describes the

45

 

46
process but does not indicate the physical parameters. This technique is therefore

empirical in nature and would be the easiest of the three techniques.

A third technique falls between the above two categories. This is semi-empirical
and gives a more realistic picture of the physical situation than the purely empirical
approach. Dimensional analysis is a method that falls under this category [21]. In order
to apply this technique, one needs to know only the variables behaved to be involved and
their dimensions. With the use of fundamental algebra, and in some cases without any
experimental data, this method can predict the effect of a variable on the model.
However, this method will not give any quantitative information and experiments must

be conducted in order to quantify the model.

Dimensional analysis tries to group all the variables involved into dimensionless

groups. There are two techniques that have been used :

a) Rayleigh method

b) Buckingham Pi method
The Rayleigh method is the simpler of the two since there is no necessity of knowing the
number of dimensionless groups beforehand [21]. It has been used in literature to model
material properties and processes [22]. Thus it was chosen for the purposes of this
modelling.

5.2 Model Development

Rayleigh’s method is based upon the premise that if n quantities, Q1, Q2, Q, Q»
. .. . , Q... are involved in a certain physical phenomenon, for the purpose of the
dimensional analysis their mutual dependance may be expressed as a power product of
the following type:

 

47
Q = K (Q2)‘2 (Q3)“ (Q4)“ (Q.)“ (3)
where K is a dimensionless constant. By introducing the fundamental units of the n
quantities and by using the condition of dimensional homogeneity, the constants a2, a3,
a., a, can be determined absolutely or related to each other. The relationships can

then rearranged to give dimensionless quantities.

Table 5 .1 Parameters involved in spreader modelling

 

Parameter Description Base units Comments
(Mass, Length. Time)
W Watts ML’I’I‘3 Related to amplitude
(rms Volts).
f0 Frequency l/T Term incorporating
applied frequency and

resonant frequency.

A Fiber adhesion MIT2 Adhesion existing bet-
ween fibers due to
sizing, etc. . Will be
zero for bare fiber.
Units of energy/ area.

11 Tow size none Number of fibers /
tow.
D Fiber diameter none Ratio of diameter of
ratio fiber to a standard fiber
(AS4 carbon fiber).
p Number of rods none - I
6 Friction none Friction between fiber
coefficient and rod.

8 Spread width L Primary variable

 

.V-_—J-II-H _‘ 5.1

48
The first step in the analysis is to list all the parameters involved in the physical situation
along with their fundamental units. The parameters for the spreader are listed in Table
5.1. The motivation for the development of the model is to relate the spread width of
the fiber tow to physically observable parameters.

By equation (3) that describes Rayleigh’s method,

S = KW‘ffA ‘n‘p ‘JD3 (4)

where K is a dimensionless constant.

Substituting the dimensional base units for each parameter in equation (4),
ML2 " l b M ‘
L= [—l H H <5
T3 T T2 . )

Using the condition for dimensional homogeneity, i.e., all the units must balance in an

equation,

EM=0 : a+c=0 (6)
EL=0 : 2a=l (7)
2T=0 : -3a-3b-2c=0 (8)

Solving equations (6), (7) and (8) for a, b, and c, a=0.5;b=c=-0.5.
Resubstituting these values in equation (4),

S = KWoJL-OJA-Ojndpeeng (9)

Rewriting equation (9),

 

49

 

s = K [f2] Mndpretot (10)

0

which is the desired relationship of spread width to the parameters.

This model predicts that the spread width goes as the square root of the amplitude
(or watts) and is inversely proportional to the square root of inter-fiber adhesion. It also
postulates that the spread width is inversely proportional to the frequency parameter, f,.
This parameter needs to be defined and must incorporate the resonant frequency, f,, of
the system. The definition also nwds to ensure that physical data collected from the
spreader fit any curve predicted by the model. A simple definition was considered as a

first approximation and is given by equation (11).

g (f-fl)2 (11)
fr

The square of the numerator was chosen to give f, the units of frequency. Substituting

 

f.

equation (11) in equation (10),

S = “[333] [(f-f} I ”new: (12)

 

Data from spreading experiments of A84 carbon fiber [23] are listed in Tables 5.2
and 5.3. The data shows a trend similar to that predicted by the model. The model
predicts that the spread width goes to infinity as the frequency approaches resonant
conditions. Physically this translates into a maximum for the spread width. The model
also describes a square root curve for amplitude. From the data it can be seen that the
spread width almost plateaus out with increasing amplitudes thus closely following the

model. Figure 5.1 graphically displays the trend of the spread width as frequency is

50
varied. The data from Table 5.2 has also been plotted. All the parameters from
equation (12) except for frequency have been combined into a single parameter which has
a value of 5.53. This value has been arrived at by visually best-fitting the data to the
model. The value of the individual parameters is however not known. A value of 35 .5
hz has been chosen as the resonating frequency of the spreader unit that was used to
obtain the data. Figure 5.2 is a similar representation of the variation of spread width
with amplitude. The single parameter is now made up of different individual parameters
and has a value of 2.29. From these two graphs it can be seen that as a first approxima-

tion the model fits the data reasonably well.

Though data from spreading of fibers with different inter-fiber adhesive strengths
has been obtained, they are not listed because of the inability to quantify inter-fiber
adhesion. The trend indicated by the model is that of decreasing fiber spread with

increasing inter-fiber adhesion which is to be expected.

Thus, a very simple model has been developed which describes some of the trends
of the spreading operation. Further data acquisition and analysis is required in order to
fit the model better to the observed values. The model is not perfect and is only an
indicator of the performance of the spreader. Further refinement of the model is
necessary to enable practical use of it. The effects of the remaining parameters, for
which coefficients have not been determined, must be investigated. Finally, in order to

complete the model scaled-up spreaders should be used to determine its effectiveness.

 

51

Table 5.2 Spreading experiments on AS4 carbon fiber at constant amplitude, varying
fr-o uenc

    

Frequency Amplitude Spread Width

 

 

 

 

 

 

 

 

 

 

 

 

 

Hz rms Volts cm
2 25.3 10.7 3-7 I
3 30.0 10.5 4-9 I
4 32.1-32.4 10.5 5-9 I
5 34.3 10.5 5-9
6 36.1 10.6 6-9 I
7 38.3-38.4 10.5 5-9
8 40.0 10.5 4-7 I
9 42.2 10.5 2-6 I
10 44.1 10.5 1.: 0.5-5 I

 

 

Table 5.3 Spreading experiments on AS4 carbon fiber at constant frequency,
varyingamplitude

 

    

    

Frequency Amplitude

         

Hz rms Volts

     
  

36.3-36.4
2 36.3

3 36.3

4 36.3-36.4
5

6

 

 

 

 

36.1
36.0

 

 

 

 

 

.8533 SE? 523 9.9% .«o 83 vs. ~32: no comma—Eco fin Baue—

 

52

 

lo
(we) 111mm peerds

O
F

 

Ané 3:262...
ov mm on mm
a.
.m.
o .m.
rm... 0 O O O
+
a. , i?§.:>mm.mum
.322
mama

 

 

 

 

 

 

53

NF

avenues 833 523 32% «a 53 Ea .82: mo 33885 «6 8:»:—

I

o.

3 2.5 83:25

m

m

N

 

 

_

—

up

_

d

fi

—

 

 

mdgvamdum
Enos.
Sad

+

 

 

 

 

OGQNCDIDVCON

P

(we) Lupw peest

Chapter 6

 

PARTICLE SIZE EFFECTS

Two different areas of the powder prepreg process were examined in this
investigation. The first was to determine the effect of particle size on the operation of
the system. The second was to develop new applications for the process. Chapter 6
describes the particle size effects while new applications are discussed in Chapter 7.

The particle size is a variable in the process. Its effect on the system could be
in two separate domains :

a) It could affect the manufacturing conditions of the prepreg tape.

b) It could change the processing characteristics of the prepreg tape.
As described in Chapter 2, the manufacture of the prepreg tape involves unwinding of
the fiber tow, spreading it into individual filaments, depositing polymer particles from
an aerosol onto the fiber, sintering the individual powder particles in place and finally
winding the prepreg tape onto a spool. The processing of the prepreg tape involves
cutting up the tape, laying it up in a mold and then applying heat and pressure in a
controlled cycle to get a final void-free composite part. It was the hypothesis of this
investigation that change in particle size would affect powder deposition and polymer
pick-up as well as sintering times. It was also thought that the consolidation pressure
needed to get a void-free part would be related to particle size. The experiments that

have been conducted were to assess these effects.

54

55

6.1 Prepregging Experiments

The experimental procedure followed for this section was as follows. The
aerosolizer chamber was removed and cleaned with alcohol to decrease the number of
high surface energy areas. The aerosolizer was replaced and charged with 225 grams
of the polymer powder under investigation. The required process variables were then
set. The operating frequency of the aerosolizer was determined by visual observation of
the diaphragm. The operating frequency corresponds to the natural frequency of the
chamber and is the point where the amplitude of oscillation of the diaphragm is
maximum. The process was started through the control software and allowed to run for
15 to 20 minutes to achieve a steady state. A piece of masking tape was used to mark
the starting point for the run on the prepreg tape. The run was allowed to proceed with
the least possible manual intervention. Manual checking was necessary from time to time
due to the variation in the line spwd control and involved slowing down the take-up
drum so that the tow tension was maintained. Fluctuations in the tow tension eaused
changes in tape width and consequently in the volume fraction of the prepreg tape. The

total run time for most experiments was approximately one hour.

The prepreg tape that was produced was cut into 32" lengths. Ten lengths were
laid together to form a batch. Three lengths (#1, #4 and #7) from every batch were
weighed accurately. These weights were entered in a spreadsheet (SuperCal4). The
average weight of an uncoated 32" length carbon fiber tow was also recorded. An
average volume fraction for the run was calculated using the technique described in

Section 2.3.1.

 

56

Table 6.1 Results of prepregging experiments

r— ,_______ _*

Size Spreader Aerosolizer V Standard ,
i Freq Amp Freq Amp m Deviation

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

: Micron Hz v Hz v % i %
5.90 17.8 3.2
7.25 14.2 4.8
5 38.5 11.0 12.4
9.00 19.0 6.5
11.15 27.2 4.3
5.90 35.0 5.9
7.25 38.9 6.4
10 38.5 11.0 12.5
9.00 39.6 5.7
11.15 47.6 4.5
4.25 10.1 3.1
5.90 14.4 4.6
7.25 23.8 4.0
20 38.5 11.0 12.5
9.00 31.3 3.1
10.00 37.6 4.5
11.15 39.5 1.4
5.90 3.1 2.0
7.25 3.4 3.7
12.3 10.00 5.4 3.9
11.15 7.6 3.4 ,
15.40 12.0 3.1

57

2. VP

63:?“ 53 ..> .6 Su§> 3 2.3..

mp

NF

3:03 85.. .026;

:

op

a

w

 

 

fi

 

:22... m
222:. or
58.8 cm
:88... mo

POO.

 

 

 

 

 

 

 

In D l0 O LO 0 ID 0 ID
v st ('3 (‘0 N or .— .—
(%) uonoeu ewmon xmew

58

The experiments were conducted on the 5, 10, 20 and 65 micron powders at
different voltage (amplitude) levels. The results of these experiments are tabulated in
Table 6.1. In Table 6.1 the amplitude and frequency have units of volts (V) and hertz
(Hz) respectively. The term V, refers to the volume fraction of matrix (polymer) present
in the prepreg tape. The results tabulated in Table 6.1 are graphically displayed in
Figure 6.1. It is seen that the data for the four different particle sizes is significantly
different. The standard deviation of each point is extremely small i.e. , any given point
can be repeated to within 5% of the stated value [24]. This fact has also been confirmed
through independent observations. The general trend is that as amplitude (and therefore
power) is increased, there is an increase in aerosol generation leading to increased
polymer pick-up. As the pick-up increases there is a increase in the matrix volume

fraction which is the y-axis of the plot in Figure 6.1.

The volume fraction of the 65 um powder shows an almost linear relationship
with amplitude. However, the pick-up is extremely small and reaches only 12 % by
volume even at significantly high amplitude levels. The 20 pm powder follows a similar
trend but attains higher pick-up percentages. The operating line is higher and a wider
range of volume fractions can be achieved with this powder. The 10 um powder
operating line is higher than the 65 um and 20 um lines and this gives much more
flexibility in operation of the powder prepreg system. The 5 micron powder shows a
trend opposite to that which is observed for the other size ranges. It is lower than the

curves for the 10 and 20 um powders.

Two important conclusions can be drawn from the above observations. First,
larger particle sizes, because of their increased weight, require considerably higher

amplitudes (and hence energy) in order to fluidize. The amplitude of the input signal

 

59
drives the speaker which in turn vibrates the diaphragm and the particles present on it.
Physically, an increased energy input to the aerosolizer translates into an increased
amplitude of oscillation of the diaphragm and hence an increased upward buoyancy force
on the powder particles. This causes larger powder particles to aerosolize and move
upwards through the chamber where they can intersect the spread fiber tow. Thus, for
the same energy, more 10 um particles intersect the fiber than 20 pm particles. This
results in approximately 10 V, % higher pick-up in the case of 10 um powder compared
to 20 um powder. In the case of the 65 pm powder this difference is approximately
30%. The operating line is much lower and hence would represent a upper limit on

particle size for the present system.

The second conclusion that can be drawn is that a lower size limit exists for the
system. This can be observed from the 5 pm curve. Theoretically, it would be expected
that the 5 pm operating line lie above the 10 pm curve since the particle size is smaller
(because less energy is needed for aerosolization for smaller particle sizes). However,
this is not the case observed in the experiments. The reason for this is the extensive
agglomeration that occurs in the aerosolization chamber due to the high charge to mass
ratio of the 5 pm powder particles. The size of agglomerates increases to approximately
3 centimeters within 10 - 15 minutes of a run. These agglomerates require higher energy
levels to fluidize and this leads to much poorer pick-up. The curve of the 5 pm therefore

lies approximately where a 35 um powder operating line would lie.

These results on the effect of particle size would indicate that a critical size range
exists with the present carbon fiber - polyamide system. Below or above this size range
the operation of the system is unsatisfactory because of poor pick-up. In the ease of the

present system this critical size range is approximately 5-20 pm. Pick-up percentages

 

60
in the range of 0-15 % by volume of matrix is undesirable since prepreg tape of this
volume fraction cannot be consolidated to give a void-free part. The theoretieal limit of
V, for a void-free consolidated part is 9.8% [25]. Therefore any prepreg tape that is to
be consolidated to get void-free parts needs to have atleast this percentage of matrix
already present. In addition a small amount of matrix is lost due to bleeding during the
consolidation process. The above factors impose limits on the powder sizes that can be
used effectively in the present system. However, these limits vary from one fiber-matrix
system to another. In order to successfully utilize the powder prepreg system, operating
windows dependent on particle size and charging properties must be established for each

fiber-matrix combination.

6 .2 Consolidation Experiments

The second part of the investigation involved the effects of particle size on
consolidation. One of the important advantages of the powder prepreg system over other
systems is that the melt viscosity of the polymer has less importance than the other
variables. The polymer has to flow over only very short distances (in the order of
Angstroms rather than millimeters) for consolidation. This would not be the case when
particle size is increased in the consolidation step when the large particles must melt and
fuse with the rest of the matrix. Thus it is important to quantify the role of particle size
in the prepreg tape as to whether processing of larger particle sizes would require higher
consolidation pressures to form void-free parts and whether the consolidation pressure
has to be higher for higher fiber volume fraction tape since the small fraction of matrix

needs more force to flow over the entire composite part.

The experimental program for this section involved making the prepreg tape with

a particular particle size. The volume fraction of the tape was estimated as described in

 

61

 

 

 

» "L . E r. ’1: 1.41;".3 " .‘ KIR'v
‘9 we); ,1- v ‘11-, 9' 3 - ~< -‘x r .,‘i, ' 1 :fi _.ti-."‘
{’1'“ ‘3. 1~ ~1/V‘I- H .7 .i ‘ “1“,.ng I \t
..1 .; . 5;, 1...“. .t ..» , .. .1.- “41,... r~
l ~'u ’l ’I ll ‘ I ~;‘ “1* ’4‘:’l 1'1
,‘i H ‘ 4,,1110‘4‘1 “ill‘ .

Easy slip fit

 

 

I 2.0- I Prepreg tape
| 7 being consolidated

 

Figure 6.2 Schematic of consolidation mold

62

Section 2.3.1. This value formed the x-axis for a plot of consolidation pressure versus
volume fraction. The entire length of tape from the run was divided into three equal
parts. Each part was cut up into 4" lengths and laid up in an unidirectional fashion in
a two-part cpen.ended mold illustrated in Figure 6.2. A base pressure was determined
from a graph of consolidation pressure and V, for the 10 micron powder [26]. This was
used as a reference point. The three parts were individually consolidated at three
different pressures - one below, one above and one at the base pressure. The consolida-
tion cycle used has already been illustrated in Figure 2.5. Sections from the composite
part were cut using a diamond saw, mounted in acrylic holders and polished down to
0.05 micron grit as described in Section 2.3.3. The volume fraction and void fraction
were determined optically using the ONVFA [8]. The consolidation pressure used and
the volume fraction are plotted as the y and x - axis respectively. An operating curve
was drawn which divided the graph into two regions. Any point above the operating
curve represented a part which showed bleeding (mattix oozes out) and was void-free.
A point below this curve represented parts that had voids. This plot is shown in Figure
6.3.

It can be seen from the graph that the operating lines for the three particle sizes
are quite close and even overlap. Within limits of experimental error all three curves are
essentially the same. This would indicate that the consolidation pressure needed to make
void-free composites is independent of the particle size over the range of 5 to 20 pm.
It must however be recalled that the melting times of polymer particles varies from
seconds to minutes depending on particle size. Thus any differences in processing that
depends on particle size will manifest themselves in this time frame. Beyond this the
polymer is basically a melt and will be a viscoelastic fluid. The consolidation times used

in the experiments are on the order of 30 minutes (heating) which exceeds the maximum

63

me

a> 3&3 2:865 85358.50 .6 83 n6 9:52

15 eczema". oE:_o> 5.8.2

cw mm

on

mm

ow

 

 

were;

 

 

:22... on I ..
5.2.: or I
:22... m . l.

 

 

 

m20>

l

 

 

com

000 _.

com F

(13d) arnssaid uouepucsuoo

64
time required for consolidation. Thus any differences between particles is wiped out,

accounting for the similarity in the three curves.

In order to study the differences in processing as a function of particle size, it
would be necessary to conduct consolidation experiments of shorter durations. The time
of consolidation should be varied at constant pressure to determine the minimum
consolidation time. Another factor that has to be considered is the thickness of the
consolidated sample. If thick samples are used for consolidation, the time for heat
transfer may be far greater than the time for melting. This would lead to uneven
consolidation in the part. So a probable experimental plan would be to choose three
different sizes of powders (preferably in the 5 - 100 pm range) and make fixed matrix
volume percent prepreg tape from them. These tapes should be laid up and consolidated
at different pressures for variable amounts of time into thin samples. A void analysis on
these samples would give data points through which an operating line for consolidation
pressure could be drawn. This line would separate the samples with voids from those

that are void-free and would quantify the effect of particle size on processing.

The above two experimental programs have showed that for the AS4 carbon fiber
- polyamide system :

a) The effect of particle size is significant and must be considered when any
sort of scale-up is made.

b) For the consolidation cycles being used, at long processing times there is
no significant differences between the prepreg tapes made from different
particle sizes. However, the relationship between processing speed and
particle size has not been established.

65
c) A critical size exists for this process with these materials for which

powder prepregging by the present process is unsatisfactory.

Chapter 7

 

NEW APPLICATIONS

The powder prepreg process is a combination of four unit operations : spreading,
aerosolization, impregnation and sintering. The individual steps of this process can also
be adapted to other applications. This chapter describes two of the new applications that

were successfully tried.

7.1 RTM Preform Manufacture

Resin Transfer Molding (RTM), also called ”liquid molding”, is a process for
obtaining a product with two relatively good appearing surfaces with better dimensional
control than other open mold processes [27]. In this process, the reinforcement, without
the resin, is loaded into the mold, the mold is closed, and the resin is injected or
transferred into the mold, in such a manner as to completely impregnate the enclosed
reinforcement. One of the problems facing this process is the shifting of the reinforce-
ment (or preform) during the injection of the resin. This leads to a bunching or nesting
of the fiber reinforcement in certain areas of the mold. This reduces the interstitial space
necessary for easy resin flow and consequently there is lesser amount of resin in these
areas of the mold. These areas are called "bald spots“ and are undesirable due to the
uneven mechanical properties they result in. Any mechanism to hold the preform in
place would therefore be beneficial. One possible way to temporarily hold the fibers
fixed in space is to use thermoplastic particles to act as welding agents between different

66

 

67
layers of reinforcement. The powder prepreg process has the advantage of being able
to produce continuous fiber tape with any desired thermoplastic matrix volume fraction

and therefore could be a method to produce RTM preforms.

The procedure to make these preforms involved operating the powder prepreg
process at low aerosolizer amplitude levels. AS4 carbon fiber and polyamide were the
materials used. A particle size of 20 microns was used in the aerosolizer and
temperatures of 160-170 °C were used in the sintering stage. The powder was sintered
to the fiber only at points of contact. Due to the low temperatures being used, the tape
remained quite flexible and tacky. This gave prepreg tape of very low matrix volume
fractions (of the order of 5 to 10% by volume) with the polymer powder almost

maintaining its original spherical shape.

The tape was then wound around a complex, three dimensional mold shape. For
the purposes of this study, a bottle with a square cross-section was used. The tape was
heated with the mold to the melt temperature of the polyamide with a heat gun or by
placing it in a convection oven for 4-5 minutes. The mold was allowed to cool and the
preform was removed. This gave a stable three dimensional reinforcement that retained
the shape of the mold. A photograph of the mold and the preform is shown in Figure
7.1.

Figure 7 .2 shows an SEM photograph at 200 X of a section of the preform. Two
important characteristics can be noticed from the micrograph. First, the polymer acts
as a uniformly distributed spacer between the individual fiber filaments holding them
firmly in space. This gives good porosity in the reinforcement allowing high mass

transfer rate and even wet out. Second, even though the fraction of thermoplastic

 

68

 

Figure 7.1 Photograph of mold with preform

69

 

 

Figure 7.2 SEM photograph of RTM preform

7O
polymer is quite small the axial distance between individual droplets of polymer is not
very large. The inherent stiffness of carbon fiber supported over such small lengths
would thus prevent any significant deformation of the fiber during high pressure

transverse flow with viscous resins.

This study has shown that thermoplastic particles can be successfully used to act
as spacers for a RTM preform. The choice of thermoplastic polymer is dependent on
the resin used during the RTM process. The thermoplastic particles are temporary and
ean be burnt off during resin impregnation. This would reduce the contamination of the
final part with thermoplastic polymer. Thus the therm0plastic particles help in holding
the preforms in place till such time the impregnated resin gels and prevents any

movement of the fibers within the part.

7.2 Impregnation of Fillers

Fillers, such as solid glass beads, have been used since the early seventies to
improve toughness of glassy SAN copolymer [28]. The addition of hollow fillers has
been of interest recently, especially in automotive applications. These fillers result in
weight savings of nearly 25 % depending on the applieation. The impregnation of hollow
glass spheres into random glass fiber mats was an application that was explored during
this study.

The properties and sources for the materials involved have already been listed in
Chapter 3. Other techniques of impregnation (such as mixing of spheres with resin
before injection into a RTM mold) have not been completely successful. The glass fiber
mat acts as a filtration device for the spheres and this results in accumulation of the

spheres near the gates of the mold. The idea behind this work was to develop a

71

   

Figure 7 .3 Schematic of aerosolizer for filler impregnation

72
technique that utilizes the impregnation capabilities of the powder prepreg system to

evenly distribute the glass spheres into the fiber mat prior to resin injection.

The setup consists of an 18" diameter aerosolizer that is capable of impregnating
12" x 12" square mats. A schematic of the aerosolizer is shown in Figure 7.3. It is
similar to the aerosolizer used in the powder prepreg process. The differences are in the
size and that there is only one slot for the fiber mat (Although a batch type operation was
used here, the process could be made continuous). The mat which is cut from a roll, is
placed in an aluminum frame that fits through the slot and rests on two supports inside
the chamber. The entire chamber can then be sealed by taping the slot.

The aerosolizer was charged with 525 grams of glass spheres. This gave a bed
height of approximately one inch. Before actual impregnation could be done, the
optimum frequency and amplitude of the system with all its components had to be
determined. To determine these parameters, a sample mat was placed in the aluminum
frame and then placed inside the chamber. The process was operated at different
frequencies and amplitudes and the fluidized bed height was recorded in each case. The
bed height was plotted against frequency for different amplitudes (voltage levels) as
shown in Figure 7 .4. The values for the amplitudes are only relative and are the dial
settings on the amplifier. It can be noticed that there is a peak in the bed height at
almost the same frequency for each curve. This frequency corresponds to the resonating
frequency of the system. The amplitude does not affect the bed height to a great extent.

The impregnation of the fiber mat with spheres involves operating the process at
the optimum frequency and amplitude with the mat placed inside the chamber.
Impregnation time depends on the weight of spheres nwded on the mat. Typically, a 5-6

 

73

8352 2a 822, 33385 .3888
bane—Se .3 Ba newsman—a meets, 3 :30: non 3&3 mucus—o2..— vfi 0.53,.—

ANIV 5:632...
9mm 05 9mm mew oéw Q: Q: mdw o.» Wm

0.0

 

. . . J . _ q _ 1 4 J _ . _ . _ d _

/

 

anaEm I
oéuaEm 4
méuaEm o

 

 

 

 

l 1

l

l 4 L

1

l

 

 

9mm
mdw
0.00
955
9mm
mN _. p
0.0m _.
mfiv P

(ww)1ufi!eH pea

9mm F

mNm P

Odom

74
minute run would give 10% by weight of spheres on the mat. This would result in a
weight savings of approximately 25%. Table 7.1 lists the parameters and weights for
a typical run.

Though the impregnation process is successful, the spheres adhering to the mat
are not fixed strongly enough to withstand normal handling. They are held to the glass
fibers by very weak forces. Figure 7.5 is an SEM photograph at 600X of the glass
spheres on a typical glass fiber. It appears that the spheres just sit on top of the spheres.
Any sudden shock or vibration to the mat will dislodge the spheres.

Many techniques were tried in order to retain the spheres on the glass fiber mat.
This included either heating the mat before or after impregnation with the hollow glass
spheres. The idea was to encapsulate the spheres in the thermoplastic binder of the mat.
However, the low density of the spheres caused them to float on the binder and remain
free. Another technique of coating the fiber mat after impregnation proved marginally
successful. Here the idea was to coat both the spheres and the mat with a thermoplastic
(e. g. polyamide) after impregnation to freeze the spheres in place. This resulted in a
high fraction of polyamide on the mat.

While a method of successfully retaining glass spheres on glass fiber mat has not
been found as yet, the above study has established that the aerosolization technique can
be used to uniformly deposit the spheres into the fiber mat. The uniform distribution of
filler will enable the final part to have uniform mechanical properties with possible

weight savings.

75

Table 7.1 Parameters and weights of typical impregnation run

 

Mass of bed, gms = 125
Relative amplitude = 4
(setting on amplifier)

Frequency, Hz = 15
Initial mat weight, gms = 45.77
Impregnation time, min = 5
Impregnated mat weight, gms = 50.1
Pick-up, weight % = 9.46

 

,} . M4 -

15m 8680' 4 - ....»«--'19"750.£E'0.981

 

Figure 7.5 SEM photograph of hollow glass sphere on glass fiber

76
In addition the study has also demonstrated that the powder prepreg system ean
be adapted to other fiber-matrix systems. The requirement is that aerosolization of the
powder should be possible and that the powder should adhere to the fiber. If these

requirements are met, commercially attractive fiber-matrix systems could be used in the

prepregging process.

 

Chapter 8

 

CONCLUSIONS

A number of conclusions can be drawn from the study presented in this thesis

specific to polyamide powder and carbon fiber :

(l) The powder prepreg process is sensitive to charge to mass ratio variations. The
smaller size particles (~ 5 pm) exhibit extensive agglomeration and do not work well
in the process. Their large charge to mass ratio causes large fluctuations in the operating
curve for the process. Larger particle sizes result in a reduction of particle pick-up due
to their poor aerosolization characteristics owing to their weight. Thus an optimum size

range exists for the system.

In order to achieve a desired volume fraction of prepreg tape from the system
with any given particle size within this range, it is necessary to determine the operating
curve for the particular conditions (e.g. , speed, fiber/ matrix system, melt temperature,
etc.). Once the operating curve is obtained, the volume fraction can be dialed in with
only 2-3 % error. The variation within each run depends on the length of the run and is
within 5 % for one hour runs. The main reason for this variation is the lack of motion

control on the fiber tow.

77

 

78
(2) Under the conditions selected for this study, the consolidation process does not
seem to be affected by particle size to the same extent as volume fraction. For the three
size ranges that were used in this investigation the consolidation curves are almost the

same.

(3) A number of flaws were found to exist in the present system. Some of these were
eliminated during the course of this study. The buzzer mechanism has reduced the
caking of the aerosolizer wall to a significant extent. Though many techniques have been
tried to break up agglomeration, none were completely successful but the size of the
agglomerates has been reduced to some degree. Problems with line speed control exist

and some recommendations have been made.

(4) New applications were explored for the process. A simple technique for RTM
preform manufacture has been developed. This preform shows good structural integrity
and has good porosity. It is envisaged that this type of preform manufacture would

eliminate porosity reduction due to shifting of the reinforcement during liquid molding.

(5) A new area of filler impregnation has been explored. This utilizes hollow glass
spheres and a random glass fiber mat. A new aerosolizer capable of operating with the
glass mats was constructed. Operating conditions for this new aerosolizer were
determined. It has thus been demonstrated that the process can be adapted to other fiber-

matrix systems.

(6) A simple model for the spreader has been developed. The approach used is semi-
empirical. The model identifies the relationship between the important factors in the
operation of the spreader.

79

The results of this investigation are by no means complete. A number of areas
of the process are still not well understood. As a result of this study, some recommen-
dations for future work can be made. These are listed below :
- Line spwd wntrol of the fiber tow needs to be improved. A new control strategy
must be used. One viable technique has been described in Section 4.1.
- The problem of agglomeration of smaller particle size must be investigated and
better methods to break up or reduce agglomeration must be developed.
- Other fiber-matrix systems must be investigated in order to make the process
commercially attractive.
- The effect of particle size on processing time has not been established. A
probable experimental program has been outlined in Section 6.2.
- A scaled-up system operating at higher speeds would make the process
economically feasible.
- Data on the spreading process is not complete and experimentation with scaled-up
units is necessary in order to develop a quantitative model for the unit operation.
- Finally, the powder prepreg process has many variables that need to be controlled
to achieve optimum quality of the composite prepreg tape. The effects of all the
variables is not entirely known. Further experimentation towards understanding the

fundamental principles and 'operation need to be accomplished.

 

APPENDIX

 

APPENDIX

The estimation of sintering time for different particle sizes of polyamide is necessary in
order to determine the operating temperature and residence time of the sintering oven.
A viscosity model developed by Iyer [l l] was used to relate the viscosity to the tempera-
ture. This is given by

T
It = 0714,17]

0

where log(a,) =2.664031(1000/T) -5.680173 (13)
log(r,) = 0.981316310g(y) + 2.702483

”'7' "’ Tr/ 7
To = reference temperature = 197 .3 "C

This model was substituted into Frenkel’s model for estimating sintering times [10]. The
model predicts sintering times based on viscosity and particle size and is given by

2:381:

where
x = radius of inter-particle bridge
r = radius of particle

17 = polymer melt surface tension

80

81
II = polymer viscosity (function of temperature)

t = sintering time

The variation is calculated using a computer program written in BASIC. A listing

of this program code follows.

10 REM : PROGRAM TO CALCULATE SINTERING TIME AS A FUNCTION
20 REM : OF TEMPERATURE AND PARTICLE SIZE
30 OPEN ”C:VISC.RES" FOR OUTPUT AS #1

40 FOR PS=5 TO 50 STEP 5

50 FOR TEMP=150 TO 230 STEP 10

60 AT= 10‘(2664/(TEMP+273)-5 .68)

7O MU =AT‘527.23*(273 +TEMP)/ (273 + 197.3)

80 TIME=2*(PS*.OOOOOl/2)*MU*.l/(3*30*.00001*100)
90 PRINT PS,TEMP,TIME

95 PRINT #1,"PS","TEMP","TIME"

100 PRINT #1,PS,TEMP,TIME

110 NEXT TEMP

120 NEXT PS

130 END

The results are shown graphically in Figure 3.4.

 

BIBLIOGRAPHY

 

BIBLIOGRAPHY

[1]

[2]

[3]

[4]

[5]

[6]

[7]

[8]

[9]
[10]

[11]
[12]

[13]

Rudinl, 913.2, The Elements of Polymer Science and Engineering, Academic Press,
Inc. , .

Carlsson, L. A. (editor), Thermoplastic Composite Materials, Elsevier Science
Publishing Company, 1991 .

I er, S. R., Drzal, L. T., "Dry Powder Processin of Thermoplastic
omposites" , Fifih Technical Conference [Proc], p 259-2 , Ameriean Society
of Composites, 1990

Iyer, S. R. "Continuous Processing of Unidirectional Prepreg" , Ph.D.
Dissertation, Michigan State University, 1990.

Iyer, S. R., Drzal, L. T., "Method and System for Spreading a Tow of Fibers",
U.S.Patent No. 5,042,111, 1991.

Iyer, S. R., Drzal, L. T. and Jayaraman, K., "Method for Fiber Coating with
Particles", U. S. Patent Application, Feb 1990.

Seborg, D. E., Edgar, T. F. and Mellichamp, D. A., Process Dynamics and
Control, John Wiley and Sons, 1989.

Waterbury, M. C., Drzal, L. T., "Determination of Fiber Volume Fractions by
Optical Numeric Volume Fraction Analysis", J. Reinforced Plastics and
Composites. 8. p 627-636, 1989.

DSC 910 manual, DuPont Co.

Frenkel, 1., J. Physics U.S.S.R.,_2, 385, 1945.

Iyer, S. R, "Continuous Processing of Unidirectional Prepreg", Ph.D.
Dissertation, Michigan State University, p 79,1990.

Iyer, S. R, "Continuous Processing of Unidirectional Prepreg" , Ph. D.
Dissertation, Michigan State University, p 89,1990.

Reist, P. C., Introduction to Aerosol Science, MacMillan Pub. Co., 1984.

82

 

[14]

[15]
[15]

[17]
[13]

[19]

[20]

[21]

[22]

[23]

[24]

[25]

[26]

[27]
[28]

83

Geldart, D., "Types of Gas Fluidization", Powder Technology, 2, p 285-292,
1973.

Rietema, K., "Powders - What are they?", Powder Technology, 31, 5, 1984.

Visser, J. , "Van der Waals and other Cohesive Forces affecting Powder
Fluidization", Powder Technology, 58, p 1-10, 1989.

Grace, J. R., Can. J. Chem. Engg., M, p 353, 1986.

Ggoré, M. , Electrostatic Hazards in Powder Handling, John Wiley and Sons Inc. ,
1 8 .

Sheehan, D. P., Carillo, M. and Heidbrink, W., "Device for the Dispersal of
Micrometer— and Submicrometer-sized Particles in Vacuum" , Rev. Sci. Instrum. , "
61 (12), p 3871-3875, 1990.

RAM-S Manual, M. LE. Inc..

Perry, R. H., Green, D. W. (editors), Perry’s Chemical Engineers’ Handbook,
Sixth Edition, McGraw Hill Book Company, 1984.

Fallstrom, K.-E. , "Determining Material Properties in Anisotropic Plates using
Rayleigh’s Method, Polymer Composites, 12, 5, p 306-314, 1991.

Iyer, S. R., "Continuous Processing of Unidirectional Prepreg", Ph.D.
Dissertation, Michigan State University, p 123, 1990.

Iyer, S. R. , "Continuous Processing of Unidirectional Prepreg", Ph.D.
Dissertation, Michigan State University, p 144-149, 1990.

C9hawla, K. K., Composite Materials Science and Engineering, Springer- Verlag,
l 87.

Iyer, S. R., "Continuous Processing of Unidirectional Prepreg", Ph.D.
Dissertation, Michigan State University, p 162, 1990.

Lubin, G. , Handbook of Composites, Van Nostrand Reinhold, 1982.
Narkis, M., Chen, E. J. H., "Unidirectional Hybrid Thermoplastic Composites

Reinforced with Continuous Fiber and Glass Spheres", SAMPE Journal, 26, 3,
p 11-15, 1990.

 

"711111111111111111111111“