Harrison, P., Thije, R.T., Akkerman, R., and Long, A.C.

(2010)
Characterising and modelling tool-ply friction of viscous textile
composites. World Journal of Engineering, 7 (1). pp. 5-22.
ISSN 1708-5284

http://eprints.gla.ac.uk/58158/

Deposited on: 6 December 2011

Enlighten – Research publications by members of the University of Glasgow

http://eprints.gla.ac.uk
 CHARACTERISING AND MODELLING TOOL-PLY FRICTION OF
VISCOUS TEXTILE COMPOSITES

Philip Harrison1*, Rene ten Thije2, Remko Akkerman2, Andrew C. Long3

*Corresponding author: [email protected]

1
Department of Mechanical Engineering, University of Glasgow, Glasgow G12 8QQ, Scotland
2
University of Twente, PO Box 217, NL-7500AE Enschede, The Netherlands
3
University of Nottingham, Faculty of Engineering, Division of Materials, Mechanics & Structures,
University Park, Nottingham NG7 2RD, UK

Keywords: Friction, Viscous Textile Composite

Abstract
The paper describes two experimental methods for measuring the tool-ply friction behaviour of
impregnated thermoplastic textile composites. These include the pull-through and pull-out tests on
the one hand and experiments conducted using a c ommercial rheometer using custom designed
platens on the other. Results from the techniques are compared and t he relative advantages and
disadvantages are discussed. A simple procedure to determine parameters in a s o-called master
curve, an e mpirical model relating the friction coefficient to normal pressure, velocity and
temperature is demonstrated. The model is convenient for implementation in a numerical code. A
predictive meso-scale model is also described that incorporates parameters such as fabric
architecture, tow geometry and matrix viscosity. The model is based on lubrication theory and can
predict steady state friction over a br oad range on c onditions. Significantly, unlike previous
attempts to model tool-ply friction using a meso-structural approach, there is no ne ed to
experimentally determine the lubrication film thickness as this is an output of the model. Meso-scale
model predictions are compared with predictions from the master equation and experimental data.

1 INTRODUCTION
Press forming of thermoplastic textile composites is potentially a fast and efficient method of
production. However, while stretch-forming and deep-drawing of sheet metal [1] are today
relatively well understood processes supported by sophisticated Computer Aided Engineering
(CAE) tools [2] the same cannot yet be said for textile composites. As such a large research effort is
underway to create equivalent CAE tools for these materials. The manufacture of textile composite
components of potentially complex double curvature geometries involves a forming stage in which
dry or pre-impregnated reinforcement takes the required shape through deep-drawing. Wrinkling of
the sheet during forming is an unwanted defect and can be inhibited via in-plane tension induced in
the sheet using a blank-holder [1-5]. Friction occurring between the composite material and metal
tooling during forming imparts tensile stresses in the sheet. These tensile stresses help counteract
the compressive stresses generated by material deformation that are responsible for wrinkling.
Several experimental methods have been employed to investigate the inter-ply [6-10] and
tool-ply [10-18] friction behaviour of impregnated textile [7, 9-11, 15-20] and uniaxial [6-8, 21]
continuous fibre reinforced composites. A heated metal sled drawn across a h eated textile
1
composite sheet was used by Murtagh et al. [21]. An early investigation by Scherer and Friedrich
[8] reported a pull-out test which could be used to characterise both tool-ply and ply-ply friction
over a r ange of temperatures. Similar pull-out designs of varying levels of sophistication and
accuracy have been used in subsequent investigations [6, 8-11, 16, 17, 20-23]. Wilks developed a
modified version of the pull-out test, a so-called pull-through test, that benefited from a constant
contact area between specimen and platens during the test [20]. The platens were electrically heated
and the rig housed entirely inside an environmental chamber allowing more accurate temperature
control of the specimen during the test [15, 18, 19]. A similar pull-through design was later used by
Gorczyca et al. [14, 24-26], though in this case without the use of an environmental chamber.
Notably, investigations by Gorczyca et al. [14, 24] aimed to reproduce conditions similar to those
experienced during actual forming operations and employed the technique of heating the preform
and tooling to different temperatures from each other before testing. More recently commercial
rheometers have been adapted to perform friction measurements [15, 18, 19, 27] . The technique,
which is reported in Section 3.2 of this paper provides a fast and efficient method to generate
experimental data, though is limited to relatively low pressures and velocities far removed from
actual processing conditions. One of the aims of this paper is to examine how useful such
experiments are in predicting more realistic friction boundary conditions.
Gorczyca et al. [24] summarise some of the different materials and test conditions that have
been investigated and reported in the literature. Materials have included: AS4/PEEK [6], carbon/PP
[8] and carbon/PEEK APC-2 unidirection composites [21] as well as both balanced and unbalanced
pre-consolidated glass/PP twill woven composites [3, 10, 11, 15, 16, 1 8, 19, 23] , glass/PP satin
weave composites [28] and glass/PP plain weave composites [14, 24, 29] . Force versus
displacement data produced by constant displacement rate tests typically follow a generic form
characterised by an initial peak force, rapidly falling to a l ower steady state value, irrespective of
the material under consideration [6, 10, 15 -19] 1. Thus, two friction coefficients can usually be
determined corresponding to a peak and steady state response. An exception to this occurs when the
displacement rate is very low (e.g. 10mm/min) [18] or the temperature is high [6, 10], under these
conditions the peak in the force curve decreases and in some cases can completely disappear.
Several factors, listed below, have been investigated to determine their effect on the
coefficient of friction between tool and composite including: normal pressure [11, 14-25], draw
velocity [11, 14-25], tool-material [21], tool-temperature [11, 14-25], fabric temperature prior to
forming [13, 14, 30], fabric orientation [6, 8, 13, 14, 21, 30], type of release agent [21] and cooling
conditions [10]. Certain factors have been found to be much more important than others. Gorczyca
et al. [14, 24, 25] list velocity, tool temperature and normal pressure as the most influential factors
in their investigation while fabric temperature prior to forming and fabric orientation seem to have
only a minor effect. In contrast, relative ply orientation has a significant effect for uniaxial
continuous fibre reinforced composites [7, 8]. Murtagh et al. [21] found release agents had a
significant effect and changing the tool material from metal to rubber also caused a large increase in
friction. Lebrun et al. [10] investigated effects arising from the significant difference between
melting temperature upon heating and re-crystallisation temperature during cooling. Several studies
confirm that the coefficient of friction is inversely proportional to normal pressure and directly
proportional to velocity [10, 11, 14, 15, 17 -19, 21, 24, 25] . Increasing temperature was
unexpectedly found to increase tool-ply friction in two studies [10, 21] an effect explained by the
action of Coulomb type friction between fibres and metal [10], but more generally has been
observed to decrease the friction coefficient due to the associated reduction in matrix viscosity [10,
11, 14, 16-19, 24, 25].

1
Controlled force-displacement tests show a yield stress rather than a peak [Scherer & Freidrich (1991); Morris & Sun
(1994)].

2
 A purely empirical approach to modelling friction was presented by Morris and Sun [6] who
determined a power-law relationship between steady-state friction, slip velocity and temperature.
Murtagh et al. [22] assumed a lubricating layer in a first modelling attempt to relate friction to the
micro-mechanics of the sliding process. Wilks et al. [23] proposed another simple model including
a combination of both Coulomb friction and hydrodynamic lubrication. In this work both the film
thickness and friction coefficient were used as fitting parameters for the model. Their model was
later modified to account for the surface area of the viscous and reinforcement fractions [31]
τ = η γ + φ µ c PN (1)

where η is the viscosity of the matrix, γ is the shear strain rate, μc is the Coulomb friction
coefficient (assumed equal to 0.3), and PN is the normal pressure. φ = A f / A , is the ratio of fibres in
contact with the tool to the overall platen area, A. The shear strain rate is simply the crosshead
displacement rate divided by the thickness of the molten polymer film separating the reinforcement
fibres and the tooling. They used a series of optical micrographs to measure the film thickness from
un-tested composite samples, giving an average value of 0.07mm. Further micrographs were
analysed from tested samples to determine the proportion of fibres in contact with the platen. Their
model compared well with experimental data generated using the pull-through test designed by
Wilks [20].
The clear dependence of friction on nor mal force, displacement rate and matrix viscosity
prompted Chow [28] to propose the use of a Stribeck curve to interpret experimental data, an
approach adopted since by others [11, 13, 14, 16-18, 28, 30] . A Stribeck Curve is a plot of the
coefficient of friction µ as a function of the multiple (Hersey number), H = ηv/P, where η is the
viscosity of the lubricating fluid layer, v is the velocity and where P was originally defined as the
normal force per unit width (for bearings) [32, 33], in which case H is a dimensionless number, but
has also been used since as the normal force [13-15, 34] or the normal pressure [11, 16]. A typical
Stribeck Curve is shown in Fig 1 which shows three distinct regions corresponding to Boundary
Lubrication, Elasto-Hydrodynamic Lubrication and Hydrodynamic or full-film lubrication.

Boundary
lubrication
Coefficient of friction, µ

Elasto-
Hydrodynamic
hydrodynamic
(full-film)
lubrication
lubrication

η v/P

Fig 1. The variation in frictional drag (expressed as the coefficient of friction, µ) with the quantity ηv/P: the
Stribeck curve. (From Hutchings [32])
In this paper P is used to indicate normal pressure as comparisons shown later in this paper,
involve data from experiments employing testing platens of different areas, and obtained over a
range of normal forces. Chow [28] originally proposed a similar model to Eq (1), assuming the
friction behaviour belonged to the Elasto-hydrodynamic region shown in Fig 1. H owever,
investigations on a p lain weave glass/PP fabric [14] indicated that friction behaviour for this
particular material fell into the Full-Film lubrication region of Fig 1. The steady increase in friction
with increasing Hersey number allowed them to propose a model including a linear relationship
between the steady state friction coefficient and the Hersey number. However, to account for
changes in temperature they had to include a linear shifting term:
3
 µ = 6.12 H + 0.27 − S tt (2)
where μ is the friction coefficient, H is the Hersey number (defined with normal force as the
denominator) and Stt is the shift factor based on t he tool temperature. So far all the models
reviewed, with the exception of [6] who used a purely empirical approach, have required an
estimate of the lubricating film thickness based on experimental measurements in order to
determine the shear rate and from there the viscosity of the non-Newtonian fluid film. Recently a
meso-scale model has been proposed based on t he fabric architecture [11, 16, 17, 19 ] and on the
fluid mechanics occurring within the lubricating film. The most interesting feature of this model is
that the film thickness is an output rather than an input of the model. The model is described later in
this paper in and its predictions compared with experimental results in Section 5.
Regarding numerical work, attempts to include Eq (2) into finite element simulations have
been published [13, 14, 30] . A simple friction rig scenario was modelled in [30] and rate
dependence of the friction coefficient was reproduced. The work was developed further in [13, 14]
to include normal force and tool temperature dependence. Simulations of the friction tool were
correct to within less than 8 percent. Thermo-stamping simulations were also conducted and results
analysed in terms of the punch reaction force. Generally the expected trends were observed though
the constitutive model used to represent the fabric was also found to influence results.

2 MATERIAL
A balanced 2x2 twill weave pre-consolidated thermoplastic textile composite, Vetrotex
Twintex®, consisting of commingled E-glass and polypropylene (PP) yarns has been tested. It has
an areal density of 710gm-2, a yarn width of 5mm in both warp and weft directions, a nominal
thickness of 0.5 mm and a fully consolidated fibre volume fraction of 0.35. The glass fibre diameter
is 16 m icrons [35]. The form of the upper surface of the tows affects the lubrication flow during
sliding friction. Thus, tow geometry is one of the inputs in the meso-scale model and can be
parameterized using a polynomial equation. The length dimensions of the longitudinal and
transverse tows together with the coefficients of the quadratic used to describe the surface profile of
the yarns are given in Table 1. The basic structure assumed in the model is shown in Fig 2 together
with a photograph of the pre-consolidated composite in Fig 2.
Table 1. Input values for the meso-scale model required to predict the empirical results.
Parameter Value
Transverse bundle length (mm) 5.0 mm
Transverse bundle approximation y = 8 ⋅ x2
Longitudinal bundle length (mm) 10.0 mm
Longitudinal bundle y = 16 ⋅ x 2
approximation

Direction of slip

20 m m

Transverse Longitudinal
bundle length bundle length

Fig 2. Left: Schematic cross section of a Twintex ply. Right: 2 x 2 Twintex® glass-polypropylene
preconsolidated sheet.

4
3 EXPERIMENTAL METHODS
In this investigation data collected using two generic experimental techniques has been
analysed. These include pull-out [23] and pull-though [18, 20] tests on the one hand and rotational
measurements using a commercial rheometer [18] on the other. Hands-on experience with the
different techniques is useful when discussing their relative merits; the advantage of the rheometer
lies in the higher rate at which test results can be produced when compared to more traditional pull-
out and pull-though methods and also the greater temperature control afforded by the much smaller
environmental chamber. The disadvantage is that the experimental conditions obtainable using the
rheometer (normal pressure, P, and velocity, v) are lower than can be expected in typical forming
situations (P up to ~ 1000KPa and v up to ~ 5000 mm/min [30]) and so any results have to be
extrapolated to the relevant processing conditions. The low velocities accessible using the
rheometer also mean that only steady state friction behaviour can be measured as the peak in a
typical force versus displacement typically disappears at low slip velocities [18]. Measurements
using the pull-through set-up showed that the temperature at the bottom of the environmental
chamber could be up to 30oC lower than at the top of the chamber, even when using the convection
fan. For slow rates this was not too problematic as the electrically heated platens had sufficient time
to heat the specimens to the correct temperature as the composite material moved against the metal.
At higher rates the heating time decreased causing large variations in the higher rate data. Thus, the
time-dependent heating of samples as they are drawn between the platens is found to be a cause of
variability and can be expected to be even more of an issue for test rigs not housed within an
environmental chamber. The test method was found typically to require between 40-60 minutes for
each test making collection of a large amount of data a laborious and costly process and test
repeatability was rather poor.

3.1 Pull-out rig

The first set of experimental data used in this investigation was collected using a pull-out rig
developed by Murtagh et al. [22] in an investigation conducted by Wilks et al. [23]. A schematic of
the set-up is shown in Fig 3. The technique involves pulling a steel shim from between two layers
of composite sheet clamped by a steel bar on three sides. A normal load was applied by two heated
platens, driven by a Dartec 100kN universal testing machine. The area of the two heated platens
was 0.012m2. Once the desired temperature was reached, the shim was pulled out by a lead screw
driven by a variable DC permanent magnet motor. A 2.2kN load cell was placed between the lead
screw and the clamping device for the steel shim to measure the force required to withdraw the
shim. Displacement of the steel shim was measured by a +/- 25mm stroke linear variable
differential transformer mounted horizontally. Testing conditions include temperatures of 180, 200
and 220oC, normal pressures of 80kPa to 2.8MPa and pull-out velocities of 0.5, 0.8 and 1.2 mms-1.
Starting at 0.5 mms-1, all three velocities occurred during a single test. Once the pull-out force
became constant the velocity was rapidly increased to 0.8 mms-1 and again to 1.2 mms-1.
3.2 Pull-through rig
Wilks [20] later designed a pull-through rig, so-called to distinguish it from similar pull-out
designs [11, 16, 21, 23]. A photograph of the rig is shown in Fig 4 together with a schematic of the
top view of the rig. The rig consists of a primary steel frame approximately 300 x 200 mm with two
steel platens, 175 x 25 x 6 mm constituting the top section of the frame. A secondary specimen
frame clamps the perimeter region of the sample and is connected to the load cell at it upper point.
The sample is the same size as the outer perimeter of the specimen frame which is guided in its
motion by grooves milled into the faces of the steel platens. The specimen frame effectively pulls
the clamped sample between the two steel platens which apply a normal pressure to the front and
back surfaces of the sample. The bottom edge of each platen is milled to prevent snagging of the
sample as it is drawn through the platens. The contact area between platen and material is 89 x 63
mm (area = 5607mm2). Two 50 W cartridge heaters heat each platen to the test temperature, which
is regulated by a feedback loop using two K-type thermocouples. The normal pressure on t he
5
platens is provided by four springs. In order to heat material initially outside of the heated platens,
the entire rig is placed in a Hounsfield Environment Chamber (fan-assisted oven) and heated to the
same temperature as the platens. The intention is that the temperature of both the oven and platens
should be identical and testing is as close to isothermal as possible. The specimen frame is
connected to the crosshead of a PC-controlled Hounsfield H25k-S Universal testing Machine, fitted
with a 2.5 kN load cell. The test specimens can only be tested in a 0° or 90° configuration otherwise
the frame is unable to clamp the specimen securely enough to prevent significant distortion of the
textile composite during testing. Each experiment was conducted at least three times.

Fig 3. Top: Photograph of experimental set-up. Bottom: Illustration of steel shim being pulled out from
between two heated composite plies to which a normal load is applied.

Fig 4. Pull-through rig. Left: Photo of pull through rig in oven. Right: Schematic of rig viewed from above.
6
 Typical results from a pull-through experiment performed at a normal pressure of 0.012 MPa
are shown in Fig 5. The temperature during each experiment was kept constant at 180°C. The force
versus displacement curves show peak values followed by steady state values. Thus, both peak and
steady state friction behaviours are evident and both follow the same general trends. These include
increasing friction with increasing rate and decreasing friction for increasing normal force. These
same trends have been reported previously for other types of Twintex [11, 14, 16, 17, 23] . In this
investigation only steady state friction has been considered. The list of experimental conditions
employed for pull-out (Table A1) and pull-through tests (Table A2) are given in Appendix A.

v= 50 / 150 / 500 mm/min, T=180C

090709 50mm/min
350
090707 50mm/min
303.2 90704 150mm/min
300
90705 150mm/min
250 090711 500mm/min
090712 500mm/min
Force (N)

200
164.3 High rate
150 158.4
Medium rate 123.3
100
85
95.75 Low rate
50

0
0 50 100 150 200
Extension (mm)

Fig 5. Typical results from the Pull-though rig tests conducted at 3 rates using a pressure of 0.012 MPa.
Numbers given by the lines indicate the average peak and steady state forces recorded during the tests.
3.3 Rheometer
An alternative method of measuring friction has been employed by adapting a co mmercial
rheometer [15, 19, 27] . Experiments were performed on a Bohlin CVOR200 Rheometer with an
Extended Temperature Cell. All tests were conducted in a nitrogen atmosphere to minimize
polymer degradation. The rheometer was fitted with a custom designed rig that allowed the textile
composite samples to be held firmly in place during testing. The rig consisted of a pair of parallel
stainless steel platens. The lower platen was a truncated cone with a diameter of 25 mm. The upper
platen was a flat disk with diameter 40 mm (Fig 6). Specimens were cut appropriately (see Fig 7)
and placed between the upper platen and a clamping ring (outer diameter of 40mm and inner
diameter of 30mm). Four small screws were used to clamp the ring and specimen in position. The
screws secured the specimen by passing through the ring and into the upper platen. The specimen
was placed in the Extended Temperature Cell and heated. After the specimen reached the required
temperature, the upper platen holding the specimen was positioned in the rheometer parallel with
the lower platen. A normal force was set on the specimen by pressing the upper platen down against
the lower platen. The value of the normal force was recorded by the computer.
Using the rheometer it was possible to generate data more efficiently than when using the pull-
though and pull-out test rigs. This meant a much larger test matrix could be completed in a
reasonable amount of time, presenting the possibility of generating a master curve incorporating
rate, normal pressure and temperature. Formulating the master curve involves shifting the data
produced under different experimental conditions such that the whole body of data can be described
using a single equation. In order to do this, suitable shifting factors must be determined. The input
data in the rheometer are normal force, shear stress and temperature. Experiments were performed
over a range of normal forces (2.5, 10, 20, 50 and 90% of the maximum force that could be applied
7
by the rheometer, i.e. 19.6 N), at various imposed shear stresses (500, 1100, 2000 and 5000Pa) and
for several temperatures (160, 180, 200 and 220°C). Each test was repeated three times and average
results were used for data processing. It was found during experiments that normal force changed
due to lateral squeeze-flow of the polymer matrix through the sample. Thus normal force was also
one of the outputs from the test and was recorded continuously. Other outputs included rotation
angle and time. A typical test result is shown in Fig 8 which shows angular displacement versus
time for a given imposed constant torque (constant shear stress).

Material

Clamping ring
Truncated cone

Fig. 6. Top: Side profile of custom made fixture with loaded sample. Bottom: Photograph of custom
designed platens.

Fig. 7. Example of a test sample following testing. The arms of the specimen are fastened under the clamping
ring.

8
 AT=50%, T=200C, 1100Pa
6.0E-01
Tangent-line
5.0E-01 fitted to data

angle (rad) 4.0E-01

3.0E-01

2.0E-01

1.0E-01
191008N
0.0E+00
0 50 100 150 200 250
time (sec)

Fig 8. Output data from an individual rheometer test. The initial tangent of the experimental curve was used
to determine angular shear rate.

3.4 Shifting Procedure for rheometer data

Clearly the data from the rheometer have to be adjusted for comparison with results from the
pull-through and pull-out tests. The normal force, n, can be converted to normal pressure, P, by
P = n Ao (3)

where Ao is the testing area ( πR 2 ) and R is the radius of the truncated platen (see Section 3.3). The
angular velocity at any radius, r, can be converted to linear velocity (mm/s) using v=ωr where ω is
the angular velocity (calculated from the initial gradient of the line shown in Fig 8) and r is the
radius. The linear velocity varies from zero at r = 0 to a maximum at r = R. The weighted average
linear velocity is used to process the rheometer data for comparison against pull-through tests, i.e.
2 (4)
v= ωR
3
Typical data generated by the rheometer tests at a temperature of 180°C, showing normal
mass (the applied load measured in grams), m, versus linear velocity, v, for different imposed
constant shear stresses are plotted in Fig 8. Similar graphs were also produced for temperatures of
160°C, 200°C and 220°C (not shown here). Trend lines were fitted through the data. Each trend line
was of exponential form as in Eq (5). The average exponent, C2, of all trend lines at different
temperatures and shear stresses was found to be -1.37 with a standard deviation of 0.4. C1 changed
according to the different experimental conditions.
P = C1 ⋅ v C2 (5)

9
 10000

1000

normal mass (gram)

100
500Pa
1100Pa
2000Pa
10
5000Pa

1
0.0001 0.001 0.01 0.1 1 10
rate (mm/sec) 2/3 from centre

Fig. 9. Normal mass versus rate data generated for different shear stresses at 180°C.

1 500Pa
1100Pa
2000Pa
normal pressure MPa

0.1 5000Pa

0.01

0.001

= 4⋅10-6 x-1.37
y = y4.00000E-06x -1.37000E+00

0.0001
0.0001 0.001 0.01 0.1 1 10
rate (mm/sec) 2/3 from centre

Fig. 10. Normal pressure versus rate data with trend lines of the form given in Eq (3) with C2 = -1.37.
Fig 9 shows the data converted to normal pressure versus rate together with trend lines with
C2 = -1.37. The general form of the final master curve is assumed to take the form
P = C3 ⋅ aτ ⋅ aT ⋅ v −1.37 (6)

where aτ is the shift factor for the shear stress and aT is the shift factor for the temperature. It is
possible to shift the data horizontally, vertically or by a combination of the two methods, the choice
here is arbitrary. A vertical shifting was chosen. In order to determine the shift factors a reference
temperature and reference shear stress had to be chosen (180oC and 500Pa). Eq (6) could then be
written as
P = 4 × 10 −6 ⋅ aτ ⋅ aT ⋅ v −1.37 (7)

when aτ and aT both equal 1, Eq (7) gives the trend line of the reference data, the lowest trend line
shown in Fig 10. Thus the factor 4x10-6 = C3ref and includes the conversion from normal mass to
normal pressure, i.e. C3ref is the normal pressure measured in MPa that produces a velocity of

10
1mm/s at 2/3 of the platen radius from the centre of the platen under the reference conditions. The
constant C3 of each trend line can be related to C3 of the reference curve, i.e., C3ref, simply by
determining the ratio between the two, as shown in see Eq (8). Thus aτ is the factor by which the
reference curve must be multiplied in order to shift it to coincide with trend lines fitted to data
produced at other shear stresses at the reference temperature. Evidently the size of aτ is determined
by the relative magnitudes of the shear stresses of the two curves. A relationship of the form shown
in Eq (8) is postulated. The aim is to determine the value of the exponent b in Eq (8). Table 3 shows
the information used to determine b.
b
C  τ 
aτ = 3 =   (8)
C3ref  τ ref 

Table 3. Information used to determine b in Eq (8)
Shear Stress τ C3
(Pa) C3 τref C3ref b
500 4.0·10-6 1 1 -
1100 4.25·10-5 2.2 10.65 3
2000 2.56·10-4 4 64 3
5000 4.0·10-3 10 1000 3

Here τ ref = 500 × 10 −6 MPa, a v alue of b = 3 was determined from the data, thus Eq (6) can be
written as
3
 τ  (9)
P = 4 × 10 −6 ⋅  −6 
⋅ aT ⋅ v −1.37
 500 × 10 
when aT = 1 Eq (9) can be used to determine P at 180oC for shear stresses between 500 and 5000
Pa. A similar equation was determined for the other temperatures though the factor C3 in each case
was different. In order to apply Eq (9) to other temperatures all that remained was to determine aT
where
P (T ) C (T )
aT = = 3 (10)
P (Tref ) C3 (Tref )
The relationship between aT and temperature was assumed to follow an Arrhenius type behaviour,
thus
1 1 
log(aT ) = A −  (11)
T T 
 ref 
The aim here is to determine A. This can be determined by plotting log(aT) versus (1/T-1/Tref).
Arrhenius type behaviour is indicated if the data follow a straight line. The data are plotted in Fig
11.

11
 0.4

0.3
log(aT)
0.2 Linear (log(aT))

log(aT)
0.1

-0.1
y = -282.85x + 0.0137
-0.2

-0.3
-0.0015 -0.001 -0.0005 0 0.0005 0.001
1/T-1/Tref (1/C)

Fig 11. Determination of the gradient of the plotted data gives A in Eq (11)

A trend line fitted to the data gives A = -282.85. Thus, Eq (11) can be written
1 1 
− 282.85 − 
aT = 10  T 180  (12)
and substituted in Eq (9) to produce a general equation including rate, normal pressure and
temperature. However, the equation requires further modification. This is because a Newtonian
assumption is made when the rheometer converts the intended input shear stress to torque for the
parallel plate geometry [36]. This problem has been addressed for non-Newtonian fluids [37]
resulting in Eq (13), which can be used to correct the friction data
M  d ln M 
τ= 3 + d ln v 
(13)
2πR 3
where M is the applied torque. For a Newtonian fluid d ln M d ln v = 1 . For non-Newtonian fluid
the term is less than 1. Using the rheometer data a value of approximately 0.39 w as found. This
results in a small modification to the Newtonian master curve, Eq (9) which can be rearranged as

v 1.37
µ = 565 × 10 −6 3 (14)
4 × 10 −6 ⋅ P 2 ⋅ aT
where aT is given by Eq (12) and µ it the coefficient of friction in N/N, v is the velocity in mm/s,
P is the normal pressure in MPa and T is the temperature in ºC. Eq (14) will be referred to as the
Master Curve in subsequent sections and is a purely empirical model. To permit comparison with
experimental measurements and meso-scale model predictions, the master curve, Eq (14), has been
used to generate results at specific experimental conditions, as noted in Table A3 (see Appendix A).
Notably the experimental conditions chosen were within the working range of the rheometer.

4 STRIBECK ANALYSIS
The Stribeck curve has been used previously in several investigations to analyse the tool-ply
friction of textiles composites [11, 13, 14, 16-18, 28, 30]. Values of 0.11 mm [17] and 0.07 mm [14,
25] have been employed in order to determine the shear rate in the fluid layer and calculate the
matrix viscosity before plotting the data as a function of the Hersey number. A similar procedure is
adopted here (see Fig 12) using a film thickness of 0.11mm. The experimental conditions of the
data points are listed in Tables A1-3.

12
 1

Friction Coefficient
0.1

Pull-out 0.01
Pull-through
Rheometer

0.001
0.00000001 0.0000001 0.000001 0.00001 0.0001 0.001
Hersey Number (m)

Fig 12. Coefficient of friction versus modified Hersey Number (m) using data from the literature. Pressure
has been used to determine the Hersey number rather than normal force per meter or normal force.
The polypropylene matrix used in Twintex has been characterised previously and fitted with a
Carreau-Yasuda model [38]. Using this information the viscosity of the fluid layer can be estimated.
It should be noted that the rheological data measured in this characterisation were reliable only for
relatively low shear rates (<10s-1) which, when using a gap of 0.11mm, corresponds to a draw
velocity of just 0.011mms-1. Above this the accuracy of the fitted Carreau-Yasuda model
predictions become less certain but should still provide a reasonable estimate. As most of the
investigations employed faster draw speeds, error in calculating the viscosity and hence the Hersey
number is inevitable when plotting the data of Fig 12.
A clear trend in the data can be seen and shows that even for the very low Hersey numbers
explored by Wilks [23] (achieved using low draw speeds combined with high normal pressures –
see Table A1) the friction coefficient increases with Hersey number. Comparison with Fig 1 may
therefore suggest full-film lubrication, a result which corresponds to that found by Gorczyca-Cole et
al. [14] when conducting tests on Twintex fabric. However, data from the different tests fail to fall
onto a single curve, indicating that the friction behaviour may be more complicated than can be
described using a Stribeck analysis. A probable source of error in performing such an analysis is the
use of a constant film thickness for all experimental conditions. Predictions from the meso-scale
model described in Section 6 s uggest the minimum film thickness can vary by several orders of
magnitude according to the experimental conditions. Thus, one option in plotting the data is to use
the film thickness predicted by the meso-scale model rather than use a constant film thickness.
However, initial attempts by Ubbink [17] to do t his failed to improve the data when interpreted
using a Stribeck approach. Use of an alternative definition of the Hersey number used in this paper
(compared to its original definition) may also be a source of error.

5 MESO-SCALE MODELLING
A meso-scale model [11, 17] based on a geometrical description of the tows within the fabric
has been developed. One of the advantages of the model is that the film thickness can be predicted
from the normal pressure and velocity. This avoids the use of the approximation of the film
thickness required in the analysis of Section 5. Fig 13 presents a schematic cross section of the
composite material in the warp direction. Hydrodyamic lubrication is assumed between the bundles
and the tool surface. The total friction force per unit width follows by integrating the surface shear
stresses over the length of the cross section, disregarding the bundle curvatures out of the plane for
the time being. The contributions of the longitudinal warp and transverse weft yarns can be

13
analysed separately and added to calculate the total friction force and hence the friction coefficient.
The Reynolds’ equation describes the relation between the pressure and thickness distributions in
thin film lubrication. The simple one dimensional steady state situation is given by
Y

Fig. 13. Schematic pressure distribution underneath a bundle.

∂  h 3 ∂P  ∂h (15)
  = 6v
∂x  η ∂x  ∂x
where v is the velocity, h is the film thickness in the Y direction and x is distance in the X direction
(see Fig 13) and η is the viscosity predicted using the Carreau-Yasuda model fitted to the actual
polypropylene viscosity data reported in [38]. Thus, the model is appropriate for predictions of
steady state friction rather than the peak friction observed at the start of the test. The pressure
distribution can be solved for a given film thickness distribution using the following boundary
conditions (see Fig 13).
∂p
P (− L ) = 0; P ( xo ) = 0; ( xo ) = 0 (16)
∂x
where the pressure is assumed to be non-negative. The bearing force per unit width is given by
xo

FB = ∫ P(x )dx
−L
(17)

whereas the friction force per unit width follows as

xo xo
h ∂P v
F f = ∫ τ (x )dx = ∫ 2 ∂x + η h dx (18)
−L −L

The one dimensional meso-scale model predicts the bearing and friction forces, FB and Ff,
with the temperature T, velocity v and minimum film thickness hmin as input parameters. The model
is used inversely, iteratively adapting hmin such that the integrated bearing force over all fibres
equals the prescribed normal load N. This procedure also leads to the pull-out force, which can be
converted to friction coefficient and compared to the experimental results. In order to compare the
meso-scale model with the master curve given by Eq (14) the tow geometry within the fabric
described in Section 2 must be modelled. The tow shape is characterised using a polynomial
function, see Table 1 and Fig 2.

6 COMPARISON OF EXPERIMENTAL DATA WITH EMPIRICAL AND MESO-

SCALE MODELS
The meso-scale model predicts a different minimum film thickness, hmin, for each
experimental condition. This film thickness and the friction coefficient predicted by the meso-scale
model are presented in Table A3. It was noted that for several experimental conditions the film
thickness prediction was less than the yarn fibre diameter (16 microns) probably rendering the
meso-scale model predictions invalid according to the underlying assumptions of the theory.

14
Nevertheless, friction predictions from the Master curve and meso-scale model are compared in Fig
14. A quantitative comparison can be drawn between the two using Eq (19).

0.35

0.3 Master Curve: Eq(14)

Friction coefficient

0.25 Meso-model

0.2

0.15

0.1

0.05

0
2

41

41

41
2

41
2

2
.0

.0

.0

.0

.0
.0

.0

.0

.0
/0

/0

/0

/0

/0
/0

/0

/0

/0
7

17

7
01

17

17

17

7
.1

.1

.1
.0

.1
/0

/0

/0
.0

.0

.0

.0
/0

/0
/0

/0

/0

/0
00

80

80
00

80
00

80

00

20
/2

/1

/1
/2

/1
/2

/1

/2

/2
3

8
2

9
1

Fig 14. Comparison of Eq (14), indicated as Master Curve in the legend, with the meso-scale model. The
experimental condition is given in the format of Number / Temperature (oC) / Velocity (mm/s) / Pressure
(MPa) to 2 significant figures.

n
100(µ MC − µ MSM ) 
relative error = ∑ abs   m (19)
i =1  µ MC 

where µ MC is the master curve prediction for the friction coefficient and µ MSM is the friction
coefficient predicted using the meso-scale model, i is the number of the data value in a given set
and m is the total number of data in the set. The meso-model correctly predicts the main trends in
the data, though underestimates the friction by a factor of approximately two (for conditions where
the film prediction is valid).
A comparison between the master curve and the experimental data produced using both pull-
out and pull-through tests has also been attempted (see Tables A1 and A2). However, it was found
that as it stands the master curve Eq (14) was unable to provide a close comparison. The reason for
this is believed to be that normal forces and velocities used in collecting the rheometer data are very
much lower than those used in conducting pull-though and pull-out tests which tend to be closer to
actual forming conditions [30]. Thus, any error involved in the shifting process used to determine
Eq (14) is magnified in extrapolating predictions to experimental conditions beyond the rheometer’s
working range. Thus, a modified version of Eq (14) is proposed:
µ MMC = K ⋅ µ MC + G (20)

where µ MC is given by Eq (14) and two extra fitting parameters K and G have been introduced.
They are easily determined by plotting the experimentally determined friction data against Eq (14)
and fitting a linear trend-line using least squares (see Fig 15). The values of K and G for the two
data sets given in Tables A1 and A2 are given in Fig 15.

15
 0.5
y = 0.35x + 0.1214

Experimentally Measured Friction

R2 = 0.9572
0.4
Rheometer
Pull-out
Coefficient

0.3 Pull-through
0.02wilks fitted
0.015Linear (Rheometer)
0.2
0.01Linear (Pull-through)
y = 0.3715x - 0.002 0.005Linear (Pull-out)

0.1 R2 = 0.9321 0
0 0.02 0.04 0.06

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1
Master Curve Prediction: Eq(14)

Fig 15. Friction coefficient versus Eq (14), pull-out and pull-through tests. Inset shows an expanded view of
the pull-out data. Trendlines fitted to the data give the values of K and G in Eq (20). K = 0.3715, G = -0.002
for the pull-out data, K = 0.35, G = 0.1214 for the pull through data.
Finally, Fig 16 and Fig 17 s how a comparison between the pull-out and pull-through
experimental data, the Modified Master Curve Eq (20) predictions and the meso-scale model
predictions.

16
 0.035

0.03
Pull-out data
Modified Master Curve
0.025

Friction coefficient
Meso-Model
0.02

0.015

0.01

0.005

0
5

.9

.5

.0

.3

.7
.4

.8

.9

.6
/1

/0

/2

/1

/2
/0

/0

/0

/0
.5

.8

.8

.2

.2
.5

.5

.8

.2
/0

/0

/0

/1

/1
/0

/0

/0

/1
80

80

80

80

80
80

80

80

80
/1

/1

/1

/1

/1
/1

/1

/1

/1
3

9
1

7
0.03 Pull-out data
Modified Master Curve
0.025 Meso-Model
Friction coefficient

0.02

0.015

0.01

0.005

0
8

7
.9

.6

.1
/1

.3

.8
.4

.9

.6
/1

/0

/2

/1

/2
/0

/0

.8

/0
.5

.8

/0

.8

.2

.2
.5

.5

.2
/0

/0

/0

/1

/1
/0

/0

00

/1
00

00

00

00

00
/2
00

00

00
/2

/2

/2

/2

/2
/2

/2

14

/2
12

13

15

17

18
10

11

16

0.02 Pull-out data

Modified Master Curve
0.018
Meso-Model
0.016
Friction coefficient

0.014
0.012
0.01
0.008
0.006
0.004
0.002
0
7

8
.8

.1

.3

.9
/1
.4

.9

.5

.6
/1

/2

/1

/2
/0

/0

/0

/0
.8
.5

/0

.8

.2

.2
.5

.5

.8

.2
/0

/0

/1

/1
/0

/0

/0

/1
20
20

20

20

20
/2
20

20

20

20
/2

/2

/2

/2
/2

/2

/2

/2
23
21

24

26

27
19

20

22

25

Fig 16. Comparison between pull-out data, the Modified Master Curve empirical model and Meso-Model
Predictions. The experimental condition is given in the format of Number / Temperature (oC) / Velocity
(mm/s) / Pressure (MPa) to 2 significant figures.

17
 0.6
Pull-though data
0.5 Modified Master Curve
Meso-Model

Friction coefficient
0.4

0.3

0.2

0.1

36

36

36

12

12

12

24
.0

.0

.0

.0

.0

.0

.0
/0

/0

/0

/0

/0

/0

/0
3

.5

.3

.5

.5
.8

.1

.8
/2

/8

/2

/2
/0

/0

/0
80

80

80

80
80

80

80
/1

/1

/1

/1
/1

/1

/1
2

7
1

5
Figure 17. Comparison between pull-though data, Modified Master Curve empirical model and Meso-Model
Predictions. The experimental condition is given in the format of Number / Temperature (oC) / Velocity
(mm/s) / Pressure (MPa) to 2 significant figures.
The average relative error expressed as a percentage for each comparison has been calculated
using Eq (19).

7 CONCLUSIONS

As a final comment, Eq (15) is written in a form that is convenient for implementation in a

finite element code. Implementation of this model will allow sensitivity studies to be conducted in
order to assess the influence of pressure, velocity and temperature and hence the friction coefficient
on the shear and wrinkling behaviour of the thermoplastic textile composite sheet during forming.

18
APPENDIX A
Table A1. Experimental conditions, pull-out friction measurement, Modified Master Curve friction
prediction, minimum film thickness, hmin, and Meso-Scale Model friction prediction. The values marked with
a * indicate that the film thickness predicted by the meso-model is less than the yarn fibre diameter,
invalidating the meso-model friction prediction. Valid results are written in bold type.
Condition T v P µ PO µ MMC hmin µ MSM
number (oC) (mm/s) (MPa) Pull-out Modified (mm) Meso-
experiment Master Model
Curve
1 180 0.5 0.45 0.01200 0.01441 0.00705* 0.03179*
2 180 0.5 0.875 0.00656 0.00853 0.00227* 0.02095*
3 180 0.5 1.863 0.00375 0.00436 0.00050* 0.01243*
4 180 0.8 0.5 0.01643 0.01696 0.00845* 0.03468*
5 180 0.8 0.988 0.00895 0.01004 0.00267* 0.02260*
6 180 0.8 2.038 0.00573 0.00543 0.00063* 0.01373*
7 180 1.2 0.638 0.01831 0.01739 0.00767* 0.03371*
8 180 1.2 1.288 0.00998 0.01014 0.00223* 0.02150*
9 180 1.2 2.675 0.00632 0.00546 0.00050* 0.01287*
10 200 0.5 0.481 0.01032 0.01191 0.00311* 0.02342*
11 200 0.5 0.93 0.00572 0.00696 0.00087* 0.01506*
12 200 0.5 1.886 0.00337 0.00360 0.00019* 0.00896*
13 200 0.8 0.595 0.01426 0.01296 0.00310* 0.02383*
14 200 0.8 1 0.00799 0.00859 0.00113* 0.01681*
15 200 0.8 2.063 0.00526 0.00453 0.00024* 0.00991*
16 200 1.2 0.658 0.01683 0.01483 0.00353* 0.02527*
17 200 1.2 1.291 0.00862 0.00874 0.00095* 0.01601*
18 200 1.2 2.785 0.00564 0.00444 0.00018* 0.00905*
19 220 0.5 0.468 0.00985 0.01083 0.00158* 0.01845*
20 220 0.5 0.924 0.00500 0.00616 0.00039* 0.01142*
21 220 0.5 1.835 0.00288 0.00316 0.00008* 0.00675*
22 220 0.8 0.538 0.01306 0.01250 0.00179* 0.01966*
23 220 0.8 1 0.00717 0.00759 0.00050* 0.01271*
24 220 0.8 2.063 0.00468 0.00392 0.00010* 0.00731*
25 220 1.2 0.684 0.01500 0.01287 0.00156* 0.01898*
26 220 1.2 1.278 0.00814 0.00780 0.00042* 0.01214*
27 220 1.2 2.873 0.00479 0.00371 0.00007* 0.00647*

Table A2. Experimental conditions, pull-through friction measurement, Modified Master Curve friction
prediction, minimum film thickness, hmin, and Meso-Scale Model friction prediction. The values marked with
a * indicate that the film thickness predicted by the meso-model is less than the yarn fibre diameter,
invalidating the meso-model friction prediction. Valid results are written in bold type.
Condition T v P µ pt µ MMC hmin µMSM
number (oC) (mm/s) (MPa) Pull- Modified (mm) Meso-
Through Master Model
Experiment Curve
1 180 0.833 0.036 0.21055 0.22653 0.11577 0.16748
2 180 2.5 0.036 0.30232 0.29503 0.18305 0.26435
3 180 8.33 0.036 0.39237 0.42229 0.26218 0.40319
4 180 0.167 0.012 0.21922 0.22626 0.09334 0.14077
5 180 0.83 0.012 0.37156 0.33968 0.21461 0.32374
6 180 2.5 0.012 0.48303 0.48256 0.33440 0.56271
7 180 2.5 0.024 0.36227 0.34892 0.23203 0.34955

19
Table A3. Experimental conditions, Master Curve friction prediction, minimum film thickness, hmin, and
Meso-Scale Model friction prediction. The values marked with a * indicate that the film thickness predicted
by the meso-model is less than the fibre diameter, invalidating the meso-model friction prediction. Valid
results are written in bold type.
Condition T v P µMC hmin µMSM
number (oC) (mm/s) (MPa) Master Curve (mm) Meso-Model
1 200 0.00167 0.02037 0.0228 0.00087* 0.01161*
2 200 0.01667 0.02037 0.0652 0.00792* 0.03255*
3 200 0.16667 0.02037 0.1866 0.05007 0.08749
4 180 0.01667 0.04074 0.0463 0.00066* 0.01021*
5 200 0.01667 0.04074 0.0411 0.00043* 0.00827*
6 220 0.01667 0.04074 0.0372 0.00028* 0.00675*
7 180 0.16667 0.01019 0.3342 0.10284 0.15320
8 180 0.16667 0.02037 0.2105 0.06699 0.10782
9 180 0.16667 0.04074 0.1326 0.04115 0.07684

20
REFERENCES
1. Kalpakjian, S.S., S., Manufacturing Engineering and T echnology. 2006, Pearson Prentice
Hall.
2. Chenot, J.L., State of the art and recent developments in the numerical modeling of metal
forming processes, in 9th International ESAFORM Conference on Materials Forming.
2005: Cluj-Napoca, Romania. p. 71-76.
3. Lin, H., Long, A.C., Clifford, M.J., Wang, J. & Harrison, P., Predictive Modelling of FE
Forming to Determine Optimum Processing Conditions, in 10th International ESAFORM
Conference on Materials Forming. 2007: Zaragosa, Spain. p. 1092-1097.
4. Lin, H., Wang, J., Long, A.C., Clifford, M.J. & Harrison, P., Predictive Modelling for
Optimisation of Textile Composite Forming. Composites Science and Technology, 2007. 67:
p. 3242-3252.
5. Dong, L., Lekakou, C. & Bader, M. G., Solid-mechanics finite element simulations of the
draping of fabrics: a sensitivity analysis. Composites Part A: Applied Science and
Manufacturing, 2000. 31(7): p. 639-652.
6. Morris, S.R., & Sun, C.T., An investigation of interply slip behaviour in AS4/PEEK at
forming temperatures. Composites Manufacturing, 1994. 5(4): p. 217-224.
7. Phung, T., Paton, R., & Mouritz, P.A., Characterisation of interply shearing resistance of
carbon-epoxy unidirectional tape and f abric prepregs, in 6th International ESAFORM
Conference on Materials Forming. 2003: Salerno, Italy. p. 867-870.
8. Scherer, R., & Friedrich, K., Inter- and intraply-slip flow processes during thermoforming of
CF/PP-laminates. Composites Manufacturing, 1991. 2(2): p. 92-96.
9. Martin, C.J., Seferis, J.C., a nd Wilhelm M.A., Frictional resistance of thermoset prepregs
and its influence on honeycomb composite processing. Comp. Part A, 1996. 27(10): p. 943-
951.
10. Lebrun, G., Bureau, M.N. & Denault, J., Thermoforming-stamping of continuous glass fibre /
PP composites: Interlaminar and tool-laminate shear properties. Journal of Thermoplastic
Composite Materials, 2004. 17: p. 137-165.
11. Akkerman, R., Ubbink, M.P., de Rooij, M.B., & ten Thije, R.H.W. Tool-ply friction in
composite forming. in 10th International ESAFORM Conference on M aterials Forming.
2007. Zaragosa, Spain.
12. Gorczyca, J.L., Sherwood, J.A. & Chen, J., Friction between the tool and the fabric during
the thermostamping of woven co-mingled glass-polypropylene composite fabrics, in Am Soc
Comp 18th Tech. Conf. 2003: Gainesville, FL.
13. Gorczyca, J.L., Sherwood, J.A. & Chen, J., A friction model for use with a commingled glass-
polypropylene plane weave fabric and t he metal tool during thermostamping. Revue
Europeeanne des Elements Finis, 2005: p. 729-751.
14. Gorczyca-Cole, J.L., Sherwood, J.A. & Chen, J., A friction model for thermostamping
commingled glass-polypropylene woven fabrics. Composites Part A, 2007. 38: p. 393-406.
15. Harrison, P., Lin, H., Ubbink, M., Akkerman, R., van de Haar, K. & Long, A.C.,
Characterising and M odelling Tool-Ply Friction of Viscous Textile Composites, in 16th
International Conference on Composite Materials. 2007: Kyoto, Japan.
16. ten Thije, R.H.W.A., R., van der Meer, L. & Ubbink, M.P., Tool-ply friction in thermoplastic
composite forming, in 11th International ESAFORM Conference on M aterials Forming.
2008: Lyon, France.
17. Ubbink, M., Tool ply friction of woven fabric composites, in Mechanical Engineering. 2006,
University of Twente.
18. Van de Haar, K., Modelling resistance at the ply/tool contact interface for Twintex®. 2005,
University of Nottingham.
19. Lin, H., Harrison, P., Van de Haar, K., Wang, J., Long, A.C., Akkerman, R. & Clifford, M.J.
Investigation of Tool-Ply Friction of Textile Composites. in 8th International Conference on
Textile Composites (TEXCOMP). 2006. Nottingham, UK.
21
20. Wilks, C.E., Processing technologies for woven glass polypropylene composites. 1999,
University of Nottingham.
21. Murtagh, A.M., Lennon, J.J. & Mallon, P.J., Surface friction effects related to pressforming of
continuous fibre thermoplastic composites. Composites Manufacturing, 1995. 6: p. 169-175.
22. Murtagh, A.M., Monaghan, M.R. & Mallon P.J., Investigation of the Interply Slip Process in
Continuous Fibre Thermoplastic Composites, in 9th ICCM. 1993: University of Zaragoza,
Spain. p. 311-318.
23. Wilks, C.E., Rudd, C.D., Long, A.C. & Johnson, C.F., Tool/Ply Friction and Consolidation
during Processing of Glass/Thermoplastic Composites, in 6th International Conference on
Automated Composites. 1999: Bristol, UK
24. Gorczyca-Cole, J.L., Sherwood, J.A., Liu, L. & Chen, J., Modeling of Friction and Shear in
Thermostamping of Composites - Part I. Journal of Composite Materials, 2004. 38(21): p.
1911-1929.
25. Gorczyca-Cole, J.L., Sherwood, J.A. & Chen, J., A friction model for use with a commingled
glass-polypropylene plane weave fabric and t he metal tool during thermostamping. Revue
Europeeanne des Elements Finis, 2005: p. 729-751.
26. Lussier, D.S., Chow, S. & Chen, J., Shear and F riction Response of Co-Mingled
Glass/Polypropylene Fabrics During Stamping, in American Society for Composites 16th
Technical Conference. 2001: Blacksburg, VA.
27. Heyer, P., & Lauger, J. A flexible platform for tribological measurements on a r heometer. in
15th International Congress on Rheology/80th Annual Meeting of the Society-of-Rheology.
2008. Monterey, CA: Amer Inst Physics.
28. Chow, S., Frictional interaction between blank holder and f abric in stamping of woven
thermoplastic composites, in Department of Mechanical Engineering. 2002, U niversity of
Massachusetts: Lowell.
29. Gorczyca-Cole, J.L., Sherwood, J.A. & Chen, J., Friction between the tool and t he fabric
during the thermostamping of woven co-mingled glass-polypropylene composite fabrics, in
Am Soc Comp 18th Tech. Conf. 2003: Gainesville, FL.
30. Gorczyca, J.L., Sherwood, J.A., Liu, L. & Chen, J., Modeling of Friction and She ar in
Thermostamping of Composites - Part I. Journal of Composite Materials, 2004. 38(21): p.
1911-1929.
31. Clifford, M.J., Long, A.C. & deLuca, P., Forming of engineering prepregs and r einforced
thermoplastics, in TMS annual meeting and e xhibition, second global symposium on
innovations in materials, processing and m anufacturing: sheet materials: composite
processing. 2001: New Orleans, LA.
32. Hutchings, I.M., Tribology: Friction and Wear of Engineering Materials. 1992, London:
Edward Arnold. 273.
33. Kumar, P.C., Menezes, P.L. & Kailas, S.V., Role of Surface Texture on F riction under
Boundary Lubricated Conditions. Tribology Online, 2008. 3(1): p. 12-18.
34. Lee, J., et al., Enhancement of lubrication properties of nano-oil by controlling the amount of
fullerene nanoparticle additives. Tribology Letters, 2007. 28(2): p. 203-208.
35. Harrison, P., Clifford, M.J., Long, A.C. and Rudd, C.D., Constitutive modelling of
impregnated continuous fibre reinforced composites: Micromechanical approach. Plastics,
Rubber and Composites, 2002. 31(2): p. 76-85.
36. Bohlin Instruments, L., User Manual for Bohlin Rheometers. 2001. p. 183.
37. Darby, R., Viscoelastic Fluids; An Introduction to Their Properties and Behaviour. in
Chemical Processing and Engineering 1976, New York and Basel: Marcel Dekker.
38. Harrison, P., Clifford, M.J., Long, A.C. & Rudd, C.D., A constituent-based predictive
approach to modelling the rheology of viscous textile composites. Composites: Part A, 2004.
35: p. 915-931.

22