Composites Part A 29A (1998) 783–793

1359-835X/98/$-see front matter

q 1998 Elsevier Science Ltd. All rights reserved.

PII: S1359-835X(98)00056-6

Macro- and micro-impregnation phenomena in

continuous manufacturing of fabric reinforced
thermoplastic composites

C. Mayer a ,*, X. Wang b and M. Neitzel a

a
Institut für Verbundwerkstoffe GmbH, Erwin-Schrödinger-Str.,
67663 Kaiserslautern, Germany
b
Department of Mechanical and Mechatronics Engineering, The University of Sydney,
Sydney, NSW 2006, Australia

Macro- and micro-impregnation of molten thermoplastic polyamide 6.6 into glass and carbon fabrics during
continuous manufacturing using a double belt press was investigated. Following the analytical approaches on
transverse permeability of fiber bundles, a comprehensive experimental set-up has been derived to verify the
effects of temperature boundary conditions, resident time, consolidation pressure and fiber bundle permeability on
the degree of impregnation. Samples with reinforcements of glass and carbon fabrics were produced at distinct
processing conditions and the degree of impregnation was investigated by microscopic observation and mirrored
by measuring the flexural stiffness in a three-point bending test. Hence, contribution of particular processing
properties on impregnation was discussed in detail and the results are to be applied to guide the manufacturing
process concerning selection of appropriate raw materials and optimized processing conditions. It could be
concluded, that the continuous manufacturing process using a double belt press might basically be simulated by
applying the current analytical approaches on the impregnation of fibers with thermoplastic polymers, but exhibits
its own characteristics with respect to intra-bundle mechanism such as microscopic adhesion and void entrapment.
q 1998 Elsevier Science Ltd. All rights reserved

(Keywords: macro- and micro-impregnation; continuous manufacturing; double belt press; reinforced fabrics)

INTRODUCTION behind the belts is consistently imposed on the laminate. At

the same time, appropriate temperature boundary conditions
Fabric reinforced thermoplastic composites belong to a new are applied in heating and cooling sections along the process
class of tailored composite intermediates consisting of direction to ensure (i) melting of the polymer without
unidirectional, multiaxial or random reinforcements of thermal degradation, (ii) impregnation of the reinforcing
glass, carbon and aramide fibers and engineering thermo- structure and (iii) consolidation and downstream cooling of
plastic polymers to meet a variety of property and the laminate2. In a second step, forming of the sheet-form
performance requirements in demanding environments. intermediates into shell-type structures or complex parts is
Compared with commonly available glass mat reinforced accomplished by automated high speed stamp-forming
thermoplastics (GMT), they exhibit superior mechanical equipment3.
properties with respect to stiffness and strength. An
economical way of manufacturing such high performance
materials is the utilization of a continuously running double THEORY
belt press (Figure 1) to achieve high speed impregnation
and consolidation of semi-finished sheets1. Most commonly In the continuous manufacturing of fabric reinforced
employed techniques to manufacture sheet-form thermo- thermoplastics, the impregnation of the porous reinforcing
plastic composite intermediates are melt-impregnation, structure by infiltration of a viscous polymer liquid is
powder-impregnation and film-stacking. In the present developed by transverse flow through fiber bundles along
report, film-stacking was used, where reinforcing layers the direction of thickness rather than in-plane. This process
and thermoplastic films are alternately combined to a stack is not only effected by pressure but also profoundly
which is then pulled into the press by the belts. Inside the dominated by the temperature and velocity controlled
press, sufficient pressure produced by a hydraulic system conductive unsteady-state heat transfer into the material
which affects (i) the onset of melting, (ii) the rheological
* Corresponding author. behavior at melt processing temperatures, (iii) the time to

783
Macro- and micro-impregnation: C. Mayer et al.

Figure 1 Double belt press at IVW; EZ: entrance zone; HTZ: high temperature zone; HZ: heating zone; CZ: cooling zone (distance between conveyor
centers: 1630 mm)

drive impregnation and (iv) the cooling induced morphol- spaced cylinders. Theoretical values have been tabulated for
ogy of the polymer matrix. both square and hexagonal arrays by Sangani and Acri-
During the last years several authors have published on vos6,7. Sangani and Yao8 presented a numerical technique
transverse flow of fluids through aligned or random arrays of to predict the permeability for longitudinal and transverse
fibers. The modeling is usually derived from Darcy’s law in flow through random arrays and exhibited, that the random
the basic form of fiber distribution may increase the permeability by a factor
K·DP of two. Bruschke9 showed that due to inherent assumptions
v< ¹ (1) of the Carman–Kozeny model, eqn (4) should not be used to
h·h
predict flow across fiber bundles and therefore provided an
and asymptotic model. In this approach, flow through an array at
h2 low fiber volume fraction can be determined by the cell model,
t< (2) whereas at high volume fractions the lubrication approxi-
h ¹ 1 K·DP mation can be used. From numerical calculations of creeping
with flow around cylinders he showed that, contrary to eqn (4), k zz is
Zt not a constant. At v f . 0.6 a large difference appears between
1 the capillary model and the numerical model.
h¹1 ¼ dt (3)
0 h Due to the reduced order of the Darcy equation several
where authors developed an extended version of the above or a
novel comprehensive approach to allow for the proper
v
specifications of boundary conditions arising in the real
is the apparent velocity, h the viscosity of a Newtonian fluid, manufacturing process of thermoplastic composites. Based
DP/h the pressure gradient in the direction of flow over a on the work of Gutowski et al.10, Lam and Kardos11,
dimension of h either thickness of sheets or radius of bun- Gebart12 and Cai and Berdichevsky13, additional effects
dles and t the time to accomplish impregnation. The perme- such as shear thinning using non-Newtonian fluids, capillary
ability K of a porous medium for a Newtonian fluid may be forces as a function of surface tension and pressure
obtained by the Carman–Kozeny equation as controlled permeability in the impregnation of fibers with
thermoplastic polymers can be ascribed using the analytical
R2 (1 ¹ vf )3
K¼ : (4) approaches of Lee and co-workers14–16, van West et al.17
4kzz v2f
and Bafna and Baird18. All models can adequately describe
Here R is the fiber radius, v f the fiber volume fraction and k zz transverse flow into ideal fiber bundles based on perfectly
the Kozeny constant in the transverse direction of the fiber aligned and spaced arrays of cylinders. However, their
bundle. The Carman–Kozeny equation is based on the application to fabrics is limited due to the non-uniformity of
capillary model of a granular bed as a bundle of tortuous woven structures and the inability to model the actual pore
channels and seems to hold reasonably accurate for isotropic structure which is significantly affected by the consolidation
porous media. However, several authors reported less satis- pressure. Consequently, Lekakou et al.19 and Simacek and
factory results for transverse flow through anisotropic por- Advani20 have published lately on a viscoelastic and
ous media (e.g. aligned fiber beds)4,5. This was often permeability model, respectively, to discuss the key issues
compensated by adjusting the Carman–Kozeny constant of non-linear elastic compression and in-plane permeability
to fit the data. Other workers predicted the permeability in impregnation of two-dimensional orthotropic woven
by idealizing an aligned fiber bed as an array of regularly structures.

784
 Macro- and micro-impregnation: C. Mayer et al.

Introducing all relevant phenomena in transverse impreg- were combined alternately to a stack with an ultimate fiber
nation of fabrics with thermoplastic polymer, modeling can content of 40 and 50 vol.%, respectively, and corresponding
get fairly complex. Additionally, the specific nature of thickness ranging from 1 to 1.4 mm.
transverse flow may not be engendered by a theoretical
approach. Whereas longitudinal flow can propagate even in
Macro- and micro-impregnation
tightly compressed fibers, transverse flow decreases drama-
tically and is ultimately stopped once the fiber volume The type of reinforcement considered in this paper is a
fraction approaches the maximum packaging density of woven structure in the form of an orthotropic, bidirectional
square (v f ¼ 0.79) and hexagonal (v f ¼ 0.91) arrays of fabric. The architecture of this porous medium can be
fibers. Furthermore, single fibers can block the flow path divided into the reinforcing network of the woven structure
while the average fiber volume fraction may be less than the and the sub-domain of fiber yarns either warp or weft which
maximum packaging efficiency. As a consequence, the together create the network. Hence, the fabric exhibits a
present investigation provides a comprehensive description dual scale pore structure of macroscopic pores among yarns
of an experimental study on the critical issues of concern for and micro-pores formed by the fine interstices of the fibers
the continuous manufacturing of fabric reinforced thermo- inside the yarn which leads to the phenomena of macro-
plastic composites using a double belt press. By estimating scopic and microscopic impregnation in the continuous
the influence of (i) polymer rheology, (ii) impregnation manufacturing. The packing density of the macroscopic
time, (iii) permeability as a function of fiber diameter and yarns normally represent the fiber volume fraction of the
pressure affected packaging density, and (iv) surface effects bulk composite as it is here between 0.4 and 0.5, whereas the
on the impregnation behavior of woven structures with non- fiber packing density of the sub-domain can approach the
Newtonian fluids, the results are to be applied to guide the level of 0.91 in hexagonal arrangements. If one assumes an
optimization of the manufacturing process with respect to ideal case, the transverse flow into the macroscopic
material selection and preselection of processing conditions. structure of the fabric can be described as Hagen–Poiseuille
flow into a channel of varying width S determined by the
diameter D of the fiber bundles as depicted in Figure 2.
EXPERIMENTAL In order to ascribe for the deformation of the ideally
circular yarns by applied pressure and the more elliptical to
rectangular cross section of weaving yarns with an aspect
Material selection
ratio up to 10, Figure 2(a) can be extended to a more
A low (polymer #1) and a medium viscosity grade realistic case. As sketched in Figure 2(b) and (c) it could be
(polymer #2) of polyamide 6.6 film (Zytel t, DuPont) (Table found that during the impregnation process there are two
1) and E-glass as well as carbon fabrics with different fiber main stages. From Carman–Kozeny the permeability is
diameters, titer of the fiber bundles and weight to area ratio proportional to the fiber diameter and increases with
(Table 2) were selected to evaluate the effect of processing reduced fiber volume fraction. Obviously, the diameter of
conditions and material configuration on the degree of the fiber bundle D is about 1–2 magnitudes larger than the
impregnation by microscopic observation and mechanical diameter of individual fibers d. The fiber volume fraction of
testing. Fabrics were supplied with commercially available the network and the sub-domains can be different by a factor
surface treatments by Verseidag Indutex, Germany. On the up to 2. It is therefore convenient to assume a significant
basis of individual pressure and temperature boundary difference in time scales associated with macro- and micro-
conditions as described in the next sections, several kinds of impregnation which can be in the order of several
fabric reinforced thermoplastics were manufactured at magnitudes following eqns (2) and (4). Due to the large
process velocities ranging from V 0 to 4V 0 where V 0 is a scale of the macro-pores among the warp and weft yarns, the
reference parameter. Following the film-stacking technique, polymer promptly infiltrates into the fabric layers and
thermoplastic films and four layers of reinforcing fabrics bypasses the yarns, whereas transverse flow into the fine

Table 1 Physical properties of PA 6.6

Polymer T m (8C) a T c (8C) a h 0 (Pa s) @ 2908C M w (g mol ¹1) M n (g mol ¹1) Additives
PA 6.6 (#1) 260 230 70 63 000 163 000 Heat stabilizer
PA 6.6 (#2) 260 220 120 67 000 161 000
a
Data were obtained by DSC at 10 K min ¹1.

Table 2 Characterization of fabrics

US-Style Type Weave m* (g m ¹2) Yarns (cm) Fil.-À (mm) Yarn (tex)
7581 Glass 8 H Satin 296 22 3 21 9 68
1113 Glass Twill 2/2 390 6 3 6.7 15 320
462 Carbon Twill 2/2 245 636 7 200

785
Macro- and micro-impregnation: C. Mayer et al.

Figure 2 Macro- and micro-impregnation: (a) ideal flow geometry; (b) real case macro-pores; (c) real case micro-pores

Figure 3 Recorded actual temperature and corresponding Newtonian viscosity profiles of polyamide 6.6 (E 0 ¼ 44.83 kJ mol ¹1; h 0 ¼ 70 Pa s at 2908C) as a
function of the temperature boundary (1.31T 0) in the HTZ and increasing process velocity

interstices of the fiber bundle is considerably delayed. accomplish impregnation without thermal degradation of
Parnas and Phelan21 described this phenomenon for in-plane the resin.
flow in manufacturing composites via RTM. Since satura- For processes exhibiting very low shear rates ( , 100 s ¹1)
tion of the macro-pores is accomplished prior to the such as calenders and double belt presses, the temperature
impregnation of individual fiber bundles, the polymer first dependent apparent viscosity of a polymer melt can be
fills all space outside the yarns and then percolates radially approximated by means of Newtonian viscosity h 0 using the
towards the center of the warp and weft domains. This Arrhenius approach in the form of
process is accompanied by a simultaneous reduction of   
E0 1 1
thickness as represented by h 0 and h 1 (Figure 2). h0 (T) ¼ h0 (Tn )exp ¹ (5)
R T Tn
with the activation energy E 0, the gas constant R (8.314 3
Non-isothermal processing conditions
10 ¹3 kJ mol ¹1 K ¹1) and the Newtonian viscosity h 0 of the
From a process point of view, temperature is the factor polymer melt at a certain temperature T n.
that determines via viscosity how readily polymer can melt Different viscosity profiles were modeled by choosing
to flow into and through the macro-pores of the fabric and constant temperature boundary conditions of 1.31T 0 in the
the minute spaces between the fibers of a yarn. In the present high temperature area, where T 0 is a reference temperature,
manufacturing process using a double belt press, the and incrementally increasing the process velocities from V 0
material exhibits a non-isothermal history of temperature to 4V 0, where V 0 is a reference line speed. Figure 3 exhibits
vs. time due to the conductive unsteady-state heat transfer in the resulting normalized temperature profile for the
the continuous process. As it can be derived from eqns (1) boundary conditions of EZ (1608C), HTZ (1.31T 0), HZ
and (2), the suitability of a particular thermoplastic polymer (2508C) and CZ (408C) which was recorded with thermo-
for a given reinforcement critically depends on its viscosity couples at the center at the composite stack. Eqn (5) was
during the process cycle. Therefore, the prediction of then applied to approximate a series of isothermal steps to
polymer viscosity as a function of temperature and time can obtain the corresponding viscosity profiles for the low
help to find the acceptable temperature and velocity range to viscosity grade of polyamide 6.6 (polymer #1) using an

786
 Macro- and micro-impregnation: C. Mayer et al.

Figure 4 Typical photomicrographs of cross sections of individual micro-impregnated fiber bundles (3k CF-Roving 200 tex, 7 mm; PA 6.6 h 0 ¼ 120 Pa s): (a)
1.31T 0, V 0, D imp: 100%; (b) 1.25T 0, 2V 0, D imp: 70%; (c) 1.15T 0, 3V 0, D imp: 55%; (d) 1.08T 0, 4V 0, D imp: 40%

activation energy of 44.83 kJ mol ¹1 and a Newtonian and impregnation time on micro-impregnation pattern,
viscosity of 70 Pa s at 2908C. Clearly from Figure 3, the photomicrographs of polished cross sections of carbon
maximum recorded temperature in the material as a function fabric reinforced PA 6.6 samples were obtained at 100 3
of 4V 0 drops down from 1.31T 0 to 1.08T 0 while the resident magnification using an optical microscope (see Figure 4).
time, the time between melting and crystallization of the The samples were fabricated according to the above process
polymer and thus the time where the molten polymer is able conditions (Figure 3) at a consolidation pressure of 25 bar.
to percolate into a porous medium, is simultaneously The pictures reveal that with reduced resident time and
shortened by a factor of 5. Once the process velocity is increasing viscosity of the polymer, the carbon yarns
reduced, the actual temperature approaches the boundary essentially exhibit an enlarged area of unimpregnated
temperature of 1.31T 0 while the resident time between fibers depicted as the large dark ellipses in the center of
melting and crystallization is extended to 100 s. This the yarns. Polymer percolation into the bundle was
particular set-up was used in the manufacturing of the negligible and can therefore only be found to some extent
samples to provide an increase in polymer viscosity along as a frozen layer at the boundary of the fiber bundle. In
with reduced impregnation time and therefore enhance the contrast, at low process velocity and high fluidity of the
effect of processing boundary conditions on the degree of polymer, the size of the dark ellipses became smaller and
impregnation. smaller, thus manifesting an enhanced degree of impregna-
From the empirical relationship of permeability and tion with increasing resident time and reduced viscosity of
velocity of the flow front as summarized in eqns (1)–(4), it the molten polymer.
is expected that a tightly packed fiber bundle with
comparably large flow distances in the radial direction
Measurement of mechanical properties
will highly resist to transverse micro-impregnation. There-
fore fabrics made from carbon fibers were found to be However, it was difficult to observe such a strong
suitable materials (see Table 2) to investigate the progress of relationship of processing conditions and impregnation
micro-impregnation as a function of viscosity and resident among as-molded samples consisting of high permeability
time. To qualitatively verify the effect of polymer rheology sub-domains, e.g. the glass fabrics of style 7581 and 1113.

Figure 5 Cross sectional view of as-molded glass fabric reinforced PA 6.6 (h 0 ¼ 70 Pa s): (a) 1.31T 0, V 0; (b) 1.08T 0, 4V 0; note the higher compression of the
fiber bundles at lower temperature and higher process velocity: (A) unimpregnated speckles; (B) region of fiber pull out due to polishing

787
Macro- and micro-impregnation: C. Mayer et al.

In this respect, the bundle areas perpendicular to the cross the resident time and the amount of heat transfered into the
section of the sample exhibited a negligible difference in the material. Since the impregnation quality observed among
degree of impregnation represented as dark minute speckles carbon fabric reinforced PA 6.6 correlates evidently with
as depicted in Figure 5(b). Though, it is assumed that the the flexural stiffness measured on the same samples, see
different process parameters must have some deep influence Figure 6(b), the flexural stiffness is an appropriate tool to
on the impregnation quality. Several authors have reported mirror the degree of impregnation. In accordance with
on quantitative methods such as thickness and density eqns (1)–(4) and the microscopic observations, glass fabrics
measurement to account for the impregnation quality. It was as used in the present experiment are much less sensitive to
found, that reduced impregnation is accompanied by an changes of temperature and resident time than carbon
increase in thickness due to unimpregnated volume fabrics. Comparing glass fabrics style 7581 and 1113, the
occupied by air while the density increases with increasing flexural stiffness of 1113–PA 6.6 (polymer #2) samples
impregnation quality. decreases considerably faster with increasing process
Analyzing the amount of fiber pull-out among the parallel velocity than that of 7581–PA 6.6 (polymer #2). This may
fiber bundles generated by the polishing process, some be ascribed to the larger diameter of the yarns used in the
evidence can be found, that the shear strength between 1113 fabric style which seems to have a stronger impact on
polymer and embedded fibers inside the bundle may be the time to impregnate the bundle than the probably more
different for different process conditions. Besides the densely packed but less thick bundle (9 mm filaments,
interface qualities, the amount of shear strength transferred 68 tex) in 7581, as expressed in eqn (2).
from the polymer to the fiber is a strong function of the real Surprisingly, the flexural stiffness of the glass fabric
area of contact between matrix and reinforcement and thus reinforced polyamide samples vs. time/temperature follows
the degree of impregnation. Therefore, the impregnation of a linear relationship whereas for the carbon fabric reinforced
thermoplastic composites with the same type of fabric samples a power fit can be applied. The curve fits are in
reinforcement manufactured at different process conditions good agreement with the experimental data since the
may be correlated with mechanical properties such as regression coefficient was calculated with 0.98 and higher.
stiffness and strength. Hence, a three-point bending test was This may be due to the lower permeability of the carbon
performed to mirror the degree of impregnation. According to fiber yarns which need more time to be fully impregnated.
a testing method for thin thermoplastic composite samples Empirically, the quality change is very quick at the primary
developed by DuPont and the Institut für Verbundwerkstoffe stage of impregnation whereas with increasing impregna-
GmbH, a span-to-depth ratio of 32/h, where h represents the tion, the change becomes weaker. Therefore, the quality
sample thickness, and a width of 25 mm was selected to change for glass is smaller than for carbon fabrics since for
evaluate the flexural property. The samples were subjected to equilibrium processing conditions glass fabric reinforced
a cross head speed of 1 mm min ¹1. thermoplastics have reached a late state of impregnation
while carbon fabric reinforced thermoplastics stay in a
primary or middle stage of the impregnation process. If the
DISCUSSION process velocity would slow down further, the same trend
could be obtained for the glass fabric reinforced samples.
For samples with incomplete impregnation, a section of
Results on temperature and impregnation time
unimpregnated or worse impregnated fibers usually remains
Obviously from Figure 6, the flexural stiffness is a strong in the center of the fiber bundle forming a weak point of
function of process velocity which subsequently effects both reduced strength (see Figure 4). Figure 7 displays a series of

Figure 6 Flexural stiffness as a function of process conditions (constant temperature boundary of 1.31T 0 in the HTZ)

788
 Macro- and micro-impregnation: C. Mayer et al.

Figure 7 Transition from tensile failure to compression failure due to incomplete micro-impregnation (a) 1.31T0 ; V0 ; (b) 1.08T0 ; 4V0 ; and detailed cross
sectional views of cavity formation after flexural loading of glass fabric (7581)-PA6.6 (h0 ¼ 70 Pa s) due to incomplete microimpregnation; (c) 1.31T0 ; V0 ; (d)
1.25T0 ; 2V0 ; (e) 1.15T0 ; 3V0 ; (f) 1.08T0 ; 4V0

cross sections after flexural loading of the glass fabric Results on pressure
(7581) reinforced PA 6.6 (polymer #1) samples presented in
Figure 5. For this particular material combination, it was As expressed by eqns (1) and (2), the pressure drop
difficult to distinguish the degree of impregnation among becomes an important factor to accelerate the impregnation
the virgin samples by means of microscopic observation. process. In the present manufacturing operation using a
After flexural loading, one can find debonding or cavities of double belt press, pressure is consistently applied on the
distinct degree occurring at the center of the yarns, whereas stack of polymer and fabrics during heating, melting and
to some extent a circular layer of impregnated fibers was crystallization of the thermoplastic polymer. Lekakou
preserved at the boundary of the bundle. Once the degree of et al.19 published on the compressibility of woven
impregnation is decreased further, the size of the cavity at reinforcements by applying the viscoelastic model of
the center of the yarns enlarges and becomes more serious as Gutowski et al.10 to the nonlinear elastic manner of fabrics
shown in the row of Figure 7(a)–(d). Hence, for the damage in compression and demonstrated the sensitivity of the
pattern a similar relationship of flexural stiffness and permeability on the degree of compression. In this model,
process conditions was found. Additionally, the formation the fibers comprised the spring element and the polymer the
of a cavity was accompanied by a change from tensile to viscoelastic damper element. However, this behavior is
compression failure in flexural loading, reduced shear phenomenologically different for the continuous manufac-
strength, strain endurance, increased energy dissipation turing of film-stacked thermoplastic composites since
and brittle/ductile transition22. Guowski et al. used low viscosity model fluids such as

789
Macro- and micro-impregnation: C. Mayer et al.

after processing at lower temperature (1.08T 0), whereas at

low viscosity using 1.31T 0 as a maximum temperature, the
fiber bundle exhibits an open structure with larger interstices
for the polymer to pass through. Hence in this stage, the
applied pressure is transferred partly to the fibers as spring
force and partly to the liquid polymer as hydrostatic drag.
Evidently, the viscosity plays a major part to reduce first the
packing density of the yarn and then benefit from a less
restricted flow into a more open structure. Depending on the
fiber volume content and the surface of the fabric structure,
it can be assumed that for low polymer contents and fabrics
exhibiting a wavy topography, a significant portion of the
consolidation pressure is carried by the network structure.
Figure 8 Compression of the reinforcing fabrics and relaxation due to
melting of the polymer (exaggerated sketch) Thus, the pressure on the resin may be reduced to such an
extent where the impregnation is mainly driven by capillary
silicon oils with a viscosity below 1 Pa s at the ambient forces as a function of interatomic and intermolecular forces
temperature of 208C. This approach does not count for the of attraction between fiber and matrix.
solid state processing of glassy or rubber like thermoplastic For experimental verification of the pressure dependent
polymers below melting temperature. impregnation, samples of glass fabric (US-Style 1113)
During heating, the solid polymeric layers essentially reinforced polyamide 6.6 (polymer #2) with a fiber volume
behave in an elastic manner and compress the reinforcing content of 50% were manufactured using 1.19T 0 and 1.08T 0,
structure to a fabric layer thickness of h s (Figure 8). The lay- respectively, as a temperature boundary in the HTZ and a
up can be described as a series of springs representing either process velocity of V 0 by stepwise increasing the consolida-
the network of fiber bundles or the fiber bundle itself. In this tion pressure from 3 to 45 bar. The lower temperatures of
stage, polymer elasticity and fabric compressibility are at 1.19T 0 and 1.08T 0 where selected to apply a higher viscous
equilibrium and the fiber bundles collapse inward by the drag on the macro and micro domains of the fabric and
applied pressure; physical contact of the fibers can take therefore enhance the effect of pressure dependent perme-
place. Once melting, the polymer generally exhibits a ability. From eqn (5) this temperature drop is associated
transition from elastic to viscous behavior where the viscous with an increase of Newtonian viscosity by a factor of 1.59
drag due to hydrostatic pressure is low compared with the and 2.62, respectively, compared to 1.31T 0.
pressure transferred by the solid elastic damper. As a Figure 9 presents the results obtained by flexural loading
consequence, the fabrics may relax to a thickness of h m of the 1113–PA 6.6 (polymer #2) samples fabricated as
larger than h s due to the reduced compression force described above. It can be seen, that the flexural stiffness for
generated by the viscous liquid. This process is accom- both temperature conditions increases remarkably once the
panied by macro-impregnation and subsequent micro- consolidation pressure is raised from 3 to 25 bar. Simulta-
impregnation. Therefore it is convenient to correlate the neously, the thickness of the samples is reduced by almost
degree of fabric network and fiber bundle relaxation to the 7% indicating that the pressure besides the viscosity of the
viscosity of the polymer. As shown previously in Figure 5, polymer in fact drives the impregnation. However, if the
the compaction of the fiber bundle is in fact much higher pressure goes further, a certain amount of the pressure has to

Figure 9 Flexural stiffness and sample thickness as a function of consolidation pressure and temperature boundary conditions at the high temperature zone.
Data obtained for 1113–PA 6.6 (polymer #1)

790
 Macro- and micro-impregnation: C. Mayer et al.

Figure 10 Microscopic adhesion mechanism

be used to force the polymer into the more densely packed

bundles since the fiber volume fraction of the yarns
increases with applied pressure. Due to this detrimental
effect, both the flexural stiffness and the thickness reveal a
plateau at consolidation pressures beyond 25 bar where a
slight decrease at 45 bar is recognizable for the samples
manufactured at 1.19T 0. It appears that at 25 bar, the
pressure dependent permeability and the driving efficiency
due to the pressure drop on the molten polymer are at an
optimum. Nonetheless, the validity of the above analysis
may be different for other material combinations, particu-
larly with respect to matrix volume fractions, viscosity, Figure 11 Tubular transport of air after macro-impregnation: (A) frozen
fabric waviness, and surface forces. bubbles, (B) warp, (C) weft

Microscopic autoadhesion
stack of polymer films and reinforcing fabrics may be
It was previously mentioned that during micro-impreg- reduced by 70–100% or even more after impregnation and
nation polymer flow fronts move radially into the fiber consolidation. This large decrease in volume can be
bundle, and impinge at the central line of the elliptical ascribed to a displacement of air within the macro- and
structure. Applying the Springer and Lee approach on micro-pores by infiltration of molten polymer into the
autoadhesion16 one can assume that after contact has been fabrics. Transport of air between the layers and within the
established, the interfaces merge by a bonding process along macro-pores is accomplished in the direction of least
this line. As a result of this microscopic autoadhesion, the resistance either to the side edges of the double belt press,
process cycle to manufacture performing composites does opposite to the process direction during compression of the
not only depend on the time to impregnate a network solid polymer or at the flow front region of the molten
structure as discussed in the previous sections, but also polymer. Since macro-impregnation performs considerably
profoundly requires some additional time first to achieve fast, every single yarn is surrounded by molten polymer
intimate contact of the flow fronts and then merging of the percolating into the fiber bundle. As a result fiber bundles
interfaces. Figure 10 displays the different stages of flow, form ‘tubes’ of decreasing diameter. In such as case,
contact and adhesion encountered in the microscopic migration of air is only possible along the longitudinal
autoadhesion. Reduction of the manufacturing time may direction of the warp and weft tubes opposite to the process
thus be followed by a weak interface of reduced shear direction or the side edges of the press rather than
strength although microscopic impregnation was com- transversely into the molten polymer (see Figure 11).
pleted. Basically, a woven structure exhibits a certain non-
uniformity of the micro-pore structure. There are two types
of micro-pore cross sections which contribute to the
Void entrapment
impregnation behavior. Type A is located at the connection
From experimental studies, there is evidence that points of warp and perpendicular aligned weft of the fabric
mechanical properties can be significantly reduced with structure and type B can be found between these points.
only small increase in the void content. Composite materials During compression, the interconnection points carry most
containing voids are also susceptible to weathering and of the pressure applied to the reinforcing network, so that
moisture absorption. Thus, minimizing and diminishing of the fiber volume fraction of type A is generally higher
voids is a critical issue in manufacturing fiber reinforced than that of type B. As a consequence, the velocity of the
materials23–25. During the continuous manufacturing pro- flow front into a more tightly packed array of fibers is
cess the thickness of an unimpregnated and unconsolidated reduced. Due to the time difference to accomplish

791
Macro- and micro-impregnation: C. Mayer et al.

Thus the governing phenomenon during the final stage of

micro-impregnation will be the generation of internal
pressure, that can reduce the radial pressure gradient to
such an extent, where the progress of flow is mainly driven by
capillary forces. As a consequence, collapsing of the voids
and the affinity of polymer to the fiber surface can be the real
limiting factors in accomplishing micro-impregnation of
reinforcing fabrics with thermoplastic polymers.

CONCLUSIONS

The impregnation of fabric structures with thermoplastic

polymers in continuous manufacturing by using a double
Figure 12 Assumed entrapment mechanism
belt press is accomplished along the direction of
thickness rather than in-plane as usual in liquid
composite manufacturing techniques. Due to the dual
scale pore structure of the reinforcing network and the
sub-domains of fiber yarns, the impregnation process
exhibits first a high volume flow into the macro-pores
among the yarns followed by a transverse micro-flow
into the fine interstices of the fiber bundles. Following
the analytical approaches employed to model the
transverse permeability, it was found that the micro-
impregnation governs the time required to saturate the
fiber bundle, and is a strong function of the polymer
viscosity, resident time between melting and crystal-
lization of the polymer, consolidation pressure and
solubility of gas in the molten polymer. The application
of an increased temperature in the high temperature zone
Figure 13 Generation of internal pressure by compression of void (HTZ) of the double belt press in combination with a
low process velocity was beneficial to the degree of
impregnation since it accelerates the melting of the
polymer, reduces its viscosity and enlarges the impreg-
micro-impregnation in domains of type A and B, air can nation time. To exhibit the stages in micro-impregnation
therefore possibly be entrapped between two sections of by means of microscopic observation, a low permeability
type B, see Figure 12. carbon fabric was selected and material was manufac-
Figure 13 displays the case, where a void is entrapped tured using different viscosity profiles as a function of a
between advancing flow fronts. The effects of such behavior constant temperature boundary in the HTZ and incre-
can have strong impacts on the degree of impregnation, flow mentally increasing process velocity. Flexural testing was
front velocity and properties of the joining interfaces. Once performed to mirror the degree of impregnation via stress
the polymer percolation into the fiber bundle continues, the transfer from the fibers to the matrix over the real area
volume of the gaseous phase is reduced and an internal of contact established by the degree of impregnation. In
pressure is generated in the entrapped air. This generates a transverse micro-impregnation, the consolidation pressure
local decrease of the overall radial pressure gradient. At the becomes an important parameter to drive the polymer
same time, a certain amount of air consisting of basically flow. The effects of the compression of the reinforcement
21% oxygen and 78% nitrogen starts to solve at the outer on the flexural stiffness and thus the degree of
surface of the liquid polymer followed by inward diffusion. impregnation are presented. It could be proved that for
Solubility and diffusivity are a function of temperature, consolidation pressures between 15 and 25 bar, the
pressure and nature of the polymer. The solubility of a gas in detrimental effects of pressure dependent permeability and
a molten polymer can be expressed by Henry’s Law26, driving force due to the pressure drop applied to the molten
where the solubility is proportional to the pressure. polymer are at optimum conditions. Microscopic impreg-
Additionally solubility and diffusivity exhibit a temperature nation involves additional effects arising inside the fiber
dependence described by an Arrhenius type equation. bundle. First, micro-impregnation goes along with a
However, these mass transfer processes are relatively merging of the radially approaching flow fronts after they
slow. Hence, a sufficiently high hydrostatic pressure contacted at the center line of the elliptical fiber bundle.
accompanied by high temperature is needed to first collapse Second, voids can possibly be entrapped inside the fiber
the void and ultimately to exceed the vapor pressure of the bundles and may generate some internal pressure. As a
gas so that the pore will simply dissolve into the polymer. consequence the manufacturing cycle does not only depend

792
 Macro- and micro-impregnation: C. Mayer et al.

on the time to accomplish impregnation, but also profoundly 8. Sangani, A. S. and Yao, C., Transport process in random arrays of
cylinders, II. Viscous flow. Physics of Fluids, 1988, 31(9), 2433–
requires some additional time to (i) collapse the voids at 2444.
reducing radial pressure gradients and to (ii) achieve 9. Bruschke, M. V., A predictive model for permeability and non-
intimate contact of the flow fronts. Although it is not isothermal flow of viscous and shear-thinning fluids in anisotropic
fibrous media. Ph.D. thesis, University of Delaware, Newark, 1992.
tested in the present paper, it is speculated that molecular 10. Gutowski, T. G., Cai, Z., Bauer, S., Boucher, D., Kingery, J. and
effects such as the affinity of molten polymers to fiber Wineman, S., Consolidation experiments for laminate composites.
surfaces and solubility and diffusivity phenomena influen- Journal of Composite Materials, 1987, 21, 650–669.
11. Lam, R. C. and Kardos, J. K., The permeability of aligned and cross-
cing the polymer properties at the vicinity of the polymer plied fiber beds during processing of continuous fiber composites. In
flow front can be the real limiting effects in accomplishing Proceedings of the Third Technical Conference, Seattle, Washing-
micro-impregnation. ton, 25–29 September 1988, pp. 3–11.
12. Gebart, B. R., Permeability of unidirectional reinforcements for
RTM. Journal of Composite Materials, 1992, 26(8), 1100–1133.
ACKNOWLEDGEMENTS 13. Cai, Z. and Berdichevsky, A. L., Estimation of the resin flow per-
meability of fiber tow preforms using the self-consistent method.
Polymer Composites, 1994, 15(3), 240–246.
The authors are grateful to the German Ministry of 14. Seo, J. W. and Lee, W. I., A model of the resin impregnation in
Education and Research (BMBF) for the financial support thermoplastic composites. Journal of Composite Materials, 1991,
25, 1127–1141.
of the project 03N 3008 (Continuous Manufacturing of 15. Kim, T. W., Jun, E. J., Um, M. K. and Lee, W. I., Effect of pressure
Tailored Textile Reinforced Thermoplastic Composite on the impregnation of thermoplastic resin into a unidirectional fiber
Intermediates). The support of X. Wang’s research (June bundle. Advances in Polymer Technology, 1989, 4, 275–279.
16. Springer, G. S. and Lee, W. I., Processing model of thermoplastic
1995–January 1997) at the Institut für Verbundwerkstoffe matrix composites. In Proceedings of the 33rd International Sampe
GmbH by the Alexander von Humboldt Foundation is very Symposium, 1–7 March 1988, pp. 661–669.
much appreciated. 17. Van West, B. P., Pipes, R. B. and Advani, S. G., The consolidation
of Commingled thermoplastic fabrics. Polymer Composites, 1991,
12(6), 417–427.
18. Bafna, S. S. and Baird, D. G., An impregnation model for the pre-
REFERENCES paration of thermoplastic prepregs. Journal of Composite Materials,
1995, 26(5), 683–707.
1. Mayer, C., Hartmann, A. and Neitzel M., Cost-conscious manufac- 19. Lekakou, C., Johari, M. A. and Bader, M. G., Compressibility and
turing of tailored thermoplastic composite intermediates using a flow permeability of two-dimensional woven reinforcements in the
double belt press. In Proceedings of the 18th International processing of composites. Polymer Composites, 1996, 17(5), 666–
SAMPE Europe Conference, Paris, 23–25 April 1997, pp. 339–351. 672.
2. Wang, X., Mayer, C. and Neitzel, M., Impregnation of thermoplas- 20. Simacek, P. and Advani, S. G., Permeability model for a woven
tic composites manufactured by double belt press technique. In fabric. Polymer Composites, 1996, 17(6), 887–899.
Proceedings of the 11th International Conference on Composite 21. Parnas, R. S. and Phelan Jr, F. R., The effects of heterogeneous
Materials, Gold Coast, Australia, 12–18 July 1997. porous media on mold filling in resin transfer molding. Sampe Qua-
3. Breuer, U. and Neitzel, M., High speed stamp forming of thermo- terly, 1991, 53–60.
plastic composite sheets. Polymer and Polymer Composites, 1996, 22. Wang, X., Mayer, C. and Neitzel, M., Some issues on impregnation
4(2), 117–123. in manufacturing of thermoplastic composites by using a double
4. Chmielewski, C., Petty, C. A. and Jayaraman, K., Crossflow of belt press. Polymer Composites, 1997, 18(6), 707–710.
elastic liquids through arrays of cylinders. Journal of Non-Newto- 23. Lundström, T. S., Bubble transport through constricted capillary
nian Fluid Mechanics, 1990, 35, 309–325. tubes with application to resin transfer molding. Polymer Compo-
5. Åström, B. T., Pipes, R. B. and Advani, S. G., On flow through sites, 1996, 17(6), 770–779.
aligned fibers beds and its application to composite processing. 24. Chan, A. W. and Morgan, R. J., Modeling preform impregnation
Journal of Composite Materials, 1992, 26(9), 1351–1373. and void formation in resin transfer molding of unidirectional com-
6. Sangani, A. S. and Acrivos, A., Slow flow past periodic arrays of posites. Sampe Quaterly, 1992, 48–52.
cylinders with application to heat transfer. International Journal of 25. Rohatgi, V., Patel, N. and Lee, L. J., Experimental investigation of
Multiphase Flow, 1992, 8(3), 193–206. flow-induced microvoids during impregnation of unidirectional
7. Sangani, A. S. and Acrivos, A., Slow flow through a periodic array stitched fiberglass mat. Polymer Composites, 1996, 17(2), 161–170.
of spheres. International Journal of Multiphase Flow, 1982, 8(4), 26. Van Krevelen, D. W. (ed.), Properties of Polymers. Elsevier,
343–360. Amsterdam, 1990, pp. 535–583.

793