Accepted Manuscript
In-Plane Permeability Characterization of Engineering Textiles Based On Radial Flow Experiments: A Benchmark Exercise
D. May, A. Aktas, S.G. Advani, D.C. Berg, A. Endruweit, E. Fauster, S.V.
Lomov, A. Long, P. Mitschang, S. Abaimov, D. Abliz, I. Akhatov, M.A. Ali,
T.D. Allen, S. Bickerton, M. Bodaghi, B. Caglar, H. Caglar, A. Chiminelli, N.
Correia, B. Cosson, M. Danzi, J. Dittmann, P. Ermanni, G. Francucci, A. George,
V. Grishaev, M. Hancioglu, M.A. Kabachi, K. Kind, M. Deléglise-Lagardère,
M. Laspalas, O.V. Lebedev, M. Lizaranzu, P.-J. Liotier, P. Middendorf, J.
Morán, C.-H. Park, R.B. Pipes, M.F. Pucci, J. Raynal, E.S. Rodriguez, R.
Schledjewski, R. Schubnel, N. Sharp, G. Sims, E.M. Sozer, P. Sousa, J. Thomas,
R. Umer, W. Wijaya, B. Willenbacher, A. Yong, S. Zaremba, G. Ziegmann
PII:
DOI:
Reference:
S1359-835X(19)30079-X
https://doi.org/10.1016/j.compositesa.2019.03.006
JCOMA 5360
To appear in:
Composites: Part A
Received Date:
Revised Date:
Accepted Date:
24 October 2018
1 March 2019
9 March 2019
Please cite this article as: May, D., Aktas, A., Advani, S.G., Berg, D.C., Endruweit, A., Fauster, E., Lomov, S.V.,
Long, A., Mitschang, P., Abaimov, S., Abliz, D., Akhatov, I., Ali, M.A., Allen, T.D., Bickerton, S., Bodaghi, M.,
Caglar, B., Caglar, H., Chiminelli, A., Correia, N., Cosson, B., Danzi, M., Dittmann, J., Ermanni, P., Francucci, G.,
George, A., Grishaev, V., Hancioglu, M., Kabachi, M.A., Kind, K., Deléglise-Lagardère, M., Laspalas, M., Lebedev,
O.V., Lizaranzu, M., Liotier, P.-J., Middendorf, P., Morán, J., Park, C.-H., Pipes, R.B., Pucci, M.F., Raynal, J.,
Rodriguez, E.S., Schledjewski, R., Schubnel, R., Sharp, N., Sims, G., Sozer, E.M., Sousa, P., Thomas, J., Umer, R.,
Wijaya, W., Willenbacher, B., Yong, A., Zaremba, S., Ziegmann, G., In-Plane Permeability Characterization of
Engineering Textiles Based On Radial Flow Experiments: A Benchmark Exercise, Composites: Part A (2019), doi:
https://doi.org/10.1016/j.compositesa.2019.03.006
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers
we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and
review of the resulting proof before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
In-Plane Permeability Characterization of Engineering
Textiles Based On Radial Flow Experiments: A Benchmark
Exercise
D. Maya,*, A. Aktasb, S. G. Advanic, D. C. Bergd, A. Endruweite, E. Fausterf,
S. V. Lomovg, A. Longe, P. Mitschanga, S. Abaimovh, D. Ablizd, I. Akhatovh,
M.A. Alii, T. D. Allenj, S. Bickertonj, M. Bodaghik, B. Caglarl, H. Caglarl,
A. Chiminellim, N. Correiak, B. Cossonn, M. Danzio, J. Dittmannp, P. Ermannio,
G. Francucciq, A. Georger, V. Grishaevh, M. Hancioglul, M. A. Kabachio, K. Kinds,
M. Deléglise-Lagardèren, M. Laspalasm, O. V. Lebedevh, M. Lizaranzum, P.-J.
Liotiert, P. Middendorfp, J. Moránq, C.-H. Parkn, R. B. Pipesu, M. F. Pucciv,
J. Raynalw, E. S. Rodriguezq, R. Schledjewskif, R. Schubnelw, N. Sharpu, G. Simsb,
E. M. Sozerl, P. Sousag, J. Thomasi, R. Umeri, W. Wijayaj, B. Willenbachera,
A. Yongb, S. Zarembas, G. Ziegmannd
a
Institut für Verbundwerkstoffe GmbH, Germany
Materials Division, National Physical Laboratory, United Kingdom
c
Department of Mechanical Engineering and Center for Composite Materials, University of
Delaware, USA
d
Department of Polymer Materials and Plastics Engineering, Technische Universität
Clausthal, Germany
e
Faculty of Engineering, University of Nottingham, United Kingdom
f
Processing of Composites Group, Montanuniversität Leoben, Austria
g
Department of Mechanical Engineering, Katholieke Universiteit Leuven, Belgium
h
Center for Design, Manufacturing and Materials, Skolkovo Institute of Science and
Technology, Russia
i
Khalifa University of Science and Technology (KUST), Abu Dhabi, UAE
j
Centre for Advanced Composite Materials, University of Auckland, New Zealand
k
Composite Materials and Structures Group, INEGI, Portugal
l
Department of Mechanical Engineering, KOÇ University, Turkey
m
ITAINNOVA Instituto Tecnológico de Aragón, Spain
n
Department of Polymers and Composites Technology & Mechanical Engineering, IMT
Lille Douai, France
o
Department of Mechanical and Process Engineering, ETH Zürich, Switzerland
p
Institute of Aircraft Design, University of Stuttgart, Germany
q
Institute of research in Materials Science and Technology, Universidad Nacional de Mar
del Plata, Argentina
r
Faculty of Manufacturing Engineering Technology, Brigham Young University, USA
s
Chair of Carbon Composites, Technische Universität München, Germany
t
Mines Saint-Etienne, Université de Lyon, CNRS, UMR 5307 LGF, Centre SMS,
Departement
MPE, F-42023 Saint-Etienne, France
u
Composites Manufacturing & Simulation Center, Purdue University, USA
v
C2MA, IMT Mines Ales, Univ. Montpellier, Ales, France
b
w
Institut de Soudure Group, France
*Corresponding author mail: david.may@ivw.uni-kl.de, phone: +49 631 31607 – 34
Abstract
Although good progress was made by two international benchmark exercises on in-plane
permeability, existing methods have not yet been standardized. This paper presents the
results of a third benchmark exercise using in-plane permeability measurement, based on
systems applying the radial unsaturated injection method. 19 participants using 20 systems
characterized a non-crimp and a woven fabric at three different fiber volume contents, using
a commercially available silicone oil as impregnating fluid. They followed a detailed
characterization procedure and also completed a questionnaire on their set-up and analysis
methods. Excluding outliers (2 of 20), the average coefficient of variation (cv) between the
participant’s results was 32% and 44% (non-crimp and woven fabric), while the average cv
for individual participants was 8% and 12%, respectively. This indicates statistically
significant variations between the measurement systems. Cavity deformation was identified
as a major influence, besides fluid pressure / viscosity measurement, textile variations, and
data analysis.
Keywords: A. Fabrics/textiles; B. permeability; D. Process Monitoring, E. Liquid
composite moulding, E. Resin flow
2
1. Introduction
Liquid Composite Molding (LCM) processes are employed for the manufacture of fiber
reinforced polymer composites (FRPC), since they allow to efficiently manufacture
components of different complexity and size at higher rates than autoclave processes. To
obtain fast and complete saturation of the reinforcement with liquid resin in LCM, a suitable
process design is desirable, which requires knowledge about material properties. The textile
permeability is particularly important. It is defined by Darcy’s law, which correlates the
phase-averaged flow velocity
with the impregnating resin pressure gradient
, its
dynamic fluid viscosity , and the textile permeability , which quantifies the conductance
of the porous media for liquid flow (Eq. 1).
(1)
The permeability of fiber structures, such as textiles, is generally direction-dependent and
therefore described by a second-order tensor. Commonly, textile symmetry conditions are
taken into account so that the tensor can be diagonalized, which leads to four remaining
values describing flow in any direction within a fiber structure (assuming absence of
coupling between in-plane and out-of-plane flow):
Highest in-plane permeability (K1), in-plane refers to the textile layer;
Lowest in-plane permeability (K2), oriented perpendicular to K1;
Orientation angle of K1 (β), relative to the production direction of the material (0°);
Out-of-plane permeability (K3), oriented perpendicular to K1 and K2.
The present paper focuses on the characterization of the in-plane permeability (K1, K2 and
β).
Despite the relevance of accurate permeability characterization for process efficiency,
existing in-plane permeability characterization methods have not yet been standardized.
Following several smaller regional benchmark studies [1-5], the results of the first truly
3
international benchmark exercise on in-plane permeability measurement were published in
2011 [6]. In this exercise, same fabric was used by all participants, but no specifications
were made regarding the measurement method and the test parameters. This resulted in a
scatter of the measured permeability values of more than one order of magnitude. A second
international benchmark exercise with a predefined measurement procedure [7] followed.
The participants were required to apply an unsaturated linear injection method. In
unsaturated linear injection of a fluid into a dry reinforcement sample, one-dimensional flow
develops. The resulting flow front movement can be tracked, and the permeability along the
specimen axis can be derived using a 1D formulation of Eq. (1). This benchmark exercise
showed - for this specific test method - that by defining minimum requirements for
equipment, measurement procedure and analysis, satisfactory reproducibility of data
obtained using different systems can be achieved [8]. In-plane permeability characterization
based on radial flow experiments is an alternative approach, where the test fluid is injected
through a central injection gate into a tool cavity containing the reinforcement sample.
Advantages of this approach are that only one test is required for full textile characterization
including K1, K2 and β and that the possible influence of race-tracking on test results is
reduced. Hence, it was agreed at the 13th International Conference on Flow Processes in
Composite Materials (FPCM) in Kyoto (2016) to perform a third international benchmark
exercise, focusing on unsaturated in-plane permeability characterization based on radial flow
experiments. This benchmark exercise was organized by the Institut für Verbundwerkstoffe
(IVW, Kaiserslautern, Germany), and strongly supported by the National Physical
Laboratory (UK), the University of Nottingham, the University of Delaware (CCM), the
Montanuniversität Leoben and KU Leuven as members of a steering committee.
Furthermore, the organizers were strongly supported by the Department of Polymer
4
Materials and Plastics Engineering at Clausthal University of Technology. Table 1 lists the
participants of the presented benchmark exercise.
2. Materials & Methods
2.1
Experimental set-up
2.1.1.
Basic requirements
The presented benchmark focused on unsaturated in-plane permeability characterization
based on radial flow. Taking into account the guidelines of the 2 nd international benchmark
on in-plane permeability characterization based on the linear flow method [7], basic
requirements for the experimental set-up, used materials and the measurement procedure
were specified for all participants:
A stack of textile layers is compressed between two rigid mold surfaces at constant
gap height (as illustrated in Figure 1).
A test fluid is injected through a central circular hole (12 mm diameter), resulting in
a two-dimensional flow pattern (typically an ellipse). The injection hole has to be
punched into the textile.
An unsaturated measurement principle is applied, i.e. flow front progression is
tracked.
2.1.2.
No vacuum is applied.
Individual set-ups of the participants
Within the constraints of the stated basic requirements, a wide variety of designs of
experimental set-ups was used by the participants. Table 2 gives an overview of the most
important characteristics of the systems. Except for system #13, which is based on constant
flow rate, all systems work with a constant injection pressure.
5
2.2 Materials
Two different reinforcement textiles were tested:
A biaxial (±45°) glass fiber non-crimp fabric (NCF) from Saertex (X-E-444g/m²)
with a nominal areal weight of 444 g/m² (217 g/m² in +45° and in -45° direction and
additionally 1 g/m2 and 2 g/m² in 0° and 90°, respectively, for stabilization) as well
as 6 g/m² polyester stitching yarn (76 dtex) with a warp pattern at a stitch length of
2.6 mm and a gauge length of 5 mm.
A twill weave (2/2) glass fiber woven fabric (WF) from Hexcel (01102) with a
nominal areal weight of 295 g/m² equally distributed in weft and warp direction.
Nominal construction is 7 yarns/cm in weft and warp direction.
Both fabrics are nominally balanced. The actual construction of the WF is somehow
different in warp and weft direction: ends (warp) count is 7.13 yarns/cm, picks (weft) count
is 7.00 yarns/cm. For the NCF the bundle count is 4.2 bundles/cm in production direction
and 4.1 bundles/cm perpendicular to it. Hence, the 45°-orientation is quite accurate. Table 3
lists further details on the geometry of the textiles. The values were averaged based on the
results of 30 single measurements, whereas the width of the yarns was measured with a
distance of 1.5 mm (WF) and 3.0 mm (NCF) respectively. For both textiles, all participants
in this study received material from the same batch, in order to minimize the influence of
potential manufacturing variations on the benchmark results. Figure 2 shows surface images
of both textiles.
The silicone oil XIAMETER® PMX-200 SILICONE FLUID 100CS supplied by Dow
Corning was used as test fluid for permeability measurement. Its viscosity is approximately
100 mPa∙s at room temperature. In order to minimize possible variations induced by the
fluid viscosity, the silicone oil was procured batch-wise. For each of the ten batches used by
6
different participants, the viscosity was centrally measured at TU Munich in a temperature
range from 15 °C to 40 °C using an Anton Paar MCR 302 rheometer. This silicone oil was
used by all participants except for participant #12, who used polymer solution in water as a
test fluid, as the silicone oil caused problems with their sensors for flow front monitoring.
2.3 Test plan
The participants were asked to perform measurements on the non-crimp fabric (NCF) and
the woven fabric (WF) using the parameters (cavity height, number of layers, injection
pressure, and number of repetitions) specified Table 4. While the number of layers can have
an influence on measured permeability, due to effects of nesting between layers and edge
effects at the fabric-tool interface [18], such influence is assumed to be negligible for this
benchmark, as eight layers or more are used. All measurements were performed at a single
cavity height, but different number of layers to minimize the effort for spacer frame
manufacturing. The specified numbers of layers, the nominal areal weights of the tested
textiles, and the specified cavity heights define the target level of fiber volume contents,
shown in Table 4.
Based on exploratory tests, the target level of fluid injection pressure (Table 4) was specified
in order to avoid possible effects of injection pressure on measured permeabilities. The
definition of target pressure is especially relevant for radial-flow experiments, because the
high pressure gradient may cause fiber displacement at the inlet hole. The specified pressure
values were also chosen to obtain reasonable test times between one and five minutes. No
vacuum was applied at the outlet during the tests.
For each of the two fabric materials, three series of experiments were specified, whereas
each series comprised five experiments at repeatable conditions in order to add statistical
significance to the results.
7
2.4 Data Analysis
Analysis of the raw data acquired in the tests, i.e. fitting of an ellipse to measured points on
the flow front and calculation of permeability values based on the process conditions and the
development of the flow ellipse geometry with time, was performed by each individual
participant for their respective data. Table 5 sums-up the analysis methods employed by the
participants. It is to be noted that all methods used here (Chan/Hwang [19], Adams
Rebenfeld [18, 20-22] and Weitzenboeck et al. [23, 24]) are based on different formulations
of the same approach, transformation of an elliptical flow front shape to an equivalent
isotropic co-ordinate system. However, these methods differ in the particular mathematical
approach chosen for this transformation.
2.5 Sample preparation
For preparation of the test specimens, all participants were asked to follow these pre-defined
steps:
In the first step, the individual fabric layers were cut out of the material at the required size
and shape (determined by each participant’s injection tool geometry) and then stacked at
identical orientation according to Figure 3. The number of layers in each test specimen was
defined by the test plan (Table 4).
In the second step, the inlet hole, a specific characteristic of radial injection tests, was
punched into the stack. A diameter of 12 mm was pre-defined. The accuracy of the inlet hole
diameter is crucial for a precise measurement and therefore special care must be taken when
preparing it. Hence, punching was defined as requirement because cutting the hole may
result in yarns being displaced. Also, punching was performed on the complete stacks,
8
because if the hole is punched into the individual layers, the following stacking process can
lead to offset and therefore insufficient accuracy.
In the third step, each specimen was weighed for calculation of the actual fiber volume
content.
2.6 Evaluation of cavity deformation
The participants were asked to evaluate experimentally the cavity deformation in their
injection tool and the corresponding deviation from the target cavity height of 3.00 mm. For
this purpose, blocks of plasticine, liquid metal filler (metal epoxy), or similar materials were
placed in the tool cavity at five points according to the scheme shown in Figure 4, where P5
is located as close as possible to the inlet. The tool was closed, compressing the material
blocks to a thickness corresponding to the cavity height. After opening of the tool, the
thickness of the material blocks was measured to determine the cavity height. Two cases
were considered. In a first test the empty and non-pressurized cavity was checked. In a
second case the tool was filled with the NCF at the highest tested fiber volume content
(58%) with cutouts for the plasticine. From compression tests it is known that the NCF at a
Vf of 58% results in the highest textile compaction pressure of all tests. It must be noted that
textile compression pressure is the dominant component of overall pressure, because it
easily exceeds the maximum injection pressure of 0.4 MPa and acts on the complete surface.
While the obtained deformations do not fully reflect the actual deformation during the
experiments, they give an good impression of the tendency of a system to deviate from the
target cavity thickness. From the five values per case, an average effective cavity height was
calculated and compared to the target cavity height and the parallelity of the upper and lower
mold halves was assessed. While more complex methods for deformation analysis, exist,
9
this simple approach exhibits a high robustness and is easily manageable for all participants
because it requires no specific technology.
3. Results
Tables 6 and 7 list the main results of all participants. It should be noted that participant #2
acquired permeability data using two different set-ups (see Table 2), while participants #7
and #17 provided two data sets, one as measured, and one with correction for deviations in
cavity height. Furthermore, participant #6 only provided data, in which Vf was corrected
with the actual cavity height, as their system provides an online cavity height measurement
during the test.
Figures 5 and 6 summarize the measured permeability values for the NCF and the WF,
respectively. Each figure contains two diagrams, showing the highest (K1) and lowest (K2)
in-plane permeability value (logarithmic scale) as a function of the Vf. In each diagram, the
blue diamonds, the red squares and the green triangles show the results for the lowest (Vf,1),
intermediate (Vf,2) and highest (Vf,3) target Vf, respectively. Each data point represents the
arithmetic average of the five repeat measurements conducted by each participant. The error
bars represent the standard deviation for permeability and Vf. Deviations from the target Vf
are induced by areal weight variations, which were taken into account by weighing every
specimen and calculating the individual Vf for every test. In these figures, calculation of Vf is
based on target values for the cavity height, not on the actually measured values (except for
participant #6, because an online-correction is implemented in their system).
In the diagrams, almost all data sets lie within a cluster at each nominal Vf. Series #1 and
#19 are exceptions for each data point, i.e. for both textiles, K1 and K2, and for each Vf (data
set #1 is incomplete, if they are not marked in the diagram then there was no data available).
The fact that the relative position of the data sets #1 and #19 to the cluster is constant
10
indicates a systematic deviation. Therefore, these data sets were excluded from further
statistical analysis since they are obvious outliers. On the other hand, this means that 18 of
the 20 data sets are included in the cluster and can be considered for statistical analysis.
Possible effects that may contribute to the observed deviation are discussed in the following
sections.
Each cluster covers a range of about two percentage points of Vf (e.g. 46% to 48%). While
these variations in Vf can have a strong effect on the permeability, there is no correlation of
permeability and Vf within each cluster, indicating that other effects causing variation in the
measured permeability values are dominant.
Within each cluster, the coefficient of variation (cv) of the permeability values was
calculated according to Eq. 2, with
being the standard deviation and
being the
arithmetic average
(2)
The results listed in Table 8 show that the average for the cv was 32.2% and 43.9% for the
NCF and the WF, respectively. The higher average cv for the WF mainly results from K2,
which shows relatively high variation. This might results from the comparably small value,
leading to errors induced by issues measurement resolution for flow front progression.
The target of the benchmarking efforts is to reach a point at which the variation between the
results gained with different systems is no larger than the variation between the results
gained on a single system. On average, the cv for individual data sets was 7.8% and 12.2%
for the NCF and the WF, respectively. Hence, there is further potential for improvement.
This leads to the question which sources of variation can be identified based on the results
and the information provided by the participants. In Section 4, different potential sources are
discussed in detail.
11
In addition to the permeability values K1 and K2, the orientation angle β of K1 relative to the
fiber directions values was determined in the tests. Figure 7 shows the results for both
textiles at each of the three nominal Vf. Each blue line in the graphs represents the β,
averaged out of the five tests for each individual participant. The red dashed line shows the
average of all participants.
Both textiles show an average orientation close to 0°. Yet, there is significant variation
between the participants, especially for the NCF. A possible explanation might be given by
the degree of anisotropy of permeability, which is defined as the ratio of K2 to K1. The closer
this ratio is to one, the more circular the flow front is. As the orientation angle is derived
from the ellipse fitted to the flow front, a near-circular shape increases the influence of
irregularities in the flow front shape. Table 9 lists the degrees of anisotropy for both textiles.
It shows low anisotropy (≥ 0.75) for the NCF. As the relative length of the half-axis
corresponds to
, a degree of anisotropy of 0.75 means that the short half-axis is
only about 13% shorter than the long half-axis. Hence, this might explain the high variation
concerning β. As there are no significant differences in fiber bundle geometry for the two
main fiber directions (see Table 3), the slight anisotropy of the NCF is likely to be caused by
the presence of stitching, oriented in the production direction. On the other side, the WF
shows relatively high anisotropy (≤ 0.21). This presumably results from the deviations from
the balanced fabric construction, which are listed in Table 3 and visible in Figure 2.
In summary, the results show that for radial flow measurements the error in orientation angle
determination increases with decreasing anisotropy. The variability is small when the
anisotropy is high. This seems acceptable, since the relevance of the orientation angle
decreases as the flow front becomes more circular.
Both orientation angle and anisotropy did not show a clear dependence on Vf.
12
4. Discussion
4.1 Influence of cavity deformation
The data listed in Table 8 clearly shows that the variation between the results of the different
institutions increases with increasing Vf. This influence of Vf is presumably related to
increasing cavity deformation resulting from increasing textile compression and also from
increasing injection pressure. While relatively stiff systems remain closer to the target cavity
height of 3.00 mm, the less stiff ones show increasing cavity height, presumably related to
tool deflection. This leads to apparently higher in-plane permeability values. As a result, the
deviations of the permeability data obtained by the participants increase with the level of Vf.
To estimate the influence of deformation in detail, Figure 8 shows the results of the cavity
height measurements. The figure contains a green, dashed line at 2 % deviation showing the
originally proposed acceptable limit for deviation, which was defined based on the guideline
for Benchmark II [7].
The results show that 10 out of 20 systems show a deviation greater than 2% when the
textile is compacted in the cavity. Relating this data to the materials used for the systems
(Table 2), one can see that out of the seven full metal systems two show deviation larger 2%
while this was the case for eight out of thirteen systems which were fully or partially made
of glass or PMMA. This indicates that neither usage of full metal system guarantees
satisfying stiffness, nor usage of glass or PMMA necessarily leads to insufficient stiffness,
although it tends to make systems more prone to deflection. Hence, it can be concluded that
appropriate tool design is the key to minimizing cavity height variations.
The impact of the deformation on the variation of the results becomes clear when only the
10 systems with a deviation from target cavity height smaller than 2% are considered for
statistical analysis: In this case, the average cv reduces to 23% and 34% for the NCF and the
WF, respectively. This gets close to the average cv found in Benchmark II which was
13
approximately 20% for systems with deviation between actual and target cavity height
smaller than 2%. Yet, it has to be noted that Benchmark II was conducted with a different
textile and hence a direct comparison is not possible.
When considering cavity height deviation, it is important to distinguish between parallel and
non-parallel deviation. The latter, which can be caused by a pressure-induced deflection or a
parallelism issue between top and bottom molds, is practically impossible to correct because
it results in a non-uniformly distributed Vf over the sample area and depends, among other
factors, on the compaction behavior of the textile in dry and wet state as well as the fluid
pressure. Both are not constant during the experiment, as the flow front propagates. It must
be noted that set-ups for radial injection tests are quite prone to pressure-induced deflection,
due to the central injection and the resulting high pressure gradient. On the other side a
known (measured) parallel deviation from the target can be accounted for when calculating
the effective Vf. Table 10 shows the normalized non-parallelity factor
for all participants (Eq. 3). It is defined as the ratio of
of the deformation
, the standard deviation of the cavity
height at the five measurement points (with textile in cavity) to
, the arithmetic average of
the five values.
(3)
Some participants observed relatively large average deviations, but only small non-parallel
deformation, so that applying a correction to the fiber volume content was possible (#7 and
#17 in Tables 6 and 7).
4.2 Influence of fluid pressure
The injection pressure to be used in the tests was pre-defined (Table 4) in order to minimize
possible influences of pressure on permeability results. Among the 20 systems used in the
benchmark, 2 provided a pressure sensor located directly at the proportional valve of the
14
pressure vessel, 11 had a sensor somewhere in the feed line between oil reservoir and tool,
and 5 had a tool-mounted sensor. The rest used the nominal pressure value to which the
proportional valve is set for calculation. Two basic possibilities for an influence of fluid
pressure on the calculated permeability are given.
Firstly, the injection pressure itself might influence the permeability: Darcy’s law assumes a
rigid porous media. However, textiles can deform under the fluid pressure. Hence, it could
make a difference which injection pressure is applied during permeability measurement,
especially because of the very strong pressure gradient at the beginning of radial-flow
experiments. All participants were asked to check the samples after testing for fiber-wash
out and send photographs – no remarkable deformation was observed. Yet, if injection
pressure is not properly set, it is presumed that the influence of fiber wash-out at the inlet
can be remarkable since it may cause locally changed porosity and permeability and strong
deviations from the expected pressure distribution. Also, as all participants were asked to
use specific injection pressures, these effects should not contribute to deviations between the
participant’s results. Yet, the actual injection pressures were indeed not fully identical.
Pressure loss in the feed line between the pressure vessel and the tool can cause deviations
of the actual injection pressure from the target pressure set at the vessel. Even if this is taken
into account for permeability calculation by using sensor data captured close to the injection
point, this means the actual injection pressures among the participants varied. Yet, as the
sensor values fairly accurately match the target values, this effect is estimated to be
negligible for this benchmark. One of the 20 systems is based on a constant flow rate
approach instead of constant injection pressure like the others. As the resulting data set was
part of the cluster, this seconds the assumption that influence of injection pressure is
negligible for this benchmark.
15
Secondly, pressure loss in the feed line between the sensor and the injection point can cause
calculation errors. Within the benchmark, feeding line diameters and lengths from pressure
sensor to injection point ranged from 4 mm to 12 mm and 50 mm to 5500 mm, respectively.
At the given injection pressures, this can cause variation between the results. Also,
analytically estimating the pressure loss in the feeding line is quite error-prone [10]. Further
influence might be given by the fact that some participants use a pressure sensor value
averaged over the complete test for permeability calculation, some use the single value of
each time step and some average up all pressure values up to each time step.
All in all, since different influences contribute to variations, the benchmark results do not
allow further statements. This would require tests with a focus on this influence.
4.3 Influence of fluid viscosity
Temperature-dependent viscosity is considered by all participants when calculating
permeability via Darcy´s law. Yet, there may be several sources of variation.
Opposing the assumptions underlying the application of Darcy`s law, differences in
viscosity could have secondary effects on the permeability, e.g. different deformation
behavior of the preform or variations in wetting behavior. However, these influences are
considered to be very small since the viscosity was in the range between 87 mPa∙s and
113 mPa∙s at temperatures between 17.3 °C and 27.4 °C, i.e. the temperature dependence is
weak.
The viscosity for each single test is calculated using a viscosity-temperature function and the
measured temperature. As can be seen in Table 2, participants measure the temperature at
different locations: in the pressure vessel, the feed line, the tool, or in the laboratory. This
can lead to differences between the temperature measured and the actual temperature of the
fluid within the tool. Also, temperature might vary during the test as not all participants have
16
air-conditioning systems that are used during the tests are performed. The data base does not
allow a detailed statement about the influence of this effect.
Regional suppliers were selected for the silicone oil and the viscosity was centrally
measured by Technische Universität München. The participants received the raw data of the
measurements and individually fitted empirical functions to the viscosity-temperature data.
At 23 °C, the measurement results of the batches showed an average variation of 1.7%. Yet,
two participants measured the dynamic viscosity of the silicone oil with their own systems.
The dynamic viscosity values at room temperature derived from the different functions
applied in the benchmark, show a cv of 3.8%. This indicates that additional uncertainty was
induced by the diverse fitting functions.
Participant #1 received oil from a different batch than the other participants and measured
the viscosity using their own equipment. Interestingly, the measured viscosity is the highest
in the benchmark study. This could be an actual difference of this specific batch, but it could
also indicate that there is a systematic difference between the measurements carried out by
TU München and by this participant. While this is speculative, it highlights a very important
issue: Variations between the viscosity measurements performed on different systems will
directly add to variations between permeability data measured at different research sites.
Even though it was tried to exclude influences of the viscosity, it presumably has an effect
that is not negligible. Therefore, efforts for standardization of permeability measurement
need to involve aspects of fluid-induced variations.
4.4 Influence of fluid wetting behavior
Silicone oil, as it was used in this benchmark, is a common substitute for resins in
experimental studies of saturated and unsaturated permeability [8]. The choice of this type
of fluid is based on its viscosity which is comparable to that of liquid epoxy resins used for
17
composite manufacturing, the non-toxicity and the availability. However, so far capillary
effects controlling wetting phenomena and potentially inducing void formation have been
mostly neglected. Silicone oil is a totally dispersive liquid with a very low surface tension
that makes it a totally wetting liquid. Uncured liquid epoxy resin, on the other hand, is a
partially wetting fluid with a very different behaviour, due to its surface tension and
components thereof [25]. It is also impossible to consider a capillary pressure [26] for
silicone oil. This could indicate that viscosity should not be the only parameter relevant to
the choice of a non-reactive fluid as a substitute for liquid resin. Further studies should focus
on the identification or formulation of a physico-chemically reliable test liquid for
permeability measurements.
4.5 Influence of textile variations
Parts of the variation between the participants’ data may be induced by textile variations.
Figure 9 shows the average areal weights measured for the test specimens by each
participant. The error bars show the respective standard deviation. Although all material was
from the same batch, some deviations between the participants exceed the variation for the
single participants. However, the differences are relatively small and they were considered
in calculating Vf. But variations in areal weight also indicate variations in the textile
structure, such as straightening of yarns that would affect the crimp of the WF and would
also influence the permeability.
Textile variation can also be seen on the level of the single yarns and bundles respectively,
as the values listed in Table 3 show. Both textiles show variability in yarn/bundle width of
about 5%. Based on common analytical models this alone can explain about 10% of the
variability in terms of permeability. The cause for these variations may be related to the
18
textile manufacturing process itself, to the rewinding procedures in the context of the
material distribution, or to the individual lay-up and cutting procedures.
4.6 Influence of data analysis
Characterization of textile permeability basically comprises three steps: (1) Acquisition of
relevant sensor data; (2) flow front modeling and allocation of pressure and viscosity values
for each time step (eventually including time-averaging of pressure and viscosity values);
and (3) computation of in-plane permeability data. Compared to linear injection tests, radial
injection tests allow by far more variation in these steps, due to the more complex flow front
shape and accordingly more complex mathematics. The 20 systems compared in the
benchmark differ in terms of type, number and location of sensors for temperature, pressure
and especially flow front monitoring (see Table 2). Accordingly, step (1) and (2) necessarily
differ. Table 5 shows that the algorithms used for step (3) are also different.
The variations induced by the differences in steps (1) and (2) depend on the flow front
shape. For an ideally homogeneous porous media, resulting in a perfectly elliptical flow
front, it does not matter if the ellipse is fitted to several thousand values (optical systems) or
only three (minimum when center is fixed), assuming that the sensor data is reliable. Yet,
imperfections in the textile lead to local variations which can have a strong impact on
measured permeability. This impact increases with decreasing number of sensors.
Inaccuracy of flow front detection also induces variation. The distance of the sensors to the
inlet can have an influence, as the inlet is circular, while the algorithms applied in step (3)
assume that it is of the same shape as the flow front. This causes an error that decreases with
increasing distance of the flow front to the inlet. Hence, it can make a difference where the
sensors are located. This, however, was not examined in detail because the superposition of
different causes of variation does not allow isolating these effects.
19
Concerning the flow front modeling (step 2), two approaches exist for fitting an ellipse
equation describing the flow front to sensor data. As illustrated in Figure 10, the center of
the fitted ellipse is either forced to coincide with the injection point, or the center is allowed
to float. The floating center approach may lead to a better fit. But the algorithms used in step
(3) are based on the assumption that flow spreads radially from the ellipse center and that
the pressure has a maximum at this point. However, this is only true when the center of the
ellipse coincides with the injection point [4]. It is evident that using different strategies can
lead to variations. 4 out of 20 have used the floating center method.
The algorithms used for step 3 are known to show some differences when applied to the
same data [4, 27]. Additional variation can be induced by the strategy with which the
algorithm is applied to the data. Four strategies can be distinguished (see also Ferland et al.
[17]:
Elementary method: One of the permeability calculation algorithms is applied to the data of
each pair of subsequent time steps and allows calculation of the permeability values based
on the differences between the data sets at both time steps (esp. flow front progression).
Hence, for each pair of subsequent time steps permeability values are obtained which can
then be averaged to receive the final measurement values (K1 and K2) of the test.
Reference time step method: As with the elementary method permeability, values are
calculated at each time step using one of the permeability calculation algorithms. Yet, not
the difference to the previous time steps is considered, but always the difference to the very
first time step (or another specific time step).
Single step method: Using one of the permeability calculation algorithms, the permeability
is calculated with the data obtained at two particular time steps (e.g. the first and the last).
Global method: One of the permeability calculation algorithms is applied to the data of all
time steps at once using a fitting procedure.
20
As listed in Table 5, 15 out of 20 stated the usage of global, 2 of elementary, 2 of reference
time step and 1 of single step method. Some researchers had observed that the ellipse
changes direction and shape based on which time step was used to collect the data, which
might be caused by local variations. Such effects can cause a difference between the results
gained with the above mentioned methods.
It is to be expected that significant variations origin from different methods for data analysis.
In order to estimate the magnitude of variations induced by analysis it was decided to
recalculate some of the results using a unified analysis approach. For this, the data sets (fluid
injection pressure, dynamic fluid viscosity, flow front data) originally used in step (2) and
(3) were collected and evaluated according to an uniform procedure: For step (2), the elliptic
paraboloid fitting method introduced by Fauster et al. [9] was applied to all of the collected
data sets, and for step (3), the Adams/Rebenfeld algorithm was used. As the paraboloid
method allows fitting an elliptic paraboloid to the entire set of flow front data acquired
during the radial flow experiments in a single step, it is a global method. Step (2) and (3)
were performed at Montanuniversität Leoben for all collected data sets in order to minimize
any influence related to data processing. This study was an additional offer to the
participants, after the measurement phase of the benchmark study was concluded. Eight data
sets (#2a, #2b, #5, #6, #7, #9, #12, #14, # 18) were recalculated this way.
To evaluate the influence of differences in step (2) and (3) on the final results, the in-plane
permeability characteristics calculated with the individual approach (
,
be compared with those calculated with the unified approach (
,
,
,
) can
). The
relative deviation was calculated for each of the 15 tests for each participant, for both
materials, and for K1, K2 and β (e.g.
). The average deviation and the
corresponding standard deviations (minimum/maximum error bar) for each individual
participant are shown in Figure 11.
21
The deviations for the orientation angle of the NCF are significantly higher than those for
the WF, which corresponds to the high variation of the orientation angle measurements
described above. No clear trend was found for the difference between K1 and K2, neither for
the NCF, nor for the WF. Also the results of the individual approach are not consistently
higher or lower compared to the uniform approach. The average deviation for all the data
shown in the diagram is 20% for K1 and K2 and 2° for β. This presumably corresponds to the
magnitude of variation between the participants which is induced by the analysis. The total
average coefficient of variation for K1 and K2 between the considered data sets is 43% when
the individual approaches are applied and 36% when the uniform approach is applied.
Hence, significant potential for further reduction of variation is possible.
5. Conclusions
The purpose of the presented benchmark exercise was to evaluate the comparability of inplane permeability characteristics obtained using different measurement systems based on
radial flow experiments, and to identify sources of variation. For this purpose, 19
participants with 20 systems measured the permeability of a non-crimp fabric and a woven
fabric.
Averaged over all 12 test cases (highest and lowest in-plane permeability of two textiles at
three levels of nominal Vf), the coefficient of variation (cv) between the permeability values
determined with the different system was 32% and 44% for NCF and WF, respectively. On
the other hand, the average cv for the individual systems was 8% and 12% for NCF and WF
respectively, so the variation between systems is significantly higher than the uncertainty for
a single system. Several causes for this difference were identified, leading to the conclusion
that strategies to minimize differences in permeability values obtained using different
systems will have to focus on these points:
22
-
Cavity deformation is presumably the largest influence and strongly varies among
participants. The results show that cavity deformation can be strongly reduced by
appropriate design of the test set-up.
-
There is no uniform strategy on where to measure injection pressure, which is might
induce variation. A pressure sensor located at the injection gate will provide more
consistency.
-
Any effort to standardize permeability must take into account the methods to determine
viscosity. Uncertainty in determination of the fluid viscosity, fitting of viscositytemperature curves and temperature measurement can induce variation in the magnitude
of several percent. Also it might help to find a model fluid whose viscosity is constant
over a range of temperature and has good wetting properties.
-
Stack-wise measurement of areal weight and calculation of corresponding fiber volume
content should be mandatory to consider areal weight variations.
-
Differences in the methods used for data analysis induce significant variation. A uniform
data analysis tool could be created and used to eliminate the variation caused by the
analysis method.
As a next step, the participants of the benchmark will derive some basic minimum
requirements for permeability measurement systems and procedures (radial flow) from these
results. Subsequently, smaller and topic-focused benchmarks will deal with remaining
questions, e.g. the influence of injection pressure and the best strategy for injection pressure
determination. Together with the first and second international benchmark exercise, the
authors are confident that this will provide sufficient data for definition of guidelines for
permeability measurement.
Acknowledgement
23
It is kindly acknowledged that the textile manufacturing companies Saertex GmbH and
Hexcel Corporation supported the benchmark by supplying the materials free of charge.
Literature
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
Parnas RS, Flynn KM, Dal‐Favero ME. A permeability database for composites
manufacturing. Polymer composites. 1997;18(5):623-33.
Parnas RS, Howard JG, Luce TL, Advani SG. Permeability characterization. 1. A
proposed standard reference fabric for permeability. Polymer Composites.
1995;16(6):429-45.
Lundström TS, Stenberg R, Bergstrom R, Partanen H, Birkeland PA. In-plane
permeability measurements: a nordic round-robin study. Composites Part A: Applied
Science and Manufacturing. 2000;31(1):29-43.
Fauster E, Berg DC, Abliz D, Grössing H, Meiners D, Ziegmann G, et al. Image
processing and data evaluation algorithms for reproducible optical in-plane
permeability characterization by radial flow experiments. Journal of composite
materials. 2017:0021998318780209.
Grössing H, Becker D, Schledjewski R, Mitschang P, Kaufmann S. An evaluation of
the reproducibility of capacitve sensor based in-plane permeability measurements: a
benchmarking study. eXPRESS Polymer Letters. 2015;9(2):129-42.
Arbter R, Beraud J, Binetruy C, Bizet L, Bréard J, Comas-Cardona S, et al.
Experimental determination of the permeability of textiles: a benchmark exercise.
Composites Part A: Applied Science and Manufacturing. 2011;42(9):1157-68.
Alms JB, Correia N, Advani SG, Ruiz E. Experimental procedures to run
longitudinal injections to measure unsaturated permeability of LCM reinforcements.
Permeability Measurement Standard, Permeability Benchmark II, http://cchp meca
polymtl ca/permeabilityBenchmarkII html. 2010.
Vernet N, Ruiz E, Advani S, Alms J, Aubert M, Barburski M, et al. Experimental
determination of the permeability of engineering textiles: Benchmark II. Composites
Part A: Applied Science and Manufacturing. 2014;61:172-84.
Fauster E, Berg DC, May D, Blößl Y, Schledjewski R. Robust evaluation of flow
front data for in-plane permeability characterization by radial flow experiments.
Advanced Manufacturing: Polymer & Composites Science. 2018;4(1):24-40.
Berg D, Fauster E, Abliz D, Grössing H, Meiners D, Schledjewski R, et al. Influence
of test rig configuration and evaluation algorithms on optical radial-flow
permeability measurement: A benchmark exercise. Proc 20th Int Conf on Composite
Materials, Copenhagen, Denmark2015.
Kissinger C, Mitschang P, Neitzel M, Roder G, Haberland R. Continuous on-line
permeability measurement of textile structures. Bridging the Centuries with Sampe's
Materials and Processes Technology, Vol 45, Books 1 and 22000. p. 2089-96.
Aranda S, Berg D, Dickert M, Drechsel M, Ziegmann G. Influence of shear on the
permeability tensor and compaction behaviour of a non-crimp fabric. Composites
Part B: Engineering. 2014;65:158-63.
Swery EE, Allen T, Comas-Cardona S, Govignon Q, Hickey C, Timms J, et al.
Efficient experimental characterisation of the permeability of fibrous textiles. Journal
of composite materials. 2016;50(28):4023-38.
24
[14]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
[22]
[23]
[24]
[25]
[26]
[27]
Louis BM, Di Fratta C, Danzi M, Zogg M, Ermanni P. Improving time effective and
robust techniques for measuring in-plane permeability of fibre preforms for LCM
processing. New material characteristics to cover new applications needs: SEICO
11; SAMPE Europe 32nd International Technical Conference & Forum; March 28th29th, 2011, Paris; proceedings 2011 SAMPE Europe International Conference Paris:
Society for the Advancement of Material and Process Engineering; 2011. p. 204-11.
Alhussein H, Umer R, Rao S, Swery E, Bickerton S, Cantwell W. Characterization
of 3D woven reinforcements for liquid composite molding processes. Journal of
materials science. 2016;51(6):3277-88.
Endruweit A, McGregor P, Long A, Johnson M. Influence of the fabric architecture
on the variations in experimentally determined in-plane permeability values.
Composites Science and Technology. 2006;66(11-12):1778-92.
Ferland P, Guittard D, Trochu F. Concurrent methods for permeability measurement
in resin transfer molding. Polymer composites. 1996;17(1):149-58.
Adams KL, Rebenfeld L. Permeability Characteristics of Multilayer Fiber
Reinforcements .1. Experimental-Observations. Polymer Composites.
1991;12(3):179-85.
Chan AW, Hwang ST. Anisotropic in‐plane permeability of fabric media. Polymer
Engineering & Science. 1991;31(16):1233-9.
Adams KL, Miller B, Rebenfeld L, Hwang ST, Chan AW. Forced Inplane Flow of
an Epoxy-Resin in Fibrous Networks. Polymer Engineering and Science.
1986;26/20(20):1434-41.
Adams KL, Rebenfeld L. Permeability Characteristics of Multilayer Fiber
Reinforcements .2. Theoretical-Model. Polymer Composites. 1991;12(3):186-90.
Adams KL, Russel WB, Rebenfeld L. Radial Penetration of a Viscous-Liquid into a
Planar Anisotropic Porous-Medium. International Journal of Multiphase Flow.
1988;14(2):203-15.
Weitzenböck J, Shenoi R, Wilson P. Radial flow permeability measurement. Part A:
Theory. Composites Part A: Applied Science and Manufacturing. 1999;30(6):78196.
Weitzenböck J, Shenoi R, Wilson P. Radial flow permeability measurement. Part B:
Application. Composites Part A: Applied Science and Manufacturing.
1999;30(6):797-813.
Pucci MF, Liotier P-J, Drapier S. Tensiometric method to reliably assess wetting
properties of single fibers with resins: Validation on cellulosic reinforcements for
composites. Colloids and Surfaces A: Physicochemical and Engineering Aspects.
2017;512:26-33.
Pucci MF, Liotier P-J, Drapier S. Capillary wicking in a fibrous reinforcement–
Orthotropic issues to determine the capillary pressure components. Composites Part
A: Applied Science and Manufacturing. 2015;77:133-41.
Fauster E, Grössing H, Schledjewski R, Materials ESoC. Comparison of algorithms
for 2D anisotropic permeability calculation in terms of uncertainty propagation.
European Society of Composite Materials), Proc 16th Eur Conf of Composite
Materials, Seville, Spain2014.
25
Isometric view:
Cross-section view:
90
Injection point & hole
Upper mold
Preform
β
Lower mold
Orientation angle of K1
0
Flow front (tracked)
Figure 1: Schematic illustration of the radial injection approach which is the focus of this benchmark exercise.
Figure 2: Images of the textiles characterized in this benchmark study.
26
0 - production direction
Layer 1
Layer 2
Layer 3
Layer 4
Layer 5
…
0 - production direction
0 - production direction
…
0 - production direction
0 - production direction
0 - production direction
Figure 3: Cutting and stacking of the samples – all layers have identical orientation.
Rectangular sample
P1
W/4
P4
L/4
W/2
P5
P3
D/4
Point of
injection
P2
P5 P1
P3
P4
D/4
P2
W = total sample width
Circular sample
W/4
Point of
injection
D/4
L/4
L/2
D/4
D/2
L = total sample length
D = total sample diameter
Figure 4: Positions for measurement of the actual cavity height.
27
NCF - K1
#19
1E-09
#19
#1
#19
#1
#1
K1 in m²
1E-10
1E-11
1E-12
1E-13
44%
50%
56%
62%
Vf
NCF - K2
#19
1E-09
#1
#19
#1
#19
#1
K2 in m²
1E-10
1E-11
1E-12
1E-13
44%
50%
56%
Vf
results for Vf,1, Vf,2 and Vf,3 respectively
Figure 5: Permeability results for the NCF.
28
62%
WF - K1
1E-09
#19
#19
#19
K1 in m²
1E-10
1E-11
1E-12
1E-13
44%
50%
56%
62%
Vf
WF - K2
1E-09
#19
#19
#19
K2 in m²
1E-10
1E-11
1E-12
1E-13
44%
50%
56%
Vf
results for Vf,1, Vf,2 and Vf,3 respectively
Figure 6: Permeability results for the WF.
29
62%
NCF orientation angles
Vf,1 = 46.4%
90
Vf,2 = 52.2%
90
90
Max: 41
Vf,3 = 58.0%
Max: 38
Max: 31
0
0
0
Min: - 29
Min: - 33
Min: - 42
WF orientation angles
Vf,1 = 46.3%
Vf,2 = 50.1%
90
90
Vf,3 = 54.0%
90
Max: 9
Max: 4
Max: 7
0
0
0
Min: - 4
Min: - 11
Min: - 14
Figure 7: Average orientation angles measured by the participants for the different target fiber volume
12%
10%
8%
6%
4%
2%
0%
-2%
11%
13
10/52
Guideline
limit
Averaged
deviation from
target thickness
contents; the dashed line shows the total average over all participants.
1
2a 2b
3
4
5
6
7
8 9 10 11 12 13 14 15 16 17 18 19
Participants #
Without textile in cavity
With textile in cavity
Figure 8: Experimentally determined deviation of the cavity height from the target height.
30
Areal weight in g/m²
Participants average areal weights for the NCF (Saertex)
465
460
455
Total average value
450
445
440
435
1
2a
2b
3
5
6
7
8
9 10 11
Participants #
12
13
14
15
16
17
18
19
15
16
17
18
19
Participants average areal weights for the WF (Hexcel)
310
Areal weight in g/m²
4
305
300
295
Total average value
290
285
280
1
2a
2b
3
4
5
6
7
8
9 10 11
Participants #
12
13
14
Figure 9: Average areal weights measured for test specimens.
Fixed center
Floating center
Point of
injection
Point of
injection
Figure 10: Basic strategies for ellipse modeling – fixed center (left) and floating center (right).
31
Figure 11: Averaged deviations between results obtained using a unified and individual approach, respectively.
32
Table 1: List of participants.
Participant
Institution
1
National University of Mar del
Plata
2
Montanuniversität Leoben
3
Institut de Soudure –
Composite Platform
4
IMT Lille Douai
Department of Polymers and
Composites Technology &
Mechanical Engineering
France
5
Institut für Verbundwerkstoffe
GmbH
Manufacturing Science
Germany
6
TU Clausthal
7
8
Technical University Munich
University of Stuttgart
9
University of Auckland
12
Institute of Science and
Innovation in Mechanical and
Industrial Engineering
Skolkovo Institute of Science
and Technology
ITAINNOVA
13
ETH Zurich
14
Koc University
10
11
16
17
Khalifa University of Science
and Technology
National Physical Laboratory
Nottingham University
18
Brigham Young University
19
Purdue University
15
Department
Institute of Material Science
and Technology
Processing of Composites
Group
Country
Argentina
Austria
France
Institute of Polymer Materials
and Polymer Technology
Chair of Carbon Composites
Institute of Aircraft Design
Centre for Advanced
Composite Materials
Composite Materials and
Structures Group
Center for Design,
Manufacturing and Materials
Materials and Components
Department of Mechanical and
Process Engineering
Department of Mechanical
Engineering
Department of Aerospace
Enigneering
Materials Division
Faculty of Engineering
Faculty of Manufacturing
Engineering Technology
Composites Manufacturing &
Simulation Center
33
Germany
Germany
Germany
New Zealand
Portugal
Russia
Spain
Switzerland
Turkey
UAE
UK
UK
USA
USA
Table 2: Details of the individual set-ups.
Participant #
Sample
size in
mm²
Length
to
width
ratio
Tool material
(top/bottom)
Flow
detection
Monitoring
injection
pressure1
1
51,129
1:1
glass/metal
optical
none
pressure
vessel
2a3
112,687
3:4
optical
feed line
feed line
[9,10]
2b3
215,812
1:1
glass+steel
reinforcement/steel
aluminum/aluminum
prop. valve
tool
[5]
3
95,586
1:1
aluminum/aluminum
tool
room
4
5
160,000
215,568
1:1
1:1
feed line
None
room
tool
[5,11]
6
62,387
1:1
optical
feed line
feed line
[10,12]
7
78,287
circular
steel/PMMA
aluminum/aluminum
glass+metal
reinforcement/metal
glass/aluminum
capacitive
pressure
(6 sensors)
optical
capacitive
optical
feed line
8
89,887
1:1
PMMA/aluminum
optical
feed line
9
72,900
1:1
glass/aluminum
feed line
[13]
10
80,384
circular
metal/metal
tool
tool
11
7,741
circular
PMMA/PMMA
feed line
feed line
12
107,187
3:4
steel/steel
feed line
tool
13
193,487
1:1
tool
n/a4
14
72,787
1:1
optical
feed line
feed line
15
31,303
circular
optical
feed line
feed line
16
89,887
1:1
metal/metal
glass-aluminum
sandwich/aluminum
metal/glass
glass+aluminum
reinforcement/metal
optical
pressure
(64
sensors)
optical
dielectrical
(22
sensors)
pressure
feed line
pressure
vessel + tool
tool
optical
tool
tool
17
125,551
circular
aluminium/aluminum
18
22,387
1:1
19
40,000
1:1
PMMA/PMMA
pressure
(6 sensors)
optical
prop. valve
PMMA/PMMA
optical
feed line
Monitoring
temperature2
pressure
vessel
Tool
pressure
vessel
tool
1,2
refers to the location of the sensor from which the values are used for
calculation
3
Participant #2 participated in the benchmark with 2 different systems
4
This was the only system in the benchmark working with constant flow rate
instead of constant injection pressure
Table 3: Variations in textile geometry.
Non-crimp
fabric
Woven fabric
Bundle width ( )
Stitches distance
Bundle gap
Warp width (
34
1.66
5.23
0.32
1.21
±
±
±
±
0.08
0.04
0.01
0.05
mm
mm
mm
mm
Literature
[14]
[15]
[16]
Weft width (
Warp gap
Table 4: Test plan.
Test series
NCF – Vf,1
NCF – Vf,2
NCF – Vf,3
WF – Vf,1
WF – Vf,2
WF – Vf,3
Material
No. of layers
1.47
0.25
±
±
0.05
0.06
NCF
8
9
10
Fiber
volume
content
(Vf) in %
46.4
52.2
58.0
Injection
pressure
(gauge) in
MPa
0.1
0.2
0.4
WF
12
13
14
46.3
50.1
54.0
0.1
0.2
0.4
mm
mm
No. of
repeats
Cavity
height in
mm
5
3.00
Table 5: Analysis methods implemented by the participants.
Participant #
1
2a
2b
3
4
5
6
7
8
Algorithm for
analysis1
Evaluation
method2
Ellipse-fitting
centered3
Method of pressure
consideration4
Chan/Hwang
Adams/Rebenfeld
Adams/Rebenfeld
n/a
Chan/Hwang
Adams/Rebenfeld
Modified5
Chan/Hwang
Adams/Rebenfeld
Adams/Rebenfeld
Global Method
Global Method
Global Method
Single Step Method
Global Method
Global Method
yes
yes
yes
no
yes
yes
average
average
average
target
average
target
Global Method
yes
average / single value
Elementary Method
yes
average
Global Method
no
average
Reference Time
9
Weitzenböck et al.
yes
single value
Step Method
10
Adams/Rebenfeld
Global Method
yes
average
11
Chan/Hwang
Global Method
no
average
12
n/a
Global Method
yes
average
13
Chan/Hwang
Elementary Method
yes
single value
14
Weitzenböck et al.
Global Method
yes
average
Reference Time
15
Weitzenböck et al.
no
average
Step Method
16
Weitzenböck et al.
Global Method
yes
average
17
Weitzenböck et al.
Elementary Method
yes
average
18
Adams/Rebenfeld
Global Method
yes
average
19
Chan/Hwang
Global Method
yes
target
1
for detailed explanation we refer to these publications: Chan/Wang: [19]; Adams/Rebenfeld: [18; 20-22];
Weitzenböck et al. [23, 24]
2
for detailed explanation we refer to Ferland et al. [17]
3
When fitting an ellipse to the flow data there are two possibilites: Either fix the ellipse-center to the
injection point (yes) or to allow the location of the ellipse center to deviate from the injection point (no)
4
Refers to the way how injection pressure is considered in permeability calculation Average: All
captured values are averaged; single value: for every time step the currently captured pressure value is
considered; target: the target injection pressure is assumed
5
Chan model was modified to correct an error: the inlet radius is stated to be falsely transformed, Fauster et
35
al. [4]
Table 6: Results for the permeability for the NCF.
Participant #
1
2a
2b
3
41
5
6
7a
7b
8
9
10
11
12
Vf,1
Vf,2
Vf,3
Vf,1
Vf,2
Vf,3
Vf,1
Vf,2
Vf,3
Vf,1
Vf,2
Vf,3
Vf,1
Vf,2
Vf,3
Vf,1
Vf,2
Vf,3
Vf,1
Vf,2
Vf,3
Vf,1
Vf,2
Vf,3
Vf,1
Vf,2
Vf,3
Vf,1
Vf,2
Vf,3
Vf,1
Vf,2
Vf,3
Vf,1
Vf,2
Vf,3
Vf,1
Vf,2
Vf,3
Vf,1
fiber volume
content in % (±cv)
46.1 (±1.4%)
51.2 (±1.3%)
58.8 (±0.5%)
47.0 (±0.1%)
52.9(±0.1%)
58.7(±0.2%)
47.0 (±0.1%)
52.9(±0.1%)
58.7(±0.0%)
47.3(±0.1%)
53.2(±0.1%)
59.2(±0.1%)
47.0(±0.2%)
52.7(±0.1%)
n/a
47.0(±0.1%)
53.0(±0.1%)
58.8(±0.1%)
46.4(±0.4%)
52.3(±0.2%)
58.0(±0.4%)
47.0(±0.3%)
52.9(±0.7%)
58.8(±0.8%)
44.2(±0.3%)
50.2(±0.7%)
54.9(±0.8%)
47.0(±0.1%)
53.1(±0.2%)
58.9(±0.1%)
47.3(±0.2%)
53.2(±0.1%)
59.1(±0.2%)
47.4(±0.3%)
53.2(±0.1%)
59.3(±0.3%)
48.0(±0.3%)
54.0(±0.3%)
60.1(±0.5%)
47.1(±0.5%)
K1
in 10-11 m2 (±cv)
K2
in 10-11 m2 (±cv)
Orientation
angle of K1
44.2 (±13.8%)
24.0 (±15.8%)
17.7 (±3.3%)
6.32(±5.9%)
3.77(±4.8%)
2.16 (±2.0%)
4.80 (±11.7%)
3.14(±1.9%)
1.39(±5.8%)
5.82 (±5.8%)
3.08(±3.2%)
1.68(±5.8%)
3.79(±3.4%)
3.08(±8.0%)
n/a
6.78(±4.2%)
3.62(±5.0%)
1.59(±8.6%)
5.75(±4.5%)
3.07(±10.6%)
1.36(±5.3%)
2.72(±7.9%)
1.53(±8.6%)
0.996(±10.4%)
2.72(±7.9%)
1.53(±8.6%)
0.996(±10.4%)
4.27(±9.1%)
2.34(±3.8%)
1.34(±7.4%)
6.55(±7.0%)
3.52(±5.4%)
1.68(±6.7%)
2.48(±11.6%)
1.65(±10.9%)
1.17(±24.0%)
5.98(±2.5%)
2.75(±5.4%)
0.953(±6.7%)
7.24(±7.3%)
38.0 (±15.8%)
21.5 (±17.8%)
13.6 (±7.5%)
4.85 (±3.0%)
2.81(±6.4%)
1.56 (±4.2%)
3.70 (±11.6%)
2.30(±4.0%)
1.02(±4.9%)
5.32(±6.1%)
3.02(±3.6%)
1.58(±10.4%)
2.99(±3.6%)
2.43(±9.7%)
n/a
5.21(±4.3%)
2.70(±7.4%)
1.12 (±9.1%)
4.63(±3.5%)
2.39(±9.3%)
1.00(±7.6%)
1.95(±9.2%)
1.09(±10.2%)
0.663(±6.7%)
1.95(±9.2%)
1.09(±10.2%)
0.663(±6.7%)
3.50 (±7.2%)
1.88(±4.7%)
1.03(±11.0%)
4.61(±11.4%)
2.46(±10.5%)
1.04(±7.9%)
1.92(±12.0%)
1.26(±5.6%)
0.832(±18.9%)
5.10(±4.2%)
2.18(±5.3%)
0.75(±7.9%)
5.53(±7.8%)
6.6
-33.0
-29.0
36
13.0
4.6
3.9
-4.97
-0.24
6.12
n/a
n/a
n/a
-3.6
-4.1
n/a
1.8
3.4
4.1
8.5
10.1
7.4
11.5
7.4
13.1
11.5
7.4
13.1
9.6
8.7
11.8
-26.3
-27.6
-26.7
29.2
20.5
29.7
20.0
18.5
13.0
-17.7
53.1(±0.5%)
Vf,2
4.22(±1.7%)
3.11(±7.1%)
-19.4
59.0(±0.6%)
Vf,3
1.94(±14.2%)
1.32(±13.2%)
-23.7
13
46.6(±1.4%)
Vf,1
5.18(±15.3%)
4.20(±14.7%)
22.8
52.8(±0.3%)
Vf,2
2.86(±5.6%)
2.41(±7.4%)
10.4
58.7(±0.1%)
Vf,3
1.56(±13.2%)
1.32(±7.4%)
4.0
14
46.6(±0.3%)
Vf,1
7.26(±3.9%)
5.65(±4.0%)
8.1
52.5(±0.1%)
Vf,2
5.59(±24.5%)
4.17(±29.5%)
9.3
58.3(±0.1%)
Vf,3
3.25(±6.3%)
2.39(±9.3%)
5.1
15
47.3(±0.1%)
Vf,1
8.11(±7.0%)
5.87(±6.8%)
16.0
53.2(±0.1%)
Vf,2
3.08 (±6.2%)
2.10(±5.9%)
13.8
59.2(±0.1%)
Vf,3
1.57(±4.3%)
1.02(±6.6%)
16.0
16
47.4(±0.1%)
Vf,1
5.46(±5.2%)
4.30(±4.7%)
-3.1
53.3(±0.1%)
Vf,2
2.89(±3.7%)
2.36(±3.1%)
9.7
59.2(±0.1%)
Vf,3
1.81(±8.5%)
1.60(±9.6%)
7.2
17a
47.4(±0.1%)
Vf,1
6.75(±5.7%)
5.12(±5.2%)
-17.2
53.3(±0.1%)
Vf,2
4.39(±3.0%)
3.30(±3.0%)
-14.2
59.3(±0.1%)
Vf,3
3.20(±2.7%)
2.40(±10.3%)
-18.2
17b
46.6(±0.1%)
Vf,1
6.85(±5.7%)
5.20(±5.2%)
-17.2
50.2(±0.1%)
Vf,2
4.68(±3.0%)
3.52(±3.0%)
-14.2
52.6(±0.1%)
Vf,3
3.73(±2.7%)
2.79(±10.2%)
-18.2
18
47.0(±0.9%)
Vf,1
3.97(7.8±%)
3.05(±8.9%)
-42.4
52.7(±0.3%)
Vf,2
2.04(9.7±%)
1.49(±13.0%)
31.3
58.6(±0.3%)
Vf,3
1.06(8.7±%)
0.758(±10.8%)
35.0
19
48.4(±0.3%)
Vf,1
196(29.0±%)
149(±25.0%)
40.7
54.4(±0.2%)
Vf,2
111(11.3±%)
91.7(±9.5%)
29.2
60.5(±0.2%)
Vf,3
74.9(12.2±%)
62.2(±20.4%)
38.0
1
These values have been revised after first results presentation at FPCM14 as a unit conversion error was
identified in the analysis software of the participant.
Table 7: Results for the permeability for the WF.
Participant #
1
2a
2b
31
42
Vf,1
Vf,2
Vf,3
Vf,1
Vf,2
Vf,3
Vf,1
Vf,2
Vf,3
Vf,1
Vf,2
Vf,3
Vf,1
fiber volume
content in % (±cv)
K1
in 10-11 m2 (±cv)
K2
in 10-11 m2 (±cv)
Orientation angle
of K1
n/a
n/a
n/a
45.5(±0.2%)
49.3(±0.1%)
53.0(±0.1%)
45.7(±0.0%)
49.5(±0.1%)
53.1(±0.1%)
45.6(±0.1%)
49.4(±0.1%)
53.2(±0.1%)
46.1(±0.1%)
n/a
n/a
n/a
5.28(±13.0%)
3.87(±21.0%)
2.86(±15.5%)
4.85(±8.1%)
2.83(±2.7%)
1.86(±17.8%)
7.24(±10.4%)
5.00(±10.1%)
2.64(±7.8%)
3.59(±10.8%)
n/a
n/a
n/a
0.993(±9.9%)
0.554(±22.7%)
0.348(±20.8%)
1.02(±23.1%)
0.422(±11.2%)
0.223(±21.0%)
1.32(±5.7%)
0.826(±12.7%)
0.424(±23.1%)
0.507(±12.9%)
n/a
n/a
n/a
0.79
1.23
0.91
-0.41
-0.31
0.94
n/a
n/a
n/a
-0.12
37
5
6
7a
7b
8
9
10
11
12
13
14
15
16
17a
17b
Vf,2
Vf,3
Vf,1
Vf,2
Vf,3
Vf,1
Vf,2
Vf,3
Vf,1
Vf,2
Vf,3
Vf,1
Vf,2
Vf,3
Vf,1
Vf,2
Vf,3
Vf,1
Vf,2
Vf,3
Vf,1
Vf,2
Vf,3
Vf,1
Vf,2
Vf,3
Vf,1
Vf,2
Vf,3
Vf,1
Vf,2
Vf,3
Vf,1
Vf,2
Vf,3
Vf,1
Vf,2
Vf,3
Vf,1
Vf,2
Vf,3
Vf,1
Vf,2
Vf,3
Vf,1
Vf,2
Vf,3
50.0(±0.1%)
n/a
45.4(±0.1%)
49.3(±0.2%)
53.1(±0.2%)
45.3(±0.3%)
49.2(±0.2%)
52.7(±0.7%)
45.7(±0.6%)
49.5(±0.4%)
53.3(±0.2%)
43.7(±0.6%)
46.3(±0.4%)
49.8(±0.2%)
45.6(±0.0%)
49.7(±0.1%)
53.3(±0.0%)
45.2(±0.1%)
49.1(±0.1%)
52.9(±0.1%)
46.0(±0.1%)
49.3(±0.3%)
53.5(±0.2%)
46.3(±0.3%)
50.3(±0.1%)
54.3(±0.3%)
45.3(±0.2%)
49.1(±0.1%)
53.1(±0.3%)
45.2(±0.1%)
48.8(±0.1%)
52.7(±0.2%)
46.0(±0.3%)
49.6(±0.4%)
53.6(±0.2%)
46.1(±0.1%)
49.8(±0.1%)
53.5(±0.2%)
46.0(±0.0%)
49.7(±0.0%)
53.7(±0.1%)
46.1(±0.0%)
49.9(±0.0%)
53.8(±0.0%)
45.9(±0.0%)
48.3(±0.0%)
49.9(±0.0%)
3.17(±10.1%)
n/a
7.26(±4.2%)
4.52(±8.8%)
2.64(±6.6%)
6.35(±10.4%)
4.03(±15.3%)
2.69(±6.2%)
3.94(±13.6%)
2.90(±12.2%)
2.26(±10.7%)
3.94(±13.6%)
2.90(±12.2%)
2.26(±10.7%)
5.36(±7.3%)
3.10(±7.5%)
1.83(±8.1%)
7.88(±30.5%)
7.09(±4.4%)
4.67(±26.5%)
2.97(±21.7%)
2.33(±7.3%)
1.38(±10.4%)
5.76(±13.7%)
3.39(±9.6%)
2.53(±19.3%)
6.64(±16.0%)
5.80(±4.0%)
4.16(±12.2%)
6.80(±8.9%)
4.20(±9.8%)
2.62(±4.9%)
10.7(±6.8%)
4.59(±6.8%)
3.57(±15.9%)
7.24(±10.1%)
4.29(±10.5%)
3.21(±8.8%)
7.19(±5.4%)
4.94(±6.8%)
3.24(±4.5%)
5.94(±3.9%)
5.32(±9.2%)
4.45(±5.4%)
5.97(±3.9%)
5.50(±9.2%)
4.83(±5.5%)
38
0.444(±16.7%)
n/a
1.05(±7.3%)
0.524(±8.9%)
0.279(±19.5%)
1.29(±10.3%)
0.616(±17.0%)
0.374(±16.3%)
0.389(±13.2%)
0.260(±7.9%)
0.176(±19.6%)
0.389(±13.2%)
0.260(±7.9%)
0.176(±19.6%)
1.08(±6.2%)
0.543(±14.9%)
0.290(±2.0%)
2.32(±28.3%)
1.52(±11.1%)
0.897(±40.0%)
0.506(±13.1%)
0.363(±9.7%)
0.211(±14.1%)
1.41(±16.5%)
0.852(±9.0%)
0.520(±12.6%)
1.70(±17.2%)
1.26(±2.4%)
0.907(±14.0%)
3.11(±16.8%)
1.64(±13.1%)
0.714(±10.6%)
1.23(±5.8%)
0.341(±17.8%)
0.304(±50.2%)
1.23(±9.9%)
0.535(±8.9%)
0.315(±14.0%)
1.16(±2.6%)
0.698(±9.9%)
0.601(±4.0%)
1.29(±5.9%)
0.926(±7.4%)
0.838(±9.8%)
1.30(±5.9%)
0.957(±7.4%)
0.909(±9.8%)
-0.67
n/a
-0.90
0.17
0.88
-0.78
-1.17
-0.01
0.36
-0.12
0.77
0.36
-0.12
0.77
0.53
1.29
1.08
-13.68
0.25
1.92
3.96
3.60
2.56
5.09
1.95
3.99
0.99
-0.03
1.30
4.29
0.87
0.70
0.96
6.59
4.38
7.21
6.01
2.39
-11.45
-9.66
-1.26
0.40
1.80
2.00
0.40
1.80
2.00
18
Vf,1
45.4(±0.2%)
2.57(±13.3%)
0.799(±15.7%)
-2.69
Vf,2
48.8(±0.2%)
1.37(±6.7%)
0.311(±10.1%)
-3.45
Vf,3
52.6(±0.1%)
0.745(±10.6%)
0.148(±9.9%)
-3.66
19
Vf,1
47.5(±0.3%)
133(±18.7%)
30.8(±59.4%)
8.52
Vf,2
51.1(±0.4%)
108(±25.%)
30.2(±79.4%)
-10.54
Vf,3
54.9(±0.7%)
126(±20.6%)
23.6(±17.4%)
0.66
1
These values have been revised after first results presentation at FPCM14 as a data transfer error was
identified by the participant.
2
These values have been revised after first results presentation at FPCM14 as a unit conversion error was
identified in the analysis software of the participant.
Table 8: Coefficients of variation based on all data sets for the in-plane permeability values.
NCF
WF
K1
Vf,1
Vf,2
Vf,3
27.9%
30.4%
38.1%
32.2%
32.2%
Average
K2
K1
K2
27.3%
29.9%
39.6%
32.3%
32.0%
32.6%
36.5%
33.7%
50.4%
56.2%
55.7%
54.1%
43.9%
38.3%
Table 9: Anisotropy values averaged over data of all participants and corresponding cv.
NCF
WF
K2/K1
cv
K2/K1
cv
0.78
0.78
0.75
6%
9%
10%
0.21
0.18
0.16
37%
39%
31%
Vf,1
Vf,2
Vf,3
Table 10: Normalized non-parallel cavity deformation.
Participant
in %
Participant
in %
1
2a
2b
3
4
5
6
7
8
9
1.1
10
2.2
11
0.5
12
0.3
13
1.9
14
0.4
15
0.3
16
0.4
17
0.7
18
0.3
19
0.5
0.1
0.3
4.7
0.6
0.7
0.0
1.0
0.6
7.7
39