Key Engineering Materials Online: 2013-06-13

ISSN: 1662-9795, Vols. 554-557, pp 465-471

doi:10.4028/www.scientific.net/KEM.554-557.465
© 2013 Trans Tech Publications Ltd, Switzerland

Validation of Macroscopic Forming Simulations

of a Unidirectional Pre-impregnated Material
Through Optical Measurements

Alexane Margossian1,a, François Dumont1,b, Uwe Beier1,c

1
Eurocopter Deutschland GmbH, 81663 Ottobrunn, Germany
a
[email protected], [email protected],
c
[email protected]

Keywords: Validation, measurement, simulation, pre-impregnated, hot forming.

Abstract. Presenting interesting aspects such as a high stiffness- and strength-to-weight ratio,
Carbon Fibre Reinforced Plastic components are frequently used in the aerospace industry. The
forming step, which conforms the reinforcement to a specific geometry, is a sensitive phase of the
manufacturing process. In order to detect the occurrence of defects prior to any trial, forming
methods are often simulated via finite element software. The presented work details the simulation
validation of a double curved helicopter frame forming out of a unidirectional carbon fibre pre-
impregnated material (M21E, Hexcel®). The finite element model is based on an explicit approach
at a macroscopic level and developed via the software Visual-Crash PAM (ESI®) [1]. The validation
was carried out on six different preforms. Measurements of the top layers were performed by an
enhanced version of a 4D-measuring system, originally developed for non-woven fabric [2], able to
make reproducible photographic and height measurements. Experimental results were then
compared to simulated ones. Due to material specificities, the photo quality reached for non-crimp
fabrics could not be achieved [2]. After hardware and software modifications, measurements and
analyses were eventually successfully completed. The validation of the simulation reached an
accuracy of 1° to 3° depending on the geometrical features of the preform.

Introduction
While new generation airplanes shall improve their onboard payload capacities due to an
expected grow of passengers in the next 20 years, the establishment of new standards to protect the
environment forces the aerospace industry to reduce the ecological footprint of the aircrafts [3, 4].
Following, new types of materials were developed to fill the demand and comply with the new
regulations. In particular, components manufactured with thermoset pre-impregnated materials
enable significant weight reduction, and so important fuel saving. These are reserved for high
performance applications, such as airframe structures [5].
For time and cost purposes, the need to develop a tool able to assess the manufacturability of a
component was early identified. In this direction, simulation techniques showed good alternatives to
trial by error approaches. For example, the modelling of the forming step helped to predict the
formability of a material over a tooling using a certain technique. If geometric methods were first
considered, finite elements programs are nowadays more extensively used since they offer a better
accuracy by taking into account parameters related to the process and the material [6, 7].
The objective of this paper is to describe the validation of the forming simulation of a double
curved helicopter frame out of a unidirectional carbon fibre pre-impregnated material. Experimental
preforms were formed by a double diaphragm machine, while a macroscopic finite element model
was developed in parallel. The validation was carried out through the comparison of both simulated
and experimental fibre orientations [2].

All rights reserved. No part of contents of this paper may be reproduced or transmitted in any form or by any means without the written permission of Trans
Tech Publications Ltd, www.scientific.net. (#615672428, Universitatsbibliothek Stuttgart, Stuttgart, Germany-19/04/23,12:42:55)
466 The Current State-of-the-Art on Material Forming

Forming process
Double diaphragm process. The double diaphragm process is an automated forming process
developed to form difficult tooling such as double curved geometries. As a high reproducible
process, it is a good candidate for the validation of a new material model.
The main steps of a forming per double diaphragms are shown Fig. 1. The machine is composed
of two deformable membranes, clamped to a movable frame, positioned on top of the tool. The to-
be-formed composite layup is placed in between these diaphragms. To ensure a good
membranes/composite contact a vacuum is applied between both diaphragms. Since then, the
composite and both membranes behave as a single element. In order to increase the
composite/membranes slippage, protective foils are often placed around the composite.

Figure 1: Essential steps of the double diaphragm forming process

As a next step, the composite is activated by the mean of infra-red heaters. Since the activation of
the material is performed prior to forming, this process is also called hot forming. The composite,
under vacuum between both membranes, is then lowered until the lower membrane enters in contact
with the tool. Thereafter, a second vacuum between the membrane and the tooling is applied. At this
step, the forming occurs. The cooling phase starts while the composite is still under vacuum by the
mean of fans. Once the cooling phase is terminated, the in-membrane vacuum is turned off so that
the higher membrane can be moved up. Subsequently, the preform can be turned out of the tooling.

Application structure. The application structure selected to support the validation process is a
generic side frame called NEXCIOS, Next Generation of Integrated Composite Aerostructures, used
in the development of helicopters and airplanes structures (see Fig. 2) [8].

Figure 2: NEXICOS – A generic side frame

In the first place, this tooling was chosen for its interesting draping aspects. Indeed, its small radii
section and global double curved shape are not easy drapable geometries. Its dimensions are
approximately of 1400x250x80mm in term of length, width and height.
This application structure has also the advantage to benefit experience from previous studies.
Indeed several works were already performed using this particular shape [9]. Therefore, valuable
time was saved during the validation process in the sense that the initialization of the 4D measuring
system used to measure the fibre orientation of the preforms did not need to be reprogrammed since
geometrical parameters were already known.
 Key Engineering Materials Vols. 554-557 467

Material. The material simulated and validated in this work is the unidirectional pre-impregnated
epoxy material, commercially available under the reference M21E (Hexcel®). Involved, for
example, in the primary structure of the new Airbus A350 XWB, this material presents a high
volume fibre fraction and as such is reserved for high performance applications.

Table 1: Outline of the manufactured used for the validation

Two Plies Six Plies Eight Plies
[0/0] [60/-60/0]S [90/0/-45/45]S
[0/90]
[-45/+45]

Note: The orientations of the material with regard to the tooling are not given with respect to the
tooling geometry but to the global horizontal direction instead.

The validation was carried out through the forming of several layups (see Table 1). Given that a
unidirectional material was used, single-ply layup could not be formed. Thus, the validation started
with two-ply layups. Then, to investigate the robustness of the simulation for multi-ply preforms for
later industrial applications, both six- and eight-ply layups were also formed via hot forming.

Forming simulation
Finite element simulation. In contrast to kinematic methods using mapping techniques based on
the fish-net algorithm, finite elements programs require more than geometrical input to compute.
Parameters related to the forming process, e.g. pressures, and the material, e.g. fibre orientations, are
also needed. This enables to increase the accuracy of the output, even if it significantly raises the
computation time [6, 7]. In particular, the post-processing capacities are extended so that the type of
defects caused by the forming process can be identified and then analysed.
The implementation of the hot forming process was achieved using the commercially available
finite element Visual-Crash PAM from ESI®. Solving was performed by the explicit PAM-FORM
2G Solver (2011).

Material modelling. The unidirectional pre-impregnated material M21-E was modelled by a

thermo-visco-elastic matrix with elastic fibre formulation applied to shell elements, referenced as
MAT 140.

Figure 3: Material card 140 description [10]

This material type is standard to model composite materials with Visual-Crash PAM. Indeed,
MAT 140 was already used for the implementation of Non-Crimped Fabric [1].
468 The Current State-of-the-Art on Material Forming

With this material type, fibres are represented by a simple linear elastic spring model. Axial,
shear and bending stiffness terms are independent from each other. The elastic modulus of the
reinforcement was defined by a constant value. The resin of the pre-impregnated material was
modelled by a Maxwell model consisting of a spring and a dashpot in a row. The in-plane resin
viscosity was defined by a constant value. Finally, as it can be seen Fig. 3, to increase the stability of
the numerical model, a parent sheet was introduced by the developer. [11]

Validation method
The validation intents to verify the compatibility of the forming simulation output with
experimental results. As the only reliable achievable measurement, the orientation of the fibres was
chosen as mean of validation. Currently in use at Eurocopter Deutschland GmbH, this method was
already used for the validation of Non-Crimped Fabric simulations [2].

Optical measurement system. The measurements of the fibre orientation were performed via a 4D
measuring system. This automated tool makes reproducible greyscale photos as well as height and
position measurements. It is composed of a 2D-camera system and a 3D-laser (see Fig. 4). The role
of the laser is essential since it enables a good placement of the pictures. Indeed, a single photo of
the entire part can not be taken due to the size and the shape of the component. Instead, several
photos are taken and then assemble aside each other to re-create the part geometry. The preform is
placed on a rotated axis that enables measurements of every side (frames and web).

Optical measurement system

(camera + laser)

Preform

Figure 4: 4D measuring system currently in use at Eurocopter Germany [2]

Fibre angle detection. The detection of the fibre orientation is performed on the photos taken by
the optical measurement system by a software developed by Eurocopter Deutschland GmbH and
FIBRE, Fasern Institute von Bremen, Germany, called EuroPAS, Eurocopter Preform Analysis
System. The setup and the development of the measurements for every side of the NEXICOS frame
have been performed in accordance to the standards detailed by the developer [12].
In a first step, both the images assembled by the 4D-measuring system and the simulation output
files are loaded on the system. Then, the mesh of the simulation file is projected on the pictures. For
every mesh element, the system detects and calculates the most frequent fibre orientation using an
optical method of edges recognition. This step is controlled by two parameters that are the maximum
standard deviation of the detected angles and the minimum edges account necessary to certify that a
fibre has a certain angle. Last step but not least, both the fibre angles extracted from the picture and
these calculated by the simulation are compared.
 Key Engineering Materials Vols. 554-557 469

The outcome is finally displayed for every mesh cell: a fibre orientation in green indicates a good
correlation between the measured and the simulated fibre angles while orange is used for a fair
correlation and red for a bad one. Blue is used for the fibre orientation detected by EuroPAS. The
mesh cell is depicted in green (see Fig. 5).

Fibre orientation
Unidirectional pre-impregnated optically detected
carbon fibre material (M21-E)

Mesh cell
Good correlation with simulation
Deviation lower than 4°

Figure 5: Outcome of the fibre angle comparison

It must be noted that the tolerance thresholds for which a correlation can be said good, fair or bad
are user defined values (see Table 2).

Table 2: Quality of the correlation

Good correlation [°] Fair correlation [°] Bad correlation [°]
X<4 4<X<8 8<X

Validation
System calibration. A good calibration of the 4D measuring system turned to be difficult due to
material specificities. Beside difficulties related to the shape of the component, the high amount of
resin present within the pre-impregnated material caused important light reflection issues. These
reflections, which appear on the composite surface, tried to be minimized as they distort the
evaluation of the fibre angle. Detection trials conducted in various conditions (with/without light
assistance, different ambient light conditions) as well as the pouring of baby powder on the
preforms or the use of a polarized lens were performed. Conditions for which best results were
obtained were for photos taken in the early morning at diffuse lighting conditions.

Threshold adjustments. Criteria for the detection of the fibre orientation have been softened
compared to these used for Non-Crimped Fabric materials. To do so, the minimum edges account
needed to determine the fibre angles was decreased and, in the same time, the standard deviation of
the fibre orientations detected within a single mesh cell was increased (see Table 3). The
combination of these two parameters led to a rather good fibre orientation detection.

Table 3: Parameters used to detect the fibre angles (M21E versus NCF)
Minimum edges account Maximum standard deviation
5/50 50/40

The modification of these two parameters intended to stabilize the optical measurements and
ease the fibre angle detection. However, to compensate these soft detection criteria and discard the
detection of absurd fibre angles, restrictive filters were set. In this direction, fibre angle values that
differ more than 4° for the web and 6° for the flanges of the simulation were not considered within
the analysis.
470 The Current State-of-the-Art on Material Forming

Results. Measurements of every side of the frame (web, inner and outer flanges) were compared to
simulation results using the EuroPAS system. Outcome of the results in term of average angle
deviations and standard deviations are detailed in Table 4.
Web. Correlations between the simulation and the web measurements are very good for all the
layups. These results were expected since this part is not much subjected to high deformations
during forming. Thus, no distinctive fibre movement happens. An average angle deviation always
lower than 1° was even calculated. Even though the standard deviation is slightly high, the results
are satisfactory.
Inner flange. Correlations between the simulation and the inner flange measurements are very
good for three of the five layups. Discrepancies were located at the peripheral regions of the tooling,
where the preforms were taped during the fibre orientation measurements. The important difference
in the angles measured for the eight layers layup could not be plausibly explained. So far, the
divergence was imputed to inaccuracies in the edges detection. Deviations occurring for the [0/0]
layup are due to fibre angle detection issues and to a severe filtering of the data. Indeed, with the
detection parameters presented in Table 3, some noteworthy values were discarded for the
comparison. The rather high standard deviations obtained for all the layups are due to the geometry
of the inner flange. Having a curved shape, this part of the preform is more exposed to deformations
than the web. Thus, a uniform measurement quality was difficult to reach.

Table 4: Validation results obtained for the NEXICOS side frame

Average angles Standard Number of cells
Preform layup Side deviation deviation of the used for the
[°] angles deviation comparisons
Web -0.36 2.19 1440
[0/0] Inner flange -1.81 2.9 491
Outer flange -2.35 2.62 345
Web 0.43 2.14 1504
[0/90] Inner flange 0.88 3.17 582
Outer flange 0.99 3.23 536
Web -0.72 2.06 1748
[-45/+45] Inner flange -0.49 2.98 660
Outer flange -1.32 3.02 537
Web -0.37 2.1 2318
[60/-60/0]S Inner flange -0.85 2.62 782
Outer flange -0.65 2.91 748
Web 0.64 2.06 1854
[90/0/-45/+45]S Inner flange 2.4 2.3 549
Outer flange 2.34 2.57 503

Outer flange. Correlations monitored for the outer flange with regard to the simulation are in the
same tendencies as these presented for the Inner Flange. In this direction, good correlations were
obtained for [0/90], [-45/+45] and [60/60/0]S even if discrepancies occurred at the peripheral
regions of the tooling. Also, higher deviations were measured for the [0/0] and [90/0/-45/+45]S
layups. Standard deviations are overall higher than for the Inner Flange. This can be imputed to the
geometry of the outer flange that is more difficult to drape. In particular, the comparisons of
simulated and experiment results become worse at the tip of the small radii part; region that induces
important fibre reorientations.
 Key Engineering Materials Vols. 554-557 471

Summary
In this paper, a validation method for forming simulations based on optical measurements of the
fibre orientations is presented. It was applied to the forming of a unidirectional pre-impregnated
composite material over a double curved side helicopter frame. In a first approach, due to material
specificities, a good photo quality could not be reached. After modifications and adaptation of the
4D measuring system as well as the software used for detection, measurements and analyses were
successfully completed. A good correlation between simulated and experimental results was
obtained. An accuracy of 1° to 3°, depending on the geometrical features of the preform, was
achieved. However, the relatively high standard deviations between the measurements and the
simulated results indicate that the robustness of the validation method needs to be improved.
Therefore, the fibre angle detection method, i.e. optical system, still needs to be ameliorated.

Acknowledgement
The authors acknowledge the contribution of Mister Maximilian Maier, who dealt with the topic
developed in this paper during his Master Thesis. Eurocopter Deutschland GmbH must also be
acknowledged for the funding of this work.

References
[1] F. Dumont, C. Weimer, Non-crimp fabric: preforming analysis for helicopter applications, in:
S.V. Lomov, Non-crimped fabric composites - Manufacturing, properties and applications,
Woodhead Publishing Ltd, 2011, pp. 449 - 460.
[2] S. Gubernatis, M. Göttinger, F. Dumont, C. Weimer, Forming simulation validation for non-
woven reinforcements, presented at TexComp10, Leuven, Belgium, 2010.
[3] U.S. Department of Transportation - Federal Aviation Administration, FAA: Aerospace Forecast
Fiscal Years 2012-2032, p. 68.
[4] Information on http://www.icao.int/environmental-protection/
[5] B.T. Åström, Manufacturing of Polymer Composites, Chapman & Hall, 1997, pp.110-121 and
209-218.
[6] E.A.D. Lamers, S. Wijskamp, R. Akkerman, Modelling of fabric draping: Finite elements versus
a geometrical method, presented at ESAFORM 04, Liège, Belgium, 2001.
[7] F. Dumont et al. 2008 Analysis of NCF-preforms for helicopter composites parts, presented
at TexComp9, Newark (DE), USA, 2008.
[8] C. Weimer, N. Bätge, F. Gaille, NEXICOS Next Generation of Integrated Composite
Aerostructures, presented at JEC, Paris, France, 2011.
[9] D. Leutz, M. Klüpfel, F. Dumont, R. Hinterhölz, K. Drechsler, C. Weimer, FE-Simulation of the
Diaphragm Draping Process for NCF on a Macro-Scale Level, presented at ESAFORM 14, Belfast,
Ireland, 2011.
[10] P. De Luca, A.K. Pickett, Industrial examples of forming nonmetallic parts using PAM-FORM,
presented at EURO-PAM’98, Tours, France, 1998.
[11] ESI-Group, PAM-FORM 2G 2012: Solver Reference - Material Type 140, 2012.
[12] A. Miene, Preform Analysis System Manual, Faserinstitut Bremen e.V., 2011.