Conference PaperPDF Available

Figures

Content may be subject to copyright.
11th European LS-DYNA Conference 2017, Salzburg, Austria
© 2017 Copyright by DYNAmore GmbH
MANUFACTURING SIMULATION AS PART OF THE
DIGITAL PROTOTYPE
P. Böhler1, J. Dittmann1, D. Michaelis1, P. Middendorf1, Christian Liebold2
1Institute of Aircraft Design (IFB), Pfaffenwaldring 31 ,70569 Stuttgart, University of Stuttgart, Germany
2DYNAmore GmbH, Industriestraße 2, 70565 Stuttgart, Germany
process simulation, braiding simulation, infiltration, data transfer,
digital prototype
Abstract
The research project Active Research Environment for the Next generation of Automobile
(ARENA2036) is a long term project funded by the Federal Ministry of Education and Research
Germany. Within this project four sub-projects are located. DigitPro, one of those sub-projects, deals
with the development of a Digital Protoype. A closed simulation process chain is built which not only
covers different simulation disciplines such as crushing or process analysis, but also various software
solutions and material models. The main goal is to use the digital prototype to decrease the weight if
an automotive structure by 10% and the development time by 50%.
In this project one of the focused manufacturing processes for composite structures is the braiding
technology followed by an infusion process. A complete numerical prediction is necessary for the
braiding as well as for the infiltration process to decrease the development time and to increase the
mechanical performance of braided structures. Within this work an overview of the newest
developments in braiding and infiltration simulation and especially in the transfer of the necessary data
from one process to the next is given.
An overview on the succeeding project “Digital Fingerprint” will be given as the results of the project
DigitPro will be used there.
1 Introduction
With respect to economic and ecological issues structures for technical applications have to be lighter,
multifunctional but still cheap. These controversial requirements can only be achieved by the
combination of different areas like the optimization of the manufacturing processes, the use of
innovative materials - such as fiber reinforced plastics (FRP) and its suitable application - and highly
automated, intelligent production processes. Especially the last request is one of the tasks within
Industry 4.0, where different production steps are intelligently connected among each other and with
the components itself.
By the use of fiber reinforced materials the amount of different variations increases as a very high
number of possible combinations exists. This high complexity leads to an increasing demand on virtual
prediction methods such as the Digital Prototype. The aim of this kind of a virtual process chain is
the numerical illustration and prediction of the complete production chain. This should lead on the one
hand to an optimized manufacturing process where the specific material behavior is considered and
on the other hand to optimized properties of the final structure.
The necessity of a virtual prediction of the structural behavior of FRP parts is based on the fact that it
consists of the fiber and resin properties and that the fiber architecture strongly influences the
properties of the final structure. The forming of textiles is limited due to forming effects. Using
simulation techniques the final fiber architecture and thus the final properties can be known bevor the
real manufacturing starts. By virtually changing the boundary conditions of the manufacturing process
the fiber architecture can be optimized and thereby the lightweight potential can be maxed. As the
forming effects might lead to a non-optimal fiber distribution the deviations are known thanks to the
simulation and can be taken into account. Both ways lead to a reduction of the material usage and
save weight and costs.
Using the virtual optimization the process boundary conditions can be adapted in a way that on the
one hand, as described above, the ideal fiber architecture can be achieved, and on the other hand, the
process itself can be designed cheaper, projectable, faster and even realizable. Provided that good
11th European LS-DYNA Conference 2017, Salzburg, Austria
© 2017 Copyright by DYNAmore GmbH
simulation models are existing, an optimization of the real manufacturing process regarding cost, time
and performance is possible.
Within the research campus ARENA2036 the project DigitPro handles the Digital Prototype for the
automotive industry. Therefore different manufacturing processes are considered and are virtually
described. As manufacturing processes the so called open reed weaving (ORW) and the draping of
such ORW-textiles is focused as well as the braiding technology and the following resin infusion.
To achieve a closed virtual process chain a new data container using the HDF5 file format is defined
[1, 2]. All input and output data of each simulation step is stored in that data container, which for this
reason serves as storage but also as documentation of the corresponding virtual process chain. In this
Digital Prototype or Digital Fingerprint - all data of the real manufacturing, the testing or the quality
management can be stored next to the simulation data. Thanks to a strict organization the whole
history of each structure can be scanned, even in the final assembly, using „near field communication
or „barcode-systems (cf. Figure1). This digital fingerprint is the direct connection to the Industry 4.0. A
log file is deposited in which all following processes in every variation is listed. The data itself is stored
in a cloud-based neutral HDF5 format.
Fig.1: The Digital Prototype as Digital Fingerprint
Using different software-tools for the different simulation steps an input deck is created by a parser
taking the necessary data out of the data container to the simulation tool and storing it back in the data
container after the simulation. The results of a forming simulation such as the braiding process
simulation for instance can be transferred directly to the resin injection simulation and those results
again to the structure simulation. The defined data container is designed in a way new processes can
be added to the Digital Prototype very easily.
2 Braiding Simulation
As said before, one of the key manufacturing processes in the Digital Prototype is the braiding
process, which is ideal for tube-like structures. Braids feature complex fibre architecture, which can be
quite well determined for simple, straight geometries with a constant, circular cross-section. The
thickness of a braiding layer, gapping effects as well as fibre undulations can either be analytically
calculated or empirically predicted [4]. The mentioned parameters are exclusively dependent on the
braiding machine setup, the number of braiding bobbins, fibre tension, linear yarn density and the
desired braiding angle. Also, the movement of the mandrel through the braiding machine has a strong
impact on the final structure behaviour.
11th European LS-DYNA Conference 2017, Salzburg, Austria
© 2017 Copyright by DYNAmore GmbH
A finite element based simulation approach for the braiding process is used to predict the complex
final fibre architecture which can then be used for the following numerical simulation steps. Due to the
mentioned effects, the resulting fibre architecture can differ largely from the expected one.
In order to predict these differences, physical effects have to be included in numerical models, e.g.
friction and occurring dynamic effects. This is the reason why an explicit, numerical finite element
simulation (PAM-CRASH V14, ESI Group) is used in the project DigitPro. On the one hand it is, due to
the high level of detail, computationally expensive. On the other hand, it has the potential to model any
kind of desired and undesired effects. The goal is to reduce the development costs, as less
preliminary experimental tests and real manufactured prototypes are needed.
A braiding simulation of a generic geometry is shown in Fehler! Verweisquelle konnte nicht
gefunden werden.. The yarns are represented by linear 2D shell elements [2]. The kinematic of the
real braiding machine is copied, and the mechanical properties of the dry carbon fibre yarns are
modelled using a specially adapted material formulation, where the bending stiffness is not coupled to
the tensile stiffness. This is very important as dry textiles always have different in-plane and out-of-
plane behaviours which has to be considered for getting suitable prediction.
Fig.2: Simulation of the braiding process of a generic geometry.
It can be seen in the figure that the effects mentioned above can be modelled. There are variations in
fibre angles as well as some gapping and bridings effects. All of them are influencing directly the
mechanical behaviour of the final structure.
An optical measurement system, co-developed by IFB and FIBRE [2; 5], can be used to determine the
fibre angle of the real manufactured component (cf. Fehler! Verweisquelle konnte nicht gefunden
werden.) based on a grey scale analysis [6]. The gained information gives a good validation of the
simulation compared to finished part. Micrographs of the braided composite part show the inside of the
fibre architecture, making a validation of the fibre undulations possible. This is necessary to improve
the simulation approach and to fit several unknown parameters in the numerical model. An approach
which is fitted to a simple geometry can then be used for predicting more complex structures which will
decrease the amount of real tests and safes a lot of money.
Although modelling the yarns with shell elements is a big simplification, a quantitative comparison is
possible. Any detachments of the braid can be detected and measured using a 3D scanning system
(ATOS by GOM), see Fehler! Verweisquelle konnte nicht gefunden werden.. All these types of
quality assessment can be saved to the digital fingerprint of the part, so that a continuous quality
monitoring is possible.
11th European LS-DYNA Conference 2017, Salzburg, Austria
© 2017 Copyright by DYNAmore GmbH
Fig.3: Recording of the fibre angle (left), real and virtual micrograph of the braid architecture
(top right) and 3D scan of detachment effects (bottom right).
Relevant parameters for the braiding process are the fundamental processibility, the mandrel path, the
choice of yarn tensions and the bobbin setup. These can be virtually determined in the braiding
simulation and transferred to the real manufacturing, reducing the number of preliminary tests.
Furthermore, mechanical properties of the finished composite part can be calculated based on the
result of the braiding simulation. Values like the local fibre angle, material thickness and pure resin
pockets due to gapping all have an influence on the stiffness and strength of the final part. Corrections
can be introduced early on in the development process. Once set-up, the digitally determined target
architecture becomes the quality reference for the production of real parts in a next generation
manufacturing line.
The combination of simulation with reality is ensured by using a CAM interface, which was developed
at the IFB. The mandrel path or rather the robot path responsible for the guidance of the mandrel
through the braiding machine is directly adopted from the simulation. A first iteration step of the path is
gained by the geometry information. Using that in the finite element simulation the path is adopted with
respect to the effects mentioned above. After some iterations the ideal mandrel path is determined
which then is interpreted as robot path and transferred to the manufacturing step afterwards.
This way, a good prediction of the part properties and an ideal robot path are achieved. Another
ambitious project is the development of an electronic bobbin which enables the change of yarn tension
during the braiding process. As a result, more degrees of freedom for the manufacturing process are
possible. This leads to better, lighter and cheaper parts.
11th European LS-DYNA Conference 2017, Salzburg, Austria
© 2017 Copyright by DYNAmore GmbH
3 Resin injection simulation
The resin injection simulation is an important part of the FRP process chain and uses the results of the
braiding simulation.
If components are not properly infiltrated or just with poor quality, previously calculated material
properties are not achieved. The most important parameter in LCM (liquid compression molding)
processes is the permeability of a technical fabric to a fluid media. The determination of this parameter
is difficult, but is essential for simulation. At the moment almost planar permeability values are used,
as near-net-shape prediction in measuring benches or tools are only realizable with increased effort
and for each new component. The transfer of the ideal, planar permeability values to near-net-shape
draping or to near-net-shape braiding structures is not adequate. Different component thicknesses in
the tool design and fiber angle variations prevent this as locally varying permeability is occurring (see
Figure 4).
Fig.4: Variation of fibre angle in a triaxial braiding process
In the project DigitPro a new numerical approach of predicting permeability values is concerned. The
textile layup of the target structure, based on the data of the braiding process simulation stored in the
Digital Prototype, is modelled mesoscopically and a flow through the textile is generated using a CFD
solver. Also the Dual-Scale-Effect can be simulated, around and through the individual yarns. This
simulation can be realized within a few hours and to this state of development a good impression on
the input parameters is given.
Fig.5: Transfer of fibre architecture to preform FEM filling simulation
11th European LS-DYNA Conference 2017, Salzburg, Austria
© 2017 Copyright by DYNAmore GmbH
The final near-net-shape, three-dimensional permeability tensor field (see Figure 5 and 6) allows the
prediction of critical points in the component and can also be used for FEM filling simulation. Therefor
a mapping algorithm such as Envyo [7] is used. Discrete determined permeabilities are mapped on a
two dimensional macroscopic finite element net and a liquid compression molding simulation is
performed. This helps defining inlets, outlets and injection pressures in the real manufacturing process
[3].
Fig.6: Permeability tensor field of a unit cell model and a near-net-shape cylindrical tube
The output parameters of virtual permeability prediction and FEM filling simulation are stored in the
Digital Prototype and are used for the follow-up processes and quality management.
The validation of the resin injection simulation at the end of the process chain is made by VARI- and
RTM-permeability measurements and is also stored in the Digital Prototype. Hence, the tracking of the
development of the real component is completely possible.
4 Summary
The Digital Prototype as a kind of Digital Fingerprint gives the possibility to store the history of each
structure and release it at any moment. Considering the manufacturing processes of braiding and
injection a virtual environment in a HDF5-format for this Digital Fingerprint is built which can be
extended by any other manufacturing process. It seems to be possible that specific manufacturing
effects are acceptable as long as they are known and can be considered in the following processes. A
strong interconnectedness of the different processes and the virtual prediction leads to individually and
automatically defined boundary conditions. A possibility to decrease the waste of material and the
costs of the structures seems to be the use of such CAM-Interfaces in the future.
5 Literature
[1] C. Liebold, A. Haufe, T. Klöppel, S. Hartmann; Recent developments and trends for composite
modeling in LS-DYNA, Fachkongress Composite Simulation, Fellbach, 2015
[2] J. Dittmann, P. Böhler, D. Michaelis, M. Vinot, C. Liebold, F. Fritz, H. Finckh, P. Middendorf;
DigitPro Digital Prototype Build-up Using the Example of a Braided Structure; 2. International Merge
Technologies Conference, Chemnitz, 2015
[3] J. Dittmann, S. Hügle and P. Middendorf; Numerical 3D Permeability Prediction Using
Computational Fluid Dynamics Techniques; FPCM - 13th International Conference on Flow Processes
in Composite Materials, Kyoto, Jul. 2016
[4] M. Bulat, H. Ahlborn, F. Gnädinger, D. Michaelis; Braided carbon fiber composites; in Y. Kyosev
(Hg.): Advances in Braiding Technology; Duxford/Cambridge(USA)/Kidlington: Woodhead Publishing,
2016
11th European LS-DYNA Conference 2017, Salzburg, Austria
© 2017 Copyright by DYNAmore GmbH
[5] P. Middendorf, D. Michaelis, P. Böhler, J. Dittmann, F. Heieck: ARENA2036 DigitPro:
Development of a Virtual Process Chain; 16. Internationales Stuttgarter Symposium 2016, Stuttgart
[6] A. Miene, M. Göttinger und A. S. Herrmann; Quality assurance by digital image analysis for the
preforming and draping process of dry carbon fiber material; in SAMPE EUROPE International
Conference and Forum, Paris, 2008
[7] C. Liebold, A. Haufe; Closing the Simulation Process Chain using a Solver Independent Data
Exchange Platform: the Digital Prototype; 14. Deutsches LS-Dyna Forum, Oktober 2016, Bamberg
ResearchGate has not been able to resolve any citations for this publication.
Presentation
Full-text available
Recent developments and trends for composite modeling in LS-DYNA
Conference Paper
Permeability measurements and prediction are one of the most critical parameters for LCM simulation and have been focused in research for many years. Experimental permeability measurements are time and material consuming, but necessary for today’s FEM simulation. Virtual permeability prediction is usually based on small mesoscopic RVE models or analytical approaches. Using these state of the art methods the transfer to large near-net-shape textiles is not applicable. In this study, based on a plain mesoscopic triaxial 12K braid model created with WiseTex, TexGen and PAM-Crash, a method for numerical permeability prediction using an open source CFD code is introduced. The dimensions of the plate are 15 x 15 x 1 mm containing one layer and a fibre volume content (FVC) of 33 %. Furthermore a second compacted mesoscopic model is introduced to show the differences in permeability results. The dimensions of the compacted plate are 15 x 15 x 2 mm containing three layers and a FVC of 50 %. A full-field fluid flow is simulated with a steady state semi implicit pressure induced solver (SIMPLE) of the software tool OpenFOAM. In a following step the 3D permeability tensor field is determined using Darcy’s equation and the calculated flow conditions. Challenges in meshing and permeability calculation are identified and possible solutions for near net shape structures are shown. The results are compared between the two mesoscopic models, to different approaches of permeability tensor field calculation and to real experiments. The latter are executed with a radial test bench, a constant infiltration pressure of 2 bars and a 12K triaxial carbon braid with a FVC of 32.7 and 49.1 %.
Prototype Build-up Using the Example of a Braided Structure
  • Digitpro -Digital
DigitPro -Digital Prototype Build-up Using the Example of a Braided Structure; 2. International Merge Technologies Conference, Chemnitz, 2015
Heieck: ARENA2036 -DigitPro: Development of a Virtual Process Chain
  • P Middendorf
  • D Michaelis
  • P Böhler
  • J Dittmann
P. Middendorf, D. Michaelis, P. Böhler, J. Dittmann, F. Heieck: ARENA2036 -DigitPro: Development of a Virtual Process Chain; 16. Internationales Stuttgarter Symposium 2016, Stuttgart
Quality assurance by digital image analysis for the preforming and draping process of dry carbon fiber material
  • A Miene
  • M Göttinger
  • A S Herrmann
A. Miene, M. Göttinger und A. S. Herrmann; Quality assurance by digital image analysis for the preforming and draping process of dry carbon fiber material; in SAMPE EUROPE International Conference and Forum, Paris, 2008
Closing the Simulation Process Chain using a Solver Independent Data Exchange Platform: the Digital Prototype
  • C Liebold
  • A Haufe
C. Liebold, A. Haufe; Closing the Simulation Process Chain using a Solver Independent Data Exchange Platform: the Digital Prototype; 14. Deutsches LS-Dyna Forum, Oktober 2016, Bamberg