A Graphical User Interface for Simulating Resin-Transfer-Molding Combining LS-DYNA and OpenFOAM M. Martins-Wagner 1, M. Wagner 1, A, Haufe 2, C. Liebold 2 1 Ostbayerische Technische Hochschule Regensburg Laboratory for Finite Element Analysis and Structural Dynamics Regensburg, Germany 2 DYNAmore GmbH Stuttgart-Vaihingen, Germany 1 Abstract The paper describes parts of the joint research project Swim-RTM including several industrial and academic partners. Its goal is to combine LS-DYNA and the open-source CFD solver OpenFOAM to simulate the production process of continuous fiber-reinforced plastics, particularly the resin-transfermolding (RTM) process, in which the layers of dry fabric (unidirectional or woven) are formed in the mold (draping) and then filled with liquid resin with high pressure at injection points. Through a combined analysis of both the structural mechanical and the fluid dynamical phases, a better prediction and thereby optimization of the textile components properties as well as injection points can be achieved, improving the manufacturing process. The draping simulation of the fabric layers is carried out with LS-DYNA, while the injection simulation of the matrix material is performed in full 3D with OpenFOAM. A key question in this research project is how local porosities can be derived from the structural computation in the draping step. The purpose of the presented subproject is to develop a graphical user interface (GUI) to enable the simulation of the entire RTM process of long-fiber-reinforced components including the transfer of results between the draping and injection phases. The complete simulation task is relatively complex and involves several software packages, meaning a high effort for the user to get familiarized with. To circumvent this, the GUI aims at requiring from the user only the minimum necessary input data, creating and running the simulation and mapping tasks in the background, and showing graphically all demanded intermediate and final results. For the draping step several current fabric materials such as *MAT_034, *MAT_234, *MAT_235, *MAT_249 are available. Several modelling techniques for the composite setup are also conceivable, including a workflow similar to metal forming applications. In the injection step the fabric is modelled as a porous medium and different transport models and liquid resin types are at hand. For the data transfer between the draping and injection models, i.e. the mapping of data between shell and volume meshes within the developed GUI, first the OpenFOAM volume mesh is converted to LS- DYNA format and the necessary passing parameters are extracted from the output files, then the mapping tool DYNAmap [3] from DYNAmore GmbH is invoked, and finally the OpenFOAM command files are created. After the injection simulation is started and successfully terminated, information, such as the distribution of air inclusions or the shear stress distribution to analyze the reorientation of component fibers, is available and can be transferred from the 3D fluid mesh to an LS-DYNA shell mesh for further computations, for instance a crash simulation. This backward data mapping between volume and shell meshes can then be performed inside the GUI.
2 Introduction Fiber-reinforced plastics are important composite materials used in almost every type of advanced engineering structure as well as biomedical devices, due to their excellent mechanical properties, as high strength and stiffness, and their light weight. Particularly in the automobile industry new possibilities are presently investigated for a large-scale production of parts made of fiber-reinforced plastics, aiming at maximum weight reduction of structural car parts, especially important in the area of electrical mobility. Many methods and tools have been already developed for analysis and optimization of the production process of short-fiber-reinforced components by injection molding. For the manufacturing of large highly stressed components such as the underfloor assembly of cars in a lightweight construction, long-fiber-reinforced preforms tailored to each particular case should be used. Their orientation is of primary importance in the design and should as good as possible match the load bearing directions. For a large-scale production of such components, the RTM process turns out to be a very promising method. Here the dry fabrics are placed in the mold (negative form of the component), which is subsequently closed and filled with liquid resin inserted with high pressure at injection points (see Fig.1.). Fig.1: Resin transfer molding of a fiber-reinforced component. The RTM process is very complex and depends on many factors like resin system and temperature, injection pressure, number and location of injection points, geometry, compaction, and reinforcement structure and draping properties of the fabric. The insertion in the mold leads to local changes in the fabric, such as fiber-orientation, permeability and porosity, which should be considered especially for complex, curved surfaces. The changes in permeability induce changes in the flow resistance parameters, whereby flow direction and velocity are remarkably affected. Moreover, the closing of the tool can cause a compaction of the fabric und hence a higher packing density of its fibers. On that account, the simulation of the draping process, determining the reorientation of the fibers, and the use of that information for the injection simulation are of major importance for the process optimization. The purpose of the presented subproject is to create a software tool to allow the complete analysis of the production of continuous fiber-reinforced components within the RTM process based on numerical simulation. The developed graphical user interface (GUI) controls the simulation of the RTM process by combining the structural solver LS-DYNA [1,2] for the draping and the CFD solver OpenFOAM [5] for the resin injection. This allows a detailed knowledge of the main sensitivities throughout the process, such as fiber position and orientation, and location and number of inlets. Hence, through this combined analysis of both the structural mechanical and the fluid dynamical phases, a more accurate prediction and thereby optimization of the textile components properties as well as injection points can be achieved, improving the manufacturing process, especially for geometrically complex, long-fiberreinforced composite parts. Initially, a numerical simulation for the draping of the fabric in the RTM process is prepared within the GUI and started with LS-DYNA. The local deformations of the fabric are computed and later used to determine the corresponding non-uniform porosities. In the next step, the fiber fabric in the mold is modelled as a porous medium and a numerical simulation of the resin injection is prepared with the GUI. The flow resistance parameters (porosities
and fiber orientations, as well as material dependent history variables) are transferred to the injection model. This data transfer between the draping and injection models, i.e. the mapping of data between the shell and the volume meshes is done also within the GUI. Firstly, the OpenFOAM volume mesh is converted to LS-DYNA format and the mapping tool DYNAmap from DYNAmore GmbH is invoked. Then the OpenFOAM command files are created and finally the injection simulation is started with a dedicated solver from OpenFOAM, particularly developed in one of the other sub-projects of the joint research project Swim-RTM for this purpose [6], which supports Newtonian and non-newtonian material behavior of the liquid resin. The computed temporal development of the flow front can be used to determine the distribution of air inclusions and thereby prevent part defects, as areas of reduced strength. Furthermore, information on the distribution of the velocity gradient can be used to analyze the shear stress distribution in the fibers and thus avoid their reorientation, for instance by providing more injection points or reducing the injection pressure. Such measures aim at optimizing the manufacturing process. For this sake, the available information from the injection simulation can be transferred to a LS-DYNA structural model for further analysis. This mapping of data between the volume and the shell meshes can be handled as well within the GUI. Firstly the OpenFOAM volume mesh is converted to LS-DYNA format and the necessary passing parameters are extracted from the output files. Then the mapping algorithm is invoked and a new structural simulation can be started with LS-DYNA, for instance for a crash analysis. To the extent of our knowledge, there is at present no available software package providing all those functionalities. However, particularly in the automobile industry, there is an increasing demand for such a numerical simulation tool comprising the complete design and optimization of RTM manufacturing of fiber-reinforced components. The software tool developed in this project offers the user those capabilities in an interactive graphical interface allowing a combined and flexible handling of the RTM production process of composites. 3 A Graphical User Interface for numerical simulation of resin transfer molding The graphical interface allows parallel computations with the FE solver LS-DYNA and the CFD solver OpenFOAM. It performs the simulation of the complete RTM process within an interactive interface, including the transfer of simulation results between the draping and injection phases. It requires only minimum user-input for the models while enabling the import from templates, generates and runs both structural and fluid simulations as well as their necessary data mapping in the background, and displays all required textual and graphical results. The GUI was programmed with the open-source widget toolkit Tk [7], with Perl as programming language for the implementation of all necessary simulation computations and data transfer. The program was written without a GUI-builder and sets up a cross-platform GUI application, which presents the user with four separate interfaces: for the draping with LS-DYNA, for the data mapping between draping and injection, for the injection with OpenFOAM, as well as for the data mapping from the injection to a strength or crash simulation with LS-DYNA. For this purpose, a software design was developed under observation of the data flux between all simulation phases and separating the GUI part from the computations with LS-DYNA and OpenFOAM. The first menu prepares the draping FEM simulation based on a continuum mechanical model on a shell mesh, writes all input files, verifies and visualizes graphically the numerical model, runs LS- DYNA and offers the post-processing of results with LS-PrePost. In the second menu, the flow resistance parameters are computed from the results of the draping simulation and transferred from the LS-DYNA shell mesh to the volume fluid mesh, which is imported for OpenFOAM. The third menu prepares the simulation of the resin injection on the fiber fabric in the mold modelled as a porous media, starting with the input data for boundary and initial conditions and for the CFD solver, then writes all input files, controls and visualizes the fluid model with ParaView, runs OpenFOAM, and offers the post-processing with ParaView. In the last menu, the distributions of shear stresses and air inclusions are extracted from the injection simulation outputs and can be transferred from the fluid
mesh to a LS-DYNA shell mesh for a further structural simulation, e.g. a crash or strength simulation with LS-DYNA. The GUI uses a configuration file to store the project directories for LS-DYNA and OpenFOAM, as well as the paths to material template files and all executable programs (solvers, post-processors and mapper). This provides every new user with an easy way of modifying or replacing the variables to his specific environment. 3.1. Draping simulation Here, a draping simulation based on a continuum mechanical model on a FEM shell mesh is prepared for LS-DYNA, computed and post-processed. For an overview of the state of the art on simulation of composite structures in LS-DYNA, see e.g. [4]. For the material description of the composite parts, several current fabric materials such as *MAT_34, *MAT_234, *MAT_235, and most important *MAT_249, newly developed for uni-directional layers as well as woven and non-crimped fabrics. For the composite setup different modelling techniques are also conceivable, including a workflow similar to metal forming applications. An extra bending stiffness in the material model can be considered if a membrane element formulation is used, to avoid excessive wrinkling and stabilize the computation. This is done by introducing an additional part containing the same nodes as the membrane part and defined via shell elements with properly chosen elastic constants. The GUI additionally presents the possibility of importing any user-defined template material card in LS-DYNA format. Compatibility of the numbering scheme is automatically ensured during the writing process of the keyword input files. Furthermore, the use of pre-defined keyword-templates provides typical material parameters for each material model while enabling changes by the user, without any deeper knowledge of the material cards. In this first module, the user is presented with interface menus and submenus, each for a specific step of the data input and writing, as well as for computing and post-processing the draping model (see Fig. 2. Fig. 4.), as follows: - FEM model setting (see Fig. 2) - Input for the draping tools: - maximum velocity and force of blankholder - tool kinematics: single action (air drawn) or double action (toggle draw) - General process settings: type of friction, friction coefficients, plot distances - Global settings: project title, unity system - FEM meshes for the blank (fabric) and for the draping tools (die, punch and blankholder) in home position - Composite description: geometric parameters of the fabric (see Fig.3) - number of composite layers - integration points per layer - thickness per integration point - fiber orientation vector, material angle per integration point, and material parameter AOPT - Material modelling (see Fig.4) - Material model - Material template file for the chosen model - Consideration of extra bending stiffness if membrane element is used - Input data specific for *MAT_249: - number of fibers per integration point - orientation of fibers per layer per integration point - Numerical settings: - type of finite element, type of contact, consideration of self-contact of fabric - type of mesh refinement for drawing (adaptive or uniform) - control parameters (time step for mass scaling, time interval between adaptive steps, system damping constant, Rayleigh damping coefficient, laminated shell theory parameter)
- Writing input file for LS-DYNA - Verifying and visualizing the generated FEM model with LS-PrePost - Computation with LS-DYNA - Input for the solver: Job-Identification name, number of CPUs - Run LS- DYNA to compute the generated FEM input deck - Post-processing with LS-PrePost Fig.2: Interface for Process setting from the GUI module Draping simulation.
Fig.3: Interface for Composite setting from the GUI module Draping simulation. Fig.4: Interface for Material setting from the GUI module Draping simulation. 3.2. Mapping of draping results to the injection model The proper data transfer between the draping and injection models is a matter of great importance. Within the GUI, this is performed in the second interface module, where specific results from the draping simulation are extracted and later on mapped to the 3D fluid mesh, with the help of the mapping tool DYNAmap. Before the data mapping can be started, the 3D fluid mesh must be
converted to the LS-DYNA format, since DYNAmap currently only reads and writes LS-DYNA formats. This is done within the GUI with a specifically developed algorithm that converts the face-oriented approach to define the mesh within OpenFOAM to a structural solid mesh in LS-DYNA format, in order to make the application independent of any other available preprocessor. The CFD mesh can contain any of the known solid element shapes in LS-Dyna. The control file for the mapper is also generated within the GUI, after the user chooses the mapping method and mapping algorithm. For special details on the mapping, the reader is referred to the manual [3]. Once the passing parameters and both shell and volume meshes in LS-DYNA format are at hand, the algorithm DYNAmap is called for performing the mapping between the meshes. Currently, strains, stresses and history variables of the chosen material model can be mapped. Moreover, the fiber directions are extracted from the history variables and transferred as one of the major parameters for determining tensorial permeabilities, which are computed later on within OpenFOAM as part of a newly developed fluid flow model [6]. After the mapping is done, the mapped strains in the volume mesh are used within the GUI to compute and set the porosities for the injection model. A proper derivation of porosities of the draped part is very important for the injection simulation and it is a difficult task, since local porosity depends on the orientation and packing density of the composite fibers. In this work, two approaches are currently implemented. Firstly, the trace of the strain tensor can be computed, since it relates to the volume change. Secondly, the initial porosity can be measured and together with the computed volume change in the element due to the draping can be used to determine the actual porosity. It is assumed that the amount of fibers in an element remains unchanged, so that the complete volume change is added to the void phase. Since these two approaches are probably not covering every effect and also to add more flexibility, the computed porosities can be multiplied with any user-defined curve to account for effects that are not represented in the above derivation. The GUI Module for transferring results from the draping simulation to the injection model presents the user with the following steps (see Fig. 5. for the mapping): - Input for determining the passing parameters: - LS-DYNA output file containing parameters to be transferred - Input data for porosity computation, to be later performed after the mapping: - Method to be used (thickness changes or invariants of strain tensor) - Initial porosity value (needed only if thickness changes was chosen) - File containing the thickness curve for porosity scaling (needed only if thickness changes was chosen - Injection model - Setting the directory containing the OpenFOAM mesh to be read in for the injection simulation - Mapping - Conversion of the OpenFOAM 3D volume mesh to LS-DYNA format - Extraction of the transfer parameters (fiber orientation vectors, history variables and strains) from the LS-DYNA output file - Input data for the mapper: mapping method, mapping algorithm, option for invariant node numbering - Running the mapper to map the extracted parameters from the shell mesh to the volume mesh, both in LS-DYNA format - Computation of porosities in the volume mesh from the mapped values of deformations - Transfer of computed porosities and mapped parameters from the volume mesh in LS-DYNA format to the OpenFOAM mesh
Fig.5: Interface for Mapping from the GUI module Interface draping - injection. 3.3 Injection simulation In the third of the GUI modules, the fiber fabric in the mold is modelled as a porous medium and a numerical simulation of the resin injection process is prepared for OpenFOAM, computed and postprocessed. The interface for setting up the injection model offers the user different transport models and liquid resin types and is based on the newly developed solver KosPorousInterfoam of OpenFOAM [6], suited for 3D flow modelling of incompressible, isothermal, immiscible fluids. The GUI module for preparing the injection simulation doesn t require from the user any knowledge of the initially complex input structure of OpenFOAM. Only the most necessary parameters can be accessed through precisely described input fields, namely maximum Courant number, write interval, and start and end simulation time. This prevents the user to change internal settings of the solver, which could lead to numerical instability of the solution process. Also changes in the discretization settings (e.g. convection model) are not allowed. After the required data for the fluid and solver are input, the boundary and initial conditions for the model must be set. In a separate interface menu, the boundary patches of the 3D fluid mesh are extracted from the mesh files and for each of them the appropriate boundary conditions (e.g. zerogradient, totalpressure) are chosen and the initial values for the respective fields (e.g. for pressure and velocity) are set. The initial values for porosities and the orientation vectors were already computed, mapped and written to the respective input files in the previous GUI menu. The fluid mesh was already imported. An automatic generation of the fluid mesh directly with OpenFOAM is extremely involved and therefore not performed through the GUI.
Finally, the input data is written to the appropriate input files for OpenFOAM, the simulation case is verified, visualized, and the CFD solver is started. The GUI presents also a possibility of restarting a simulation or starting it new by changing the solver input parameters (time step, start time, end time) and running the solver again. For setting up and running the injection simulation, the user is presented with the following interface menus (see Fig. 6.): - Fluid flow model setting - Global settings: case title and run directory - Fluid flow setting: surface tension between phases and transport model (Newtonian or non- Newtonian). For non-newtonian fluids, the resin type should be chosen. - Solver settings: maximum Courant number, write interval, and start and end simulation times - Boundary and initial conditions for the boundary patches - Writing the input files to OpenFOAM - Verifying and visualizing the generated injection model with ParaView - Computation with OpenFOAM - Input for the solver: case name, number of CPUs - Run OpenFOAM to compute the generated fluid flow model - Post-processing with ParaView - Restarting with OpenFOAM Fig.6: Interface for Fluid from the GUI module Injection. 3.4 Mapping of injection results to a crash/strength model As result from the RTM simulation, important information, such as the distribution of air inclusions and shear stresses in the draped part, can be transferred to a LS-DYNA model for a further crash or strength simulation, which then incorporates the influence of the draping and injection in the final part. The mapping of data between the volume and the shell meshes is done as well within the GUI. After the selected mapping parameters are extracted from the OpenFOAM output files and the volume mesh is converted to LS-DYNA format, the mapping algorithm is invoked and the mapped data can be visualized in LS-PrePost. The data is converted in a DYNAIN-file for further usage in crash or strength simulations. The last GUI Module for transferring results from the injection simulation to a crash or strength model presents the user with the following steps (see Fig. 7.):
- Input for determining the passing parameters: - Setting the OpenFOAM working directory, which contains the data to be mapped - Data to be mapped: - distribution of air inclusions - shear stresses. In case shear stresses are of interest, a maximum value of allowed shear stress must be input - Setting the directory containing the LS-DYNA mesh to be read in for the crash or strength simulation - Mapping - Conversion of OpenFOAM fluid mesh to LS-DYNA format - Extraction of the transfer parameters (distribution of air inclusions and shear stresses) from the respective OpenFOAM output files - Input data for the mapper: mapping method and number of cells through the thickness - Running the mapper to map the extracted parameters from the volume mesh to the shell mesh, both in LS-DYNA format Fig.7: Interface for Mapping from the GUI module Interface injection strength/crash. 4 Summary and outlook A software tool for the numerical simulation of the design and optimization of RTM manufacturing of long-fiber-reinforced components was implemented. It features a flexible and combined handling of both the structural mechanical and the fluid dynamical simulations. The developed GUI allows the simulation of the complete RTM process within an interactive interface, which requires only a minimum user-input for the models while enabling the import from templates. It generates and runs both structural and fluid simulations, with LS-DYNA and OpenFOAM, respectively, as well as all necessary data mapping in the background, and displays all required textual and graphical results. Several modelling techniques for the composite and material models for the draping simulation are available and proved to work well. In this work, the porosities are determined from the initial porosity values and the element volumes before and after deformation, with the assumption of volume constancy of the fibers. The developed algorithm for transferring the computed porosities from the draping simulation to CFD computation is rather simple. Improvements should be aspired in the modelling of the inhomogeneity and misalignments of the fiber fabric after the draping. Especially the change in volume due to fiber interaction should be implemented in the current material models.
Further enhancements can also be made towards better understanding the infiltration process, especially the influence of fiber orientations and misalignments onto the porosity of draped parts. Homogenization strategies for fiber orientations and porosities can be developed and investigated through crash and strength simulations of the final RTM parts. The presented GUI serves as a demonstrator for the capabilities of this RTM simulation process. Additional testing on the accuracy with different material data and part geometries coming from industry should be further carried out. Last but not least, a still open question is the contemplation of the coupling between fabric and resin at the contact surface. It should be investigated whether at the interface the fluid stresses need to be transferred to the structure and the structure displacements transferred to the fluid. If a coupling is necessary, a fluid-structure interaction (FSI) approach could be an alternative to the RTM simulation process. A partitioned FSI solver could be implemented, such that the flow and structural solvers are maintained and advanced separately. 5 Literature [1] LS-DYNA Theory Manual, Version 4778, 2014. [2] LS-DYNA Keyword User s Manual, Version 4782, 2014. [3] Dynamore GmbH: DYNAmap User s Manual, Volume 1, March 2015. [4] Knust, G., Klöppel, T., Haufe, A., Liebold, C.: New developments to capture the manufacturing process of composite structures in LS-DYNA, Oasys LS-DYNA User s Meeting, 2014. [5] OpenFOAM User Guide, Version 2.3.1., 2014 [6] Wegh, N.; Bachschuster, S.; Gaudlitz, D.; Klein, M.: Three-dimensional simulation of the impregnation stage of the RTM process considering the local structure of the fiber layup, In 8th International Conference on Computational and Experimental Methods in Multiphase and Complex Flow, Valencia, Spain, April 2015. [7] Lidie, S., Walsh N.; Mastering Perl/Tk Graphical User Interfaces in Perl, 2002.