Digimat material model for short fiber reinforced plastics at Volvo Car Corporation Mats Landervik DYNAmore Nordic Johan Jergeus Volvo Car Corporation June 2, 2015 26TH SICOMP CONFERENCE, GOTHENBURG 1
Contents DYNAmore Nordic Scope of study Short fiber reinforced plastics Digimat Material data & calibration Results Conclusions 2
DYNAmore Nordic Offers Software LS-DYNA DIGIMAT Developed by e-xstream Engineering an MSC company ANSA OASYS PRIMER Etc. FE models Consulting Code development Hardware Nordic locations Linköping Gothenburg Mother company DYNAmore, Germany Composites research and development Sweden Matisse FFI Crash in Composites Germany Composites process modeling Etc. 26th SICOMP Conference, Gothenburg June 2, 2015 3
Scope of study CAE assessment of requirements connected to various vehicle attributes Model performance Predictability CPU cost Material data from suppliers? Material test needs Process data/simulation needs Analyses Safety Strength Thermal creep Not finished Fatigue MSc thesis Fatigue Analysis of Anisotropic Short Fibre Reinforced Polymer by Use of Digimat and ncode DesignLife Jan-Anders Lindhult & Miranda Ljungberg Presented on June 5 2015, Chalmers FE codes LS-DYNA Abaqus Simulations done by Magnus Andreasson (MSC Software Sweden) Vehicle loadcases Pedestrian safety Strength loadcase Component tests Quasistatic strength 2 loadcases Dynamic impact bending 2 loadcases Thermal creep Fatigue failure Cyclic loading at different maximum force levels 4
short fiber reinforced plastics Semi-structural parts Injection molded Popular for Cost Short production cycle time Weight Design freedom etc. Example component is this study Front End Carrier (FEC) From New Volvo XC90 Balancing of performance in Pedestrian safety Static strength 5
short fiber reinforced plastics, Fiber orientation Properties depending on injection process Fiber orientation variation In-plane Through-thickness Skin & Core Predictions from process simulations Porosity (variation) Fiber length (variation) etc. In-plane variation of first principal orientation value close to skin Thickness variation of element vertical orientation component 6
Digimat principles Microstructural properties Ellipsoids of revolution (fibers) Aspect ratio Short fibers Continuous fibers Fraction Orientation Material models for the separate phases (Damage and) failure models Mean field homogenization of micromechanical properties Pre-homogenization to macroscopic model dependent on fiber orientation possible Parameters determined by reverse engineering 7
Applied material models Two phase micromechanical model Linear elasticity for glass fiber phase Different models for matrix material depending analysis type Elastoplastic: J2-plasticity model Yield function f σ, ε p = 3 2 s: s 1/2 σ Y R(ε p ) 0 Hardening R ε p = kε p + R 1 exp mε p Flow rule f ε p = ε p, ε σ p = 2 ε 3 p: ε p Elastoviscoplastic Current yield Norton law ε p = σ m Y f η σ Y + R(ε p ) Failure Tsai-Hill 3D Transversely Isotropic f = ε 11 2 X 2 ε 11 ε 22 + ε 33 X 2 + ε 22 2 2 + ε 33 Y 2 + 1 X 2 2 Y 2 ε 22ε 33 + 2ε 12 2 + 2ε 13 2 S 2 + 4 Y 2 1 X 2 2ε 23 2 No tension-compression differentiation Applied on pseudo grain level 8
The Pseudo grain concept A representative volume element (RVE) in each material point Including an ensemble of fibers Orientations according to tensor Orientation space discretized into pseudo grains Chosen number of grains Unidirectional Axisymmetric Two-step homogenization 9
Available Material data for component LANXESS Durethan BKV30 H2.0 PA6 GF30% Crash data Tension 0 and 90 specimens cut from plaques Abaqus input Stress vs. plastic strain 14 rates (0 to 1000 /s), scaled Failure strain in tension only Strength data Temperatures 0, 23, 40, 60, 90, 120, 140 C Failure from diagram Directly injected dumbbells Creep isochronous stress/strain curves for 23 C / ISO1110 conditioned 23, 40, 60, 90, 120 C / dry Fatigue S-N curve for directly injected dumbbells Orientation missing Moldflow estimation Estimation from experience Limitations No raw test data, only Abaqus input Data extent sufficient although not ideal Lower confidence than ideal All material data received from supplier 10
Calibration of material parameters Quasistatic tension tests Dynamic tension tests at different rates Specimens dpp 11
Quasistatic loadcases Case 1 implicit Case 2 implicit 12
1, center 3-point bending Test 1 Test 2 Digimat anisotropic, LS- DYNA explicit, crash mesh CrachFEM isotropic, LS- DYNA explicit, crash mesh, NOTE! Different PA6+GF30% Digimat anisotropic, LS- DYNA implicit, crash mesh Digimat anisotropic, Abaqus implicit, strength mesh Current method isotropic, Abaqus implicit, 2, off-center 3-point bending Quasistatic results strength mesh 13
Dynamic loadcase 1 June 2, 2015 26TH SICOMP CONFERENCE, GOTHENBURG 14
Dynamic loadcase 2 June 2, 2015 26TH SICOMP CONFERENCE, GOTHENBURG 15
Dynamic results Test Digimat anisotropic, LS-DYNA explicit CrachFEM isotropic, LS-DYNA explicit, NOTE! Different PA6+GF30% 1, center 3-point bending 2, off-center 3-point bending 16
Discussion Perfect agreement of stiffness and initial failure in quasistatic loadcases Overprediction of failure in compression Early loss of force in quasistatic progressive failure Too tough Digimat material in dynamic tests Lack of reliable failure material data for higher rates Rate independent failure applied Limitations Material test data limited in general No tension-compression differentiation of failure Failure realized by erosion of elements Failure paths dependent on discretization Coupling from failed integration points not unambiguous Moldflow predictions not validated Not considered Pores Fiber length and fraction variations (Weld lines) Orientation considered Convergence of meshes 17
Conclusions Digimat parameters for Durethan BKV 30 has been determined from the limited test data available A more complete testing program would yield higher confidence in parameters The predictability for quasi-static stiffness and initial failure in Abaqus has been shown Simulation of full strength loadcase (Fat Farmer) has been shown Decrease of simulation time due to quicker convergence The predictability in terms of stiffness and progressive failure in LS-DYNA has been shown 2 quasi-static loadcases 2 dynamic loadcases 129 % increasing computational cost (cycle time) for replacing isotropic CrachFEM with anisotropic Digimat in the component test Simulation of full crash loadcase (Pedestrian) has been shown 3-5 % increasing computational cost for replacing isotropic CrachFEM with anisotropic Digimat in one component Paper for European LS-DYNA Conference 2015, 15-17 June 2015, Würzburg, Germany to be published online on http://www.dynalook.com/ 18