Performance Modeling of Polymer Composites

1
Performance Modeling of Polymer Composites

• Pete Foss
• Materials and Processes Lab
• General Motors Research and Development Center

2
GM
DELPHI VISION
PREDICT DIMENSIONAL AND MECHANICAL PERFORMANCE OF
POLYMER COMPOSITES BY UTILIZING PROCESS
SIMULATIONS TO PREDICT MICROSTRUCTURE AND THUS
PROPERTIES FOR ANY PART GEOMETRY
3
Areas of Interest - Thermoplastics

• Short fiber reinforced thermoplastics
• Fiber orientation measurements and predictions by
  commercial codes (Moldflow, Moldex, Sigmasoft)
• Utilization of microstructural predictions
  coupled to FEA to predict mechanical performance
  including dimensions, strength, stiffness,
  fatigue and creep.
• Currently use nonlinear orthotropic plaque data
  in conjunction with 2D flow analysis to predict
  performance.
• Future would like to use 3D but must include
  material nonlinearity.
• Long fiber reinforced thermoplastics
• Prediction of initial mechanical properties
• Long term durability
• Rate dependence

4
Areas of Interest - Thermosets

• Automotive Composites Consortium
• Crash modeling of composites.
• Material properties for durability predictions.
• Carbon fiber cost reduction
• Class A structural composites
• Paintability of SMC
• Cost effective low volume solutions
• Glass fiber
• Carbon fiber

5
Barriers To Expanded Use Of Composites

• Cosmetic
• Powder prime
• SMC and RRIM are incompatible with the powder
  prime at our assembly plants.
• Need materials that Paint like steel.
• CLTE
• Thermal expansion coefficient of thermoplastics
  forces inter panel gaps to be much larger than
  steel.
• Structural
• Cost vs. mass savings
• Predictability of crash behavior
• Automotive is fairly unique in that we must model
  progressive damage behavior not just failure
  initiation.
• Assembly and joining methods

6
Gaps

• Accuracy of 2 and 3D fiber orientation
  predictions
• Short fiber injection molding
• Long fiber injection molding
• Long fiber extrusion-compression molding
• 3D structural analysis
• Current solvers cannot handle millions of
  materials - one for every element
• Rate dependence
• Data, material models, testing procedures
• Progressive damage material models
• Software validation