Skip to main content

Multiscale and Data-Driven Modeling Platform Development

Multiscale and Data-driven Modeling Platform Development with Applications in Advanced Material Simulation
-Milestone 1 Update

Funded by BICI

Jiaying Gao, Cheng Yu, Modesar Shakoor, Wing Kam Liu, Jian Cao, Gregory Wagner

POTENTIAL MARKET

Composite design and property evaluation

  • Lightweight composite design
  • Impact absorption structure
  • Tire composition design

Metallic materials manufacturing process and material evaluation

  • Material corrosion fatigue evaluation
  • Image based material property prediction
  • Advanced manufacturing product property analysis
  • Multiscale modeling software package

CURRENT FUTURE GRAPHIC

LIMITATION OF EXISTING METHODS IN INDUSTRY

Limitations of finite element (FE) software used in industry:

  • Simulation results are mesh sensitive.
  • Requires complicated material law. For example, orthotropic elasto-plastic material law[1]describes behaviors of composite, such as Unidirectional (UD) composite.
  • The fracture envelope is complicated and requires many experiments to calibrated:

MESH GRAPHIC

WHAT WE CAN DO THAT OTHERS CANNOT

In the developing Multiscale and Data-driven Modeling platform, the material responses will be computed on-the-fly

WHAT WE CAN DO GRAPHIC

 

WHAT WE CAN DO, FAST AND ACCURATE

  • Multiscale modeling method captures intrinsic behavior of microstructure (modeled by Representative Volume Element, RVE ).
  • The Multiscale and Data-driven Modeling platform under development enables concurrent capture of macro nonlinear responses and micro-scale evolutions. No other existing method has achieved both features.
  • The cost of modeling of microstructure is inexpensive, compared to a full-scale model, as shown in following composite coupon example:

HOW WE DO IT BETTER GRAPHIC

SAMPLE UD COUPON MODELING

SAMPLE UDP COUPON GRAPHIC

MULTI-RESOLUTION CONTINUUM THEORY (MCT)

  • A hierarchical multiscale modeling approach.
  • Microstructure is embedded into macroscale as extra-scale.
  • Material length scale to represent microstructure.

PLASTIC STRAIN GRAPHIC

WHAT WE CAN DO THAT OTHERS CANNOT DO FOR MCT

One model, Multiple scales:

  • One macroscale and multiple extra-scales.
  • Material length scale, such as size of the void, is embedded into the model.
  • Material length scale regularized the model for less mesh sensitivity.
  • Homogenized microstructure properties as inputs for extra-scale.
  • Effect of microstructure observed on macroscale model.
  • Identify causes of fracture using microstructure information.

BENCHMARK PROBLEMS

Verify MCT model software interface in LS-DYNA using:

  1. Simple shear:
    1. Verify MCT is not sensitive to mesh size.
    2. Capturing shear formation due to voids growth and coalescence.
  2. Cutting model:
    1. Material length scale based on literature results.
    2. Avoid mesh distortion and mesh sensitivity by length scale.

BENCHMARK PROBLEM 1 – SIMPLE SHEAR PROBLEM

Boundary Condition:

  • Top surface fixed in y, bottom surface fixed in x and y.
  • Apply displacement towards right on top surface.
  • Material length scale: 0.4 mm

MCT MODEL GRAPHIC

BENCHMARK PROBLEM 2 – CUTTING PROBLEM DEMONSTRATION

Goal: model shear band formation during cutting process with MCT model

MATERIAL LENGTH GRAPHIC

DEMONSTRATION

  • MCT is implemented into LSDYNA as user-defined element.
  • Multiscale and Data-driven Modeling platform is a proprietary software (we deliver the executable file but not the source code).
  • The executable file has to be compiled and run in the same environment.
  • The macroscale finite element model is built in LS-PrePost. Users can design their macroscale model.
  • The microscale RVE model is built in Matlab. Users can design their microstructure model.
  • Demonstration:

FUTURE DIRECTIONS

Concurrent modeling of Self-Piercing Riveting Process

FUTURE DIRECTIONS GRAPHIC

Material Design Enabled by Multiscale and Data-driven Modeling Platform

  • Material microstructure responses captured in macroscale by MCT
  • SCA provide fast computation of RVE responses on-the-fly

 

  • MCT and SCA enabled mechanistic multiscale modeling with microstructural evolution.
  • Effect of microstructure and constituents can be measured quantitatively.
  • Material design is possible by iterating through different material combinations and microstructures.
  • Minimum experiments required.

CONCLUSION

  • MCT has been implemented into the Multiscale and Data-driven Modeling platform.
  • Two benchmark problems are demonstrated.
  • The future development of Multiscale and Data-driven Modeling platform will allow change of microstructures for various applications.
  • The potential market is broad, including automotive, medical, shipbuilding, etc.

Download the PDF for Research Multi-Scale

Principal Investigators

Wing Kam Liu

Wing Kam Liu