HomeContact Login
ENDE

Automated Generation of Digital Twins of Granular Media

GeoApp: Generate Granular Statistical Digital Twin

The analysis of complex granular microstructures is a central element in the engineering of certain digital materials – and presents many companies with the challenge of how to create a reliable digital model with realistic physical properties from limited imagery.

This is precisely where the Generate Granular Statistical Twin GeoApp is used: Based on 3D-, or even 2D-image data, it automatically generates statistical digital twins (SDTs) of granular microstructures. Not only do these digital twins replicate the stochastic geometric properties of the original - an integrated optimization process also ensures that the physical properties are precisely matched to those of the real material.

With SDTs, different material structures can be digitally analyzed, varied and further developed - a decisive advantage for anyone who wants to make material development more efficient and data-driven. Most innovatively, 3D geometries can be generated and properties such as permeability can be accurately predicted even from 2D data such as thin sections or SEM images.

Main features:

  • Selection of physical and geometrical properties: Specify which properties should match between the SDT and the original, such as grain size distribution, pore size distribution, specific surface area, permeability, or tortuosity.
  • Optimization algorithm: The Nelder-Mead algorithm iteratively adjusts the structural parameters to achieve the target values.
  • Generation of the digital twin: At each optimization step, a new structure is generated through three process steps: grain packing, distance field calculation, and thresholding with up to two Gaussian Random Fields (GRFs).
  • Error analysis and adjustment: Deviations from the reference structure are automatically analyzed and the SDT is optimized accordingly.

The result is a highly accurate digital representation of the microstructure - adaptable, flexible, and simulation-ready. The combination of GeoDict and GRFs forms the basis for a modern, automated modeling concept.

GeoApp Features

Workflow setup

The Generate Granular Statistical Twin GeoApp provides a complete, automated workflow to generate statistical digital twins of granular microstructures:

  • The GeoApp generates a digital image of the microstructure by loading a 2D or 3D data set.
  • A statistical digital twin is then created, and its properties are compared with those of the original – including pore size distribution, chord length distribution, permeability or diffusivity.
  • The reconstruction can be carried out in different resolutions and volume sizes – also independently of the optimization loop.

The result is a powerful tool for data-driven granular design and optimization.

Feature summary

  • Reference structure analysis: Physical parameters of the original structure are automatically recognized and prepared for modeling.
  • 2D-to-3D: 2D slices, such as SEM scans, are sufficient to calculate 3D properties such as permeability, taking into account 2D-specific constraints.
  • Integration of GeoDict: The GeoApp is fully integrated into GeoDict and uses GRFs for realistic modeling.
  • Error analysis: Each iteration is checked for deviations based on an error function.
  • Automatic adjustment: Structural parameters such as grain sizes or GRFs are dynamically adjusted.
  • New modeling methods: Combination of Gaussian random field with distance map for realistic grain structures.

Generating statistically and physically consistent microstructures even from 2D image data has never been easier - with the Generate Granular Statistical Twin GeoApp.

A license is required for the following modules:

Required module GrainGeo
GeoApp: 2Dto3D (if this setting is activated)

Following modules are often used in combination with the "Generate Granular Statistical Digital Twin" GeoApp:

Import & Image Processing ImportGeo-Vol    
Image Analysis MatDict PoroDict GrainFind
Material Modeling GrainGeo    
Simulation & Prediction FlowDict DiffuDict
Interfaces      

Suitable modules depend on the specific application.