KI4MaterialModeling

AI to generate simulation models for material development from CT data.

In modern material development, geometry models (GM) are used to numerically predict material properties for component or process simulations. The exact modeling of the material structure and microscopic objects is essential for quantitative simulation.

The GM are currently generated by two methods:

Method A: 3D imaging techniques, such as Computed Tomography (CT), are used to generate 3D images of the material structure, which are segmented based on the gray values. In this process, the limited resolution and image artifacts lead to erroneous GM.

In method B, the GM are generated synthetically based on statistical data from real structures. This requires a high number of generation parameters to represent real structure variations. The advantage here is that all objects are known and can be varied selectively.

The project goal is the development of an AI-based workflow using the example of fiber-polymer composites which closes the gap between the methods and enables the generation of a precise GM from CT data.

The workflow is based on three pillars:

  • the generation of synthetic CT data with realistic artifacts,
  • an AI for information enrichment,
  • and an AI for object detection in CT data.

This AI-based workflow enables tremendous increase in efficiency of material analysis and development by creating high-quality GM from real CT data through object recognition (optimized method A). Structural information from CT data is automatically extracted to greatly simplify synthetic GM generation (optimized method B). Real and synthetic CT data is combined to create a larger hybrid GM with high resolution (synergistically combined method A+B). The textile permeability use case promises broad transferability to other classes of materials, such as filtration media, batteries, and rock physics.


Our project partners:

Institut für Verbundwerkstoffe (IVW)

Start: October 2023
End: March 2026

Project members

Dr. Oliver Rimmel
Dr. Dominik Michel
Dr. Rolf Westerteiger
Dr. Dennis Mosbach
Andreas Grießer

Project members

PD Dr.-Ing. habil. David May
Tim Schmidt
Benedikt Boos
PD Dr. rer. nat. habil. Martin Gurka

Latest developments

Kick-off meeting for the KI4MaterialModeling project

The KI4MaterialModeling project journey has officially begun!

Our kick-off meeting took place with our valued project partners at the Leibniz-Institut für Verbundwerkstoffe GmbH - PD Dr.-Ing. habil. David May, Tim Schmidt, Benedikt Boos, PD Dr. rer. nat. habil. Martin Gurka. Together, we are dedicated to enhancing Artificial Intelligence for the generation of simulation models for material development from CT data.

The KI4MaterialModeling project starts

The KI4MaterialModeling project has officially started and aims to bridge the gap between synthetically generated geometry models and geometry models derived from CT data through the targeted use of AI methods. We look forward to working with our partners on this exciting two and a half year journey.

Funding

We thank the Federal Ministry of Education and Research (BMBF) for the ongoing funding under the funding code 01IS23054 A