Development of an AI for the generation of representative microstructures

News

07.04.2025 | Awards
The 29ᵗʰ national SAMPE Symposium highlighted composite technologies for a sustainable future

Around 80 experts from science and industry came together in the university’s main building to discuss current developments in the field of composite materials. The event was organized by RWTH Aachen University in cooperation with SAMPE

read more

Bachelor- or Masterthesis

Develop a neural network for AI-based generation of microstructures for simulation.

Plastics are often multiphase materials whose properties, whether mechanical, thermomechanical, etc., are strongly influenced by the internal microscopic structure. To investigate the range of properties, substitute models of these microstructures, so-called representative volume elements (RVE), are often created, which can then be used to carry out simulations. However, the generation of such macrostructures is complex and can be very time-consuming. In addition, the algorithms for generating RVE structures are limited to special use cases. There may also be restrictions within the use cases. For example, microstructures of fiber-reinforced plastics can only be generated for a small proportion of fibers without significant computing costs.

Stuwi.-KI-fuer-Mikrostrukturen — AI-generated image illustrating the schematic explanation of an AI-based RVE generator using thermoplastic vulcanizates as an example. On the left, you can see a microscopy image of the material, and on the right, the generated RVE structure.

One way to efficiently generate new structures is to use artificial intelligence (AI) in the form of neural networks. Several examples of the applicability of AI for RVE generation can already be found in the literature. The aim of your work will be to develop an AI-based framework using artificial neural networks (ANN), which generates RVEs for micromechanical simulation based on image recordings. This will initially be worked out using the example of a class of materials, e.g. foam structures or thermoplastic vulcanizates. The transferability of the framework to other microstructures can be checked.

The work is related to this research project:

The Structural Calculation department carries out a wide range of studies on models at the micro level. This ranges from micromechanical considerations of fiber composites to crystal structures. IKV also carries out permeation simulations of coatings on RVE. The development of an AI-based approach to RVE generation contributes in every respect to expanding the modeling potential at IKV.

Objective:

Develop a methodology and algorithm for AI-based generation of microstructures for RVE modeling.

Your task

For a Bachelor's thesis you work on the following tasks	For a Master's thesis you work on the following tasks
Literature research on existing KIRVE generators	Literature research on existing KIRVE generators
Development and implementation of a methodology for the AI generation of RVE structures	Development and implementation of a methodology for the AI generation of RVE structures
Validation of the generated microstructures and micromechanics simulations	Validation of the generated microstructures and micromechanics simulations
	Investigation of transferability to other microstructures

Your profile:

You already have knowledge and experience in programming (Python, or similar).
You are interested in AI and ideally already have experience in the development of
neural networks.
You are interested in the mechanics and micromechanics of plastics.
Knowledge of the finite element method and simulation with ABAQUS is advantageous

If you are interested in writing your thesis at IKV and in this task, please apply.