AI-supported prediction of the permeability of variable-axial FRP structures

Motivation

The development of macroscopic models for the infiltration of fiber-reinforced plastic composites (FRP) requires a large amount of permeability data, which is traditionally obtained by complex simulations or experiments. In order to accelerate this process, deep learning methods are to be used that can directly infer the permeability tensor from voxelized microstructures.

The aim of the master thesis is to develop a Convolutional Neural Network (CNN) that predicts the permeability tensor based on voxelized 3D images of FRP structures. The voxelized structures come from virtual models and permeability data are calculated using numerical simulations in OpenFOAM.

Task definition:

• Literature research on infiltration behavior
• Creation/further development of an automated workflow for data generation
• Creation/further development and training of a CNN for tensor prediction
• Evaluation and data analysis

Requirements profile:

• Degree in mechanical engineering, mechatronics or similar.
• Interest in machine learning
• Structured, goal-oriented way of working
• Programming experience in Python is a strong advantage

Start: immediately

Application: Please send your CV and transcript of records to the contact email address

Contact:

M.Sc. Dennis Weitze
Email: dennis weitze ∂does-not-exist.kit edu
Tel: +49 721 608-45383