Back to TUM-PREP Projects overview

Project Overview

Project Code: ED 16

Project name:

Machine Learning of Material Models

TUM Department:

ED - Mechanical Engineering

TUM Chair / Institute:

Chair of Metal Forming and Casting

Research area:

Data-driven Material Modeling in Solid Mechanics,

Student background:

Computer ScienceComputer Science/ InformaticsMathematicsMechanical EngineeringPhysics

Further disciplines:

Participation also possible online only:

Planned project location:

TUM, Chair of Metal Forming and Casting
Walther-Meissner-Strasse 4
85748 Garching near Munich
Germany

Project Supervisor - Contact Details


Title:

Dr.-Ing.

Given name:

Christoph

Family name:

Hartmann

E-mail:

christoph.hartmann@utg.de

Phone:

+49 89 289 13769

Additional Project Supervisor - Contact Details


Title:

Given name:

Family name:

E-mail:

Phone:

Additional Project Supervisor - Contact Details


Title:

Given name:

Family name:

E-mail:

Phone:

Project Description


Project description:

Introduction:
For numerical analysis of solid mechanics problems, materials must be adequately represented, for example to simulate the elasto-plastic processes during metal forming or or crash cases of mechanical structures. Currently, this is usually done by parameterised analytical descriptions, where the parameters are determined by a large number of standardised experiments for each material. The project will take a different approach. The aim is to either replace the analytical material models with a data-driven material model or to determine the parameters of the analytical models using a machine learning model. For the training, a so called information-rich experiment is to be designed using finite element simulation, where the objective is to get as much information about material behaviour out of a single test as possible. This virtual experiment is then used to sample training data for the machine learning model. Finally real experiments will be performed to validate and verify the results.

Tasks:
- Setting up a machine learning model that is capable of representing elasto-plastic material behaviour (e.g. Python, MATLAB)
- Design of an information-rich experiment and optimize it with finite element simulation
- Sample data for various material configurations to train the machine learning model
- Set up adequate analytical material models with identified parameters for simulation or implement the machine learning model into the siumlation
- Perform validation experiments and analyse the results

Outcomes:
- Trained machine learning model representing material behaviour of a chosen material
- Validated information-rich experiment for model training

Working hours per week planned:

40

Prerequisites


Required study level minimum (at time of TUM PREP project start):

3 years of bachelor studies completed

Subject related:

Other:

  • Keine Stichwörter