Physics-informed neural networks for predicting fatigue life of flax fiber reinforced polymers

Institute

Lehrstuhl für Hubschraubertechnologie

Type

Semester Thesis / Master's Thesis /

Content

theoretical /  

Description

Project Outline

Accurate fatigue modeling is essential for ensuring the safety, reliability, and lifetime prediction of aerospace materials operating under complex loading conditions. Data-driven models can identify complex patterns but often lack physical context, while classical physics-based models capture governing mechanisms but struggle with variability and real-world complexity. Physics-informed neural networks (PINNs) bridge this gap by embedding physical information directly into the learning process. With an experimental fatigue dataset of fiber-reinforced polymers – an emerging alternative to traditional carbon fiber composites for aerospace applications – PINN-based approaches shall be developed and evaluated to assess their ability to generalize, capture underlying degradation mechanisms, and provide robust fatigue life predictions.

Work Packages

1.Literature review

2.Data Preparation and Analysis

3.Development of Physics-Informed Neural Network Approaches

4.Training and Validation

5.Evaluation and Optimization

Requirements

• Confident handling of python and common data science tools

• Experience in machine learning methods

• Experience in structural analysis and fatigue modeling

Possible start

sofort

Contact

Christoph Hackner
Room: MW2701
Phone: +49 (0)89 / 289-16362
christoph.hacknertum.de

Announcement