Industrial AI Podcast Numerical Machine Learning: Where Physics Meets AI
11 snips
Nov 26, 2025 Oliver Niggemann, a Professor of Computer Science in Mechanical Engineering at Helmut Schmidt University, delves into the fascinating intersection of numerical machine learning and engineering. He reveals how this fusion is transforming everything from diagnosing issues in the International Space Station to optimizing bridge designs. The conversation highlights the revolutionary role of surrogate models in speeding up simulations, AI-driven automated design workflows, and the critical need for interdisciplinary collaboration in today's engineering education.
AI Snips
Chapters
Transcript
Episode notes
Diagnosing The ISS Columbus Module
- Oliver describes using ISS Columbus sensor data to detect anomalies and support diagnosis at Airbus Defense.
- The project leverages years of space-station recordings to help second-level support teams find faults faster.
Marry Physics And Data For Robust Models
- Numerical machine learning combines symbolic physical models with data-driven neural methods to improve robustness and reduce data needs.
- Embedding known math into models makes them more trustworthy and better at out-of-distribution generalization.
Use Surrogates To Speed Design Exploration
- Surrogate models replace expensive simulations with fast neural approximations to explore many designs quickly.
- Use surrogates for broad search and then validate promising candidates with high-fidelity simulation.