Data-Driven Solutions in Engineering
summary
Presents data-driven engineering solutions using neural-network surrogates and physics-informed models. Describes a virtual twin for a vibrating wind turbine where a 20-lag black-box NN fails at resonance but adding physics (an autoregressive 2-lag filter) enables correctly predicting resonance; shows CFD combined with POD to train a surrogate NN that predicts POD coefficients, giving ~1000× cheaper aerodynamic snapshots with good subsonic and usable transonic accuracy; and discusses costly full-physics laser-welding simulation (weeks per millimeter) motivating physics-informed loss functions and surrogate modeling to reduce data and compute.
AI-written summary · about
chapters (7)
- 0:00Data science in industry
- 3:12Industry 4.0 overview
- 8:30Wind turbine resonance
- 11:23Math modeling vs ML
- 17:51ML in aerospace
- 28:17Neural network for pressure
- 30:27Q&A
related talks
01h 11mHidden Physics Models: Machine Learning of Nonlinear Partial Differential Equations
01h 23mReal-Time Structural Health Monitoring Using Physics-Driven Digital Twins
45mSolving Inverse Problems with Deep Learning
32mData-Driven Inverse Modeling with Incomplete Observations
31mScientific Machine Learning With NVIDIA PhysicsNeMo
37mNonlocal Physics-Informed Neural Networks - A Unified Framework for Nonlocal Models
22mAI Systems Engineering - Challenges of a new professional discipline
29mHow Does Data Science Revolutionize the World of Machines?