Planes, Trains, and Skateboard Shoes - Bayesian methods in engineering and product design
summary
Uses Bayesian optimization with Gaussian processes and an acquisition function to tune real designs: a ducted‑fan with eight geometry parameters evaluated by CFD in star CCM+ (each run ~4 hours, ~10 runs finds a minima ~40 hours; full campaign hits ~120 iterations over two weeks) with a human‑in‑the‑loop web app; a skateboard sole optimized for grip and Voronoi‑based cushioning via Rhino/Grasshopper and HyperWorks; and an aesthetic smoke animation optimized by pairwise comparisons driving Blender renders, converging to preferred results.
AI-written summary · about
chapters (8)
- 0:00Bayesian methods in design
- 2:37Bayesian optimization for design
- 7:18Gaussian processes and uncertainty
- 15:42Acquisition functions for decisions
- 18:09Engineering case ducting drones
- 24:11Product case skateboard shoes
- 28:10Optimizing smoke aesthetics
- 33:04Q&A
related talks
28mBayesian Optimization- Fundamentals, Implementation, and Practice
48mBayesian Optimization of Gaussian Processes with Applications to Performance Tuning
24mUsing Bayesian Optimization to Tune Machine Learning Models
46mBayesian Optimization Accelerator
34mBayesian Optimization and it's application to Neural Networks
39mBayesian optimisation with scikit-learn
01h 30mBayBE: A Bayesian Back End for Experimental Planning in the Low-To-No-Data Regime
42mSolving Real-World Business Problems with Bayesian Modeling