Design loops
What a trusted surrogate is for: sweeping and optimising designs in seconds, choosing the next simulation where it helps most, and always verifying the winner with the solver.
Start the unit Using the surrogate to choose and optimise
- 8.1Putting a surrogate to workMillisecond answers turn design studies, sensitivity analysis and real-time tools into routine work.
- 8.2Choosing the next simulationRun the solver where the surrogate is unsure or a better design may hide.
- 8.3Optimising on a surrogateFollow the surrogate’s gradients to a better design, then re-simulate it: optimisers exploit errors.
You can thenUse a surrogate to choose and optimise designs safely.
Read and build
ReadActive operator learningOn this site · arXiv 2025Bayesian OptimizationGarnett · Cambridge University Press 2023 ↗
BuildWrite an active-learning loop around a Gaussian process, then optimise a design on the surrogate and check it with the solver.
Basics Core Advanced Frontier