Rescale has outlined two approaches for running Design of Experiments with its AI assistant, Rescale Edison, and Workflows. The existing DOE job type does not support Custom Fields or Automations in child runs, and offers limited algorithms. Edison aims to generate the study matrix, including Full Factorial, Latin Hypercube Sampling, Fractional Factorial, Box-Behnken, Central Composite Design, and Monte Carlo, then submit it through Workflows. Options include:
- Workflow Runs: each design point runs independently, with per-run pre- and post-processing; a failed point can be rerun alone.
- Workflow Template: all design points sit on one canvas for review before submission, with centralized monitoring and concurrency limits.
It works with any solver, and results land in Custom Fields, aiming to support AI Physics model training.
Image courtesy: Rescale
