CyRxnOpt: Generalized Interface for Process Optimization

CMI researchers at Ames National Laboratory conducted the research for this highlight

Innovations
We developed CyRxnOpt, a software library to train and deploy optimization algorithms to accelerate materials discovery and process optimization.

Achievement
A free, open-source library was created to provide a general interface for multiple optimization algorithms and digitally benchmark these optimizers.

Significance and Impact

  • Allows using available process data to virtually benchmark AI/ML optimizers quickly and cost-effectively without running actual experiments.
  • Speed up the down-selection of optimizers for a specific process or materials formula to accelerate material manufacturing.
  • Next steps include inclusion of additional optimizers and integration of CyRxnOpt into automation software for open- and closed-loop experimental studies.

Hub Target Addressed
Crosscutting tool to accelerate critical materials technology development.

Comparison of search patterns for four optimizers, helping understand algorithm behavior and efficiency. The top right may funnel to a suboptimal solution, the bottom right may favor the corners of the space, and the left two optimizers densely search the reaction space.
Comparison of search patterns for four optimizers, helping understand algorithm behavior and efficiency. The top right may funnel to a suboptimal solution, the bottom right may favor the corners of the space, and the left two optimizers densely search the reaction space.