Abstract

Enhancing Optimization on a Robotic-Flow Platform: Integrating Statistical Filtering with Bayesian Methods

Francois Xavier Felpin, Professor, University of Nantes

We recently developed a simple and comprehensive mixed variable optimization strategy. Our approach, which includes sequential sampling, statistical filtering, and black box optimization, is coupled with an automated micromole scale flow platform to perform complex optimizations with limited chemical expense and minimal human intervention. We conducted a comparative analysis of three mixed-variable optimization approaches: filter-assisted Nelder-Mead and Bayesian optimization, along with a standard Bayesian optimization approach. Regarding the filter-assisted methods, we built upon our recent Sampling-Filtering-Optimization (SFO) concept, which involves: i) sampling the continuous domain, for all the discrete possibilities, through a Design of Experiments (DoE), ii) filtering relevant discrete variables through statistical analysis, and iii) optimizing the reaction with the filtered variables. The efficiency of the sampling, filtering and optimization (SFO) strategy was demonstrated with the development of ultra-fast, sustainable and mild formal cycloadditions [3 + 3] that usually require in batch prolonged reaction times and/or high temperatures. Successful scaling experiments demonstrated the transferability from the micromole scale flow platform to a standard flow chemistry reactor.