FlowBO: A Flow Chemistry Bayesian Optimization Framework Benchmarked by Kinetic Models

14 February 2023, Version 1
This content is a preprint and has not undergone peer review at the time of posting.

Abstract

The applications of flow chemistry (continuous flow reactions) in the synthesis of pharmaceuticals and fine chemicals require more advanced optimization algorithms to guide laboratory-scale and industry-scale optimization. Although several Bayesian Optimization (BO) frameworks have been developed, they are rarely equipped with state-of-the-art noise-handling acquisition functions and have not been benchmarked by multiple real-world continuous flow kinetic models. In this study, we developed FlowBO for flow chemistry, equipped with the recently-developed MOO algorithm qNEHVI that can better handle experimental noise and make parallel recommendations. Also, five kinetic models built from experimental results, including four series reactions, were used as benchmarks for FlowBO and two other recognized BO frameworks. The results show that FlowBO outperforms in all four series reaction cases with optimization results >99.9% for conversion and selectivity. At the same time, FlowBO offers a range of optimum advantages with a wide choice of temperature, residence time, and reactant concentration, facilitating process optimization for subsequent steps (i.e. separation).

Keywords

Bayesian optimization
flow chemistry
continuous flow
self-optimization
multi-objective optimization
kinetic models
acquisition function
machine learning

Supplementary materials

Title
Description
Actions
Title
Supplenmental Information
Description
Supplemental Tables and Figures
Actions

Comments

Comments are not moderated before they are posted, but they can be removed by the site moderators if they are found to be in contravention of our Commenting Policy [opens in a new tab] - please read this policy before you post. Comments should be used for scholarly discussion of the content in question. You can find more information about how to use the commenting feature here [opens in a new tab] .
This site is protected by reCAPTCHA and the Google Privacy Policy [opens in a new tab] and Terms of Service [opens in a new tab] apply.