Melodie Deniz Christensen, University of British Columbia

Autonomous process optimization involves the human-intervention free search of an input parameter space in order to minimize or maximize output parameters of interest. These systems automate the repetitive and manual approach to process optimization that currently dominates the field. The field of autonomous process optimization is in its infancy and a majority of examples have focused on flow reactor-based systems, however, not all processes can be executed in flow reactors due to heterogeneity or long reaction times. We extended this novel technology to high-throughput microscale batch reactors through the integration of commercial Chemspeed robotics platforms with online analytical instruments and optimization algorithms using a Python interface. The first system was applied to the optimization of a stereoselective Suzuki-Miyaura process, and the second to the optimization of a Wohl-Ziegler photobromination process. We found that optimization performance was impacted by the definition of the search space, the appropriate representation of chemical structure, the optimization algorithm, the acquisition function and the selection of the appropriate reactor and analysis technology. These aspects are discussed in detail for each case study. Finally, the analysis of multivariate optimization data proved to be a challenging proposition, and we found that qualitative machine learning approaches such as random forest modelling allowed for the determination of parameter influences in an effective way. The strategy that we have implemented for autonomous process optimization has the potential to be scaled to additional robotic and analytical instruments, as well as optimization algorithms, through its modular Python interface. We envision a future where the utilization of such systems is common place.

University of British Columbia Library
DOI: 10.14288/1.0422385

For more information about Chemspeed solutions:

FLEX ISYNTH for Library Synthesis Small Scale

FLEX ISYNTH Library Synthesis with online NMR




Contact us to learn more about this exciting article: