Comparison of Derivative‐Free Algorithms for their Applicability in Self‐Optimization of Chemical Processes

In this work, several implementations of different derivative‐free optimization algorithms are compared for the usage in chemical process optimization. As such, a benchmarking process is carried out, using optimization problems of different types to compare reliability, accuracy, and performance. Finally, using an automated reaction setup and a bespoke Python‐based script featuring a graphical user interface, all algorithms are tested in an optimization of a Suzuki‐Miyaura cross‐coupling reaction in continuous flow. To increase the scope of comparison, a model function based on the reaction is also used, to allow for a more in‐depth comparison without the use of physical resources. Currently unused black box optimization algorithms are compared against two popular self‐optimization methods, the Nelder‐Mead algorithm and the SNOBFIT method. Using various test problems and a real‐life continuous flow Suzuki‐Miyaura coupling reaction, two promising methods which surpassed the predominant methods in terms of reliability and speed of convergence were identified.