DoE framework for catalyst development based on soft computing techniques

Abstract

[EN] Soft computing techniques have recently been successfully applied in the chemical engineering field for reducing both high temporal costs and financial costs. A generalist, configurable and parameterizable experimental design framework has been developed for the discovery and optimization of catalytic materials when exploring a high-dimensional space. This framework is based on a Soft Computing architecture in which neural networks and a genetic algorithm are combined to optimize the discovery of new materials; and process conditions in catalytic reactors at the industrial scale. Users can configure the parameters of the soft computing technique, adapting it to the specific features of each experiment. Moreover, a new problem codification for genetic algorithms has been developed, which deals with optimizations simultaneously considering complex catalyst formulations and different synthesis/testing conditions as variables.

The proposed framework has been employed in several examples, in which both the developed codification and the framework parametrization have been tested.