Random Forests for mapping and analysis of microkinetics models

We introduce the application of an ensemble learning method known as Random Forests to microkinetics modeling and the computationally efficient integration of microkinetics into reaction engineering models. First, we show how Random Forests can be used for mapping pre-computed microkinetics data. Random Forests can be used to predict new datasets while keeping the prediction accuracy high and the computational load low. The method is also used to identify the important variables in the mechanism in regard to overall reaction rate and selectivity. The results are compared with results from a similar study using the Campbell's Degree of Rate Control approach and it is shown that the Random Forests method could be used to identify important features of the mechanism over a wide range of reacting conditions. Finally, the inclusion of the suggested method into reaction engineering models such as Computational Fluid Dynamics (CFD) resolved-particle simulations of fixed bed reactors is presented.