Dresden 2026 – wissenschaftliches Programm

Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

O: Fachverband Oberflächenphysik

O 23: Catalysis and surface reactions – Poster

O 23.5: Poster

Montag, 9. März 2026, 18:00–20:00, P2

Automatic Kinetic Equation Discovery from Experimental Data — •Maryke Kouyate, Gianmarco Ducci, Karsten Reuter, and Christoph Scheurer — Fritz-Haber-Institut der MPG, Berlin

Optimizing and controlling industrial catalytic processes requires effective kinetic models that are both robust and interpretable, capable of predicting how key operating variables affect reactor behavior. We introduce a model-based, adaptive design-of-experiments (DoE) framework that efficiently learns parsimonious, mechanistically meaningful rate expressions from data. At each step, it selects feed conditions to maximize expected information gain.[1,2] Profile reactors further boost per-run information by providing spatio-temporally resolved composition profiles along the reactor axis. Applied to CO oxidation on Pt, the framework automatically constructs a robust, chemically interpretable kinetic model that reproduces the characteristic mid-reactor transition in concentration profiles and reveals the need for a nonlinear term to capture the observed shift from CO poisoning to a non-poisoned regime.

[1] G. Ducci et al., J. Chem. Phys. 162, 114118 (2025).
[2] M. Kouyate et al., J. Chem. Phys. (in press), DOI: 10.1063/5.0289751

Keywords: Effective Kinetics; Kinetic Model Identification; Profile Reactors; Adaptive Design of Experiments; Hidden Variables