Understanding the NH3 adsorption mechanism on a vanadium-based SCR catalyst: A data-driven modeling approach

[Display omitted] •A data-driven modeling framework was developed for ammonia adsorption over a vanadium-based SCR catalyst.•The proposed model, based on the Langmuir adsorption framework, involves 5 adsorption sites.•Model parameters, such as the enthalpy and entropy of adsorption, have physical significance comparable with other studies.•The data-driven modeling framework maximizes the utilization of information from experiments.•Besides the model parameters, the adsorption model structure is considered as a variable which can be optimized. Ammonia adsorption is a precondition for the selective catalytic reduction (SCR) of nitrogen oxides (NOx) to take place and it influences catalyst performance under transient conditions. For a vanadium-based SCR catalyst NH3 adsorption takes place on multiple adsorption sites over the catalyst surface with different behaviours depending on temperature, gas concentration and catalyst oxidation state. In this study, a mechanistic NH3 adsorption model within the framework of Langmuir adsorption models was developed for describing the NH3 adsorption isotherms obtained with a gas flow reactor for a vanadium-based SCR. The model was created by a data-driven modeling process, which involves different steps. First, a large set of candidate models was created systematically by combining multiple feasible adsorption mechanisms. Then, a parameter estimation workflow was performed using three different objective functions with increased complexity. Finally, a model reconciliation step was executed and a quality assessment was done for creating a unified robust model with a high degree of validity. As a result of this method, an NH3 adsorption model with five adsorption sites with different mechanisms was obtained that captures the main features from the experimental data. Furthermore, the model parameters have physical significance and relate to the adsorption strength and spatial arrangement for NH3 and water molecules. The proposed model can be used in the development of transient models with increased validity over a wide experimental region.