Modelling of Catalytic Monolith Reactors for Exhaust Aftertreatment

Abstract: The incomplete combustion of liquid or gaseous fuels in the internal combustion engine inherently produces several toxic emissions that need to be  removed. This is done through a series of catalytic converters, referred to as the exhaust aftertreatment system (EATS), each catalyst with its own purpose. To cope with increasingly stringent emission legislation for the automotive industry, the performance of these catalysts needs to improve. For their development, modelling is an indispensable tool. This thesis presents a mathematical model, a so-called single-channel 1+1D reactor model,  that describes the reactions that occur inside a diesel oxidation catalyst (DOC) - along with the heat and mass transport. The purpose was to improve an already existing model using relevant experiments such as kinetic experiments using a synthetic catalyst activity test (SCAT) bench, gravimetric analysis (GA), temperature programmed desorption (TPD) as well as scanning electron microscopy (SEM). The inlet conditions for the used catalyst sample in a injection-based SCAT bench along with important model assumptions was validated rigorously. A  smart experiment, using a small active DOC core positioned in various radial positions within an inert, larger monolith revealed that there was a radial concentration maldistribution within the SCAT bench. After developing a modified SMX mixer and running the experiments again, it was concluded that the SCAT bench fulfilled the assumptions for single channel models, while still maintaining its transient benefits compared to the alternative premixed design. After the SCAT bench inlet conditions were verified, an efficient experimental method was developed for identifying mass transfer limitations for a monolith reactor. This experimental method constituted the basis for the experimental observations in paper III and IV. It was shown, using ratios of various timescales that NO oxidation on Pt/Al2O3 can be internally mass transfer limited already at 175◦C. This agreed fairly well with the modelling findings of paper IV using the same timescales. Another important modelling assumption that was investigated was the one regarding the assumed uniformly distributed washcoat. This assumption was challenged by developing a parallel, tangentially resolved mass transfer model with local effective diffusivities established using a combination of SEM and IGA. The model shows great promise for monolith reactors with high washcoat loading, however, it showed to be redundant for the catalysts used with uniform washcoat distributions. It might still find uses in reactors that utilize differences in internal mass transfer to promote selectivity for series reactions. To accompany the experiments performed, without having to perform dedicated experiments for each unknown model parameter, a robust parameter estimation algorithm using response surface methodology (RSM) was integrated into the 1+1D reactor model in Matlab. The tool utilized a design of experiment (DoE) to numerically approximate the objective function for gradient and step size determination. It was shown that different DoEs require different step sizes but a more sophisticated DoE can allow for a faster optimization. Other important aspects for the tuning was the weight function to weigh together the temperature resolved residual - a temperature dependent weight function improved the fit and boosted the importance for external and internal mass transfer parameters. Lastly, the use of an additional, inert washcoat in the experiments showed to be a vital observation for tuning the internal mass transfer. These findings are relevant for accurate and predictive modelling of catalysts for the automotive industry.