Fixed bed catalytic reactors (FBRs) are the most commonly used reactors in industrial practice. A commonly used type of FBR is a tubular reactor packed with spherical catalyst particles, and often small tube diameter-to-particle diameter ratios are realized. This makes an analytical description of the phenomena in these FBRs difficult, and computational or experimental techniques must be used to access reactor performance. In this talk we will focus on a combined computational (using Direct Numerical Simulations, DNS) and experimental investigation of a heterogeneous reaction relevant for fine chemicals production.
As a model reaction, an exothermic esterification to produce acetylsalicylic acid was investigated. This reaction takes place on the surface of an organic catalyst (Amberlite IR120). It was found that due to the organic nature of the catalyst particles, solvent can diffuse into the solid phase which causes particle swelling. The swelling leads to a significant bed compaction which (i) on the one hand influences the transport processes, and (ii) on the other hand complicates the computing of characteristic geometric bed properties such as the particle volume fraction or the surface area. The particle bed geometry was simulated applying a Discrete Elements Method (DEM). By Introducing a Monte-Carlo integration method it was possible to determine the particle volume fraction and the (for chemical reaction available) particle surface area. Investigations on a single sphere were used to quantify effects due to heat transfer outside and inside the catalyst particles, and its influence on the reaction rate. In subsequent DNS studies the reactive flow inside a short section of the particle bed was investigated and process relevant quantities (i.e., pressure drop and conversion) were computed for different reactor operation conditions. Extreme spatial grid refinement near the particle surface was employed to resolve concentration gradients in this high Schmidt-number flow (Figure 1). An analytical model was calibrated by use of the DNS data which makes it possible to precisely predict the overall reactor performance of packed bed reactors.
The authors acknowledge funding through the “NAWI Graz” project by providing access to dcluster.tugraz.at. SR acknowledges funding through the NanoSim project (http://www.sintef.no/projectweb/nanosim).
Figure 1: Concentration distribution on the surface of a reacting particle bed at high Schmidt numbers (the velocity profile is illustrated at the outlet of the domain).