Towards realistic large-scale simulations of fixed bed chemical reactors: Bridging the gap between discrete element and porous media computational fluid dynamics models

Abstract

Meeting our climate goals and achieving sustainable development requires the bulk production of renewable and carbon-neutral hydrocarbon fuels and chemicals. Industrial-scale heterogeneous fixed bed chemical reactors will play a key role in this. Their overall performance optimisation, however, requires deep knowledge and understanding of all length-scales and physicochemical phenomena involved in their operation. Such knowledge can be acquired through multidisciplinary studies, where experimental investigations are combined with validated computational tools. For the latter, Computational Fluid Dynamics (CFD) models hold great potential. Their unique ability to couple the internal bed structure with flow-related parameters can prove invaluable to advance the technology readiness level (TRL) and aid in the further optimisation of industrial-scale fixed bed reactors.
Experimental investigations of methanol synthesis from mixtures of CO2/CO/H2, using a Cu/ZnO/Al2O3 catalyst, have been used throughout the literature to understand and describe the kinetic mechanisms of the reaction. From such studies, several kinetic models have been produced, each considering unique interactions between the species and the catalyst. The accuracy, adaptability, and potential of CFD models was explored here by investigating and validating the predictions made by two kinetic models against experimental data, yielding excellent accuracy. These two models considered distinct mechanisms and roles for CO, with only one of them considering a direct pathway between CO and methanol. By comparing the flow profiles predicted by them, the key differences in the predictions made by the two kinetic models were highlighted. Specifically, the inclusion or exclusion of a direct CO hydrogenation pathway in the kinetic model significantly altered the behaviour of the involved species, with one model even predicting the loss of methanol for CO production. These observations can act as guidelines for further experimental investigations to prove or disprove the observed behaviours.
Throughout this investigation the utilised CFD model treated the bed as a porous continuum, where the bed structure was approached through effective properties and momentum sinks. Treating the catalytic bed either as a porous continuum or as a structure formed by homogeneous spherical particles, however, is an oversimplification. Realistic, lab-scale packed beds are formed by sieving, a process which offers limited control over the size and shape of the catalytic particles. To quantify this, the internal structure of six catalytic beds was reproduced through micro-Computed Tomography scans, and the size and form of each individual particle was quantified. It was observed for the first time that repeated passes through sieves of pre-determined sizes offered a higher level of control over the particle size range within the bed, promoting homogeneity, while highly irregular particles were filtered out. Moreover, repeated passes were highly efficient in greatly reducing the population of dust particles, <100 µm in size, which was prominent within the beds formed after a single sieve pass. The bulk, radial, and axial porosities of the catalytic beds were also analysed. The range of particle sizes, shapes, and orientations existing within the bed create highly heterogeneous and random structures. Their radial porosity profiles, when compared with those predicted for homogeneous spherical beds, appear smooth, as any heterogeneities are averaged out. Reproducing the radial porosity profiles seen in the actual beds would require correlations that take into account both size and shape heterogeneities. Using ethanol dehydration as a case study, it was also observed that multiple sieving passes accelerated the reaction rate and the product formation. Rationalising this behaviour simply through the outlet product formation, however, is not possible. A higher level of investigation is required, which would combine the actual bed structure with CFD simulations.
Meshing the scanned bed structure and using it as a computational geometry for CFD models to simulate the flow field through it would yield unique observations. Due to the bed complexity, however, the computational resources required for this task would quickly become prohibitive. As a result, novel computational approaches are needed, able to reproduce particle arrangements and their contribution to the flow while keeping the computational demands to feasible levels. Such a novel approach is presented here, referred to as the Semi-Realistic (SR) model. In the SR approach, the effective properties considered by porous continuum models were spatially localised. This allowed the creation of distinct interparticle and intraparticle zones with unique physicochemical mechanisms for each. With this method, the SR model was able to reproduce particle structures within the bed and replicate their impact in the flow profiles as accurately as particle-resolved CFD models. Combined with its significantly reduced computational demands, the SR model is a unique, highly flexible, and adaptable tool, which can be tailored to either the study of heterogeneous particle beds or of industrial-scale fixed bed reactors.
As list of novelties in this thesis, the following are highlighted:
1) The application of CFD models as detailed investigative tools to compare and to evaluate the predictions of kinetic models. With CFD models revealing the species behaviour and interactions within the catalytic bed, the feasibility and accuracy of the considered mechanisms can be quantified. This approach then guides experimental setups towards a deeper understanding of reaction mechanisms, promoting both novel catalyst and reactor engineering.
2) A unique insight into the morphology of realistic catalytic beds, formed by particles produced by sieving. Specifically, the Computed Tomography scans identified in-situ catalytic bed parameters, under real-world experimental setups. The produced scans provided for the first time insights into how sieving affected the shapes and sizes of the produced catalytic particles, thus identifying their highly heterogeneous nature. Through this analysis, the accuracy of semi-empirical correlations to describe the interparticle porosity of polydispersed beds was quantified. Furthermore, the scanned geometry can be meshed and used for CFD simulations, thus offering a direct coupling between experiments and simulations, and promoting the accuracy of CFD models.
3) A novel CFD method to simulate the catalytic particles within fixed bed chemical reactors without explicitly resolving them. This new method significantly reduced the computational demands required by particle-resolved CFD models, while also being highly flexible in its approach of catalytic particles. Specifically, the particles were approached through two distinct porosity terms, a macro-porosity term responsible for the hydrodynamic profile of the particle, and a micro-porosity term responsible for its physicochemical (i.e., diffusion and reaction) phenomena. Consequently, this enables CFD models to more accurately describe the intraparticle structure of particles with multi-pore-scale porosities. By showcasing the impact of dual-scale porosity particles in the full bed scale, the need for a deeper understanding and experimental characterisation of the intraparticle structure of porous catalytic particles was highlighted.

More information

Identifiers

ORCID for Robert Raja: orcid.org/0000-0002-4161-7053

Catalogue record

Export record