Date of Award
Doctor of Philosophy (PhD)
Combustion of pulverized coal and biomass in furnaces and boilers involves millions of particles with a distribution of sizes and morphologies. Because char particles typically react under zone II conditions, in which pore-diffusion and heterogenous reaction both influence the rate of conversion, the morphology of the porous particles affects reactor-scale outputs. To understand the impacts of complex morphology, 3-D pore-resolving simulations employing real char geometries obtained from high-resolution X-ray microtomography are first used to study combustion of 150 coal char particles and 30 biomass char particles. Localized reactant penetration into the innermost regions of the particles is observed, facilitated by the presence of large macropores connected to the external surface, resulting in non-monotonic and non-uniform reactant distributions. In contrast, temperature distributions are nearly spatially uniform throughout both the large pores and microporous char regions.Existing analytical effectiveness factor models, which are often used as sub-models in reactor-scale simulations, are then assessed by comparison to the effectiveness factors obtained from the 3-D geometrically-faithful simulations. Conventional, frequently used uniform sphere and cylinder models significantly underpredict effectiveness factors for real coal and biomass char particles, whereas an accessible hollow cylinder model achieves good accuracy for the biomass char. Low connectivity coal char particles can be reasonably modeled using an inaccessible hollow sphere model, while combustion of coal char particles with higher connectivity can be well-represented with an accessible hollow sphere model. To facilitate modeling large distributions of particles in reactor-scale codes, machine learning algorithms are trained to classify highly porous char particles according to their expected combustion behavior and to apply an appropriate, computationally efficient, analytical particle-scale combustion model. Whereas existing approaches have classified particles solely according to their morphology and used 2-D measurements based on particle cross-sections, the present approach classifies particles according to their combustion behavior, using 3-D morphology data as input and 3-D pore-resolving simulation data for classifier training. Finally, to facilitate application of the workflow to other highly porous char particles, an automated 3-D image analysis routine is developed to segment carbonaceous regions from resolved pores and to measure the morphological parameters required by the classifiers.