(671h) Capturing Mesoscale Structures in Multiphase Reactor Simulations

Multiphase reactors, such as fluidized beds and bubble columns, are the workhorse of the chemical industry. The performance of these reactors is determined by the mesoscale structures that form during their operation, such as gas bubbles in a fluidized bed or a bubble column. These mesoscale structures determine the residence time of the reactant and product species, pressure drop, and heat and mass transfer rates. Thus, accurate quantification of the mesoscale structures present in multiphase reactors is essential. Obtaining this information experimentally is expensive and challenging. The recent advances in supercomputing facilities and numerical techniques for multiphase flows allow high-fidelity CFD simulations to investigate multiphase reactors. However, fast postprocessing of the simulation data to obtain quantitative characteristics of the mesoscale structures is challenging. In this regard, we develop a methodology based on the commonly used DBSCAN algorithm - an unsupervised machine learning algorithm. The methodology’s predictions are assessed against various analytical shapes. Moreover, CFD simulations of two-dimensional and three-dimensional multiphase reactors are performed, and the simulation data is analyzed using our methodology. The obtained mesoscale characteristics, such as bubble size as a function of reactor length, are assessed against existing correlations in the literature. The computational complexity of the methodology is calculated by applying it to simulation data generated on up to 100 million mesh elements. The computational complexity is found to be , which is much faster than the existing algorithms for quantifying mesoscale structures.