Heat transfer and flow in packed beds with nuclear magnetic resonance microscopy and computational fluid dynamics

Abstract

Fluid flow and heat transfer characteristics in packed beds are studied extensively due to its importance in different fields. The macroscopic and continuum approaches used for analysis require a degree of empiricism and theoretical assumptions. Pore-scale models drive out the need for empiricism and theoretical assumptions but cannot be validated due the lack of accurate pore-scale experimental methods. This thesis presents a novel method that utilizes Nuclear Magnetic Resonance (NMR) techniques to map the pore scale melt fraction and velocities within packed beds, non-invasively. An initial experiment was conducted where heated Nitrogen was flowed through a packed bed filled with PCMs. The increasing signal intensities due to the melting of these PCMs were captured using a 1H tuned coil. Another experiment was conducted where heated Fluorinert was flowed through a packed bed filled with PCMs. The melt front of the PCMs and the velocity of the Fluorinert was imaged using a 1 H/19 F dual tuned coil. Discrete Element Modelling (DEM) was used for the generation of randomly packed beds that mimic the experimental packed beds. These numerical packed beds were modelled under the same inlet conditions as in experimental work to yield models that showed similarities to the processes seen in experimental results. Numerical work analyzed the effects of particle size and geometry on flow, heat transfer and pore structure. Three models were developed: a packed bed of monodisperse spheres, a bed of spherical particles with a Gaussian distribution in diameters and a bed of non-spherical particles with a Gaussian distribution in diameters. It was concluded that the beds of spherical and non-spherical particles with a Gaussian distribution in diameters yielded the best complementary results to the experimental work. These numerical models and the experimental work yielded maximum velocities in the range of 6 mm/s to 8 mm/s, while showing similar attributes such as intra-particle melt gradients, preferential flow pathways and channeling effect. Experimental work shows a melt front of 60 mm in 41 minutes while models yielded a melt front of 18 mm in the same time.