Comparing two IBM implementations for the simulation of uniform packed beds

Nowadays, the design of fixed packed bed reactors still relies on empirical correlations, which, especially for small tube to particle diameter ratios, are mostly too inaccurate because of the presence of wall effects. Therefore, the simulation of fixed packed bed reactors plays an important role to predict and control the flow and process parameters in such, nowadays and in the future. Because of its straightforward applicability to non-uniform packings with particles of arbitrary shapes, the immersed boundary method (IBM) has advantages over other numerical methods and is used more and more frequently. This paper compares two approaches of IBMs for the simulation of fixed bed reactors with spherical shaped particles. The classic, smooth approach is compared to the straightforward to implement blocked-off method for velocity fields above the fixed bed for particle Reynolds numbers of 300 and 500. Results from experimental inline PIV-measurements of the reactor to be simulated serve as a basis for comparison. Very good agreement with the experiment is found for both simulation methodologies with higher resolutions, considering the more stable flow at a particle Reynolds number of 300. Differences in the different IBM approaches occurred for the more unsteady flow at a particle Reynolds number of 500. Compared to the blocked-off method, the smooth IBM reflects the formation of additional jets and recirculation zones better right above the bed, though increasing the fluid mesh resolution improves the accuracy of the blocked-off method. Overall, a more diffusive behaviour is found for the blocked-off simulations due to the stairstep representation, which is avoided by using interpolation stencils as in the smooth IBM. With higher mesh refinement in the blocked-off IBM this effect can be reduced, but this also increases the computational effort.