(98c) CFD Modeling of Flue Gas Separation with Hollow Fiber Modules and Experimental Verification Via 3D Printing

CO₂ removal from flue gas is an important emerging application for membrane technology. A salient challenge is the low concentration of CO₂ in flue gas that coupled with use of highly permeable and selective membranes leads to rapid driving force loss within modules.

CFD is a powerful tool for optimizing module performance. Simulations are reported for a laboratory scale hollow fiber module that are the first to include the full module geometry for comparison to experiment. Geometries were selected to emphasize potential detrimental effects from poor flow distribution (at fiber bundle/case interface and between shell inlet/outlet ports) and axial concentration gradients that arise from permeation through highly selective membranes.

Module cases were 3D printed to control internal dimensions and external port locations. Thin film composite hollow fiber membranes were formed with a Pebax 2533 coating and axially aligned using buttons to control spatial location. Figure 1 illustrates one module. Mixed gas permeation experiments were performed with feed flowrates, pressures, and compositions for CO₂/N₂ mixtures representative of carbon capture applications.

Module performance was evaluated using COMSOL Multiphysics®. Simulations evaluated velocity, pressure, and concentration fields in: 1) shell (external to fibers), 2) lumen (internal to fibers), and 3) membrane domains. Two performance metrics were calculated: 1) retentate recovery (ratio of retentate to feed flowrate) and 2) dimensionless feed flowrate (ratio of feed flowrate to product of CO₂ permeance, membrane area, and feed pressure).

Experiment and simulation were in good agreement (Figure 1). As the external ports are moved away from the module tubesheets, simulations capture the reductions in performance metrics that occur. Moreover, simulations capture the subtle effect of axial concentration gradients that occur due to selective permeation and are detrimental to performance.

Figure 1. Sample experimental module (left) and comparison of simulation to experiment (right).