Model-based performance analysis of filtration devices and membrane adsorbers

Filtration devices and membrane adsorbers are commonly applied in biotechnology for separating target biomolecules from impurities. Filtration devices typically contain one or two membrane layers that are arranged in pleats around a cylindrical core. These membrane layers are held in place by fleece layers and a plastic cage. The water flow rate at different pressure drops is an important performance indicator for such filtration devices. This performance has been studied using Computational Fluid Dynamics (CFD), and the results were validated against experimental data. Four different methods for reconstructing the geometry of the pleated membrane were developed and compared. The impact of the plastic cage on the water flow rate was assessed using 3D simulations. The results reveal that this plastic cage has a negligible influence on the water flow rate except for very high permeabilities of the porous region that are not relevant for practical applications. However, the fleece layer has a significant impact on the water flow rate. The permeability of fleece layers is complicated to measure and direct experimental information was not available for this study. Hence, the unknown fleece permeability had to be determined from the measured water flow rates of the cartridges. The estimated values depend on the flow rate because they are affected by compression of the fleece layers towards the center of the device which is not explicitly accounted for in the geometry of the CFD simulations. The degree of compression also depends on the number of pleats and the core diameter. A linear dependency was found between the resistance (1/permeability) and the pressure drop, which allows predicting the full range from only two data points of a cartridge. Moreover, CFD simulations allow to transfer estimated fleece permeabilities between filtration devices with similar membrane properties. Membrane adsorbers (MA)s typically contain many more membrane layers than pleated filters in order to achieve larger bed height. Processes are often developed at small scales and then transferred to larger scales. This transfer is generally complicated by differences in the scaling behaviour of flow and binding related non-idealities. CFD simulations allow to consistently transfer model parameters across scales by independently capturing and practically decoupling the impacts of these non-idealities. In this study four commercial MAs are analyzed which contain the same membrane. They have membrane volumes from 0.08 mL to 1.6 L. The largest device has recently been introduced to the market. In addition, these so-called Cassettes can be parallelized to process volumes of up to 20.8 L and above. The estimated adsorption and desorption rates are flow rate dependent, which can be explained by a combination of convective and diffusive transport in the micro-pores of the membrane. However, the equilibrium constant and all other binding parameters are unaffected by that. Scale-up can be additionally complicated by batch-to-batch variations in the membrane capacity. This problem can be practically resolved by calibrating the CFD simulation to just one experiment with membrane from the same production batch