Simulation of shear induced diffusion based separation processes

Abstract

Although microfiltration is used frequently because of the ease of upscaling and the relatively low operational costs, it also has its challenges. Concentration polarization and fouling will (ultimately) decrease the efficiency of the process and at the same time increase operational costs, and costs related to cleaning. This also implies that if these effects can be mitigated, the separation can be carried out in a more sustainable fashion. To achieve this, we have considered flow-based segregation of particles as a starting point of design, and used shear induced diffusion effects that will make particles that are typically between 0.5 and 10 micrometre migrate towards lower shear regions, which is away from the area where the pores are positioned. The migration rate depends amongst others on the particle size, and particle concentration. This allows for concentration and fractionation inside a channel, that when combined with an area with large pores, can be used to remove a permeate with different composition as the feed. Since the particles move away from the pores, it was expected that the abovementioned complications related to fouling can be mitigated, leading to an innovative separation process. In order to understand this process in detail, the thesis revolved around a computer model that was developed as a tool to elucidate the effect of process and membrane design parameters on separation efficiency. In Chapter 2 we present a computational fluid dynamics (CFD) model that describes the concentration profile of a monodisperse suspension in laminar flow, based on shear induced diffusion in a nonporous channel. The model is compared to literature and differences with experimental data are most likely due to interactions of particles with the wall. When comparing with a highly idealized experiment carried out within our group we find excellent agreement with the model predictions that were obtained. This made us conclude that the model could serve as a good foundation for further process and membrane design studies. Chapter 3 focuses on process design, and different parameters (bulk particle volume fraction, channel height, particle diameter and ratio between the height and the diameter) were evaluated on their effect on both the development of the shear induced diffusion profile as well as its steady state profile. We have incorporated one single sided pore in the model, and investigated the effect of channel height and bulk particle volume fraction on transmission and the recovery length of the shear induced diffusion profile. The transmissions found are low considering that a monodisperse suspensions was modelled, and also the recovery lengths are very acceptable; therefore we can take the next step toward more detailed membrane design. In Chapter 4 we consider different membrane designs: the number of pores, and pores on one or two sides of the channel are investigated for their effect on transmission. Double sided pores are clearly the preferred option since the shear induced diffusion profile gets less disturbed and therefore the profile also recovers faster after extracting liquid from a pore. When using multiple pores at the same flow split, the transmission decreases with the number of pores because the layer near the pores contains less particles due to the fast recovery of the profile. When looking at 10% flow split the differences in transmission are small, because the concentration profile is very flat, but as later discussed in chapter 6, this is very dependent of the flow split and the concentration of particles used. In Chapter 5 we investigate experimentally the transmission of fat from a polydisperse cream and compare that to the monodisperse model of previous chapters. At a permeate split ratio below 5% there is a clear distinction between the transmission of small and large fat globules in the cream permeate; therefore it can be concluded that the shear induced diffusion effects described earlier for monodisperse dispersions, also hold for this polydisperse suspension. The model shows a similar trend which confirms our previous findings that it can help in setting up process and membrane design We conclude with Chapter 6 where we put the findings of previous chapters in a broader perspective. The transmissions are extrapolated to higher permeate split ratios and other particle concentrations. These advanced insights are essential to design this innovative process, which is expected to be intrinsically more energy efficient than standard membrane filtration. This is due to the fact that, the novel process revolves around the use of shear-induced diffusion (laminar conditions) and the specific design of a small number of pores, contrary to regular membranes that contain many less defined pores.