Search

Driven by the current insights in sustainability and technological development in biorefining natural renewable resources, the food industry has taken an interest in fractionation of agrofood materials, like milk and cereal crops. The purpose of fractionation is to split the raw material in several functional ingredients. For example, milk can be split in fractions containing milk fat, casein micelles, and whey proteins. Traditionally, separation processes in food industry are mainly aimed at separating fluid from a suspension stream. Frequently membrane technology is used this type of separation; membranes seem an obvious choice because they are able to sieve components during mild fractionation of many foods, which are suspensions by nature, like milk, or are suspended in liquid during processing (such as starch granule suspensions). However, membrane separation is hindered by fouling of the pores by the food ingredients and accumulation of these components in front of the pore, which makes fractionation with membranes more challenging than plain separation of fluid and solids. That is why we have investigated the possibilities of alternative technologies such as microfluidic devices, and evaluated them under conditions required for food applications. Microfluidic devices are currently investigated for fractionation in biological applications, like sorting of DNA or cells. Due to the large degree of freedom in design, these devices are very suited for innovative fractionation technologies. First, we have evaluated various designs available in literature in chapter 2, which concludes that so-called deterministic ratchets are the most promising technology for fractionation of food suspensions. This conclusion is based on the high yield, compactness of equipment, and high selectivity that can be reached with such devices. In chapters 3 6, we report on detailed investigations on deterministic ratchets through 2D simulation (chapter 3), image analysis in comparison with simulation results (chapter 4), and full 3D simulations in combination with the previously mentioned methods (chapter 5). In the last chapter, our findings are summarized in classification and design rules, and an outlook for future developments is given. Deterministic ratchets are microchannels, containing a regularly spaced array of obstacles, through which the particle suspension flows. The essential property of these ratchets is that each obstacle row is displaced slightly laterally with respect to the previous row. Small particles follow the streamlines of the fluid, and zigzag around the obstacles, while particles larger than a certain critical size bump into the obstacles, and are consequently displaced from their streamline. The larger particles will continuously be displaced in a direction in which the obstacles are placed, and have a certain angle with the flow direction. The small particles are moving in the direction of the liquid flow, which implies under an angle of zero degrees. Via the difference in migration angle of the zigzag and displacement motion, particles can be fractionated, and collected from different outlets. An important property of deterministic ratchets is the size of the particles relative to the width of the so-called flow lane, which determines whether it will show zigzag motion or not. This we have investigated intensively in chapter 3 by means of 2-D flow field simulation. The critical particle size is related to the width of the flow lanes, within which the zigzagging particles will move, and we have determined the flow lane widths for various designs. The distribution of the flow lane width is found to depend strongly on the design of the ratchets. For a limited number of designs the original hypothesis of the inventors of the deterministic ratchets holds, and the flow lanes are symmetrically distributed over the space in between obstacles in one single row. In general, ratchets have an asymmetric flow lane distribution, and typically, ratchet designs suitable for food applications show a strong asymmetric flow lane distribution. An asymmetric flow lane distribution implies that there is not one critical flow lane width but two that determine the type of motion of particles inside the ratchets. As a first approach we have taken these as the first and last (and largest) flow lane width, df,1 and df,N. Consequently, particles are expected to show alternative motions that are in between zigzag and displacement motion. Its existence has become evident in the experiments described in chapter 4, and we have named it mixed motion. The mixed motion is irregular, in contrast to the zigzag and displacement motion, and has a migration angle which is intermediate between the angles corresponding to zigzag and displacement motion, 0 < _ < _max. The particles moving in the ratchets we have tracked by high speed recording, and the migration angle were quantified through tailor-made image analysis. As expected, the transitions between the different types of particle motion seem to occur on the basis of the critical length scales, df,1 and df,N. However, this conclusion can not be stated with high certainty because of the large experimental error due to the wide particle size distribution of the used suspensions. Because the ratchets used in chapter 4 has not been specifically designed to investigate various particle behaviors, we have designed new ratchets based on the critical length scales, df,1 and df,N, via 2D flow simulations, in order to allow detailed investigation. Although these critical length scales do not take all aspects that play a role during particle movement in a ratchet into account, we have stated that they can be used as an initial guideline for ratchet designs. Next, we have performed detailed and computationally intensive, 3D simulations, that include the particles. These 3D simulations are performed to check the validity of the classification rules, derived from the 2D simulations, that only include fluid flow. The simulation results show that the transition between zigzag and mixed motion occurs indeed at the critical length scale, df,1, being the width of the first flow lane. However, the length scale determining the occurrence of displacement motion is larger than the last lane width, df,N, and might even be uncorrelated with it. We have concluded that this second critical length scale, df,c, can only be determined via 3D simulations. The thus obtained classification rules are investigated experimentally and we have been able to correlate the migration angle of many observed particles exhibiting mixed motion, to the critical length scales. This makes us confident, that we now have identified the relevant critical length scales in deterministic ratchets. In the concluding chapter, we discuss the approach that we chose to ultimately derive the classification rules, and discuss the implications of the corrected length scales on the key performance indicators of ratchets, that are relevant to food applications. We find that obtaining the correct critical length scales requires computationally intensive 3D simulations. Specifically for compact ratchet designs, which are relevant for food application, the critical lane width df,c is not much different from df,N, obtained via 2D flow simulations - and 2D simulation may thus offer a more time-efficient way of estimating df,c. Further, we have discussed the existence of mixed motion in terms of selectivity during fractionation for polydisperse suspensions, and have found that the yield, compactness, and selectivity, all decrease, but at the same time it also opens possibilities for fractionation in multiple streams in one step.