Shear stress on membrane surface varies depending on the location in a module since direct air scouring effect, upflow velocity, particle concentrations, etc. are site specific. For instance, even in a very short hollow fiber module with 120 mm length, some bubbles escape from the fiber bundle even though air nozzle was installed in the center of the module. This means only small number of bubbles directly contact the top portion of the fibers (Chan, 2007). Liquid upflow velocities are also quite different inside and outside the fiber bundle due to the differences in flow resistances. In addition, particle concentration inside the fiber bundle is higher than outside since mixed liquor looses permeate while it penetrates into the fiber bundle. Overall, the fibers located near the center of membrane bundle are exposed to higher membrane fouling potential than other fibers.
When nine hollow fiber membranes were mounted tightly as Fig. 1 and the flux of each individual fiber was measured at an average flux of 18 LMH, the highest fluxes were found in the four corner fibers while the lowest flux was in the center fiber (Yeo, 2005, 2006). It is clear that the bubbles and liquid flow passing near the corner fibers experience the least resistance since the fibers are neighboring with only two fibers. Therefore, the membranes in four corners are exposed to the fasted flow movement that causes least membrane fouling.
On the contrary, the upflow velocity and the turbulence around the center fiber is the lowest due to the hydrodynamic resistances inside the bundle. In addition, particle concentrations are the highest in the center since particles are concentrated while they move into the center. The concentration effect becomes more significant as membrane bundles become larger. As a result, membrane bundles have evolved smaller in general in the last decade.
If bubbles are supplied from the external space of the bundle, eddies are formed in the bottom of the fiber bundle due to the upflow passing near the bottom permeate header as shown in Fig. 2. Due to the low flow rate in the area, dead zones can be developed, where local fluxes are substantially lower than in other places. However, the dead zone effect is not significant in full scale modules since it is a small portion of the large membrane surface area.
Fig. 1. Position of nine fibers and needles for aeration (Yeo, 2006).
Fig. 2. Flow pattern in the bottom of hollow fiber bundle.
One mechanism called “permeate competition” has been suggested to explain an experimental finding that the center fiber had a lower flux than other fibers even in clean water, where no membrane fouling supposedly occurs (Yeo, 2006). The apparent filtration resistance of center fiber was measured at 3.4 x 10-12 m-1 while the average resistance of other fibers was 2×10-12 m-1 using the same fiber bundle illustrated in Fig. 1. The permeate competition mechanism suggests that obtaining permeate from the center fiber is harder than obtaining the same volume permeate from the other fibers because the permeate must travel through the fiber bundle that can act as a hydraulic resistance. As a result of the pressure loss during the permeate transport to the center of the bundle, pressure near the center fiber is lower than external spaces, which leads to a lower TMP. This mechanism was also cited positively by Braak (2011).
Although it is hard to explain perfectly why the center fiber had lower flux than other fibers in clean water, the mechanism suggested does not seem to be reasonable. As shown in Fig. 1., there are plenty of spaces among the surrounding fibers through which very small net amount of permeate travels. The transversal flow velocity on membrane surface is only 1.4×10-5 m/s even if flux is as high as 50 LMH (=50/1000/3600). Since there are eight 0.5 mm entrances that makes 4 mm spaces total, the effective net water velocity in the spaces is only 1.1×10-5 m/s (=1.4×10-5x 3.14/4). The exact pressure loss thorough the fiber bundle is not easy to estimate, but it is easy to come to a conclusion that such slow flow velocity cannot generate a pressure loss of even 1 Pa based on Hagen-Poiseulle euqation, which is a negligible fraction of TMP.
© Seong Hoon Yoon