Distribution of foulants in fouled and chemically cleaned modules

Based on the literature presented in section 3.10, the amount of fouling was expected to vary vertically and horizontally in the module. The fouling on the membranes was analysed based on the ATR-FTIR spectra. Because the spectra were normalized, the amount of fouling can be expected to correlate with the intensities of the peaks. This is demonstrated by Fig. 6, where spectra of three samples with different cake contents (in w-%) are presented. The cake contents of the samples were determined in conjunction with the extraction experiment.

Figure 6 FTIR spectra of three samples with different cake contents. The cake contents and sample names are shown in the legend in parentheses. Samples were dried in oven at 65 °C. Reference sample is represented by the black line.

As seen from the spectra in Fig. 6, the intensities of the three distinctive peaks p1, p2 and p3 correlated with the cake mass fraction of the fiber samples. Thus, while the intensities can’t be directly translated to foulant concentration, they can be used to estimate foulant amounts on the samples. Based on the previous section, the three peaks have been linked to different chemicals compounds:

• Intensity of p1(at 3380 cm^-1) caused by O-H stretching indicating organic fouling

• Intensity of p2 (at 1600 cm^-1) caused by COO^- or C=C stretching indicating humic substances

• Intensity of p3 (at 487 cm^-1) indicating possibly metal and/or organic fouling.

Comparison of these peak intensities has been applied in Fig. 7 and Fig. 8, which compare the horizontal (variation in the distance to the center line, r) and vertical (variation in distance to the open-end, d) fouling gradients in the fouled module and a chemically cleaned module. The intensities presented in Fig. 7 and Fig. 8 were calculated by extracting the intensity of the clean membrane spectrum from the fouled membrane spectra at specific wavelengths of p1, p2 and p3.

Figure 7 Surface fouling distribution in A) fouled membrane module (Mf) and B) chemically cleaned membrane module (Mc2) based on the intensity difference of ATR-FTIR spectra between the fouled samples and the clean sample at three specific wavelengths.

The foulant concentrations and placement (presented in Fig. 7) varied remarkably between the two modules. Near the open-end of the module (d=15 cm), the fouled module (Mf) was the most fouled at middle ranges of r (estimated 4–5 cm), while the chemically cleaned module (Mc2) was more fouled near the center (r < 2.6 cm). In other area, either at the middle of the module (d=60 cm) or near the closed-end (d=15 cm), module Mf had clearly more surface fouling than Mc2 indicating that the chemical cleaning was effective at these areas. In general, the chemically cleaned module did not have any visible cake layer in these areas, while the Mf had some. Fig. 8 presents the vertical fouling gradient at fixed distance away from the center (r=5.6 cm).

Figure 8 Vertical fouling gradient at in two modules at r=5.6 cm: A) Mf, and B) Mc1. Based on the intensity difference of ATR-FTIR spectra between the fouled samples and the clean sample at three specific wavelengths.

The vertical fouling gradients (presented in Fig. 8) varied between the two modules. In the fouled module (Mf), surface fouling increased while approaching the open-end from the middle.

However, in the chemically cleaned module (Mc2), the fibers were most fouled at an area between the open-end and the middle of the module (d=15–30 cm). Another difference was that Mf was slightly more fouled near the closed-end than at the middle (also seen in Fig. 7).

Overall, the chemical cleaning procedure was ineffective at removing cake layer from the modules but rather it seems likely that the cleaning process pushed foulants towards the center of the modules, both horizontally and vertically. The results also indicate, that when the filtration was restarted after a chemical cleaning, foulants may have started to move away from the clogged center towards areas, where the permeate flux was the highest. This might be the best explanation for why the areas, which were badly fouled in the chemically cleaned modules, had less fouling in Mf. Thus, the fouling profiles of Mf can be used as some indication about where

the permeate flux has been the highest during filtration, because areas were permeate flux is high can be expected to foul more rapidly than areas with low permeate flux (as mentioned in section 3.10). However, it should also be noted that there are other factors to fouling besides the permeate flux, such as hydrodynamic conditions. The areas near the outer border of the modules most likely had increased crossflow velocity and were mixed better, which could be why they were generally a bit cleaner in Mf than areas couple centimetres towards the center.

The increased fouling of areas near the closed-end compared to middle parts of the module in Mf could also be explained by uneven hydrodynamic conditions, due to the bottom of module restricting water flow. Overall, it appears likely that areas at the outer border near the open-end had the highest local permeate flux of all areas. It should be noted that while the chemically cleaned modules and Mf both contained high amounts of cake layer, the total fouling resistance (Rt) still recovered a lot during the chemical cleanings, as seen in Fig. 4. It could be that by pushing the cake layer away from the most active filtration areas, the chemical cleanings were still somewhat effective at recovering permeate flux. It appears that the fibers in the center of the module might have been ineffective at filtrating the feed water at all points of the filtration periods, because they were either clogged by the cake layer (start of filtration) or surrounded by it (end of filtration).

As stated in the section 5.3 concerning sampling, the horizontal fouling gradient was assumed be symmetrical around the centreline. The three chemically cleaned modules were also assumed to be mostly identical regarding fouling. If these assumptions were not correct, the reliability of the produced results would suffer and the sampling could not be considered comprehensive. To test this assumption, samples with same coordinates (r,d) were selected from different modules and from different sides of the same module, and ATR-FTIR analysis was conducted for these samples. The results are presented in Fig. 9. Overall, the compared samples had similar spectra, which suggests that the assumptions made for the sampling were correct. There is still some deviation in the badly fouled samples, most likely due to the non-uniform distribution of the dried cake layer. Also because the long fibers in the modules were only attached to both ends of the module in a loose manner, there might have existed some variation to local packing density, due to fibers being able to move past each other.

Figure 9 Comparison of ATR-spectra of fouled membrane samples from different modules at same coordinates: A) r=5.6 cm, d=15 cm, B) r=5.6 cm, d=60 cm and C) r=3.6 cm, d=15 cm (Op. indicates that the sample was taken from the opposite side to the other sample from the same module).