3.1. UF and NF Membrane Characterisation
The mean permeate flux values obtained in each experiment at different temperatures regarding the UF ceramic membrane used can be observed in
Table 3. They show that the permeate flux increases with temperature, which is a logical behavior in membrane filtration processes.
Afterwards, each value is taken, and a lineal regression is applied, plotting the permeate flux versus temperature (
Figure 3). This adjustment is made by forcing the intersection through the origin, thus determining the water permeability of the membrane at a fixed TMP of 2 bar with a value of 8.18 ± 0.37 L/°C·m
2·h and a R
2 of 0.994.
As is evident, the resulting permeability parameter exhibits a regression coefficient of 0.994, indicating a good fit. Regarding the resulting value, information has yet to be found in the literature employing this type of membrane in similar experiments, making it impossible to validate the obtained result. However, studies that used ceramic membranes of 15 kDa, similar to the one used in this work [
39,
40,
41], show flux values inferior to the ones obtained experimentally in this study, presenting values ranging from 30 L/m
2·h (at 1 bar of TMP and 20 °C) to 40 L/m
2·h and 84 L/m
2·h (at 2 bar and 20 °C and 25 °C, respectively). The permeability experiments in NF resulted in the permeate values compiled in
Table 4.
A linear regression is performed, plotting the permeate flux versus the TMP (
Figure 4). Once again, this adjustment is made by forcing the intersection through the origin, determining the water permeability of the membrane at a fixed temperature of 20 °C, with a value of 8.23 ± 0.24 L/bar·m
2·h and a R
2 of 0.997.
The resulting permeability parameter presents an elevated regression coefficient of 0.997, which indicates a good fit. This value is similar to those reported by other authors [
27,
42,
43,
44], maintaining the same order of magnitude. Values of permeability 8.15 L/bar·m
2·h were observed and the same operating conditions, whereas other experiments presented values of 7.04 and 8.7 L/bar·m
2·h at an operating temperature of 25 °C. These differences could be explained by the volumetric flow applied in the experiment, as they were inferior to the experiments conducted in this study.
3.3. UF and NF Membrane Recovery Analysis
To study the effectiveness and state of the membranes used in the UF and NF processes, several experiments were conducted using the same feed solution in each step to analyze the fouling effect in each membrane as well as how the permeate flux can be recovered after the filtration of said solutions.
Regarding the UF filtration experiments, the resulting permeate flux in each is represented in
Figure 5.
Test (1) corresponds to the first filtration experiment of the lees, where it can be observed that the flux decreases drastically from 60 L/m2·h to approximately 20 m2·h. The filtration process was paused between filtration tests (2) and (3). Considering the initial flux value obtained in the experiment (1), the recovery of the membrane was calculated. In this case, it is observed that the membrane recovers up to 52.31% in test (2)
Afterwards, a cleaning with distilled water was performed to observe the membrane recovery after two filtration trials without intermediate cleaning. The subsequent results of the filtration correspond to test (4). In this case, the recovery increases to 48.79%, indicating that the water cleaning can remove some of the residual particles present in the membrane after two filtration processes. Despite recovering some of the initial permeate flux, more would be needed to consider this cleaning step effective in recovering the membrane.
Regarding test (5), a distilled water wash was performed before this trial to verify whether the permeate flux could decrease below the initially observed 20 m2·h after the solution filtration. In this case, the flux reached a similar flux value, and the obtained recovery is 44.28%. It is observed that the second water cleaning was not as beneficial as the first water cleaning due to the decrease in the membrane recovery value from 52.31% to 44.28%.
Due to the decrease in flux obtained in the trial compared to the initial one in test (1), a chemical wash with 1% NaOH was subsequently performed, as previously explained, to verify if it was possible to recover the initial flux fully.
With the collected data from test (6), the effectiveness of the chemical cleaning of the membrane is evident, as a recovery of 83.68% has been achieved. Consequently, periodic cleaning with caustic soda allows the wine lees filtration process to be repeated. The flux difference between the first and last tests recorded could be explained by the large molecules in the solution attached to the membrane breaking down into smaller ones after the washings and filtration cycles.
It is worth noting that the difference between the initial flux recorded and that obtained after chemical cleaning indicates that a percentage cannot be recovered due to the generation of irreversible fouling. This would require a more thorough washing process, increasing the maintenance cost. The flux recovery values obtained for each test are compiled in
Table 6.
The NF filtration experiments are shown in
Figure 6, in which the filtration of the UF permeate is constantly recirculated through the system.
Test (1) corresponds to the first filtration process of the permeate extracted from the previous UF stage. A decrease in the resulting permeate flux is observed, from 27 L/m2·h to approximately 25 L/m2·h after 2 h of filtration.
Once the initial test was completed, the subsequent filtration was carried out, simulating a discontinuous filtration process without a prior washing and after turning off the pump to check whether the permeate x would continue to decrease from the flux value measured in test (1). As seen in test (2), a significant decrease in permeate flux occurs, reaching 19 L/m2·h.
Following the initial tests, cleaning with caustic soda was necessary upon the reduction from 27 to 19 L/m2·h. In this way, the efficacy of caustic soda can be verified. The effect can be visualized in test (3), where the flux is considerably increased to 33 L/m2·h, surpassing the flux measured in test (1).
With the completion of test (3), feed filtration was again carried out without prior cleaning. As a result, the data corresponding to test (4) are obtained. They show a dissimilar behavior compared to the registered tests, where the flux initially increases slightly. This event can be justified due to the effect of temperature, where, initially, the feed was at a lower temperature. As the test progressed, the temperature stabilized at the desired value. With this test, the importance of temperature in membrane filtration processes can be appreciated.
Before test (5), the membrane was cleaned with caustic soda. As a result, a new increase of similar proportions to that recorded in test (3) is evident. This phenomenon is caused by the caustic soda on the membrane, which could break down part of its dense layer, and in the membrane module, meaning a higher permeate flux can be extracted. Nevertheless, this is only observed in the first cleaning cycle, as posterior chemical cleaning cycles did not further improve the permeate flux obtained.
Finally, in test (6), no posterior cleaning step was performed, resulting in a reasonably stabilized flux of 22 L/m2·h, similar to the one recorded in test (4). Therefore, chemical cleaning provokes the permeate flux to reach stable values of 20–22 L/m2·h. This behavior contrasts with that observed in previous runs due to the plant being idle for a shorter period between tests.
Overall, it can be observed that chemical cleaning is required to maintain a high permeate extraction rate, which implies that a previous filtration step would be necessary to remove as many large molecules as possible to ensure the lifespan of the membrane can be prolonged. The resulting values of membrane recovery after both cleanings compared to the initial flux of test (1) are collected in
Table 7.
To study the effectiveness of each washing process, the permeate flux using distilled water was reobtained after each cleaning step. Using the initial flux of the membrane before the filtration experiments, the recovery of the membrane is calculated, resulting in the values presented in
Table 8 for both UF and NF processes.
As can be seen, cleaning with NaOH offers the best results in terms of membrane recovery in both filtration steps, recovering up to 95% and 135% of the original permeate flux in the UF and NF processes, respectively. However, cleaning with NaOH results in permeate flux in the NF membrane being significantly higher than the measured initial flux due to a possible damaging effect on the membrane. In contrast, cleaning with distilled water is less effective, as only 36% of the original permeate flux can be recovered with this type of wash in the UF step and up to 75% in the NF step.