I was working on a pilot-scale investigation of a hybrid coagulation-ceramic membrane treatment system for my master's work. Three coagulants were considered in the study, aluminum sulfate, ferric chloride, and polyaluminum chloride (PACl). We were able to successfully optimize the pretreatment stage and significantly improve the performance of the system with both aluminum sulfate and ferric chloride, but not with PACl. This was very aggravating since almost all of the past studies using ceramic membranes successfully used PACl for their coagulation pretreatment stage. We decided to try aluminum sulfate and ferric chloride since they are common coagulants in the US. In some cases, the system that used polyaluminum chloride performed worse than the control system that did not receive any coagulant. The figure below shows a schematic of the system:
Our first theory was that perhaps the detention time between the coagulant addition point and the membrane was not sufficient. So the detention time was significantly increased by placing a 200 ft coil before the membrane thinking that the coagulant would have extra time to react with the contaminants. However, this did nothing at all to help with the problem. We then ordered a new batch of PACl thinking that the age of the coagulant might be a factor, however, that didn't seem to help either. Even after moving the pilot to a new water source, the problem persisted. Thinking that the problem might be related to the chemistry of PACl, we even consulted some PACl experts and we still couldn't figure out what was going on. At that point I was ready to accept the fact that for some odd reason PACl was just incompatible with our system.
After many months of reflecting, I thought well what if the problem was not the coagulant itself but rather how the coagulant was being dosed. Unlike aluminum sulfate, PACl comes prehydrolyzed and has to be dosed "neat", i.e. in the concentration they are supplied in (approximately 12.5% as Al). For this reason we had to use a syringe pump in order to achieve the extremely low feed rates (0.7 - 2 mL/hr). Also, since the experiment occurred over an extended time, a large syringe (30 mL) had to be used. Aluminum sulfate and ferric chloride were dosed using a peristaltic pump since we could control the concentration of the stock solutions. During the preliminary phase of the project, it was confirmed that the syringe pump can accurately dispense the target volume calculated for the duration of the experiment; therefore we didn't suspect it to be the problem.
Hypothesis: The dosing rate of the syringe pump could be inaccurate, especially under pressurized conditions and small volumes
Approach: Place conductivity sensors before and after the coagulation addition point and measure the conductivity of the water to determine if the coagulant is actually being fed
Factors: 1) Syringe Size 2) Coagulant Concentration 3) Dosing Method
The figure above shows conductivity measurements before (raw water) and after coagulant addition (membrane influent). The blue line shows the membrane pressure versus time. The membrane is backwashed every 30 minutes (as indicated by the sudden loss of pressure). All parameters were kept constant except for the syringe size. The results showed that with the larger syringe, almost no coagulant was fed in the first 5 cycles of the experiment. The membrane influent conductivity then increased slightly for 3 cycles after an approximate 15 minute lag after each backwash. The membrane influent conductivity then returned to ambient conditions for the remainder of the experiment. The experiment was repeated under the same conditions using a smaller syringe to determine if forcing the syringe motor to go faster would have any effect. The results clearly show an improvement in terms of membrane pressure and dosing of the coagulant. Apart from a 5 minute lag time at the beginning of each cycle, the coagulant was fed consistently at each cycle. However, it should be noted that this does not necessarily imply that the correct dose was being fed, but simply proves that the reason why PACl was ineffective with our system was because it was being inaccurately fed as a result of the relatively low feed rates using a relatively large syringe. Conductivity spikes were observed during each backwash cycle due to the coagulant being dispensed due to the sudden loss of pressure in the main line or by the coagulant diffusing out of the line. The lag time from using a larger syringe was found to be approximately 1.2 hours which would suggest that the syringe pump would need to overcome the back-pressure from the main line before it can start dosing the coagulant correctly. What was most interesting was the fact that the pressure immediately stabilized as soon as the coagulant was properly injected into the main line.
The effects of diluting the coagulant in order to increase the feed rate also investigated. All other parameters, including the syringe size were kept constant. The feed-rate was approximately 2.5x faster for the diluted coagulant. The results illustrate that increasing the feeding rate while keeping the PACl dose constant has the same effect as reducing the size of the syringe. A lag time of approximately 5 to 10 minutes was observed after each backwash cycle. Comparable results were observed when the same diluted coagulant was dosed using a peristaltic pump at the same feed-rate.
Since the coagulant couldn't be diluted, we thought what if we could improve the design of the coagulant feed point. The design at the time was a simple tee with quick connect fittings. We fitted a barbed fitting into the quick fitting in order to create a nozzle into the main line. The figure below shows the results a simulation of the velocity streamlines inside the old and modified tee fittings. It is clear from the streamlines that the back-pressure from main line hinders the flow of the coagulant in the old configuration. In the modified fitting the effects of the main line back-pressure is significantly reduced. The results of improving the design of the coagulant feed had the same effect as reducing the size of the syringe or increasing the feed rate for the 30 mL syringe.
Conclusions: It was apparent that the combination of a relatively large syringe dosing relatively small volumes might explain the reason PACl did not perform as expected. Although it appeared as though the 24 hour target volume was injected, the results presented above proved that either inaccurate or no feeding of PACl occurred during the filtration cycles when a 30 mL syringe was used. Diluting the coagulant was not considered as a practical solution since PACl is supposed to be dosed neat and using a smaller syringe at the scale of the pilot experiments would have limited the duration of each experiment to approximately 4 hours.
The findings of this project were published in a Water Research Foundation report "Coagulation-Ceramic Membrane Filtration For U.S. Surface Water Treatment".