alum dosing

Alum Dose

Alum dosing has been an important variable throughout flocculation history. It is considered an art form and is generally picked up through practice and experience. At the Cornell Water treatment plant the operators rely on past data and a streaming current director to establish their alum dosing. They also rely on some rules of thumb that are affected by the temperature and the turbidity of the water. It appears that the higher the temperature the less alum is needed. This data is displayed in Table 1.
Table 1. Rule of thumb data used by Cornell University's Water Treatment Plant Operators.
|| || Temperature > 10°C || Temperature < 10°C ||
|| NTU || Alum dose (mg/L) || Alum dose (mg/L) ||
| 1 | 17 | 10 |
| 10 | 27 | 20 |
| 50 | 43 | 34 |
| 100 | 60 | 46 |
| 200 | 77 | 60 |
In an attempt to automate and decrease the mystery behind alum dosing the log relationship equation (Y = A + B*log(NTU)) was used to test its impact on flocculation in the pilot plant flocculator. After the value of A was lowered and the tube settlers stopped clogging there appeared to be good floc formation and clean water being produced in the flocculator. While the third section appeared to break floc ups the alum dosing was sufficient so that by the end of the second section the turbidity was usually around 1 NTU and was almost always below 2 NTU (see Figure 12 and Figure 18). The raw water turbidity coming into the flocculator stayed between 2 and 6 NTU during most tests that were run. While the turbidity is low equation 18 allows good floc formation and for clean water to be produced by AguaClara technology.
Alum dosing was also investigated by watching floc formation in the flocculator at different alum doses. This was done in an attempt to note if it was possible to visually discern when the alum dose needed to be changed. Doses of 0, 5, 20, and 50 mg/L were noted.
Upon observing 0 alum dose, it was very clear that there wasn't any improvement through the tank. In fact the whole way through the flocculator it looked as though there were tiny particles floating along with big particles and this never changed.
When observing 5 mg/L the difference was hard to discern. It appeared that there was improvement through the tank. The last two sections didn't appear to have as many small particles in between some of the larger particles were but the flocs that were seen appeared to be smaller than those observed when using equation 18.
When I raised the alum dose to 20 at first there appeared to be a rush of floc at the beginning that moved up the first section of the tank. It is not clear if this was due to the introduction of the higher alum dose or not. After the initial rush of flocs the water entering the tank didn't seem as dirty as the raw water entering the tank without any alum. The floc formation appears earlier in the tank, after about a third of the first section. They are still small at this point but the improvement over the end of the first section and the middle of the second section is rapid. Most of the particles are in floc and there are not as many particles in between flocs.
When I finally increased the dose to 50 mg/L the water in the first section appeared to have the same small particles in between larger particles as the raw water when there was no alum added.. Through the tank it appeared that there lots of medium to small sized flocs, almost as though there were more medium sized flocs and not any large flocs. It does not look like the flocs are large. They are small and never get bigger throughout the tank. There are a few rare flocs that could be considered large. This is contrary to what I had expected to see, I thought I would see the same type of floc as when the tube settlers were clogging. It appeared almost as though by overdoing it negated the effect of the alum, by creating the reverse charge on the particles which again had a repelling affect.

Preliminary Results from Testing of Tapered Spacing

Overall data was collected from two flocculator set-ups. We tested the original set-up where there where 79 baffles spaced equally throughout the flocculator at 6.4 cm apart. And then we switched the flocculator set-up and collected more data for a tapered flocculation set-up that consisted of 39 baffles. The purpose of these experiments was to test whether or not our assumption that better and more efficient flocculation could be achieved with a tapered baffle configuration was accurate.
A head loss analysis was done on the tapered set-up to determine if the theoretical parameters being used were accurately describing the system. The collected head loss data was found by comparing water height in a free surface tube to the water height in the tank. A hole was drilled between the third and fourth baffles a tube was inserted into a quik-connect fitting attached there. This tube was open to the atmosphere at the other end and the water height in this tube is a direct measure of head loss through the plant when compared to water levels at other points. The graph seen below compared the theoretical head loss for the tapered set-up with data that was collected from the first section of the tapered set-up. Data was not collected from the section because method used for collecting the data in the first section could not be repeated in the second section. The graph of our measured headloss versus the theoretical head loss can be seen below.

Based off of looking at the graph it seems that at least for the first three sections we are over calculating the amount of head loss that the system is creating.
Further analysis will include changing the theoretical parameters such as K to try and get the theory curve to match the data curve. This analysis should give us a better representation of what K is for the AguaClara plants.

Conclusions from the Testing of the Uniform Flocculation Configuration

After the initial design and construction of the flocculator was complete, attention focused on getting the flocculator running. The beginning of the summer was spent ensuring that all the individual parts of the flocculator and tube settler setup were working and that all Process Controller methods were setup. Once this was finished data collection started and it was at this time that flaws in the design became apparent. Attention then shifted to fixing those flaws. The design of the dividers was the most problematic portion of the tank. The dividers were made as a separate piece and then lowered into the tank. This created several spaces and large gaps between the tank and the divider where water could skip sections and head directly towards the outlet. Several steps were made to fill the gaps and stop the leaking caused by the gaps. This caused more problems as the dividers were not straight and easily deformed, which caused bowing and other deformations when the Kwik Foam was used. The divider design thus caused skipping around the end of sections as well as around baffles. The leaks in the tank were discovered through observation, and head loss measurements. The first leak was discovered by observation of the tank and the subsequent leaks discovered and fixed through use of the head loss tubes and sampling in the tank.

Alum Dose

For the majority of testing, alum dose was set by equation 18. After the change of A from 15 to 10 was made this approach was effective for the low turbidities that the flocculator experienced this summer. Hopefully in the future the raw water turbidity will change enabling testing of higher turbidities. Through use of equation 18, observing the floc tank and conversations with the operators at the water treatment plant it has become apparent that there is still a lot of research that needs to be done regarding alum dose. Observing the floc tank was helpful in being able to identify different kinds of floc and what different alum doses looked like in the water entering the water treatment plant. The water treatment plant has now switched to a different coagulant but if they had to go back to alum they said they would use past experience and alum doses as well as jar tests to set their doses. This suggests that for each water treatment plant an equation, formula or at least a rule of thumb could be developed off of past water treatment for future dosing. If this formula would be translatable to other water treatment plants and different water types is uncertain. The run increment alum dose test should help to shed light on alum dosing as it allows the alum dose to be changed while at a relatively constant raw water turbidity. Hopefully the data from this test will show either an optimum dose or a small range of optimal doses for specific settled water turbidity.

Design Suggestions (for future plants)

If a serpentine flow is again used, future designs should include a way to make the dividers a more central part of the design, and a material that does not deform easily but holds its structure and can be sealed to the tank should be chosen. If this is not a possibility then re-enforcement to the dividers should be added to ensure that deformation does not cause problems with baffle skipping. This way the width of the sections can be easily controlled as well as locations where leaks could be problematic could be observed during installation. The problem with making the divider and the tank not one central piece is that it is difficult to make the two pieces fit together and make them water tight.
Another suggestion that would make maintenance of the tank easier would be to include an outlet that could be opened and closed nearer to the bottom of the tank than the outlet pipe currently is. The current design leaves water at a height of a few inches above the bottom that needs to be pumped out before the tank is fully drained. Even then with the design of the dividers the bottom of the tank can never be fully drained because the bottom piece of the divider covers most of the bottom of the tank. This is problematic if repairs to the caulk or Kwik Foam need to be made, as they seal best on dry surfaces.
When dealing inside the tank the current configuration of the modules is sturdy and provides a structure that allows the baffles to move as a whole maintaining baffle spacing. This is an advantage as it ensures that they are evenly spaced and that the value of G is constant throughout each module. One of the problems encountered this summer was that in the pilot plant configuration the inlet and outlet are both pipes that were added that decreased the space in the first and last section of the flocculator. If this design is to be replicated the space that the inlet and outlet occupy should be taken into account when designing the number of baffles. This is due to the fact that the connector pipes that are used to keep the baffles from drifting to the end could become caught on the exit pipe and cause the baffles to be pushed against the inlet. When lowering the modules into the tank they need to be lowered very carefully and each portion of the section needs to be lowered at the same time, necessitating at least two people, usually three. If the modules were bent the connector pipes would pop out of the caps and it would be hard to replace them. The connectors are important because of the forces that they carry.

Experiment Suggestions

It is assumed that all leaks due to dividers have been fixed either with sand, or with Kwik Foam and caulk. This idea is supported by the increase in head loss and the visual appearance of the velocity in all three sections being equal. The major concern that should be further evaluated is skipping around the edge of the baffles due to divider bowing. After skipping is either confirmed to be minimal or fixed future experiments should include focusing on the third section. Data collected and analyzed up to this point suggests that the third section is breaking floc up. In building this pilot plant it was hoped to test tapered flocculation where changing the value of G over the three sections is explored. It is suggested that future experiments change the spacing of the third section from a G of 29 s-1 to a G of 15 s-1. In order to ensure that the design for G does not actually yield a lower value of G than wanted it is suggested that the head loss for the third section is measured and the K value back calculated. This new K value should take into account any skipping that is occurring around the side of the baffles. It is assumed that the skipping in this section is coming from between the baffles and the divider. Thus when G is decreased only the baffle spacing and thus number of baffles will be changed. This new K value assumes that the skipping is occurring uniformly over each baffle and thus should account for the skipping around the baffles that will be left. It should yield a more accurate calculation of head loss and G as it takes into account the space between the dividers and the baffles, which should stay constant as the same baffles will be used, just less of them. When this is done the focus should be on collecting data on the effect of settled water turbidity and floc formation. It is hypothesized that a lower G will not break up larger flocs but still provide gentle mixing allowing smaller particles to be integrated into floc.

Developing Turbidity Profiles along the Flocculator

This testing procedure has two parts: finding the optimal alum dose on the day of testing and running the Profile Test. This procedure was developed and implemented in the Spring '08 semester.

Optimal Alum Dose Testing

Because the pilot plant takes water directly from the stream, environmental conditions change all the time and affect the incoming turbidity to the plant as well as the chemical composition of the particles causing turbidity. It is therefore necessary to determine the best alum dose for each day of testing to ensure the formation of good flocs. This requirement was implemented in the following way:
The Process Controller was used under the "Increment Alum" setting, which starts the Alum Dose at 0 mg/L, and increments at 5 mg/L until it reaches a maximum of 40 mg/L. Each of the alum doses ran for a 30-minute period, or for about 3 times the residence time of the tank. After the completion of the test, data processing was performed to select the data from the last 10 minutes of each individual alum increment. The first 20 minutes of data were rejected because the residence time in the flocculator was 10 minutes, and the residence time in the tube settler was also about 10 minutes. Therefore, in order to get readings from the final turbidimeter that were representative of the alum dose that we were testing, we discarded the first 20 minutes of data at each alum dose. The outgoing turbidity (from turbidimeter 4) was then analyzed for each increment, and the alum dose achieving the lowest turbidity was selected for the second part of the experiment.

Profile Test Procedure

The purpose of this experiment was to develop a profile of flocculation at different places along the flocculator. In order to do this, we moved the tube settlers to different points along the length of the flocculator and tested the turbidities of the water after it passed through the tube settler and reached the turbidimeter. In Experimental Set-up of the Tube Settler Placement above, you can see a schematic of the experimental set-up. A photograph of the same set-up is also shown. The experiment has three parts, A, B, and C, each lasting for 45 minutes. During the experiment, the location of tube settlers 2 and 3 were moved to different places along the flocculator as shown above. Turbidimeter 1 was always testing the incoming water, and turbidimeter 4 was always testing the turbidity of the water at the end of the flocculator (location 4 above). Along with moving tube settlers 2 and 3, we emptied tube settler 4 of water between parts A, B, and C of the experiment. This is because when the tube settler is filling with water, plug-flow conditions exist in which velocity gradients cannot develop, and flow up the tube settler is more even. So, by emptying tube settler 4 of water, we ensured that the potential effects of this condition in the tube settler were even across all the tube settlers.
Likewise, when performing the data analysis after the experiment, we found an increase in the turbidity to unreasonably high levels (on the order of 100 NTU) for about 10 minutes after moving the tube settlers. This was because the air in the tube settlers which was being pumped through the turbidimeters. Therefore, only the data at the end of each part of the experiment was used (approximately after 10 minutes of running). The removal of the air can be easily observed in the data when the system appears to have reached a steady-state.