Yoshika and I have spent the past two weeks field testing our prototype at a hand pump in Dhaka! We quickly discovered that the field conditions here are quite different from the conditions that we had created in the lab at Stanford, and have had to make a couple of changes to our device.
First, we noticed that the pressure at the hand pump we are testing at in Dhaka is much higher than the testing setup we used in the lab. This higher pressure has meant that the same volume of water flowing through the pump receives a much higher chlorine dose than what we saw in the lab. To correct this, we have reduced the concentration of chlorine in our chlorine feed and dialed down our flow regulator to a lower dosing level.
Also, when we initially installed our all-plastic prototype on the hand pump, it raised the height of the hand pump significantly and made the hand pump quite wobbly. To fix this, we have changed the pipe section that we are installing back to metal and removed both large non-return valves. It turns out that the large non-return valves were only necessary to maintain chlorine consistency in the lab to prevent chlorine back flow into the testing water. In the field, we found that removing the large non-return valves has no affect on consistency. Which is great news for us, because it reduces that cost of our device by about $25/each, bringing the total capital cost of our device down to about $5!
We’ve also have seen good accuracy of our device at the one hand pump that we been testing it at, and have been able to meet our chlorine dosing range at a number of flow regulator settings. Our next step is to try it out at other hand pumps, particularity those with lower pressure, and see if we can repeat the results.
All-in-all, we’re quite happy with the results of our testing over the past few weeks!
To test how corrosive the chlorine is on the pump, we cut off two small chunks of the pump and placed one in a beaker with 0.001% chlorine solution (10 ppm) and the other in a beaker with DI water. From visual observation so far, it appears that DI water is just as corrosive as the chlorine solution.
This week we gave a presentation to the CEE 177S class. It was nice to hear about what the other teams have accomplished over the course of the past two quarters. Moreover, ist was a great opportunity for us to take a step back from the product development and see where we stand overall. I think, of course, we have a lot of testing to do in the next 2 weeks. Luckily, we have a number of testing sessions planned out already. Also, we now have a more systematic description of how our device works and what variables affect it, so our testing protocol seems to be more efficient and effective. Previously, there were simply too many variables at play to really understand how our device was working. I expect that we will primarily be looking for the appropriate chlorine concentration, head difference, and IV regulator setting. Also, we’ll have to see how pumping speed has an effect, while also considering any other remaining confounding variables.
We revived our chlorine tests this week and found the following:
1. The non-return valves we installed about a month ago are working properly—we no longer see chlorine contamination in our water reservoir!
2. After dropping our starting chlorine concentration to 0.06% and closing our IV regulator halfway, we finally started seeing downstream chlorine concentrations that made sense given our setup. (~0.25 ppm average across three or so trials) These numbers are also closer to the model that the prototyping team has been developing based on Bernoulli’s principle.
To maintain our improved consistency (25% error or less) in future tests, we have determined that we should:
1. Flush out our system with DI water at the beginning of each testing session.
2. Discard the first 4 strokes for each trial. This effectively serves as a “re-priming” step between trials.
3. Collect 1 L samples (instead of ~100 mL samples) to account for the possibility of poor mixing inside the pump, where the concentrated chlorine first comes into contact with the water source.
Now that we’ve determined our current setup (IV = 1/2 open, starting Cl conc. = 0.06%) provides decently reproducible results, our next goal is to use this setup to compare the tube-in-pipe concept with the venturi to finally settle on one prototype for on-site installation in the summer. We will eventually modify our chosen starting concentration of chlorine to yield a final dose within our target range (1.0 +/- 0.5 ppm), but for now the goal is to verify the reproducibility of our promising findings so that we can be confident in whichever prototype we choose.
This week we evaluated how various pumping speeds affect the dosing. It appears first that the dosing is variable across pumping speeds. For slow speeds, we’ve determined that dosing is fairly inconsistent. For medium and fast speeds, the dosing is fairly consistent, especially for our original non-drilled out Venturi device. Our next approach will be to actually use chlorine to determine the dosing across our devices and then come to a conclusion on which device is best to use moving forward in our pilot tests this summer. Furthermore, we’re hoping to begin using a flow meter because our current classification of pumping speed is highly arbitrary so our data may not be as consistent as we could be getting.
This week, we devoted most of our testing time to investigating the effect of varying pumping speed, as well as the height of our chlorine source relative to our water source. We found that slower pumping speeds resulted in higher downstream chlorine doses across our three prototypes (the venturi, the T, and the tube-in-pipe). Our tests in which we varied the chlorine height were less conclusive.
Our objective with these tests was to determine which of our three models provided the most consistent dosing despite anticipated variations in our system, such as differences in pumping speed across different users. Surprisingly, we found that there was little difference between the tube-in-pipe and T models. The only difference between these designs is an extra T-pipe located where the chlorine tube feeds into the water source. We therefore concluded that this T has a negligible effect on the dosing capabilities of our device, and was an unnecessary added cost for the final product. The venturi dosed slightly higher than the other two models, but unfortunately was not any more accurate/consistent despite its sleeker design.
Next week, the plan is to begin testing with chlorine to see if the patterns we’ve been seeing with our simplified setup actually correlate to realistic dosing levels. We will use these tests to evaluate the reliability of both the tube-in-pipe prototype (simplest) and the venturi (most advanced design) and decide from here which would be best to implement on-site in Dhaka for the summer.
This week we began to take a more serious look at the math behind our device. Specifically, we are trying to discover (1) if the chlorine dosing is in fact linearly related to the flow rate of the water through the pipe, (2) which height differences matter in dosing, and (3) how much head changes dosing. We began tackling these question with Bernoulli’s equation. We will have a working excel model to test different parameters by next week. Hopefully with a better understanding of the physics of our situation, we will be able to make more informed design decisions.