Impeller Water Pump
The Design for Manufacturing class that I took during my sophomore year in college introduced students to a broad range of manufacturing processes and techniques. As a final test of our familiarity with the topics that we learned, students formed four-person groups in which we would design and prototype small water pumps. We were given starting materials that were required to be used in our final pumps so that on the final day we could test the head pressure and flow rate of every pump in a competition. The switch, battery type, motors and inlet/outlet hose fittings were standardized between groups to provide an even playing field and allow the competition to really focus on the differences between mechanical designs in efficiency.
A small amount of research led most groups to use an impeller-based design, which I ended up using, while a select few thought instead to use piston-based pumps and fewer still looked into gear pumps. I had never heard of gear pumps before already settling on an impeller-based centrifugal pump design, but with the knowledge that I have now, I would have attempted to model a gear pump instead. I had the most experience with 3D modeling of any member in my group, so after discussing design and material ideas, I threw together a basic version of the pump that we imagined in Creo Parametric.
Over time, I improved the details of the components in my model and modified specific features to work towards a more efficient pump, as well as changing a few dimensions that were not easily manufacturable. Over the course of a few weeks, we spent a large amount of time in the lab machining and fine-tuning the components of our pump. We tested it a few times once it was fully assembled to make sure that it would work during the final test and experienced great results, as our pump was able to push 3.4 liters in one minute.
On testing day, our pump only pushed a flow rate of 2.25 liters per minute and achieved a head pressure of 6.25 feet. Of the 27 groups, we placed in 6th for head pressure and 4th for flow rate. However, the pump severely underperformed, as after the 40 second mark, the battery was unable to provide sufficient power to rotate the impeller and the pump shut off entirely. The flow rate that we had achieved in our own tests with a fresh battery would have moved us to second place for flow rate.
This project introduced me to a few new manufacturing techniques and gave me insight into the material selection process of mechanism design, as well as giving me direct experience with operating the machinery that we used to form our pump. I learned to convert our pump chamber into cutting tool paths and then to g-code, and how to run that g-code on the mill which will undoubtedly be useful in engineering design. The impeller of our pump was also the first piece that I ever designed and had 3D printed.
The component of our pump that took the most effort to get right was definitely the impeller. At first, I thought that an impeller with eight “fins” as I called them at the time, or vanes as they’re actually called, would be the most efficient in terms of weight/volume/surface area/friction. Upon modeling it, I felt that it looked too tightly packed, especially after increasing the vane thickness to be successfully 3D-printed by an FDM 3D printer. I then modeled a six-vane design, but that looked like there was too much open space, so I moved to a seven-vane model which seemed to have a very nice balance between the vanes and the gaps between them.
I made sure that the vanes were thick enough to print properly, added rounds on sharp edges and corners, and built up reinforcement in the center to prevent the rotor from shearing away from the impeller and to give it structural reinforcement. Although the rounds that I added were barely noticeable with the surface finish of the 3D printer that was used, in a real scenario they would have been beneficial in providing smooth fluid flow and preventing stress concentrations from occurring in corners.
We began the manufacturing process by submitting an stl file of our impeller to be 3D printed by the university’s engineering shop, which used a fused deposition modeling printer. Groups were given access to a CNC mill which most, including us, used to form their pump chambers. The primary materials that we were encouraged to use for our pump components were clear acrylic and polyethylene. While acrylic is brittle and not well suited for CNC milling, we decided to use it as the material of both our chamber top and the chamber itself.
It took a few attempts to machine the components of our pump without some kind of setup or runtime failure, but once we had it completed, the pieces fit together easily. Mainly, we had issues with brittle failure in thin sections of the acrylic due to high cutting forces in the CNC machine. Small modifications to the input file and g-code greatly improved this issue, and we were able to successfully mill our pump chamber from acrylic despite being told that it would never work. We only settled on acrylic because of its clear nature as we wanted to be able to see within the pump from any angle.
Polyethylene would have definitely made the manufacturing process simpler, but we had time to modify the design until it worked the way that we had intended. In a practical situation, we would have used polyethylene to save time and to obtain a better internal surface finish to reduce friction. Ten M5 machine screws were used to fasten the two sections of the pump together thus encasing the impeller to ensure that there was a strong seal around the entire interface. We filled the gaps with silicone before screwing everything down for the last time, then installed and wired together the switch, motor and battery. To attach the brass hose fittings to the chamber, we tapped threads directly into the acrylic and screwed them into the tapped holes, again using silicone to prevent water from leaking through small gaps.
There was a testing station set up in the lab at which groups could test the flow rate of their pump with the same conditions that would be present in the final test. Two chambers were set up with a one-foot height difference in water level that our pumps were expected to overcome. During our testing, the impeller’s center structure sheared off, rendering the pump useless until repaired, but we did find before the failure that with a full nine-volt battery, the pump was able to carry 3.4 liters up from one container to the other in one minute, which was the time that would be allotted for each pump during the final test. I repaired the impeller eye that night using epoxy, reconnecting it to the remainder of the impeller which had sheared off during dry testing, likely because of the much higher angular velocity that the motor could achieve. Any error in the perpendicularity of the axis of the impeller to the plane that the impeller was built on would show itself later as wobble, so I made sure that they were as perpendicular as possible using the bottom of the chamber and the motor shaft as references while the epoxy set.
When we ran the pump following the repairs, though, it was for the final test, as we had no other lab time before the final day available to ensure that the repairs would hold and function well. The epoxy held just fine, but the battery, which we had just purchased earlier that day and not used for any runs prior to the final test, failed 40 seconds into our only run. As a result, our pump was able to push only 2.25 liters in that time, for which the judges of the assignment compensated by boosting us to 2.7 liters per minute as an approximation of how it would have performed. If we had used a truly fresh battery that could last through the whole trial, as we incorrectly thought that we had, I’m certain that our pump would have performed far better.
The team that achieved the highest flow rate of 4.5 liters per minute used a gear pump design, and its performance led me later to be very interested in the mechanisms behind different gear pump models. The efficiency of gear pumps is superior to centrifugal pumps as there is far less wasted motion in comparison. Because of this, and the fact that they contain solid material preventing backflow, they can achieve both a higher flow rate and a much higher head pressure, as was seen in the performance of the first-place group’s model. The gear pump achieved 19 feet of head pressure and 4.5 liter per minute flow rate with the same input power as the best impeller pump, which could only pump 3 liters per minute and fell short of the gear pump’s head pressure at only 10 feet and 11 inches.
Just for fun, when I have my own SLA 3D printer, my plan is to one-up my group’s pump from my Design for Manufacturing class and attempt to outperform the other pumps in the class. I will most likely use or improve upon one of the gear pump models that I have designed. I am certain that I can make a pump that will outperform my original prototype with the knowledge that I’ve acquired since I built the original, and I love that kind of challenge.