At Ocean Networks Canada (ONC) there is a relatively small saltwater cylinder tank that they use to test their scientific instruments. The main type of instrument tested in this tank is called a CTD, which measures conductivity, temperature, and pressure (depth). They sometimes also measure turbidity, oxygen content, or other data.
These instruments are calibrated by the manufacturer, and then each instrument is tested at ONC to verify that they were properly calibrated. The method most often used is to put multiple instruments in the tank at the same time, and ensure that all instruments are in agreement. However, if the saltwater in the tank is not evenly mixed (homogeneous), then the measurements between the devices will not be consistent. This is not a good situation. ONC’s data scientists have noticed this problem occurring and passed this information on to the test and development team. The test and development team lead tasked me to quantify how well the tank was currently mixed, then to improve it. Ideally, improvements made would remove this problem entirely.
After doing some research, I selected 4 direction-adjustable powerhead pumps, which are generally used in reef tanks that need strong directional flow. These pumps were meant to be mounted with magnets through the outside wall of a 1/2″ flat glass tank, but the walls of this tank are 4″ thick and curved. To mount them, I designed a 3D-printed “twist to lock” mount and machined some spare plastic pieces to make the arm.
The four pumps were placed in the tank to create two counter-rotating vortices, the upper one clockwise and the lower one counter-clockwise. This reduces centrifugal action (where there is a radial density separation) and creates more turbulent flow by having flow intersection. Laminar flow needs to be avoided, since by definition this means the fluid is not mixing.
Measurements were taken at two depths (12″ and 30″ below the water surface), at 3 radii from the centre (0″, 7″, 14″), and at four angles (0°, 90°, 180°, 270°) for a total of 18 measurement locations. ~30 data points were taken at each location. These measurements were taken concurrently with a stationary reference probe to account for the average temperature of the tank increasing or decreasing as the water chiller turns on and off.
After taking measurements with the pumps off, and then taking measurements with the pumps on, the following results were found.
The maximum ΔT between any two points in the tank decreased by 53%, which is very significant. This represents a ΔT decrease of 0.012°C. The resolution of this instrument has a resolution of 0.0001°C, so this is very noticeable in the data and will improve future verification testing. There is still room for improvement. The maximum measured ΔT with the pumps on is 0.01°C, which is 5 times greater than the instrument’s initial accuracy of 0.002°C.
Further analysis of the data revealed that there is a radial temperature gradient when the pumps were not present, which was fixed with the addition of the pumps. There was a measurable “dead zone” in bottom corners of the tank, which should either be addressed in the future, or else the test and development team will know to avoid putting sensors in those locations.
I presented these finding in the form of a PowerPoint presentation to the data scientists and the test and development team. You can see that presentation here: