Research: Optimize, optimize, optimize

  I have been showing you pics of our experimental setup and some of the details about it. One thing I have not talked about is how we determine aspects of the system that are working or are not working. We have been optimizing the system for the past week in order to get the best results.
  The kind of movement we expect in a beaker with a stir bar is a vortex. In the pictures below, you will see vector field lines determined by the analysis of the data from Insight 4G. The flood maps are what you would expect from a vortex structure. The speed of the magnetic stir bar is approximately about 60 revolutions per minute.

U Bar graph in pixels

V bar graph in pixels

Vorticity in pixels

U root mean square

V root mean square

  Root mean square is a statistical measure of the magnitude. It is the square root of the average speed, squared. For the moment, I just wanted to show you what we have been up to for the past few days. One of the things we realized is that when we analyze an area, it needs to be the same for all the experimental runs that day. If you notice, the center of the vortex is not centered. Therefore we should make sure that the vortex is in the center of our post-processing window which will make a lot of analysis code run smoother. We also need to convert our graphs to a meaningful unit. Pixels don't mean much to the scientific world and it is not a conversion anyone wants to do in their head so we will convert the pixels to centimeters per second.

  There are a few things we have noticed from the data and from observations while running the experiment. The first is that the magnetic stir bar gets weighed down easily by the electrical tape. I have been changing the tape on the stir bar at the beginning of the day. Convergence tells us how accurate and precise our statistical data set is over the course of the experiment. However, the experiment was run again when I was not there and we had poor convergence from that one versus the experiments I ran earlier that day.

70 revolutions per minute from the morning, good convergence

70 revolutions per minute from the afternoon, poor convergence

  One of the things that might be happening is that the covered stir bar is sitting in the water between runs. Water has some very cool properties but these properties may also affect the setup. The solution is to take the stir bar out of water between extended breaks, dry it off, and check it before putting it back in the water. We will also have to check the motion of the stir bar in between the individual runs to make sure that it is not "jumping" due to friction issues, such as the tape loosening up in the corners giving the bar a jagged circular motion. 

  We have also noted that we have had to wait longer than three minutes between increasing and decreasing the speeds. Therefore we are waiting five minutes. The majority of the graphs are showing that the system reaches convergence by 150 to 200 frames by doing this which is what we want.

  During the analysis of the convergence graphs from yesterday, we noticed that the green line was a consistent outlier. We tracked down the position of the point and realized it was close to the center of the vortex, which should have an average speed of zero, and realized that the vector readings for this point might have larger errors if it is averaging the data around a zero point. The good thing is that we can go back into the code, pick another pixel point close by that will give us x and y readings and not just x readings. 

  I am going to be a picky bugger in the morning so we can do our last set of data collection for freshwater now that we know what we are looking for that might be problematic in the system before the other team runs their 18 experiment runs in the afternoon. We should be testing the saltwater setup by Friday. We get algae next week! And remember the picture that included the big tank next to the beaker from yesterday's post? We will be putting everything in there the following week!