Week Two - Model Checkout in the Wind Tunnelby Fanny Zuniga
January 30, 1998
We are in the most demanding phase of our test. Getting started! This week everything was going on at once as we try to get everything ready. This is one time where I wish there were at least three of me. Some people are getting the model ready. Others, the tunnel. Others, setting up drawing programs on the computer so we can plot data. Still others, setting up cameras. Still others, safety training before we turn on the tunnel. Everyone is checking everything. You get the idea. I want to describe at least some of all this, so this will probably be the longest journal entry I write. By the way, there is a spot in this web site for you to ask questions about my wind tunnel test.
The name of the game this week was making sure our measurements are accurate and repeatable before we spend a lot of time collecting data. Why spend so much time on this, you ask? Let's start with accuracy. This means we want to make sure that if we repeat a run in the wind tunnel three times in a row, we will get the same results each time. This tells us that all instruments (balance, etc) are working properly. What if your bathroom scale was off by up to 10 pounds each time you stepped on it? How about if it had only 1 pound of error? 1/20 of a pound? In a test like ours, we have to make sure we get data that is accurate enough to answer the questions we are researching. For example, which flap angle works the best? Which type of flap works the best? What angle of attack is the best for landing this type of plane? Since this is a scale model, each pound of Lift on the model is equal to a lot of Lift (up to 40 pounds!) on the real airplane, so you can imagine we have to be pretty accurate in measuring aerodynamic forces on this model.
And what about repeatablity? Sometimes our sensitive electrical equipment fails or changes over time. So we occasionally repeat certain standard conditions (called "baselines") throughout the entire test to make sure nothing has changed. We have to be very sure our data is reliable so that we can compare runs made at different times during a test. For example, we may want to compare the Lift and Drag of two different types of flaps but the actual runs could be made weeks apart, depending on when it is sensible for us to change the flaps. We can come rerun baseline conditions any time we want to make sure all our electrical systems are healthy.
Let me put this another way. If you set two watches to the same time, and a week later they read differently, which one would YOU trust? How would you make sure which one, if either, was correct? Hopefully you can see how much work it takes to make sure the data we collect is good data. When its possible, we make several measurements of the same thing to help make sure everything is working. We can't do this with the balance, though. So we'll spend part of this second week making sure we are ready to start exploring some of our questions about this airplane. We hoped it would go quickly so we could get on to the interesting stuff.
Monday (January 26): We ran into some trouble right from the start. See, there's this electric motor that moves the rear post to tilt the model up and down. Well, this motor creates electronic noise when it runs, which then interferes with the signal coming from the balance (like when you get a bad connection on your phone). We found this out because with the wind off, and nobody touching the model, our balance ouputs (Lift and Drag) were drifting all over. Big problem! This is has to get fixed before we start research runs.
We finally closed up the model and ran the wind tunnel near Midnight for a "shakedown" run at low speed! This is one of my favorite moments in a test, because this was the first time I got to see our model look like a real airplane. I sure hope to see these flying someday soon. Maybe I'll fly in it too! We are still making last-minute changes to our computer programs, and still having trouble making our tufts stick to the model.
Tuesday: We spent most of the two shifts today fixing the noise problem. Meanwhile, there's plenty of other work going on. Some of the team are not sorry we have a problem with the balance, it gives them time to catch up on other stuff.
Wednesday: We were able to minimize the interference problem, at least enough to move on in our test. We decided to do another checkload now that the problem had been fixed, so we hung weights off the model one last time (hopefully). A bunch of other minor problems popped up, got solved, and we went on to run the tunnel again. This time we got our first decent data. Now we can begin the process of gaining enough confidence in all of our measurements that we feel comfortable proceeding with the rest of the experiment.
Thursday: Today we made a bunch of runs to check our accuracy. We checked out various speeds and model angles of attack. We uncovered a bunch of new minor problems. For example, there are little electric heaters that keep the pressure modules at a constant temperature. Well, the heaters died and we had to fix them. But our balance data was looking OK, so we pressed on to establish that "baseline" I told you about.
Another part of getting our baseline, and being confident in our accuracy, is figuring out how long we should "sample" our balance signal to record a data point. Let me explain this English. At every speed and angle of attack, we will record the balance signal to measure Lift and Drag (and a few other forces as well). But the wind in the tunnel is turbulent (mostly from the big fan that pushes the air around the tunnel) so the model is bouncing around a little. You can actually see the model wiggling if you watch it on the video cameras. So we "sample" the electrical signal for a few seconds and average it out to get a single data point to store on the computer. We are making several runs to determine how long we need to sample the data to get a good average.
Friday: As part of our baselines, we were running at higher and higher load conditions; like higher speed and higher angle of attack. We had a new problem. This time its where the model was shaking enough to rub up against the supporting posts. We call this "fouling". This means we can't trust our force measurements. Its like you are standing on a bathroom scale to get weighed, but you're leaning on a counter while you are doing it. Not good. Eventually, we decided to file away some metal to enlarge the opening in the model where the posts go inside to hold onto the balance.
Friday night, we ran the tunnel again and still got some fouling of the model against the mounting posts. We didn't know where the fouling was occuring, so we decided to simulate the loads that caused the foul when the wind was on. To do this, we rotated the test section, rolled the yellow crane back into the test section, and used the crane to "lift" up on the model to simulate Lift loads. We found the problem, which was fixed while most of us went to dinner. We then loaded the model with the crane again, and it looked like we fixed the problem and were ready to go.
Saturday Night (Saturday Night?!?!): By Friday we were pretty far behind on our testing; we wanted to be done with this checkout phase early in the week. We got some volunteers to run tonight to make up some time. We had trouble with tunnel instruments and data system for the first 5 hours into our shift. We finally got running and finished Thursday's effort to determine how long we should sample the data . We picked 4 seconds for every data point (every angle of attack).