AERO 307 Logbook

Noah Bachir

[simulit-blog]

Week 10- The Last One

Week 10: dead week. Contrary to popular belief it’s not because there’s nothing to do, but because you actually wish you were dead. We had one last test to get done during dead week, and it was the full scale smoke-car setup. Because we couldn’t do much during the day, on Monday we decided to play around with the smoke in the maintenance road between the aero labs to see what the best setup would be for the test later that night.

We tried deploying the smoke closer to the wall:

Closer to the ground:

And we tried shining the sheet laser through it to see if we got any interesting flow viz during the testing/verify if the laser would work later that night

Because it was so bright and windy during the day, we actually didn’t get any more useful information about how to set up the smoke machine and laser that we wouldn’t have already done. We did, however, set up the mounting brackets for the lasers that would allow them to be taped to a window (which, sadly, did not end up as the final setup).

Once we had gathered all we thought we needed from testing different setups, we put the smoke machine and lasers away in the props lab and worked on the presentation some more.

When night fell, it came time to test. The winds were much calmer and the light was much...darker, but it still wasn’t perfect. This is what it looked like once we had all the lasers and smoke machine set up

It was still a little bright, so we shut down the lights in the props lab and covered up the walkway lights closest to the test.

We also tried different methods of lighting up the smoke, like the Ryobi spotlights, since it was somewhat difficult to see the lasers from every viewing angle (an explanation for that can be found below). The spotlight wasn’t much better, so we decided to stick with the laser sheet.

Who knew a CR-V could be menacing?

The smoke was hanging around much better than it was during the day, but it still needed a little assistance. Thus Chase was designated as the smoke-scooper, wherein he used a box to redirect some of the smoke into the ground as it exited the machine to bleed off some of the momentum. It worked pretty well, to my surprise.

Once the smoke and lasers were set up, then came the admittedly pretty fun task of driving through the smoke attempting to not be blinded by the laser sheet. I found that the faster I drove, the better Dr. Doig said the flow viz was, so I tried to speed up as quickly as possible through the smoke without peeling out. This was the best run we had that night, slowed down a little as the car passed through it, with an annoying unexplainable ten seconds of black screen at the end of it. I think I was too hasty when I exported the video to YouTube, and didn’t make sure I had cut it all the way.

The smoke looks like these awesome green flames, credit due to cameraman Dr. Doig. The vortex due to the RVSM is admittedly pretty difficult to distinguish from the rest of the vehicle, but if I squint really closely at the right area, I can convince myself there’s a little bit of movement in the smoke before the rest of it churns up.

With the final videos taken, all that was left was to embed them in the presentation and prepare for Wednesday.

Presenting on Wednesday went smoothly and I was especially excited to see what other groups had come up with. Things that I found especially cool included the active flow control wing and the shock tube project. I think the active flow control in particular is something that will absolutely be widely used in the future in some form of implementation or another, the benefits are way too good to look over.

If you were wondering why the smoke and lasers only looked good from a specific angle, I have these sources for you:

https://www.wyatt.com/library/theory/multi-angle-light-scattering-theory.html

https://www.azom.com/article.aspx?ArticleID=9821

A quick rundown:

source: azom.com

When you shine a light (laser) through a medium, you’ll get two types of scattering: isotropic and anisotropic. As the names suggest, isotropic is more even scattering, and anisotropic is uneven. Anisotropic scattering occurs when the light impinges on particles that are larger than it’s wavelength. The particles create zones of constructive and destructive interference that manifest themselves in certain “islands”, if you will, of local maximum intensity. The schematic below provides a simple overview of this.

Cursory observation of the Deconvolution and Scattering Pattern charts show pretty clearly that the intensity of the light will gather at certain deflection angles. You’ll notice that the most intense of these deflection angles are mostly small angles, however. As you move further away from the center, or focus, of the laser, it gets progressively less intense. This means that looking at the medium the laser passes through from a small angle means you’ll get a fairly sharp image due to the relatively high intensity of scattered light. However, if you are looking, say, normal to the medium, you won’t see much of anything. This explains why the laser sheet only looked good through the smoke when you were on the opposite side from the lasers, because you were able to capture a smaller deflection angle from the laser.

Course Reflection:

a) If you were to pick one thing that you feel like you understand pretty well about aerodynamics now that you didn't at the start of 307, what is it and why?

I feel I understand the actual mechanism of things like turbulators, trip strips, and vortex generators. Previously I didn’t exactly understand how roughness influenced the boundary layer, but now I know that surface roughness and vortex generators can influence tiny vortices into the flow, localized within the boundary layer, and that is how the boundary layer becomes energized.

b) If you were to pick one thing about aerodynamics that still confuses you, what is it and why?

I still don’t fully understand how you can aim a dye needle directly at a model car mirror, in the freestream and still have it push right off of the car. I guess that means I don’t understand streaklines. We got good enough results by moving the dye needle closer, but that shouldn’t have been necessary. Is there a theoretical spot that the streaklines around the car will direct the dye flow into the mirror? I think there should be but we certainly didn’t find it.

c) What was your 307 highlight?

My 307 highlight was definitely the UV flow viz of the Tesla. It was the perfect combination of good dye mixing, the dye actually flowing, and getting the needle in the perfect spot.

d) What was your 307 lowlight?

307 lowlight was probably either the force balance lab or the wake survey. The force balance lab because it was incredibly strange how drag numbers were high across the board but we couldn’t seem to figure out why. The necessary numbers were being subtracted from each other, but numbers were still way too high. I am tempted to chalk it up to bad calibration practice, even though it’s a lazy answer.

The wake survey deserves honorable mention, in my opinion, because of the run where we adjusted the angle of attack even though we hadn’t hit the wake yet with the probe position. It was a mad dash to gather the correct data before the end of the lab, and though stressful at the time, it’s good to laugh about it now.

e) For many of you, this'll be the last time you really engage with aerodynamics, since you prefer structures or controls or design or just anything else... for others, this course will have been a springboard to many future aerodynamic adventures. What do you think the future holds for you in aerospace engineering?

There’s a metaphor (simile? analogy?) I heard a while ago. I think it’s from an Asian philosophy like Taoism or Buddhism but I’m not actually sure. There’s all sorts of things in the river, like sand, which goes with the flow no matter what, it has no choice. There are large rocks, which sink to the bottom and move no matter what. Pebbles, however, can stay if they fall in the right spot, but once they get kicked up, they’re along for the ride.

Why the cheesy metaphor? I feel like I’m the pebble. I don’t have a strong desire to do one thing or another in aerospace engineering, because there’s so many things to know and gain mastery on. So, I don’t know what I think the future holds for me in aerospace engineering. Once I get kicked up by graduation, I’m going to follow the opportunities I see to land in the perfect spot to settle down.

(Except structures. I hate that stuff)

[simulit-blog]

Week 9- Finally, Results

With week 9 being an abbreviated week, I didn’t expect there to be much to report, but, as a pleasant surprise, we were able to get some great flow visualization of the Tesla model side mirrors. We got out the fluorescent dye and set up the blacklights over the water channel. It took some minor adjustments, but eventually we figured out the best spot for the dye needle to be to produce optimum flow viz results. In order to avoid the dye catching a streamline that ran up the A-pillar and down the length of the car, we had to place the dye needle a short ways upstream of the mirror. For a little while we were able to get some great shots. More on the failures of the test later on.

Here’s the flow viz captured at 4K auto settings, so the camera AF causes the dye to flash in and out of focus (That’s a little disingenuous, since YouTube only plays back at 1080p):

And here’s the flow viz captured at 1080p@60fps. This one was taken on pro settings so the dye is not terribly in focus. This should allow better analysis of the film in a video editor, however:

There are three things I think really stand out from these videos. The first is that there’s no perfect dilution possible between the dye and water, as the dye is nonpolar, as evidenced by the fact that it must be diluted in mineral or paraffin oil, while water is polar. The end result is that instead of a clean bit of glowing fluid, there are distinct particles of dye that swirl around in the flow. Any shedding vortices or other flow features are still fairly visible, so this isn’t too concerning.

The second thing that stands out to me is the apparent downward component to the wash behind the mirror. Dr. Doig theorizes (you might be able to hear this in one of the videos) that this is because the flow coming off of the A-pillar whips down and drags some of the flow behind the mirror with it. This is a fairly likely explanation, but I would want to know if the turbulence having that downwash effect increases or decreases the drag from the mirror.

The final thing that I notice from these shots is the sheer size of the wake. The main part of the recirculation zone reaches almost a third of the way down the driver’s window, and every so often you can see licks of the dye reach all the way back to the B-pillar. This is about 4 times the characteristic length of the mirror (the thickness). Given that the wake of a car is anywhere from 1-2 characteristic lengths long, this leads me to believe that the mirror has proportionally a larger amount of drag associated with it than the rest of the vehicle. Thus, I think that simple frontal area calculations will not be sufficient to calculate the effect that removing the side mirrors will have on the drag of the car.

Speaking of frontal area calculations, we took rough measurements of the Tesla to figure out what percentage of the frontal area that the mirrors were. Approximating the main chassis as rectangular (a bad assumption) and the mirror as also rectangular (an even worse assumption), we got measurements of 108*75 mm for the chassis and 13*9 mm for the mirror. This comes out to 117 mm^2 for the mirror and 8100 mm^2 for the chassis. That’s about 1.44 percent of the frontal area per mirror. Times two mirrors means that in total,  the RVSMs on a Tesla Model S make up 2.88% percent of the total frontal area. It’s worth noting that as a sedan, the mirrors on the Model S are much smaller than they are on, say, a pickup truck or a semi truck. When we eventually test with the truck and smoke (hopefully on Monday), we should be sure to take measurements of the truck and its mirrors to get a sense of how big the mirrors are on other types of vehicles. Running back around to my theory about the drag on the mirrors, just because the mirrors make up 2.88 percent of the frontal area on the Model S, I predict that if they were to be removed, the drag on the vehicle would be improved by more than that value. However, in the absence of a way to test this prediction, I’ll leave this problem to future scholars who revisit this problem using CFD or even practical testing.

Edit: I realized I didn’t talk about the failures we experienced using the fluorescent dye. Unfortunately we didn’t grab any pictures as we had our hands full trying to clean up the mess, but after a very long time using the fluorescent dye without any hiccups, the dye suddenly stopped coming out of the needle. Upon opening the bottle up, it was found that the dye had congealed into a goopy mess. We figured that we had been a little lax on making sure the bottle was shaken up, so we washed the bottle out and tried again, making sure to shake the bottle vigorously this time. Still, nothing came out. We found that again the dye had congealed into a mess around the outgoing tube.

So clearly the problem with the dye was that it was solidifying very quickly. We tried all sorts of things to remedy this. We tried diluting with cold water, with hot water. We tried shaking the bottle, and vigorously stirring the bottle. Nothing worked, and as it was late, we decided to call testing for the day. While cleaning up, Kenny had the idea that because the dye was solidifying and settling on the bottom of the bottle during testing, if we had a way to make sure that the outgoing tube didn’t rest on the bottom of the bottle, we could avoid grabbing all that solid material and instead grab the liquid that won’t clog the tubes. We’ll see if we manage to get another run of testing in to verify this.

[simulit-blog]

Week 8- Down and Dirty

The first lab session featured the Car Mirrors team doing the awfully glamorous work of cleaning out the crusty old water channel dye setup. I have no idea how long that crust had been developing, but luckily it came off fairly easily with some hot water (no elbow grease required!)

The water channel in its original state looked like a bad elementary school art project, with black stains all over the bottom of the bucket. Rinsing that out was a treat, and the bubbling black water reminded me of a horror film. Like I said, though, it all washed off fairly easily with some hot water, and only minor scrubbing was necessary to remove a clot of dye that had stuck to the bottom of the bucket.

This final dial setup is shown above. Somewhere among the mess of tubes you can see a dirty tube. That one appeared to have mold in it, and try as we might, we couldn’t clear the apparent blockage in the tube. So we decided the best course of action was to chalk up that one bottle as a loss and limit ourselves to two bottles.

All clean!

On Wednesday, Dr. Doig was absent, so we set out using what we had discussed last with him about where to go with the project. Another team was using the water channel, so we decided to scout out potential places to run a truck through at moderate speeds. We started pretty close to the wind tunnel to minimize the amount of extension-cord tomfoolery needed to set the smoke machine up. This is the view from down the service road:

So I was fairly surprised that the path was actually wide enough to fit a car (or truck) through. Of course, moving at neighborhood speeds like we’re planning to makes this a safety risk for people leaving their labs. The doors won’t suddenly open into the vehicle, but anyone walking out would be in for a surprise. An alternative to running the vehicle down the length of the walkway is to have it start from that reddish bush behind the spacecraft environment students. It probably wouldn’t be possible to get much higher than 20-25 mph (~40 kmh) before needing to stop before the bus/service road, and if we mount the laser sheet on the wall of the prop lab yard, shown below:

There wouldn’t be much time to stop before said service road.

One other option is the yard in between all the aero labs. There’s a more open space for the vehicle to run, and if the laser sheet is mounted to the fence seen, there should be slightly more time to stop. So, we have three options to drive the vehicle down during testing time that would be of at minimum dubious safety standards. If none of those work, the “wind tunnel option” (stationary body and moving air) shouldn’t be too hard to recreate with a leafblower, but I don’t know how the turbulence from a leafblower would impact the smoke. Perhaps if there was some leftover flow straightener honeycomb? Decisions decisions...

Of the many idle conversations we have in the conference room, one involved none other than Santa Claus

Specifically, we were discussing the rules of the Santa Clause (the Tim Allen movie) and his magical weight gain, but I can turn this discussion to sciency talk pretty easily.

Suppose there are just under 2 billion children under age 14 in the world. Now, Christians make up about 30 percent of the world population. At that clip, there would be about 600 million Christian kiddos on Christmas Eve. Lets assume this is a very generous year for St. Nick, and 9/10 of the children make the Nice List™. If all 540 million children were split at 2 to a household, that’s 270 million households to visit. Upon trying to look this distance up, I realized I’ve run into the Traveling Salesman Problem, and I didn’t really want to go down that rabbit hole for just a Tumblr post, so I’m going to cheat, and use these numbers for how far Mr. Claus has to travel in a night. So he’s got to travel 3 million miles (~5 million km) in 32 hours, and I will be generous to Santa’s delivery skills and say that he spends only 25% of his time delivering, and the rest traveling. That leaves him 24 hours to travel those 5 million km. That translates to a speed of 5787 m/s-at STP that’s Mach 170!

I wanted to translate this number into power requirements using the weight of 9 Reindeer + 1 Sleigh + 1 Obese Male (ignoring gifts of course) but then I realized I’d have to make assumptions about the shape and drag of Santa’s sleigh so I decided against that. What I can tell you is that traveling at this Mach number, design considerations must be made to prevent both the vaporization of your reindeer, and whatever heat shield you try to put in front of them. There’s also the issue of the shock wave. Traveling at Mach 170 would create a Mach wave behind the, uh, craft, at an angle of .33 degrees to the horizontal. If Santa wants to avoid hitting airliners, he should fly at an altitude of 20000 m (the U-2 had an “operational ceiling” of “70000 ft” *wink wink*. This is just over 21000 m, but Santa shouldn’t run into another super secret spy aircraft. Until he flies over Russia, anyway). Anyway, at an altitude of 20000 m, a Mach wave at .33 degrees to the horizontal should strike the ground 3400 km behind Santa. This is just about a quarter of the diameter of the Earth. I think that even when you consider the spherical nature of the Earth, the Mach wave should actually hit the ground, instead of careening off into space, guaranteeing that Santa blasts out every window on every building around the world once a year. Of course, he’s probably not traveling that high to just jump between neighboring houses, but to go across the Atlantic Ocean (which takes anywhere from half a second to a full second at these speeds, by the way), he’ll probably be able to reach his cruising altitude.

One final note, because this thought exercise has gone on way too long already, Santa has got to be inhuman in some way to be able to perceive delivering almost 20000 presents a second. New movie idea: Santa Claus is actually an alien, one of 365 from different planets (or universes, if we want to go off the rails here) that each celebrate Christmas on a different day. From here, we could enter any genre: comedy, science fiction, political thriller, even Alien-style gory horror-action in which a human infiltrates the North Pole. We call it: Nightmare at the North Pole

[simulit-blog]

Week 7- How Much is a Pound Again?

Week 7, without fail, is one of the hardest parts of the quarter, because any cheerful energy you had at the outset is gone, and you’re still basically only halfway done what with all the homework assignments, midterms, and finals left to come. Life goes on, though.

On Monday I spent my day analyzing the data we got from the force balance lab, some of it better than the rest. The first thing I decided to do was convert the lift and drag data we calculated from pounds to Newtons. A simple calculation, just multiply by 2.204 pounds in a kilo, and multiply by 9.81 for gravity. This is the plot I got from that conversion:

Nothing broke thankfully, but I realized pretty quickly that I don’t actually have a sense for how much a Newton is, let alone a pound. My sense for the performance of wings comes from CL and CD numbers. That was the next step of conversions. It turns out that my original math was wrong somewhere, because I got a peak CL of around 6. I’m an optimist, but I’m at least realistic in my expectations. It turns out instead of multiplying by 2.2 to turn pounds into kilos, I should have divided. Oops. These are the correct plots (with some extra data, which I will explain)

The circular points represent the experimental data, while the solid lines represent theoretical data obtained by Neiman from XFLR5. Neiman also modified my code to produce these plots.

The lift data is fairly close to what we expect from a 4412. The drag, and thus literally ever other plot, however, is not. I don’t really have a theory for why this is. It seems like a systemic error, because every other group that I’ve talked to has the same information. It can’t be the sting, though, because we’re subtracting wing-off data  from the rest of the data we have. Maybe we’re not doing this correctly, but if that were the case the lift data would be off too, wouldn’t it?

In any case, the trends appear correct, and there are still a few useful bits of data that can be gleaned from these sets. For example, the high AR wing has a steeper lift curve and stalls earlier than the low AR wing, which matches the expected data. The drag trends are flipped, as the high AR wing experiences more drag, not less, than the low AR wing. The drag polar matches the expected trend (barely), and the L/D plot doesn’t actually have enough resolution to see what we want to see. Errors are still fuel for the report though!

The second lab session was when we had been assigned our mini projects and broke into teams to research the projects. It turns out that Dr. Doig was right (who would have thought?), removing the rear view side mirrors- RVSM, as they’re called in the biz- is a relatively untouched topic (although I did find an article from 2014 reporting on Tesla’s attempt to change automobile mirror laws). For some reason, people seem more interested in changing the design of RVSMs using active flow control, among other things, rather than doing away with them altogether.

Perhaps it’s the strength and intimidation factor of dealing with challenging a federal entity, or perhaps it’s the fear that the public will not be receptive to cameras instead of mirrors. Personally, I’m still a bit skeptical of the camera solution. For one, I do believe there is some merit to the hardware malfunction concerns (if only because the backup camera on my first car had some serious issues just through normal use).

More importantly though, I think that the ergonomics of replacing mirrors with cameras pose some issues. The Honda Civic has an augmented RVSM with a camera, but only on the passenger side. Why? Isn’t the blind spot on the left side just as important? The explanation that I accept is that it doesn’t make much sense intuitively to turn your head right to the center dash, to then move your car to the left. If both mirrors get replaced with cameras, the driver will have no choice but to move their head right to move the car to the left, assuming the technology we have doesn’t update too much. There are transparent OLEDs that could maybe somehow be integrated into the window or windshield, but what happens when the windows are down? Now the driver has to turn their head all the way back to look at their blind spot to change lanes, which doesn’t alleviate my concerns about keeping one’s eyes at least a little bit on the road. If these cameras are activated by the turn signal, what happens when BMW drivers want to change lanes?

In spite of those concerns, I do feel that replacing RVSMs with cameras is more likely to happen sooner rather than later. The economy of scale argument is hard to ignore: there are 190MM drivers licenses issued in the US. At an average of 12k miles (20k km) per year, that comes out to 2.3e12 miles (3.8e12 km) driven on American roads alone (while America is big (>4MM miles (>6.7MM km)), it has far from all the roads in the world (>20MM km)). At those scales, even a 1% change in any value for a vehicle gets multiplied over a million-fold. The first step is generating interest in the movement!

[simulit-blog]

Week 6- Back to Basics

So, my title game definitely peaked in my Week 4 post. I’m not sure how I’m going to come back to that perfect mix of rhyme and information contained in a single title. So like the activities we’re undertaking this week, I’m going back to basics with my title (eh? eh? It’s not great, I know).

Anyway, this week was fairly uneventful. We watched some flow viz presentations, which were very cool. I quite liked the supermileage car project. The very premise, that of a laminar separation bubble, was a concept completely foreign to me. I also wonder if the supermileage team had considered adding some sort of trip strip to their car to reduce the effect of this laminar separation bubble. I think it’s at least worth a mention to the team right?

The other, more interesting part of the week (to me) was spent back in the lab, once again getting back to basics with a simple force balance test lab. The premise being that we test two different aspect ratio wings and measure the lift and drag they generate to make sure the force balance is working correctly.

There were a couple other pieces of data at work though, so it wasn’t totally boring. One of the things we were looking for was when the wing stalled completely.

One thing that absolutely amazes me is that Tumblr has had the same terrible gif support for like ten years. Anyway, as you can see from the gif (or maybe you can’t, I don’t know), the most outboard tuft still seems to be in attached flow, but the inboard tufts are flapping wildly about, indicating they are in stalled flow. This peculiar behavior is a consequence of the wingtip vortices and thus is impacted by the aspect ratio of the wing. So, theoretically, the higher aspect ratio wing should stall at a different angle of attack than the lower aspect ratio wing.

So does the higher aspect ratio wing stall earlier or later than the low aspect ratio wing? I initially thought later, using the logic that a higher AR is closer to an infinite (ideal) wing, so a higher AR gives you more ideal performance and thus a later stall angle.

I was wrong in that assumption. I should have actually thought through what I had experienced with the flow viz experiment, where we noticed that the tip of the wing was experiencing attached flow, but the middle was stalled due to the wingtip vortices. The higher AR wing, being closer to an infinite wing, will have less influence due to wingtip vortices, and thus will stall completely earlier than the lower AR wing.

Besides stall, I was again the code monkey for my group, using my quick thinking and fast fingers to whip up a MATLAB script to generate lift and drag curves. This one was a bit trickier, because the data files I’m reading are not .mat files anymore, they are .xls files. MATLAB has xlsread(), but it’s not as intuitive to use as load(), and because these .xls files have different axis data in different sheets, it’s a whole other mess. I got it all sorted though, and I was able to generate some okay-looking lift and drag curves.

These matched what I expected, but upon further thought, I remembered how we had to correct for angle of attack when getting the load cell data and translate it to lift and drag. I wondered if I had to do the same here, but upon applying the requisite corrections...

I determined that wasn’t the case. Neiman and I then figured that the load cell was probably contained in that 3D printed housing above the sting, and that would be why the load cell data was in terms of lift and drag, and not normal and axial forces.

Though the trend of the data matches what is to be expected, the actual units are still a little off. The measurements are still in pounds, and not only do I still have to convert to SI, I still have to nondimensionalize to CL and CD. But that will happen later.

Something odd happened while I was grilling the other day (though I forgot to grab a picture). The propane tank had completely iced over, but only on the bottom of the tank. I figured there was some gas trickery at work here, so I started looking into the nitty-gritty of propane storage. I learned that propane is stored in a highly pressurized form to keep it mostly liquid. However, it’s not completely liquid. That extra space is propane gas, which gets drawn into the grill to be burned and contribute to global warming.

Drawing the gas into the grill means that more of the liquid propane must evaporate. The process of evaporation draws heat from the propane’s environment, which happens to be the tank. The tank gets colder, which draws moisture to condense on the tank. If it gets cold enough, the condensation that the tank is drawing can actually freeze over, which is what I observed.

[simulit-blog]

Week 5-Bon Jovi Lyrics Here

Week 5, the halfway point! Only one wind tunnel test this week, the rest was spent in the conference room starting to put together a presentation, and painstakingly stepping through a video frame-by-frame trying to calculate the frequency of vortex shedding on our wing after the flow detaches (preliminary estimates place this number at 40 Hz).

After the Great Wing Bending of April 2018, we decided to fix the wing flexibility problem by performing the ever-elegant solution of shoving a carbon rod in the wing. We figured that placing the carbon rod at about the quarter chord of the wing would allow it to counteract most of the bending the wing experienced. We drilled a hole out of the wing, then tried to stuff the carbon rod in further. All told, we probably managed to get the carbon rod in to a depth of about 4 inches. That should be good enough, right?

That’s still not ideal, and even though the reinforced wing did actually bend less than the original wing, Dr. Doig still likened it to a banana. Such a strong curve definitely affected the angle that we were able to observe stall (it’s safe to say that most airfoils, let alone the 4412, stall well before 22°).

The reason that the wing, reinforced or not, stalled at such a late angle of attack is due to changes in the local angle of attack of the wing. At first I thought the bending would cause such changes, but thinking it over further, I realized that this wouldn’t change the local angle of attack that much. I realized that if anything, the wing must be twisting in a sort of weathervane effect, causing it to have a much lower angle of attack than what is indicated on the winch. This effect must have been stronger on the unreinforced wing, considering we never actually observed stall on that wing.

This twist comes from two likely places: one being that the wing we made didn’t fit in the turntable perfectly, and two that the wing was made of EPP foam, which is a lot less stiff than typical wing materials.

The final thing that we did was attempt to calculate the Strouhal number for the wing once the flow had separated. This necessitated finding the frequency of vortex shedding off the wings. We didn’t actually know how fast the slow-mo was on a Galaxy S8, so some trickery was required: by filming a timer in slow motion, we were able to find out how many seconds passed with each frame, which allowed an fps calculation. (The Galaxy S8 takes slow mo video at 240 frames per second). Then came the meticulous flipping frame by frame through the video to find the exact moment the vortex gets shed off the wing. We found that a similar image showed up every six or so frames, meaning that the vortices were being shed at a pace of about 40 Hz. What does this mean? A little more interpretation is needed on that question.

Today’s fact of the week: did you know a pangram, or a holoalphabetic sentence, is a sentence that uses every letter in the English alphabet? They’re often used to display typefaces, as you can see we did on our slides below:

My favorite pangram is “sphinx of black quartz, judge my vow”, which I think we can all agree is objectively the far cooler way to display an alphabet. There are far more examples, but I think perhaps the most applicable to aerospace engineers is: “ Pack my box with five dozen liquor jugs.”

Source for this adorable image

[simulit-blog]

Week 4- Eggy Rings and Bending Wings

It’s week 4 and First Team=Best Team is back in the wind tunnel, getting a disproportionate amount of wind tunnel time in (if the tunnel is open, we might as well use it right?). We hadn’t made our wing yet, so we decided to use one of Dr. Doig’s pre-made wings to make sure we knew how to set up the sting assembly and the smoke machine. We didn’t actually end up using the sting, but that’s a minor detail that I can discuss later. Below is a picture from that lab period.

Besides the fact that it looks like a still from a Michael Bay film, there isn’t anything particularly notable about this shot. I suppose it’s just proof that we actually know how to set the sting and pitch assembly up, sans wiring.

Triumphant in having set up the sting, we now set to our original plan of making our own NACA 4412 wing out of materials we had laying around the DBF lab. Luckily we found a wing already made out of EPP in the shape of what looks like a 4412. It had a chord of just over 6 inches, which is a little bit shorter than we wanted but we figured it was worth the time savings. All that was left to do was to cover it in Microlite to take away any surface roughness so that our VGs were working with as clean of air as possible.

(insert picture of covered wing here)

I didn’t actually grab a picture of the wing by itself, oops. There’s still a lot of other pictures and videos that show the wing pretty well.

The next step was to design some VGs. We found this nice website that very helpfully laid out how we should go about choosing a size for our VGs. Upon doing a boundary layer height calculation, we found that if we made our VGs 80% of the size of the boundary layer, they would be on the order of 10^-5 m tall, which is pretty difficult to measure, so we decided to make them 1.5 mm tall, since that was about 1% of our chord, an arbitrary number we found a lot of actual planes using for the size of their VGs (of course there’s quite the difference in Reynolds numbers between a 777 and some dinky foam wing in a wind tunnel). We made the length of the VGs 5% of the chord, per the recommendation of the website, so they ended up about 8mm long. Then, using this handy dandy equation:

We calculated a vortex radius (and thus vortex generator spacing) of like half a meter. Well, that clearly wasn’t right given our total span couldn’t have been much longer than .5m, so where did all the error come from? First of all, I blame whoever wrote this website because they chose variable names that are literally the exact same as the rest of the aerospace industry uses to describe wings. So in this calculation I was listing the planform area of the wing and the span of the wing, which, if you’ll take a gander at that handy “Variable Meaning” section, was not correct. Plugging the proper values into the equation spit out a spacing of 1 VG per 1 cm, which would have been about 50 VGs on our wing. We decided that was excessive, and decided to bump that spacing to 1 per inch (2.54 cm for you commies out there), and that would yield about 20 VGs on our wing, a much more reasonable number.

We didn’t jam the laser-cut VGs into the wing just yet, because we only had one wing and we wanted to test out the smooth wing first. I grabbed some video that shows roughly what we were trying to capture. The first is this profile shot that shows the flow maybe kinda sorta detaching toward the trailing edge? Either way, there’s not much to actually compare this smoke to, because we ran out of time before we were able to attach the VGs. We also had some, uh, wing flexibility issues. (excuse the rotation on the gif)

The picture below gives a pretty good idea what I mean by wing flexibility.

So if you can’t tell, that wing is actually bending incredibly strongly toward the glass, and not because it’s trying to escape the test section. It appears that when crafting the wing, we didn’t take into account that it would actually be experiencing lift (that’s a hell of a thing to forget, right?). And that that lift would be strong enough to bend the wing. Because of this, I think that we weren’t able to reliably detach the flow by changing the angle of attack. The solution to this bending problem would be to jam a carbon rod in the wing and affix it to the roof so it stays upright. One of the other reasons I included this picture is because it shows very clearly the effects that the wingtip vortex has on the flow, as shown by the smoke splitting off upwards (I thought it looked super cool in person).

As a bonus for sitting through my posts, here’s a neat thing that I saw when I was making some hardboiled eggs. Aerodynamics? No. But fluid mechanics is close enough right?

To start, I pinky promise I didn’t arrange the eggs that way to begin with, they did that themselves. I think the pot was just centered enough that there was hotter water in the middle than the outside, and this made a convection current that pushed the eggs to the outside of the pot. This would have had to happen while the water was still heating up of course, since water is a constant temperature when boiling. The eggs were no less delicious.

[simulit-blog]

Week 3- Oops

Week 3 was an amalgamation of testing and planning, though I must say the interesting parts only seemed to occur during Group 4′s time in the wind tunnel. I was somewhat intrigued by this portion of the lab, given that we were finally putting a quantifiable number to the size of the wake behind a wing, as well as the amount of drag that friction actually contributes to the overall drag number. I was back on data duty again, which means I was also on “write a MATLAB script as quick as possible to verify the data” duty. Once the data for the first set of probe measurements had come in, I was able to test it out. The result was something like what is shown below:

I don’t know about you, but that looks like a pretty good wake profile to me, though the data on the left side seems a little sparse. Having developed that code, I decided to take a bit of time to clean my code up so that instead of four for-loops I stow all the repeated calculations in a separate function to be called as needed. Once that was all coded up, I got some pretty good data except for one angle of attack in particular, α=6°.

Uhh, that’s definitely not right. I noticed pretty quickly, however, that the values for total pressure were all within 4 or 5 Pa of each other, which, upon inspection, was also incredibly close to the freestream total pressure. Clearly, we missed the wake somehow. Looking at the probe position, I can see that the probe swept from a position of 316 mm to 331 mm, which was actually pretty far away from the measured location of the trailing edge using the probe (about 308 mm, as I recall). It wasn’t until I had finished my code toward the end of the testing period that I noticed this mistake, so there was what some may call a mad dash to collect the rest of that data.

There are some pretty obvious signs that we had an issue with this run. There’s no obvious lowest point in the data, which could have been solved by moving the probe a little less through the wake (if we had time). There is also a pretty obvious grouping of data points toward the right of the plot, which indicates where we were taking our data before we noticed the mistake.

The rest of the wake plots look much better, however, and it should be pretty easy to grab accurate drag data using one of the other Josh’s code.

Part 2 of Week 3 tested our creative skills as First Team=Best Team had to decide what to do for our flow visualization project. A sort of running joke in DBF is to just chuck some VGs--vortex generators--on a wing to improve the flying characteristics. We decided to see if we could use flow visualization to analyze and quantify the impact (if any) that VGs have on the performance of a wing. Namely, we want to look at the change in the α that the flow detaches halfway (at x/c =.5) from a NACA 4412 wing, where the flow starts to detach, and (if we have time) the frequency of vortex shedding, all of those things compared between a wing with and without VGs. In dividing the labor for this project, I was saddled with the task of researching vortex generators.

Probably the most important things to do with VGs on a wing are their location and size. Too big and they will poke out of the boundary layer and add drag, and too small and they won’t energize the boundary layer enough to obtain the desired effect. The location is also important, though I have a feeling we won’t be able to get too detailed of data, and will have to be satisfied sticking the VGs on our wing in one location and seeing how that impacts performance.

So, how big will our VGs be? Will they be like those on a plane, small relative to the size of the body?

Or will they be like those on a Subaru WRX or STI that won’t come close to sniffing a track?

Some basic hand calcs will give a good estimate of the boundary layer at various points on our wing, and we will use that number to determine how big to make our VGs. Wing construction isn’t supposed to happen until Sunday, so we have a bit of time to work out those numbers.

[simulit-blog]

Week 2 Part 2-Dude, What’s Up With Those Lift Curves?

Week 2 Part 2, and Group 4 is back in the famed Aero Conference Room. In the meantime I have discovered a pace of 2 posts per week is probably unsustainable, but we’ll see what next week has in store.

Fresh out of the lab with our data, I and one of the many Joshes set out attempting to analyze that data with some lift code he developed two meetings ago. It took a few tries, but eventually we got it working

Wait, that’s not right. Never mind the completely erratic data for the lower speed data, the peak Cl value for the high-speed run is just about 4 times higher than one would expect it to be.

It took a bit of digging through the code to find the issue, but basically what had happened was when figuring out the lift on the airfoil, we had figured it out for a chord length of 1. When nondimensionalizing this lift into Cl, we divided by the chord length, which we did not need to do. Given that the chord length is about .25 m, this resulted in Cl values about 4 times higher than they should be.

That’s better, as long as we, again, ignore the low speed run. Somewhere along the way, we got some strange pressure data or lift data that severely affects the Cl values we calculate. But still, analyzing one set of usable data can still provide some insight into the accuracy of that data. Two easy ways to do this for a lift curve are looking at the zero-lift α and the stall α. The most noticeable feature of our calculated lift curve is that it does not actually reach Cl=0, which means that unless our method is wrong, we failed to capture the zero lift α. Some preliminary thoughts on how this happened: we decided as a group to test from α=-5 to -3 in order to capture the zero-lift α, but α=-5° was the lower limit for zero lift for lower Reynolds number flow. When we actually tested, we made a quick decision to change the speed of our second run from 15 m/s to 25 m/s.

Simulating this new regime in XFLR5 yields a zero-lift α ranging from just above -5° to just below -5° (The group of lines on the left). So, it is likely that speeding the flow up caused our zero-lift α to move outside our test matrix.

The peak Cl is also higher than one would expect. For a higher Re flow, the simulated data shows a peak Cl of about 1.3, not 1.4. Thus, it is possible that for some reason our data has shifted upward on the order of ~.1 Cl.

Referring back to the fixed Cl/α plot, the Cl values for the flow that has separated but not yet reattached is depicted by the black X’s. One thing I mentioned in my last post was wanting to see how the characteristics of the flow changed at equivalent angles of attack for flow that has (a)not yet separated and (b) separated but not yet reattached. The results almost universally showed a Cl greater by ~.05 on the flow that had separated but not reattached. It will take some more thinking to figure why this is, but I initially think that the answer lies in the recirculation occurring in the separated flow, and that this region has a slightly lower pressure than the flow does before it separates, thus imparting a higher lift force. I’m not entirely certain this is the right answer, however.

As I sat, ruminating about the right answer to the question above, as well as trying to figure out our issues with the lift code, I had an interesting (to me) question. How big is the boundary layer on the surface of the Earth? That question is easily answered by the Windows To The Universe website. I thought the website seemed less than reputable at first, but it appears it is written by Earth Science teachers, as well as sponsored through NASA and NOAA. So, we have an answer (50m to 2km, if you don’t want to click the link), but now my question becomes, how would one calculate that? What characteristic length would you use? It’s not like I can Google the equations either. In our classes, we as students are able to simplify things by looking at the closed system only. We assume uniform and simple inlet conditions, and things like that. But on the surface of the Earth, there is no “inlet”. The “wall” that the flow rides on has no beginning and no end. If you wanted to define the starting point for a wall, how would you define the ending point? Sadly, the Earth isn’t flat, so at some point you’re bound to circle around and reach your original starting point. When does the characteristic length “reset”? Suffice it to say, the Earth is a substantially more complex system than a flat plate.

[simulit-blog]

Week 2 Part 1-Hello MATLAB My Old Friend

The first meeting of class this week saw Group 4 spending the second half in the wind tunnel. We were looking for some pretty basic things- zero-lift angle of attack, stall angle, and the reattachment angle. Staring at a computer screen as data flows in isn’t a great way to find this out, though, and in the absence of time to write panel code I instead decided to try and verify the numbers flowing in. One quick way to do this is to look at the pressure data around where we predict stall to occur.

(A quick note, the first column is supposed to read 0, as that is the “reference pressure” for all the other pressures, so when I calculated the Cp I subtracted the average of the first column from the average of all the other columns.)

The data shown above is the pressure data taken at V=10 m/s and α=15°. Column 3 has the lowest pressure data, so that’s a good place to observe changes in pressure. Looking at α=16°...

Column 3 no longer has the lowest pressure data, but it is fairly close in value to Column 4, which is the lowest. What is notable, however, is that Column 3 has now dropped over 60 percent. This indicates that the pressure is now much higher at that port, indicating that the flow has stalled near the leading edge, thus the flow is completely stalled.

Looking at the theoretical data, 14°<α<16° is about when we expect flow to stall, so this pressure data appears to be correct.

Next is to make a Cp plot to verify that the scale of the numbers is correct. By the time I finished writing the script, we had finished with the V=10 m/s, α=6° run, so that was the first thing I tested the script on.

Ouch.

There are some pretty obvious fixes here, though. It was made very clear that port 7 (Column 8 in the original data) was not working, so the best course of action is to average the two data points around the faulty point. This is also missing a trailing edge pressure point, so I averaged the two points at the back of the airfoil for that number. The result:

This looks much better, but there are a few iffy data points (marked) that seem out of place. These indicate a partial blockage, and should probably be averaged out. Otherwise, the data looks about right. It should be noted that as the speed goes up, the fussy data points start to become less apparent. Below is the data from the V=25m/s, α=6° run.

Higher angles of attack smooth out the data even more (V=25m/s, α=13°):

Overall I’m thinking that this Cp code will be a good place to start getting some useful numbers from. It should only require a few additions to get lift and drag data from.

Bonus: Something not very apparent from the pressure data is the reattachment point once the flow has separated. We found that reattachment occurs almost exactly at α=14° when V=10 m/s (if you’ll excuse the commentary)

Because reattachment occurs at a lower α than separation, I would be interested to compare the pressure data before separation and the data after separation but before reattachment. As far as I know only the 15° data point would be in this range, but it should still be an interesting comparison.

[simulit-blog]

Week 1 Summary

I was placed in Group 4, which means that I won’t be in the wind tunnel lab until next week, so my work this week was pre-lab analysis.

As I had some familiarity with the XFLR-5 program, I was in charge of that analysis for my group. As a group, we decided that some of the things we would be looking for during the experiment was the zero-lift angle of attack, the stall angle of attack, and the reattachment angle. I figured that using XLFR-5 to generate a Cl/alpha plot for the NACA 4412 would be a good place to start to determine what angles of attack we want to test during the experiment.

Above is what was generated when I ran an analysis on the NACA 4412. I simulated Reynolds numbers 100k to 300k in increments of 25k, through a range of angles of attack from -10 to 25 degrees, in increments of .25 degrees. The specific Reynolds number range was chosen because our group wanted to stay within the laminar flow regime. The angle of attack range was determined using the limits specified in the lab manual. As is strongly evident in the drag polar and the lift curve, XFLR generates some erratic data for the 4412 at low Reynolds number and high angle of attack, so comparing experimental data to the XFLR data in those conditions will probably not generate any useful insight.

However, given the kind of information we are looking for, looking at only the lift curve is necessary. The lift curves for most of the Reynolds numbers simulated reach 0 lift between angles of attack -5 to -3 degrees, so we focused a few of our test matrix points around these angles. The lift curves peak, indicating stall, anywhere from 10 to 15 degrees angle of attack, depending on the Reynolds number, so we focused some more points in the test matrix in that range of angles.

Using the information gained from XFLR, as well as some intuition, we decided to test the wing at a range of angles of attack from -5 to 18 degrees, and then back down to about 9 degrees to attempt to capture the reattachment angle. We will be testing at speeds of 10m/s and 15 m/s in an attempt to remain within the laminar regime.