Compact method of generating and imaging shockwaves of aerodynamic models
Advisor: Mr. Kevyn Adams
Internship Location: Remote
Onsite Mentor: Dr. Elizabeth K. Benitez, DR-02 USAF AFMC AFRL/RQHA
The research will be novel experiments conducted in a Mach 1.5 shock tube built in-house. Shock tubes are a type of wind tunnel that generate 10-20 ms of high-speed supersonic flow and are used to test the aerodynamic conditions of supersonic models. They consist of a) a driver tube that holds higher pressure air; b) a driven tube that holds lower pressure air, c) a test section with the model, and d) operating equipment such as vacuum pumps, measuring equipment, and remotely-operated valves. When the diaphragm between the driver and driven tube is burst, high-pressure air will run through the driven tube and strike the model. In our case, the driver tube is held at atmospheric pressure and the driven tube is depressurized. The shock tube will have a maximum operating speed of Mach 3, an outer diameter of 2.5 inches, and a length of about 18 feet. Schlieren imagery will be the primary measurement system used to collect data on the vertical displacement of the model and shockwave formations. The final product will be the completed wind tunnel, as well as graphs of these quantities against certain independent variables such as model shape, model angle, and shockwave speed.
Project Blog
Post 1: Compact method of generating and imaging shockwaves of aerodynamic models(Blog #1)
Hi all! Welcome to my senior project: building a compact shock tube wind tunnel to generate and image shockwaves off aerodynamic models. A little bit about me: I'm planning to major in aerospace engineering, going to probably Cal Poly or Purdue. I love Minecraft, chess, cards, music, and art. I've attempted to build dozens of personal engineering projects, such as an electric arc welder, a laptop cooler, an algae harvester, a spring-loaded ball kicker, an RC car, and a small pinhole camera. A lot of these, of course, proved way too ambitious for me at the time, so you might be wondering why this project isn't similar. I've come out of these experiences, though, with skills and resources that should hopefully make this project a success:
Engineering philosophy: I've made many of my projects unnecessarily complicated. While this was okay in terms of building skills, complicated modifications usually cause far more issues than they solve, and the more complicated the modification, the harder those issues are to predict. Especially with ambitious projects, focusing on the minimum viable product can be extremely beneficial.
Physical resources and skills: I basically already have a workshop in my room, with tools, raw materials, and space. I've also picked up a thing or two, such as how hot glue is horrible at attaching to metal.
Advisors: I have two fantastic advisors helping me out for this project. My external advisor, Dr. Elizabeth Benitez, is a research aerospace engineer currently working for the Aerospace Systems Directorate of the USAF. She has extensive and specialized experience, having gone to both MIT, Purdue, and Georgia Tech. She was the one who convinced me that this project was possible while I was working under her last summer.
Okay, enough about me; let's discuss the actual project. Broadly, wind tunnels are devices that blow air towards a test section with a model inside it, with the aim of extracting useful aerodynamic characteristics of that model. Wind tunnel analysis of shockwaves and aerodynamic characteristics is extremely important, since real-world flight testing is often cost-prohibitive, and CFD(computational fluid dynamics, where you simulate the airflow and the model) can be inaccurate and requires computational time. In our case, we'll be building a special kind of tunnel known as a shock tube. This tunnel doesn't provide continuous airflow– in fact, it only provides a few milliseconds of it. However, that air is going supersonic, whereas the vast majority of wind tunnels operate subsonically. The few tunnels that operate supersonically, while having a large test section, are prohibitively expensive. Therefore, this research project aims to examine the feasibility of a low-cost, rapid-fire supersonic tunnel. We think that with the massive upfront cost with operating traditional tunnels, our tunnel can fill a niche, providing low-resolution data with unconventional designs that inform whether or not to proceed with more accurate study.
Design
I'd love to attach images of drawings but I haven't yet figured out how to without them becoming horribly pixelated, so bear with me. Our shock tube can be split into four sections, (1) the driver tube, (2) the driven tube, (3) the test section, and (4) the Schlieren system and other accessory systems:
Driver tube: A long 5ft, 2.5 in OD steel pipe with an endcap on one end and a diaphragm on the other. It holds air at atmospheric pressure.
Driven tube: A longer 10ft, 2.5 in OD steel pipe connected to the other side of the diaphragm and to the test section. This holds air depressurized to around 14.4 kPa for Mach 1.5.
Test section: A pipe assembly containing the model mount, valve to atmosphere, connection to the vacuum pump, and sights for the Schlieren.
Schlieren and other subsystems:
Schlieren: An imaging system which works by passing light through the test section, manipulating said light, and projecting it onto an image plane with a high-speed camera pointing at it. It accentuates the light's distortions as it travels through the test section filled with air of varying densities, and therefore, of varying refractive indices.
Diaphragm: A single-use 3D-printed heat-triggered diaphragm separating the driver and driven tubes. When the heat is triggered, it will rupture and the test will quickly commence, with the air being allowed to rush into the driven tube at supersonic speeds determined by the pressure difference.
Vacuum pump: The vacuum pump and its associated plumbing.
Control motors: Used to remotely trigger all non-diaphragm operations, such as vacuum pump and valve operations.
Electronics Control Board: Houses all the 12V high current batteries and fuses, and connects to the control motors and diaphragm.
To shrink the size, and therefore cost, of the tunnel, we can take advantage of the Reynolds number. Reynolds number is a function of size, air density, and airspeed. Objects of the same shape will have the same airflow surrounding them if the Reynolds number and Mach number of the air matches, even if one of the objects(such as the one in our test section) is scaled down.
Timeline
This project is hands-down the most ambitious I've ever attempted(and that's saying a lot) so I started early. I've been working on this project with my advisors since last October, drawing diagrams, ordering parts, and optimizing the Schlieren. Again, I'd love to show an image here but I can't. Here is our current timeline:
- Week 1: Finish modifying the 3D-printer, connect the vacuum pump, and start work on the JB-Welds
- Week 2: Finish the JB-Welds and start work on safety equipment
- Week 3: Finish work on safety equipment, such as window blockers, the pump enclosure, and a potential hallway door.
- Week 4: Conduct a depressurization test, noting down and improving operational procedures. If time left over, optimize Schlieren.
- Week 5: Create a rough model and conduct a full system test, noting down and further improving operational procedures.
- Week 6: If full system test failed, find out why(e.g. slow cameras, slow or nonexistent shockwave, etc.) and address the problem.
- Week 7: Finish addressing the problem and test again. If time left over, optimize Schlieren further using the data from the first and second test.
- Week 8: Further optimize Schlieren and develop a rigourous procedure for rapid-fire testing.
- Week 9: Work on the model processing, try to make a smoother model that is more realistic. Test once more with the smoother model.
- Week 10: Demonstrate rapid-fire testing capability by testing as many times in a two-hour period as possible.
- Week 11: Spare week in case any of the above weeks take too much time. If no extra time is needed, this can be used to show the versatility of the tunnel by testing unconventional objects(such as liquid droplets) in the tunnel, or even used to convert the shock tube into a Ludweig Tube with a makeshift 3D-printed smoothened nozzle.
- Week 12: Wrap up, collect and analyze data, and work on final presentation.
Outro
Thank you all for reading!! I'd love to go into more detail about any part of this. Outside of this blog, you can reach me at:
email: yashaskhattar@gmail.com
discord: the_tiny_lizard
Post 2: Compact method of generating and imaging shockwaves of aerodynamic models(Blog #2)
Intro
Hello again! This is my second senior project blog post after my first week of dedicated work. I caught an annoying sickness between Sat and Tues, so I couldn't really make as much progress as I'd wanted this week, but I still think I did some interesting stuff. If you haven't, I'd advise reading the first blog post for a bit of context about the design and how far we've gotten, but it's not necessary. We are essentially building a type of supersonic wind tunnel called a shock tube.
3D-printer modification
To do this, we absolutely need a good 3D printer. Not only would making good custom models and diaphragms be incredibly difficult without one, but a printer allows us to precisely print parts, holders, and fittings such as those in the Schlieren imaging system. I initally underestimated how important this was, going with a cheap(~170$) Anycubic Kobra 2 Neo printer. Although it got the job done for a lot of this project, and was easy to access, it was slow, buggy, and noisy. Knowing just how much I had used it(for context, I've used more than 5kg/11lbs of filament by now), and how much I was likely to continue using it, I eventually decided to return and replace it with the Flsun T1 Pro. This printer cost more than three times as much, but the benefits were tantalizing; it was one of the fastest printers on the market, with a max print speed of 1000 mm/s(four times faster), had a larger build volume, had an enclosure, and had remote printing capability. It proved to be a fantastic printer, without any serious bugs. Its noise, while slightly worse than the Kobra's, was easily mitigated by moving it into the attic, where I also planned to build and test the shock tube.
However, I have one main problem with my current setup. My main workspace where I develop 3D designs, experiment with small parts, and fit-test many printed parts, is unfortunately pretty separated from the attic space where I put the printer. It takes a couple of minutes to go upstairs and back for each model, which can add up. However, pretty much all printer operations can be triggered remotely, so I got an idea: what if we modified the printer with the capability to automatically remove models off the build plate? This way, I could print a bunch of different models and come upstairs to bring down and fit-test them all at once. I could potentially even automate it, successively printing as many files as I wanted to; the only limitation would be the filament. Printing issues such as bed detachment and stringing are also largely due to model design, so they can be debugged remotely as well. Lastly, this would give me experience working with the motors and batteries I'd use for the valves.
My planned setup was actually surpisingly simple; you might expect to need some extremely complicated robotic arm, but I figured all I needed was a small collector ring wrapping around the print bed attached to the printer's door. When the door was opened, the ring would move with it across the bed, collecting the model and dispensing it outside. All I needed, then, was an automated way to open and close the door, which I figured was possible with a rubber band system to apply tension and a motorized string system.
That motorized string system, however, proved extremely difficult to wire electronically. I started on this whole thing a few weeks ago, having constructed the structural framework with little issue. However, I'd also realized that my motors were kind of crappy and didn't work, so I swapped them out. This week, I began working on the electronics, and after three hours of wiring, rewiring, and referencing, I came to the conclusion that my four-year-old RC radio circuit boards might not have been up to the task. I'm considering ordering a new one, but I think I should probably move on for time's sake; radio-based remote controls aren't needed for any other system anyways.
Vacuum Pump Attachment
So I promptly moved on to working on the vacuum pump connection. The vacuum pump is used to lower the pressure inside the driven tube to the necessary 14 kPa(for Mach 1.5). My vacuum pump came with a manifold gauge set, which is a seperate part with a valve and pressure gauge used for control. Connecting the gauge set to the test section inlet was really annoying, though, and I ended up ordering a part that was useless to me because of different pipe standards, but this week I tried a new adapter piece and it fit perfectly.
I then plugged it in, with it connected to a 4-way connector; one end was connected to an end cap and another connected to the pump, so two air inlets remained open. However, when I flicked it on and put my finger on one of the inlets, I didn't really feel any air. Even when I covered it with my hand, I still didn't feel much. Without thinking, I put my hands over both inlets, and I immediately felt the suction. I yanked my hands off and quickly realized how stupid I'd just been. There was plenty of time to react, but if I hadn't reacted in time, my hands could've gotten stuck to the inlets, which I don't imagine would have been very healthy. But in any case, it had worked, so it was time to move on.
JB-welding
JB-weld is a type of adhesive meant to connect metal parts together. I've heard it's strong enough for vacuum use, and taking the entire thing to a welding shop would've been nearly impossible, so I decided to use it to connect the shock tube's parts. I'm kind procrastinating JB-welding the tubes, because it was really important to ensure a good-quality weld. If the weld had any leaks, air could enter the system, and if it wasn't strong enough, it could potentially implode and ruin hundreds of dollars worth of vacuum parts. That being said, I think I came up with a good system for doing it. I 3D-printed some short plastic tubes that fit snugly over the 2.5 inch OD pipe. Then, I would spread some JB-weld on the inner plastic/metal interface, fit the other pipe inside, and periodically use some paper to wipe off the JB-weld on the inner part of the joint. After drying, the joint should be strong and leakproof, with the plastic offering additional reinforcement. Although I printed a few of the plastic tubes this week, I haven't gotten around to testing the JB-weld, but I'll hopefully be finished with most of them by next blog.
Electronic Control Board and Remotely Operated Valves
As I mentioned, the shock tube had an inherent small risk of implosion, because to my knowledge, none of my parts or manufacturing techniques have been explicitly certified for use in an actual shock tube(such parts would likely be nearly impossible to find, be more expensive, and result in a non-novel design). For safety reasons, therefore, I'll be remotely operating it, and not even be able to see the shock tube. This means we need a remote way to reliably operate the system. Specifically, there are four parts that need remote input: the valve input to the vacuum pump, valve to atmosphere, on/off switch for the vacuum pump, and diaphragm trigger. The first three of these are controlled by high-torque 10rpm motors, and the last is controlled by sending electricity to resistance wires inside the diaphragm. All of these are controlled by an electronics control board physically wired to these systems. I worked on the motor valve mounts this week, designing 3D-printed plastic attachment points for the motors(attaching them directly would be difficult because hot glue is very weak when attaching to metal) and working on the control board. I didn't have the chance to print them though, because the printer was busy printing the tubes mentioned above.
Outro
Thanks for reading!! I hope to get a lot more done next week since I won't be sick, but we'll see. As always you can contact me outside this blog at yashaskhattar@gmail.com or the_tiny_lizard on Discord. Have a wonderful day and see you next time.
Post 3: Compact method of generating and imaging shockwaves of aerodynamic models(Blog #3)
Intro
Hi all,
It's been about two weeks since my last blog, and we've made quite a lot of progress. There are four general areas which I worked on during this time: electronics, valve control, safety equipment, and JB-welds. Each of these is essentially its own sub-project; the main goal of the electronics project is to create a visually intuitive system to control the valves and other equipment during operation, the main goal of valve control is to use these signals to control valves using motorized systems, the main goal of safety equipment is to mitigate risk to the environment and of fire, and the main goal of JB-welds is to connect all metal parts together with a vacuum seal. I've decided to try and use pictures in this blog to explain things better, but things might not format so well so bear with me.
Electronics
There are four main systems needing remote control during normal operation; the heat-triggered diaphragm, the vacuum pump switch, the valve to the vacuum pump, and the valve to atmosphere. The former is relatively trivial to set up; one simply needs a 12V battery able to provide sufficient current, a switch on the control board, and a fuse for safety. The vacuum pump switch is also similar; you might expect to need to control the motor both ways, but we can 3D-print an attachment such that when the motor is continuously rotated, the vacuum pump switch gets flicked on and off. The valves, however, are more analog devices that need precise control, so I don't think this method is viable for them; I'd have to have two separate open/close switches. In hindsight, I could have just used two wires with separate switches and reverse polarities, but I'm kind of afraid of what'll happen if I accidentally turn on both the forward and backwards switches. So, I resolved to electronically use a control circuit called an H-bridge to send signals, shown below. I believe the control board is pretty much finished now, which is a major milestone.
Valve control
Closely related to electronics is how one would actually control the valves using motors. I've already spent way too much money, so I ordered regular valves and resolved to try and and attach motors to said valves. One of these valves is quite loose, but the other is somewhat difficult to turn even by hand. For the latter valve, I planned to have an attached motor pulling on a string which would be connected to the valve handle. I also had to attach rubber bands to it to have it return to its closed position when the axle was rotated the other way and the string was fed back out. This one ended up being extremely difficult, and despite using 12V motors that rotated at only 10 RPM, it still wasn't enough torque to rotate the valve with enough rubber bands to have it return. I had to extend the valve handle to even get close to the required torque, and the motor sounded like it was dying. However, my batteries were kind of crappy and they kept running out, so hopefully once I replace them with an actual 12V power supply, it'll run a lot better. This experience genuinely taught me a non-negligible amount of all my structural engineering knowledge up to this point, as I spent hours testing it and adding struts where they were necessary. Shown below is the scuffed, overbuilt, and yet necessary contraption I built around the valve.
Safety equipment Safety equipment is, of course, extremely important when working with large and ambitious projects like this. I built two main structures to address these concerns; the vacuum pump enclosure and the window blockers. The vacuum pump enclosure was made of refractory fire-resistant bricks which would hopefully somewhat contain any explosion(for example, due to the diaphragm breaking prematurely with the valve to the pump still open and the shockwave able to travel into it). To connect these, I used refractory cement, but my first attempt was kind of a disaster. Although it did hold the bricks together, I don't think it was very strong, and there were huge gaps in between where I applied the cement. The main problem was that I poured in the cement from a tiny hole in the packaging(taking even more time in the meanwhile trying to use a knife to progressively widen the hole in impatience) and by the time I'd added all of it, the cement only had a few minutes left before it became too thick and hard to work with. I decided to have a second go at it, so I ordered more cement and cut a massive hole in the cement container using a saw blade. I then quickly poured it all out, mixed it, and applied it, this time using different and better tools. I even used my (gloved) finger to spread everything out along the gaps, and even though it trashed the gloves at the end of it, the results were pretty worth it.
Simultaneously, I began work on the window blockers. I plan to use these to block two windows in the room, primarily for noise reasons but also to stop any sunlight from messing up the Schlieren(and reminding me to go outside). These windows were around 28x61 inches, and there were two of them, so I'd have to work at a large scale. Last Friday, I went to Home Depot with my mom and got eight 6 ft long, 7.5 in wide, 3/4 in thick pieces of wood(using four for each window). After hauling them upstairs, I cut them to the required size and connected them using screws and smaller wood pieces, using some wood glue for good measure. Many of those wooded pieces cracked while being screwed in, but I had a limited amount of them, so for some I just had to deal with it and use wood glue to fill the void as best as possible; I also cut some more small wood pieces from what I had cut off earlier.
I also plan to use spare wood to create the blocker's "feet", since I'd ideally want them to be able to stand at least partially on their own without too much wall support. I tried to 3D-print the said feet, but I was running into issues due to the large print size, which requires bed glue due to adhesion issues; I kept accidentally waiting too long, so the glue often dried before the print started. I also think large models are generally harder to get right, as the one time it did keep adhesion, it had another strange misalignment/defect. Therefore, I sandwitched mostly empty PLA plastic spools with spare wood to make up the feet, and plan to just lay them on the floor to prop up the blockers. They aren't pretty, but they're precise where they need to be, so they should work. I also plan to secure them to the wall using some screws, and detach them when the tunnel isn't in use, but I'm not sure how that'll work just yet.
JB-welds I believe I described the process for making these welds in the last blog, but essentially, thick plastic 3D-printed pieces were used to align and connect the metal pipes together using the epoxy JB-weld. I think this might prove to be a holdup; since these are precise welds that need to hold vacuum pressures, and messing up means ruining expensive metal parts, I keep procrastinating actually making them, which isn't great.
Outro/Future Progress But in any case, thanks for reading!! I believe we might be able to do a depressurization test(no model, diaphragm, or Schlieren) by the end of this week, but I'm not sure. If I do, I'll let you know the results in the next blog. Stay tuned and see you then! -Yashas
Post 4: Preparing for depressurization test(Shock Tube Blog #4)
Hi all,
Happy Pi Day! This week I continued to work on the systems needed for the depressurization test. To recap, we need to finish work on electronics, plumbing control, safety equipment, and JB-welds. The electronics are pretty much done; I added a DC power supply to power the motors, which provided the required voltage much more consistently than my low-capacity batteries. The plumbing control is 2/3 done, I just need to wait for the epoxy on my motor to dry and then I can attach it to the vacuum pump to finish the 3rd system. The safety equipment is mostly finished now, but there are a few small tasks I still need to do; I might add an attachment to make window blockers easier to mount flush, add some plastic pieces to protect the drywall, add the noise-reducing covers which I've made, and bring up a fire extinguisher and ear protection.
However, I haven't made as much progress on the JB-welds, and they'll likely be holding me up for a few days. The four horizontal JB-welds that I have to do are especially tricky. I tried my hand at one of them on Wednesday, and it was kind of a mess. First of all, aligning and mounting proved to be quite hard, since small variations in my 3D-printed mounts and the pipes themselves made centering difficult. Also, they were along a wall, and any JB-weld on the lower surface of the pipe would just drip off, so I could only really apply it on a quarter of the surface and hope that rotating the other piece with the plastic would spread it around. Lastly, with the pipes oriented horizontally, gravity slowly pulled the JB-weld downwards. I attempted to compensate for this by spreading it around where the metal met the plastic, but I might have made some air bubbles. Some of it also got on my hand, which was not great. However, I think I've learned some important lessons that'll make it much easier next time.
I'm very hopeful that we'll finish the JB-welds and the vacuum pump controller by the end of next week, so we can potentially do a depressurization test then. Basically, I'll omit the diaphragm and evacuate the entire system to expected operating pressures. If I do, I'll be sure to let you know the results!
Post 5: Depressurization test soon/ hopefully full test next week! (Shock Tube Blog #5)
Hi again!
This week we finished the main pressure vessel of the shock tube and will soon do a depressurization test(hopefully tomorrow)! This represents a major milestone, arguably the most major milestone after a successful full test. The main thing we needed to do was finally finish the J-B welds that had been holding me up for quite a while at this point. It took a while(about 30 mins for every vertical connection and an hour for every horizontal connection), and I had to wait about a day for it to cure before I could move the part again to align for another test, but I eventually finished the entire pressure vessel.
While I was waiting for the J-B welds to dry, I wanted to design a few silly experiments. I was actually really curious about how different things would hold up to near-vacuum pressures(but above the 3 kPa vapor pressure of water, below which it starts to boil even though it's at room temperature). I also think it'd be kind of cool if I could use them as actual models, and see how they held up to shockwaves. Specifically, I decided to test a piece of ravioli and a strawberry. I figured these would be fun guinea pigs to use before I figured out an actual way to 3D-print and smoothen models, but they might actually carry some scientific value; they might provide insight into how humans, animals, and fruits respond to a vacuum(such as in a spacecraft leaking air) or to a shockwave(such as in a nuclear blast). However, I didn't exactly want strawberry juice strewn all over my brand-new wind tunnel, so I designed and 3D-printed a container to hopefully let air escape while also containing any splatter. This proved quite difficult to print; I tried a circular design, but it kept detaching from the support and bed, so I switched to an equilateral triangle model. This picked up a severe and worrying warp, but the model surprisingly stayed attached. Unfortunately for me, I didn't know how basic geometry worked, and I had to print two more versions before it actually fit in the tunnel.
Unfortunately, I don't think I'll use these in the first depressurization test. I was originally planning to use just normal Scotch tape to attach the covers to the chamber, but last-minute I decided to research if that was safe, and they apparently release X-rays when peeled in a vacuum(and [duct tape] outgasses formaldehyde according to ChatGPT). This was really random, but also well-evidenced and scary, and I couldn't find info about the other tapes I had on hand. Therefore I'll probably have to order special vacuum-certified tape or reprint with bolt mounts.
I also worked a bit on the diaphragm while waiting for the J-B welds. This is essentially a part that blocks air from transferring from the driver to the driven tube until electricity is sent through it, heating it up and decreasing its strength to the point where it bursts. Originally, I planned to use 0.1mm nichrome resistance wire in the configuration shown to the right(black = nichrome, grey = aluminum). However, finding the electrical resistance of this(which I needed to see if my power supply would break) mathematically is quite tricky. I didn't have the knowledge to do it, and when I asked three different ChatGPT models to hopefully save me, two of them gave me completely different formulas, and the third errored on me. I then realized that I had a multimeter, and I could just build the setup and test its resistance that way. However, when I tried to actually attach the wires on the 3D-printed board, I had a substantial skill issue. The wires were really thin and hard to work with, and I wasn't sure how well I could consistently make diaphragms like this. It was also really time-consuming, and I was doubtful about how well the wire connections would actually conduct electricity. Also, since the wires were so thin, I was worried about heat conduction between the wire and the diaphragm plastic being less than ideal.
I figured that instead of this, I could use stainless steel mesh. I originally bought this mesh to filter the air going into my vacuum pump, but I figured it would work really well here. It would just be a single piece, and I would only have to connect external power to specific places of the mesh, which should be way easier, faster, and more consistent. However, I had to cut it in such a way that it a) focused heating on only the areas I want to heat up, b) did not go under the required resistance, and c) would easily break. It should also be as consistent as possible, to make sure they were always triggerable in the same general pressure range. My solution to these requirements?
The humble apple slicer, it turns out, matches the dimensions of the cutting I need pretty well, and it's also really cheap. My plan is to 3D-print a mount for it, and slide it down guillotine-style onto a precut circle of mesh on another 3D-printed cone-shaped mount. The mesh would fold down, allowing me to control exactly how far out radially the cut goes. Once it reached a predetermined spot, I could lift up the slicer and make a few more cuts with scissors. The to-scale final product should look something like the image on the right, but this is still a work in progress, and I need to see if the resistance is good enough.
But in any case, I shouldn't need to do much else before a depressurization test. Most of the safety equipment is up and the parts are currently being bolted together. Thanks for reading and see you next time with the results!
Post 6: 99% of engineers quit before they finish (Shock Tube Blog #6)
Hi all,
Okay, let me explain. I didn't actually do a diaphragm test this week like I had been planning to for the past couple of weeks. I'd run into issues which were honestly kind of minor and simple to fix, but the solutions take a lot of downtime to implement. Despite what the title(and being like two weeks behind) might suggest though, I'm nowhere near quitting or scaling back, and I still intend to complete at least a few tests of real aerodynamic models to prove the viability of this concept.
Let's start with the diaphragm. I thought I'd created a diaphragm that would fit perfectly in with the O-ring, but that was quite tricky to get right, especially on a horizontal connection. I was originally testing it with mounting it vertically, but a horizontal connection meant it was way more shifty and hard to align, so I ended up redesigning it a few times. Also, I was planning to just use glue to secure it to the buildplate like I did with the other models, but some of this glue remains on the model and likely will outgas in the vacuum environment, leading to all sorts of issues(mainly safety ones). Without this glue, I had to scrub the print bed with alcohol in order for the print to even have a chance of sticking. Despite the fast speed of the printer, these considerations, along with the fact that I used 100% infill for maximum strength, delayed the process.
Another delay was the actual bolting itself. The parts are really close to the table and next to a wall, making some of the bolts extremely difficult to work with. I also 3D-printed some wrenches for tightening, but they weren't really strong enough, and I needed more of them at different angles. This is definitely something I look forward to improving, because replacing the diaphragm and model should be a routine process.
I also realized that the actual electrical connections needed to be made between the motors and the control board, mapped to the correct inputs. This took a fair amount of time to test since I had to keep running back and forth, but it wasn't too bad, and everything worked as designed. I also needed to attach the phone mount to the table, but when I did, I realized my phone was too heavy and the mount was way too flexible. I thought I had fixed this by reinforcing the main structural wire with aluminum, and using zip ties and tape to secure it, but apparently not enough. I might have to try with a temporary fixed phone mount to continue, but we'll have to see.
The main issue, however, that completely doomed the prospect of a depressurization test this week, was with the bolts. The bolts on other connections always had a bit of extra length, and I had always assumed that it would be enough to accommodate a thin diaphragm in the middle. However, this assumption turned out to be completely wrong, and the diaphragm was way fatter than I pictured it to be. It wasn't nearly enough to fit, and since it was Wednesday by this point, I was pretty much out of time. I was leaving to visit Purdue later that day, and in hindsight I maybe should've just went out to physically buy longer bolts, but I'm not sure. In any case, that'll be it for now!
Post 7: Depressurization test!(Shock Tube Blog #7)
Hi all,
As the title suggests, we finally conducted a series of depressurization tests! Before we get to those though, we have to go back to last Sunday when I got back from visiting Purdue. I'd finally gotten the longer bolts that I needed, so I figured that it was just a matter of attaching them, but then I realized that I had made a huge mistake:
I had J-B welded it such that the original bolts were essentially locked in place between the plastic alignment tube and the flange. I now had the glorious task ahead of me of manually shaving off bits of the alignment tube to create enough room to take out the bolts. I couldn't use any cutting or sanding, because that would create a lot of plastic dust riddled with trace J-B weld, which wouldn't be great to breathe in. Thankfully, I had some foresight and offset the tube a bit in case something like this happened, but it still took almost five hours of filing. It was made better by YouTube videos playing in the background, and me making my own knives that were more convenient for something like this, but it was still far from ideal.
Two days later, though, I had actually finished the entire thing, and it was time to finally test it!
I started off by setting up my cameras. I had my old phone(before y'all yell at me for upgrading, this was an iPhone 7 and didn't charge half the time) looking at the gauges, and my new phone propped up against my gloves box overlooking the test section and diaphragm. I hopped on a Zoom call with the two phones in the other room where my control board was, and put on AirPods and a set of drumset earphones over those. This way, I could still get the audio from one of my phones and be able to hear stuff, while also being protected from any excessive noise caused by an implosion. When I felt ready, I flicked on the power supplies to the vacuum pump and the motor controller, opened the valve to the vacuum pump, and turned on said vacuum pump.
The pressure gauge quickly decreased to around 95 kPa absolute pressure, but then just stayed there and never got lower. I initially thought that it was just taking its time due to the large volume, but even after waiting for a while, it never got below that 95 kPa mark, and as soon as I released the vacuum, it quickly returned back to atmospheric pressure. I then just attached the vacuum pump directly to an end cap, and the pressure quickly lowered to 0, so I figured that I probably just had a bunch of leaks. I did a bit of research and found that the connectors really needed to be tightened down, so I retightened all the bolts again and tested it, and it got to around 90kPa this time. This was obviously better, but still nowhere near what we needed, so I 3D-printed custom wrenches to allow myself to tighten some of the annoying bolts down.
However, it still didn't get much lower than around 85 kPa, and I began to worry a bit. During that test, however, I ended up doing something that would completely change everything. I decided to take off all my hearing protection and go into the testing room to check for leaks. This was pretty reckless, since if it decided to implode my eardrums might get cooked, but I figured it would probably be fine. I got really rewarded by this, though, because I heard a really clear and loud leak at the end cap connected to the driver tube.
When I tightened that down and tested again, I was amazed at the results. It seemed to go all the way down to 15 kPa absolute pressure, which is almost what I need to actually trigger a shockwave! Not only that, but it also repressurized way slower, so I'd have time to close the valve to the vacuum pump and trigger the diaphragm. After this test, I lowered the lights and turned off the 3D printer to lower background noise for next time, because it was clear finding leaks this way was incredibly useful. I found another leak stemming from an improper installation of the vacuum pump mesh filter I had cut, where it got in between the O-ring and the pipes, visible in the imprint of the mesh on the O-ring(might be hard to see on camera).
After fixing this, I was surprised to see it not go even lower, but in hindsight I think it was actually lower than what the gauge might suggest. The camera was looking at it from the side, so I think the needle was more to the left than it suggests.
In any case, thanks for reading! I know it seems like I'm getting a bit exited over a gauge needle moving, but this is actually really relieving for me. It validates that my whole strategy of using J-B weld to connect weld-on parts together did actually work, and I think this is probably the second biggest milestone of the project(second to the full test). See you in the next one!
Post 8: I kind of hate a specific plastic now(Shock Tube Blog #8)
Hi all,
Since the recent successful depressurization test, I've been really hopeful in the future of this project, because that actually validated the viability of J-B welding joints like this, as well as the viability of using the 3D printed plastic diaphragms to hold a vacuum. The general next step is conducting a full test, and there are three main things that we need to work on before then: Schlieren, diaphragm, and model mounting. However, I ended up only working on the model mounting due to a combination of just losing important stuff and other issues we're about to get to...
Early in the week, I began designing and printing a few parts to hold the model in the tunnel. My idea is to use a stainless steel rod connected to a 3D printed part which fits snugly in the tunnel. However, I'd realized that all my planned mounting mechanisms from a few months ago used some kind of adhesive which would probably outgas in vacuum and cause all sorts of problems. I also really didn't want to J-B weld it, because that was a strong and permanent epoxy which would probably be near-impossible to undo. I was pretty much left with only one real option, which was to fix it inside the tunnel using just stainless steel nuts and bolts. I designed the following setup, where the half-cylinder would fit neatly inside a cross-connector, and the partial cone would slot into the KF-CF adapter piece. Then, new models could be attached to a rod connected to the third expendable piece, which would be temporarily bolted to the partial cone. All parts were designed with air ducts to allow air to be sucked out of the driven tube.
I'd initially had a lot of trouble printing these; I thought I couldn't use bed glue because the residue would likely outgas in vacuum, so I resorted to just cleaning the bed with isopropyl alcohol and hoping for the best. That was really inconsistent though, and these parts in particular just couldn't stop detaching from the bed. However, I then had the idea to use a sacrificial plate between the model and the bed. This way, I could use glue to secure the model to the bed without having the glue residue remain on the model itself.
This actually worked out really well, and despite some surface roughness, I'm really happy with how the parts turned out. All parts were made using high infill PLA plastic, which should be strong enough to resist the forces given that they're mounted securely. However, what it wasn't strong enough for were my wrenches:
Many of my CF connector bolts were in really weird positions, and I'd need to constantly tighten and untighten them to put in new diaphragms and models. I tried making custom wrenches for these out of PLA, but even after making them ridiculously thick and cranking up the infill to 70%, they kept getting deformed, especially after multiple uses. I really didn't want to have to print a new wrench for every two bolts to be tightened, so after some thought, I decided to try and look for a new material. My printer, the Flsun T1 Pro, is quite advanced, and the manufacturer stated it could print PLA, PETG, PVA, PET, ABS, ASA, PA, and PC. After some research, I found that PC(polycarbonate) was probably best from these options, being stronger and harder than PLA. However, it was apparently really tricky to print compared to PLA, and is is also twice as expensive. I figured, though, it was about time that I graduated to a better filament in general, and even if it didn't work for wrenches, it could be a handy skill to know.
With that, I ended up buying a 1kg spool and it arrived a few days later. It felt bendier and more elastic, which was odd considering the stress-strain graphs I had consulted earlier while researching PC, but I was still really excited to try it out. I immediately cracked it open and tried to print a bunch of wrenches. Considering how many I was trying to print, and the fact that it was my first attempt, it went surprisingly well. However, after the first layer the filament broke off:
and resuming it probably messed it up. I think it was also really struggling to print the infill pattern, so I just went with 100% infill from then on. However, it was still insanely finicky, and I spent the next two days just solely focused on trying different things. A lot of research and many attempts later, this was the best procedure I could come up with:
1) Wipe down the print bed with water and isopropyl alcohol.
2) Set the temperature settings to PLA and wait a few minutes with the door open to vent alcohol fumes.
2) Add some PVA glue stick and spread it around the bed with napkins or a microfiber cloth.
3) Bed level.
4) Wait 10 minutes.
5) Start print with hotend at 280˚C, bed temp 60˚C, min fan speed 0%, max fan speed 65%, temp control w/ 40˚C chamber.
I ended up getting something that was about 30 layers thick as my best attempt, but it was really warped. The layers also got misaligned, so despite some cleanup with an Exacto knife, it still couldn't even fit a bolt all the way through.
I was really stuck, and despite wanting to figure out how to print PC, I knew it was taking up way too much time this late in the game. I could order a 3D print online out of a different material, but that might not even be strong enough, and I might get the dimensions mixed up. Also, depending on the infill and material it would be somewhat expensive, and it would take a while to ship. My girlfriend then suggested just bending an off-the-shelf socket wrench to fit the angles that I need, which she'd successfully done for a robotics project. I thought that this'd be difficult, but probably way easier than whatever I was trying to do with the printer, so I looked on Amazon for some socket wrenches. I then found one with a hinge, so ideally I wouldn't even have to bend it at all, so I ordered it along with a bunch of bits(hopefully one of which can fit the bolts I have). It wasn't easy, but I think I can get all the bolts with this new wrench without having to bend it, so I should be good on that front.
I also switched my 3D printer filament back to PLA, and put the PC filament in a storage box along with some drying agent in case I ever got the courage to try again. However, you are only really done with PC once PC decides it is done with you; I afterwards accidentally tried to print PLA on PC temperature settings, which was not only embarrassing, but also a huge pain safety-wise. I had caught it really early after it had only printed a few grams, which was good, but you really aren't supposed to try to print PLA at a temperature 60˚C higher than normal. According to ChatGPT(but notably not really anything else), when heated excessively, PLA can break down into various smaller molecules including formaldehyde! I also noticed a slight smell when opening the printer door, which wasn't a great sign. I decided to just clean up the print bed with a mask and gloves, and come back tomorrow to leave any potential vapors to just vent out.
This week was definitely one of the less productive ones, mostly thanks to my admittedly strange decision to try and print a way more difficult material to chase a marginal performance improvement, but I still think we can build out the rest of the system in time. Thanks for reading and see you all next time!
Post 9: Full test soon! (Shock Tube Blog #9)
Hi all,
I think we're getting close to a full test! At a good time too, because a month from now is presentation day. My main priorities were finishing the model mounting, integrating the Schlieren setup that I had tested, and trying to figure out how to cut the diaphragm. Let's start off with the model mounting. Because it was kind of cool, and because I wanted to start off the full test soon, I decided to just go with a piece of pasta. It was a simple cylinder with a diameter of about 9 mm, and I designed and printed this little mount for it(using aluminum wires to hold the pasta in place). I kind of messed up and forgot to cut the metal rod to make it such that the pasta was lined up with the test section, but after doing that it was looking pretty good.
My next priority was engineering the diaphragm. I had to cut the a stainless steel mesh in a very specific pattern, so that the diaphragm would only heat up and break in specific places. I designed and printed two halves of a cutter using the apple slicer that I was talking about a few weeks ago. After destroying said apple slicer to get the blade section and intentionally dulling the parts of the blade I didn't want to cut the diaphragm mesh, I was left with this:
With my Spongebob arms and weak hammer though, I couldn't make it actually go through and cut. What was worse was that the clamps I had got to hold the steel mesh in place just barely didn't fit, so the mesh curved upwards where it wasn't being cut, which would be really bad for the heat conductivity. I made some stencils, however, to cut the diaphragm into the initial circle, and they had worked really well. I could use the clamps to secure both of them to either side of the mesh, and use an Exacto knife to cut it somewhat precisely. I figured I could just use those stencils to cut out the whole pattern, so I designed and printed these ones here(still have yet to test them):
While I was working on this, I'd also picked up moving the Schlieren system components upstairs to the shock tube room and integrating it. My prior attempts while developing and testing this system a few months ago were kind of disappointing, because they were really dim and blurry(the following is footage of the hot exhaust of a candle, inverted):
Video: https://www.datocms-assets.com/221631/1790630437-schlieren-attempt-4-noaud-shortened.mp4
Although picking up normally invisible air disturbances like this was kind of cool, I really wanted to figure out a way to make it brighter and sharper. Midway through setting it up upstairs, though, I figured I may as well just stick my iPhone camera directly looking at the light(instead of looking at the light reflecting off of a piece of paper) and I was kind of shocked by what I saw:
It was really clear and bright compared to last time with the candle, and you can see fine details of the model like the aluminum wiring. I then stuck my actual high speed camera on there, and finished the setup by aligning a razor blade in front of it at the focal point, and that worked pretty well too. To figure out if it could actually catch any air disturbances, though, I'd have to run a few tests. My first idea was to just use my mom's hair dryer positioned in front of the lens:
However, I couldn't really distinguish whether the slight disturbances on the image were from actual air movements or from the vibrations caused by the hair dryer. The camera was focused on the model, which was further back in the setup than the hair dryer's air, so I think it was really out of focus and blurry. I tried cranking up the heating, because that wouldn't affect the vibrations but would affect the air disturbances, but it was still really difficult to tell. After a while, the dryer was apparently consuming way too much power, and it triggered the circuit breaker on my power adapter. I also smelled smoke, which isn't something you usually want to have with electronics. I figured it was about time to call it quits with this method, and instead, I decided to use a candle in front of the lens:
This way, I could avoid exploding a hair dryer, but also hopefully get some better results. I used a focusing object where the candle was, so I could tune the camera to focus on that instead of the model, and it was actually focused at the correct distance. Unfortunately, I forgot to record the footage from this, but it was looking okay, and I could see the candle was definitely causing some air disturbances beyond the visual flame. I figured that if it worked okay in this situation, it would probably work as well or better when it was actually focused on the model, so I called it done for now.
That's about all I have for you! Thanks for reading and I think we're getting pretty close to a full test with the Schlieren and model mounting done, but I'm not sure how long it'll take to make and test the diaphragm. See you all next time!
Post 10: Stuck in diaphragm hell (Shock Tube Blog #10)
Hi all!
Sorry for the lack of comms last week, I didn't feel like I'd made enough progress to really show off. As it turns out, designing a diaphragm with appropriate thermal and electrical conductivity that's safe for vacuum is quite difficult. The goal of the diaphragm is to hold the pressure from the driver tube until the time comes to trigger the testing run, at which point electrical current is passed through a heating element in the diaphragm, causing it to break and allowing the air to burst through the driven tube. It would probably be fine if I'd allowed myself to use hot glue, tape, solder, etc., but I wanted to be safe; it was still possible that even under the very rough vacuum I was pulling, these could outgas some potentially dangerous substances. Therefore, along with the PLA plastic, I could only really use materials like copper, stainless steel, aluminum, and J-B weld. I was even more limited by the fact that my aluminum wire was difficult to connect through electrical contact(potentially due to some thin non-aluminum coating?) and that J-B weld is difficult to consistently work with on a small scale and is likely an electrical insulator. After several iterations, I think I came up with a relatively easy and fast process to manufacture diaphragms with consistent properties. This is quite long and honestly as much for myself as for this blog, so feel free to skip over it if you aren't interested. 1. I cut out a small piece of stainless steel mesh. This is easy to do with just normal scissors, but the mesh fibers are extremely thin and carelessness can lead to cuts. Although stainless steel is a good electrical conductor, the low thickness gives it a slight resistance of about 0.9 ohms, which provides adequate power when connected to the 12V 12.5A power supply(I believe the wiring to and from the testing room will provide >0.1 ohms of additional resistance).
2. I printed out two cutting stencils without bed glue. After aligning them and clamping them down against the mesh, I got out my exacto knife and cut along all the stencil lines. I then put away the stencils and cut out the final pieces of connecting mesh. Lastly, I bent the fins upwards slightly to ensure that holding the fins down also held the central bit against the diaphragm plate. These stencils were really useful in allowing me to consistently cut diaphragm heating elements in the same shape, which is instrumental when trying to predict what pressure the diaphragm will fail at given a certain heating time.
3. With the heating element done, I printed out two more cutting stencils and grabbed a 3 inch silicone O-ring. I then clamped it in, poked four holes in it using a sharp screwdriver, and inserted four 12+ inch 1.3 mm pure copper wires. These wires serve a dual purpose. Anchored to the outside of the tunnel, they kept the four major fragments of the diaphragm from being flung through the tunnel by the shockwave. They also serve as electrical connections to the heating element. It is true that these wires might introduce leaks through deformation of the O-ring, but since the O-ring would be on the side of the system with atmospheric pressure until after the diaphragm is triggered, these leaks are likely inconsequential.
4. I then printed out the diaphragm plate, removed all the 3D support, and placed the O-ring into the slot. For each wire, I pulled it out until it met the other side of the O-ring, and then wrapped it around the 3D supports. I had initially planned to use J-B weld for this, but as I mentioned before, it was hard to work with on this small scale, and a small slip-up could leave the copper wire covered in a thin layer of it, making the electrical connection extremely difficult and inconsistent. It also required PPE and at least a day to dry, and after testing, the connection wasn't very strong. Using supports like this made things way easier, and the only real downside would be that I'd have to be more careful about the copper touching the inner part of the diaphragm.
5. I then got the heating element from earlier, carefully positioned it in the middle, and used some thin copper wire pieces to electrically connect the ends of the diaphragm to the copper wire. One of these was also a mechanical connection, to ensure that the heating element also doesn't get flung across the tunnel. I didn't need to solder anything, since I made the second diaphragm plate compress both electrical connections.
6. Finally, I printed the other diaphragm piece and aligned it such that the grooves fit into the thick copper wire. This is a bit finicky, but since copper is so flexible, it's usually not a problem. The final assembly can be clamped until ready to bolt in.
7. Final installation just consists of unbolting, taking out the previous diaphragm, adding the other two O-rings, and rebolting the new diaphragm in. I also wrapped the copper wire around the connection anchor on the driven tube, and connected them to the positive and negative wires from the control room. As a quick point of safety, the flange is also wrapped in electrical tape to ensure none of the bare metal wires make contact with it, and I made sure my power supply had short-circuit protection.
In hindsight, I probably should have at least tested the diaphragm, but I don't have many doubts that it will at least heat up. This represents the last major component I still had to design and build! I'm currently making finishing touches and should be ready to finally do a full test of the system on around Tuesday(fingers crossed). Thanks for reading and see you later, hopefully with the results!
Post 11: First two tests failed (Shock Tube Blog #11)
Hi all,
We've had a crazy few days of testing; I'll try not to go into too much technical detail because it gets really confusing really quicky, but feel free to post a comment if you're more interested in a certain part!
We left off finished with making and installing our first diaphragm, so after a few more finishing touches, we tested the shock tube for the first time on Tuesday. However, it was actually pretty uneventful. The diaphragm just didn't pop, even when I repeatedly depressurized the driven tube and powered the heating element for a long time. The battery was heating up, so I figured current must be making its way through the circuit and heating up the diaphragm as well, but when I opened up the system after venting the pressure, the diaphragm was just at room temperature.
At first, this really didn't make sense to me, but after a lot of multimeter testing, I found two potential reasons why the test failed. The first is the inconsistency of my power supply. I had used 12 volt batteries which claimed to have a max current of 23 amps, but I found that in practice, when given a <1 ohm load, they provided almost no voltage. This was very disappointing, and it would probably explain all the issues I had had with those batteries before. I decided to just switch to using my power supply, which actually gave a consistent 12V voltage.
After fixing this and testing the diaphragm, it was heating up, but it was really inconsistent due to the second reason why the first test likely failed: the finicky nature of the diaphragm wires. Sometimes, when the diaphragm was moved or handled, they would get out of alignment and cause the resistance to spike. I eventually decided to just cut grooves into the mesh where the wire could sit. This made things more robust and resistant to movement, and combined with more careful handling procedures and the use of hot glue on the edges, I was able to mount the new diaphragm tightly in the tunnel. After tightening all the bolts, I measured the resistance again, but it was only around 0.2-0.3 ohms instead of the 0.4-0.5 ohms I was getting before. I was a bit worried that there might be some kind of short in there, but I figured the extra compression from being mounted in the tunnel would lower the resistance a lot. After measuring the resistance of the diaphragm with the long connecting wires to my remote setup, however, it was about 1.4 ohms. This was quite surprising, but I figured that's just how long wires work sometimes, so I went ahead and tried to test again. However, the diaphragm again failed to pop, and this really confused me. With more multimeter testing, I found that the current was way lower than anticipated, and it also decreased during the test. When I was opening up the multimeter leads after the test, however, I was in for a surprise:
The connection apparently had an insane amount of resistance, and it heated up enough to melt the electrical insulation of the wire and the plastic covers on the multimeter lead. There was also a nasty bit of oxide on the copper, which would explain why the current decreased. After even more testing, I grew confident that these connections were the main issue, especially after I found the theoretical resistance of the entire circuit was around 0.65 ohms. The current, however, indicated a resistance of about 2 ohms, meaning our connections just ate about 1.35 ohms of resistance. They not only ate up the majority of my power, but they also lowered the total power delivered, and the heat they generated was a major safety problem. I replaced the wire with one that was better at electrical contact, and tried to minimize wire connections. I then tried to test the resistance again, but my multimeter was giving me absurd and fluctuating readings of over 100 kilohms. I was very confused, and I couldn't find anything wrong with the circuit path, so I tested the resistance of the diaphragm itself and it gave the same results. I could tighten the bolts to lower the resistance to around 200 ohms, but this was still extremely high, so I figured I should probably open up the diaphragm to see what was up. Almost immediately, though, I could tell something went horribly wrong. I was hit with a really strong burnt plastic smell, and it only got stronger the more I opened up the diaphragm. My dust mask barely seemed to help, so going as fast as possible I opened up the bolts, snapped some pictures, and left the testing room's potentially carcinogenic air:
The diaphragm clearly made a bunch of heat from all my testing, and at some point it just gave out. Clearly, I was wrong and there was actually a short, and it looked like the top right structural copper wire had arced onto or touched the stainless steel mesh where it wasn't supposed to. This caused enough heat to melt and discolor the plastic, and it looked like the stainless steel got a fair bit of oxidation as well. Right now, I'm working on a new diaphragm design that should hopefully avoid the shorting issues by positioning the copper holders farther from the center. I also made it thinner by my advisors' request. Hopefully I can get a third test off today or tomorrow, but I'm not sure. In any case see you next time!