02/20/2014

I believe Thomas Edison said, "I have not failed. I've just found 10000 ways that won't work." Well, I've found at least one way to build a flight weight tank and a valve that don't work at cryo temperatures. I ran cold tests a couple of weeks ago with LN2 and the tank leaked like Niagara Falls around the seal when pressurized. I knew it would leak some but I didn't expect that much. So, it's back to the drawing board for the LOX tank. The o-ring design should work fine for the fuel tank.

I looked into what it would take to make welded tanks and decided it's not the way to go right now. There are just too many issues involved with making safe pressure vessels out of aluminum, at least with my skill set. I could send them out and have them professionally made, inspected, and heat treated to restore the strength but that would 1) almost certainly cost a lot than I would want to spend, and 2) take some of the fun away from the project. A major part of my satisfaction in this project is that I get to do almost all of it myself. I would like to learn how to weld (and I still may get a TIG welder for other non-pressurized parts of the airframe) but my gut tells me not to do it for the tanks.

Here's my list of tank ideas along with the pros and cons of each:

  • Welded tanks
    + least weight solution
    + no leaks (in theory)
    + no issues with reassembly (because you can't take it apart)
    - can be difficult/expensive to get a quality weld with high strength
    - if not done properly, can rupture without warning due to fatigue, etc.
    - warping after welding
    - can't easily clean or inspect inside of tank
  • Tapered threads with PTFE tape (like a threaded pipe)
    + would probably work
    - can't cut tapered threads on my lathe
    - a 1/8 tube may not be enough wall thickness for long tapered threads
    - poor position repeatability for reassembly
    - potential contamination issues with lots of tape used
  • Tube/plug design with Astra Seal (or other engineered seal)
    + should work with existing design
    + good position repeatability for reassembly
    - expensive
  • Boss-type o-ring arrangement with the tube and a plug
    + may work with a viton or PTFE o-ring (but probably not) - an Astra Seal would though
    - minor variability of position repeatability
  • Aluminum tube with stainless bulkhead
    + differences in thermal expansion would tend to improve sealing (might could use a viton or PTFE o-ring)
    - heaviest solution - the yield strength of SS304 is about the same as 6061-T6 but almost 3x the weight, stiffness is better though
    - more expensive

The aluminum tube with a stainless plug bulkhead is an intriguing concept. Between ambient and LOX temperatures, aluminum shrinks about 0.38% in length and stainless steel about 0.26%. For a nominal 5 inch tube, this means when it cools down, instead of a gap, there would be an interference fit of around 0.006 inches which should dramatically improve the sealing capability of a Viton or PTFE o-ring. I'm probably going to try that option first, then if it doesn't work I'll try the Astra Seals (but they're $235 each in small quantities for a 2-246 sized seal). Even though this is a hobby for me, my time isn't completely free but just purchasing 3 Astra Seals (2 plus a spare) would be $700 and that is right in the range where I start to trade off cost vs. time.

I had better luck with the custom propellant valve but it needs a little more work. I took the ball and seats out of a 3/8 inch Swagelok 62-series ball valve and built an all-new aluminum body, end caps, and new stem. I found that most of the torque required to turn a ball valve came from the stem packing so I wanted to use some PTFE spring energized shaft seals to seal at cryo temperatures with lower torque. I've had good luck with these seals in static cryo applications but this is the first time I've tried them in dynamic use. The valve is actuated by a Dynamixel AX-12A servo in a direct drive arrangement. The valve and servo together weigh about 0.35 lb.

I don't have a real mounting bracket to connect the servo and valve so they're just clamped to a plate for now. The ball valve is connected to the servo using a flanged sleeve with a spring pin through the shaft. Since the servo has a 300 deg range and I only need 90 degrees, I plan to use a 3:1 external gear reduction to give it more torque and so I can drive both the fuel and oxidizer valves with a single AX-12A servo. I had to adjust the compliance slope on the servo from 32 down to 5 or it wouldn't make it all the way to the destination without getting stuck. This gives it a bigger kick at the start of the move and a more abrupt stop. During the "steady state" portion of the move, the load measured by the servo was about 70% worst case (at cold conditions).

The only problem was that the valve leaked at the stem seal when it got cold because it seems I didn't do my homework on the seal design. It worked fine at ambient temps with water but not good enough for cryo use. Between ambient and LN2 temperatures, aluminum shrinks about 0.38% in length and PTFE about 2%. At ambient temperature, the seal OD has a design squeeze of about 0.014 inches. However, I used an end mill to make the bore (a bit too large) and at cold temps the PTFE seal shrank so there was only a 0.003 inch squeeze when cold. Even when the seal was energized by pressure it still leaked. For the cryo version of the valve, I should probably cut the bore diameter around about 0.010 less to maintain the same squeeze. The normal solution for this is to use a flanged seal which keeps the seal OD near the gland wall. But those are custom parts, not the cheaper off-the-shelf McMaster PTFE spring energized seals.

To try and mitigate the leakage problem when cold, I added a second spring energized seal and put a hole in between them to create a pressurized cavity. I then routed a 1/8 inch tube from the helium supply upstream of the tank to the cavity between the two seals. With no flow, the pressure across the stem seals should be zero and when flowing liquid, the upstream pressure will be slightly higher because of pressure drops in the lines between the tank and valve. This worked a lot better (no LN2 leaks) but there was a small amount of gas leaking past the outer stem seal after 4 open/close cycles. I suspect if I clean up the tolerances a bit, this will be a workable solution.

Since cryo ball valves need a vent hole to prevent trapped fluid, I drilled a small hole but found that it worked better when the ball was in the downstream position when closed (normally the ball points upstream). Even when chilled down and after sitting for a while, there was no leakage around the upstream seat since the Swagelok design is spring loaded with a cone washer. Putting the vent upstream just lets the stem get cold quicker. Since I only need this valve to work for one open and one close cycle without leaking and it's OK for the valve body to discharge downstream after closing, I'll leave the ball vent hole facing downstream.


11/17/2013

Work has begun on my first rocket, named "Rocket 1" for now until I can come up with a cool name for it. While there are a lot of design decisions to make, I've settled on a few so far: 1) it will use the 250 lbf LOX/kerosene engine recently tested, 2) the propellant tanks will be integral to the airframe, 3) the plumbing and instrumentation lines will be routed through the interior of the tanks, and 4) the tanks (and vehicle diameter) will be a 5 inch OD. The first rocket will be a bit short and fat since it will only burn about 10 seconds but the 5 inch OD will allow a more reasonably-shaped vehicle later for a 30-40 second burn. This way I only have to make one set of tooling for a range of vehicles.

I just completed hydro testing at 750 psi of a prototype propellant tank. This particular tank is 6 inches long but only has a interior length of 3 inches since the bulkheads extend down 1.5 inches into each end. The tanks are made from 6061-T6, 5 inch OD, 1/8 inch wall and were turned true the lathe to a wall thickness of about 0.1 inch. I'll need to make a large diameter steady rest for the lathe to true up the real tanks but for now, a 6 inch part was all I felt comfortable cutting without a steady rest. The tank bulkheads are 0.3 inch thick flat plates to allow enough material for a port for the SAE o-ring fittings. A 1 inch OD 6061-T6 tube runs through the middle of the tank with spring energized PTFE shaft seals in each bulkhead. An aluminum washer with some #4-40 fasteners holds the shaft seal in place at each end and retaining rings keep the tube in place. An o-ring is used to seal between the side of the bulkhead and the tank wall. It should work fine for the fuel tank but I expect some small leakage for the LOX tank which should be OK if planned for. Later I'll run some tests with LN2 to quantify how much leakage will actually occur. The tube running through the middle will also provide a convenient way to mount anti-slosh baffles by just using perforated circular plates with retaining rings on each side of the plate.

The fuel tank pressure of 500 psi drives the design point. I went with a conservative safety factor of 2.0 since this is my first pressure vessel and it will also be taking flight loads. Of all the stress analysis points (fastener shear stress, bearing stress in tank skin, shear tearout, net area tension, edge margin), the bearing stress was the limiting factor and required 16 #10-32 NAS623 fasteners to maintain the 2.0 safety factor. The load from the tank pressure is entirely in shear for the fasteners but any bending loads during flight and ground handling will put a small tension load on the fasteners. This initial prototype is a bit heavier than I had wanted but it should work fine for the first vehicle and provide an opportunity for optimizations later if needed.

The next step is to start working on the main flow valves. The Peterson aluminum ball valves look attractive but I believe they are only rated to 200 psi. I may end up modifying an existing valve or making my own LOX valve since regular cryogenic ball valves are both too large and too heavy. The valves would be actuated with a small air cylinder supplied from the onboard helium supply.


9/22/2013

I ran my 250 lbf LOX/Kerosene engine last week to test out the new leak-free injector. Everything worked as expected so I'm *finally* done with engine development for a while. Three runs were performed:


Run Pc (psia) F (lbf) mDot (lbm/s) r Isp (s)
1 237 221 1.14 1.95 194
2 273 258 1.23 1.61 210
3 275 261 1.24 1.85 210

When I tested this engine back in March, the performance was lower than expected (Isp = 193) but I had a LOX leak in the injector so I wasn't sure about the mass flow numbers from those tests. But after the first run these recent tests, I came to the conclusion that this engine is just not as good as I had hoped from an Isp standpoint. The design point was an Isp of 234 but due to poor mixing or too low of an L*, I was only getting about 194. So the goal became to try and feed it with enough propellants to hit the other design point of 250 lbf of thrust and a mixture ratio of 1.8 (chosen for slightly cooler gas temperature). On the 2nd run, the both pressure regulators decided to mysteriously change from their setpoint so the mixture ratio was only 1.6. The 3rd and final run came pretty close (r = 1.85) and the Isp was actually a bit higher (210 s). There was still some deformation of the injector face this time but it was much smaller (about 0.003 to 0.004 inches at the center). It did not cause any leaks due to the redesigned seal arrangement (see notes from 7/24/2013).

The load load cell setup on the test stand has too much static friction along with preload issues from the flex hoses (around 10 lbf). The thrust computed from Pc, Cf, and throat area matches up very well with the load cell so the thrust numbers above are based on Pc since there's less uncertainty in that measurement. Another thing I noticed was that the injector Cd from hot fire testing was consistently about 10% higher than the cold tests with water. I was convinced this was due to running at ambient back pressure so after the hot runs, I ran a quick test with kerosene through the flowmeter and measured about 3% lower than with I got with water. I didn't expect the viscosity difference between water and kerosene to account for 3% (the flowmeter data sheet indicates it should be within 1% for that amount of viscosity variation). I didn't actually capture and weigh the kerosene squirting through the injector but if you assume for a moment that the 3% is due to viscosity effects in the flowmeter instead of the orifice, this means that the fuel flowmeter calibration reads higher than expected. That in turn implies a lower mass flow (at least for the fuel) which would point toward a better Isp than originally calculated. On the other hand, if you believe the water flowmeter calibration is also good for kerosene, then there are orifice Cd effects due to viscosity which I'm sure is also plausible. Based on what I've read, turbine flowmeters tend to be pretty accurate and repeatable when used at reasonable flow rates so I tend to believe the flowmeter data.

One of the design parameters I was interested in was the thermal margin. I had predicted a worst case fuel temperature of up to 445 degF but in all the runs this weekend, I only measured up to 270 degF. Also, the embedded thermocouples in the wall only got up to 434 degF vs. a worst case estimate of about 550-590 degF or so. I used extrapolated data from my 100 lbf regen engine for this engine but evidently I overpredicted the heat load. This means there is a higher pressure drop than I really need in the cooling jacket. I've been just using red-dye K1 kerosene and I haven't noticed any deposits in the cooling passages. On all three recent runs, the walls were heavily sooted after the run except for one spot where there was some bare metal showing, presumably due to a misaligned jet.

For this test, I made some upgrades to the test stand. I decided to go with a 40 micron sintered bronze air line filter for the LOX fill line (Grainger 1EJV7), installed in reverse. When I ran water through it in the forward direction, the velocity of the resultant jet looked a lot faster than it did when I ran it in reverse so I decided to go with the reverse direction to keep the velocity down. After doing some reading, it appears the reason for the direction indicated on filters with tapered elements is to maximize the surface area for debris to collect so it doesn't clog. I cut one open with a band saw and it looked like the tapered element was secured no matter which way it is oriented. For the fuel flow line, I went with a Swagelok SS-6F filter and a 40 micron stainless mesh element. Both the igniter lines have miniature Beswick 43 micron stainless mesh filters. I also added an accumulator (an E-sized aluminum cylinder that was my old propellant tank) to the test stand for the compressed air that controls the propellant valves. I got tired of dragging the air compressor back and forth 150 feet just to charge it up and I didn't want to purchase a bunch of expensive 10 gauge extension cords. But even when it was located 50 feet from the test stand, the data showed about a 15 psi pressure drop when the valves actuated so it would have only been worse when I moved the compressor farther away. A check valve on the inlet of the accumulator ensures at least some reserve air pressure to try and close the main valves in case a fire damages the compressed air hose. Spring loaded actuators would be better but with the size I'm using, after the spring force is overcome there wouldn't be enough torque left to move the LOX ball valve when cold. Even with running a higher air pressure on this test (120 psi), I had one instance of the LOX vent valve sticking.

The plan now is to switch gears and start developing a passively stabilized rocket around this 250 lbf engine, probably only running for about 10 seconds to ensure a high T/W at liftoff. Once I get the kinks worked out with launch, recovery, etc. of that, I'll move on to a longer burn (30+ seconds) with active control. I figure it will take several years to get there at my present rate of progress.


7/24/2013

I've spent the past few months working on a new injector. After looking more closely at the injector after the last run, I noticed a 0.010 inch permanent deformation in the center so the face seal is definitely a bad idea because it will leak. I redesigned the injector to use a second spring loaded PTFE shaft seal (photo, drawing, entire assembly) and ran cold water tests this past weekend. The new V5 injector also has increased thickness for the injector face and fuel cap. However, an FEA run even at elevated temperatures didn't produce the permanent deformation observed during the last test so I'm not sure what was going on there. But the new seal arrangement should be fairly tolerant of any relative movement between the injector and LOX cap making a leak unlikely. I also made some small tweaks to the orifice sizes and angles to hopefully improve mixing performance. I wonder if the 250 lb motor is in a size class where unlike impinging elements would need to be too small for practical use to get optimum mixing.

I'm definitely going to add filters to the propellant lines. The fuel filter will be located between the main fuel valve and the chamber. A high pressure LOX filter is quite a bit more expensive so I plan to use a sintered bronze air line filter on the LOX fill line only. Sintered bronze is evidently much safer than stainless steel mesh for oxygen use.

I also re-ran cold tests of the V4 injector and the Cd of the LOX orifices changed from 0.68 (original) to 0.66 (new) which is the opposite direction from expected. It also doesn't agree with the calculated Cd from the hot-fire test based on pressure vs. flowmeter data so that needs to be investigated. The Cd for the fuel orifices from these recent cold tests match the original cold tests pretty closely.

During cold tests of the new V5 injector, I observed some instability of one of the orifices on the opposite side of the manifold inlet. It looked good for the first inch or so (the impingement distance is ~0.25 inches) but when run by itself, there is visible instability several inches out. I originally thought it was a problem with just that one orifice but as I rotated the LOX cap around, the instability moved to the next orifice that was opposite the inlet. I suspect this is occurring because the orifice opposite the manifold input can be supplied from either direction and it is randomly switching back and forth. A better design would be to not have an annular manifold and instead use a partition at the far side so each orifice can only be supplied from a single direction. Even though the fuel manifold is also annular, it probably doesn't have the same issue because of the large pressure drop across the face cooling passages and also because the manifold has a larger cross-section.

An updated drawing of the test stand is also up that shows the recent changes for the ASI (Augmented Spark Igniter) and the valve position sensors.

I've also started work on the flight computer. So far, I have a Netburner MOD54415 running the uCOS RTOS with an Analog Devices ADIS16350 MEMS IMU. This is an older IMU (without temperature compensation) but it sufficiently representative of the newer models to perform initial experiments with. Coding up a Kalman filter has been interesting - one article that really helped me grasp the concept is Kalman Filter for Dummies. In my simple one-axis experiments with the IMU, the filter seemed fairly insensitive to the values chosen for process noise and measurement noise. However, this particular IMU has so much drift that it makes any long term measurements difficult. I can rotate the cube back and forth a few times and the integrated angular position will track the motion but then if I let it sit for a bit, it sometimes just takes off at a rate up to a degree per minute. I expect that I'll need to set up some sort of IMU test rig using servos with encoders to properly tune the filters. In any event, the first flight vehicle will most likely not use any active control but just gather data that can be used to develop the models in preparation for a subsequent flight with active control.


3/25/2013

After a long delay, I had two good runs this weekend on the 250 lbf regen motor.

regen3 test

This is the one that had the hard start a couple of years ago so I switched from using direct spark ignition to an augmented spark igniter which seemed to work great. The first run was a 5 second run to check the mixture ratio and the run the next day was an full-duration 18-second run.

There doesn't appear to be any damage or erosion in the chamber and you can still see the tooling marks from the lathe. I was worried that the Viton o-ring seals wouldn't stand up to the chamber temperature but they seemed to work OK. Note this motor is designed to run fuel rich (r=1.8) and have a lower chamber temperature to help with cooling. Also the soot deposits from kerosene make a good insulator and help with cooling.

The performance was a bit lower than I had hoped - based on the measured mass flow, I should have seen about 300 lbf but the load cell only measured 243 lbf. I suspect poor mixing in the injector. The good news is the chamber pressure was right where it should be (~270 psia) and it seemed to have plenty of thermal margin.

On both runs, there was a bit too much LOX lead so there was a bit of hesitation for about a half second until the fuel flow picked up. The extra LOX really enhanced the effectiveness of the igniter though. On the first run, the igniter ran on GOX from a cylinder but the 2nd run used a heat exchanger to generate GOX from the LOX tank. The LOX and kerosene tanks were pressurized with helium. Although I didn't touch the pressure regulators in between the runs, there was more mass flow recorded on the 2nd run. I cleaned up the injector exit orifices with a bit of sandpaper in between runs and the change in mass flow corresponds to changing the discharge coefficient from about 0.72 to 0.77. I measured 0.65 during the cold water tests but I had cleaned up the inlet holes a bit after that test and the hot fire. I plan to run another set of cold test (without touching the orifices!) to validate this theory.

I was hoping for 20 seconds on the 2nd run (I filled enough for 30) but it ran out of LOX at 18.5 seconds due to a LOX leak, probably at the LOX manifold at the top of the chamber. The video, still photos, and flowmeter data show it in both runs although in both cases, it seemed to have sealed itself up by the end of the run. After the 2nd run, when I was inspecting the motor, the nut that secures the LOX manifold to the igniter body seemed to be a bit looser than I remember so that's the likely place for the leak. Other than tightening the nut more, a better fix would be to use a spring loaded face seal instead of a Teflon o-ring ($85/ea!).

The really interesting data was from the embedded thermocouples in the chamber walls for the 2nd run. I wouldn't have expected the temperature data to rise and then fall in the middle of the run. I believe what happened is that there was a clogged fuel orifice for a couple of seconds that corrected itself. A clogged fuel orifice would have caused the oxidizer to spray directly on the hot wall. On the 5 second run, the two chamber TCs are pretty close but on the 18 second run, one of them shoots up really high, then drops down near the other one. The only other time I've seen a jump in conditions during a run like this was when I had a Grafoil gasket flaking away during the run and clogging up fuel holes on my 100 lbf motor.

Aside from the LOX leak, I had a problem with a sticky LOX vent ball valve prior to the first run. I had to use a heat gun on it and loosen the body bolts to get it to unstick before the test. I cleaned out all the Krytox and loosened the bolts which seemed to work well for the 2nd run. I don't think this was the source of the leak however.

The next goal for me is to fly a liquid rocket. There are a few issues with this motor but I think it will work for what I want it to do.


For archived articles, check out the News section.