Ox Tank (2024)

TLDR: I redesigned our oxidizer tank to be 50% larger while weighing 7% less. Hydrostatically proof tested to 1,350 PSI.

The 2023 oxidizer tank (left) and 2024 oxidizer tank (right).
Side-by-side comparison of the first top end cap and the latest iteration. Clear improvements in total mass, finish, and tolerances.

The 2024 ox tank is a substantial iterative improvement on the 2023 oxidizer tank. There were several key areas that I thought could be improved:

  1. Mass: The 2023 oxidizer tank was redesigned to use a 3/16″ tank wall after hydrostatic failure. Since we were sticking with the same rocket diameter (6″), a 1/8″ tank wall is thick enough to withstand up to 1800 psi without yielding (hoop stress). Therefore, the design tradeoff for using a thicker wall was bearing stress. However, after performing a design optimization analysis, I determined that the added mass of more 1/4-20 bolts was significantly less than using a 0.1875″ tank wall with fewer bolts. So, I changed the tank design to use 26 1/4-20 bolts on each end cap with a 0.125″ tank wall (as opposed to 16 bolts with 0.1875″ wall).
  2. End Cap Design: For the 2023 oxidizer tank, I used a hemispherical end cap design. The rationale for this choice was that hemispheres minimize wall thickness since they are the most efficient at distributing pressure over a given area. However, I realized that based on the tooling and geometric constraints of the end caps, it would significantly decrease the overall part mass if I used an elliptical, rather than hemispherical, geometry. I think this will be pretty clear in the cross-sections.
  3. Fill Architecture: In PleaseGoUp, we used a single inlet/outlet port on the bottom end cap. This included a 3/4″ tube stub with Swagelok fittings interfacing to a fill line and feed valve (servo-actuated ball valve). However, this system architecture had a couple of drawbacks. It was significantly heavier than necessary because it used stainless steel tees and reducers, and it added to the overall rocket length. Since the oxidizer tank is now 54″ long and the thrust chamber also increased in length, I must try to minimize the length of the propulsion stack. Keeping this in mind, a marginal increase in machining complexity would pay dividends back in reducing overall vehicle length by 6″ or more.
  4. Lightweighting: To further reduce component mass, we can remove material from low-load surfaces. On the curved elements of the bulkheads, the load through the retaining bolts of the airframe is relatively low and some mass can be removed.

For the design parameters, I used our internal pressure vessel design spreadsheet that solves for hoop stress, bearing stress, bolt tear-out stress, and casing tensile stress, which were the constraining forces.

The 2023 oxidizer tank used 3/16″ aluminum for the wall and had 16 1/4-20 bolts. However, after performing a mass optimization study I determined that reducing the wall thickness to 1/8″, which has a hoop stress FOS of 2.04, then increasing the number of bolts from 16 to 26, which gives a bearing stress FOS of 2.06, would result in a lighter overall tank (the bolt mass was second order). Another advantage of using 26 bolts is that the distance between the edge of the tank wall and the bolts can be reduced to 0.375″.

I also decided to change the geometry from hemispherical to elliptical. I used a A=B/2 ellipse, i.e. the radius of the major diameter is the minor diameter, since this was geometrically elegant and well-documented in pressure vessel design. These two cross-sections (2023 vs. 2024 oxidizer tank) show how much material can be removed just with this change:

I also added curved lightweight elements to both the top and bottom end caps. It contributed to the overall mass reduction, was a marginal increase in machining complexity (I already needed to run radial operations with a 1/4 flat end mill to clear the strut channels), and made the parts look extra cool.

Before sizing the tank length, I first had several discussions with our FAR-OUT safety leads, Drew Nickel and Peter Tarle. They had previously been on the UTK hybrid rocketry team and shared their experiences with nitrous oxide tanks. One key consideration I took away from those discussions was ensuring the tank was sufficiently sized for the maximum hydraulic pressure experienced. When approaching its critical point, nitrous oxide’s density rapidly decreases as a function of temperature (this behavior is approximately quadratic). Thus, if our tank is full of liquid nitrous and is sitting on the pad for some time then the hydraulic pressure from the liquid nitrous could destroy the tank.

For safety, I then took our nominal oxidizer mass (30.6 lbs), added some margin in case our performance parameters increase during the testing campaign after consulting with our thrust chamber team (35 lbs), then used the maximum temperature (and minimum density) our nitrous oxide would experience (1000 psi, 5.14 lbs/gal) [obtained using REFPROP] to size the tank.

Solving for the tank volume, this gave a value of 6.8 gallons. With a 10% ullage volume, the minimum value before the steady-state nitrous oxide boiloff rate cannot replace rapidly increasing ullage volume in the first few moments of the burn, this gave a tank volume of 7.4 gallons, which translates to a 54″ tank.

This is exactly a 50% increase in length compared to PleaseGoUp, however, the return to a 1/8″ tank wall means the oxidizer tank tube weighs the same. Including the redesigned end caps, my new oxidizer tank weighed 17.3 lbs, compared to the PleaseGoUp oxidizer tank which weighed 18.5 lbs. The tank had a 50% increase in volume while decreasing 7% in total mass.

To summarize, I now have some of the critical dimensions for this pressure vessel (bolted closure design):

  • 6″ tank OD, 5.75″ tank ID, 0.125″ wall thickness
  • 26 radial 1/4-20 bolts per end cap
  • Hemi-ellipsoidal geometry
  • Redundant 254 static o-ring seals

For the remaining parameters, scope and consideration of the full fluids system (valve, venting, and integration) were necessary.

The first of these considerations is integration with the pneumatic valve. I decided to use piston seals rather than face seals for the interface because it allowed me to build in 60 thou of axial tolerance. This gives us some leeway in machining the struts such that no load is transferred through the valve (although I have run FEAs of this edge case and it should be able to support the entire thrust of the rocket).

In addition to pneumatic valve integration, as discussed earlier I decided to change the configuration such that nitrous load flowed through a separate radially offset NPT fitting. I also added an additional port for a pressure transducer on the bottom end cap. This serves several purposes:

  1. Provides redundant tank pressure data ( + the pgh of the nitrous oxide which can fluctuate based on mass and temperature but is on the order of 7.5 psi).
  2. Provides a direct pressure drop across the pneumatic valve (we also have a transducer at the injector manifold).
  3. Also allows us to observe any changes in the compressibility effects.

The bottom-end cap design is shown below:

The top-end cap was also modified to be easier to integrate. One of the lessons I learned trying to assemble the fitting last year was that adequately tightening fittings was remarkably difficult with the bulkhead rails. So, I radially offset the pressure transducer and venting 1/4 NPT from the center to make this easier.

These parts were designed to be machined on our Haas ST-20Y 4-axis CNC lathe. A sped-up video of the CAM of these components is shown below:

On the first version of the top end cap, we discovered in October that when the lathe was serviced over the summer the technicians set the wedge angle to 0 degrees. So it was machining parts like this:

Fortunately, I caught this before any damage was done and we diagnosed the problem as the lathe’s Y-axis malfunctioning. This was due to the machine thinking the wedge angle of the turret was 0 degrees, rather than 45 degrees, and was the fault of the technician that serviced the lathe over the summer.

After Michael Sheehan (prop-cc lead) purchased the lathe key to correct the wedge angle (which unfortunately consumed most of the fall semester due to delays with Columbia) we could resume machining the top end cap with a new piece of stock.

We went on to successfully machine the top end cap:

Power tapping! It was challenging to get quality footage of other operations due to the coolant spray blocking the windows (we used coolant for the rest of the tapping as well).

The final product:

Overall I am extremely pleased with the finish on this part and I think the machining is significantly better compared to the first set of end caps, however, there was one glaring issue with this part: The inner feature had a rotation of roughly 1.5 degrees.

We are not sure why this happened and I’ve been working with our prop-cc lead Michael Sheehan to find the root cause of the issue. First, we used gauge blocks to tighten the tolerances in the X and Y axes. Then we successfully replicated the issue using a small test piece:

We also performed other diagnostics to determine tram in the X, Y axes.

In this diagnostic, we observed some waviness in the feature (around 4 thou) which is likely due to a bent rail from some past crash (this machine has been through a lot over the years).

We managed to dial in the lathe to extremely tight tolerances, however the axial milling offset persisted and is still a mystery. If you’re reading this and you think you know why, please shoot me an email at aka2205@columbia.edu!

At this point, we had run into finals week and couldn’t continue work, but I decided as a workaround it is probably best to just perform turning and radial milling operations on the lathe, then use a custom nylon jig to work hold the bottom end cap, then use the 3-axis Haas Mini-Mill to perform the axial clears.

Returning from the break, I led the team in machining the bottom end cap on the 4-axis mill turn lathe. Due to the aforementioned issues with the axial live tool (we now believe it is because the actual turret of the lathe is slightly tilted from crashes by other users), I designed a custom jig to allow us to do the milling on a 3-axis machine.

This jig was 3D printed out of Nylon and was exceedingly strong. The CAM for the bottom end cap was very similar to that of the top end cap.

Verified the offset issue was still present.
Workholding for the bottom end cap. It ensured the axes were aligned and provided additional stability with the radial bolts.
Nice roughing passes
The final finish of the part. Far and away the most beautiful part I’ve ever programmed and machined.

Next up, milling the elliptical geometry for mass savings:

Finishing on the top end cap. CAM for these two parts is almost identical.
Finished piece!

Both parts were remarkably light! The last part to be machined was the flange adapter. This part serves two purposes: seal the bottom end cap for hydrostatic testing and work as an adapter for the ball valve is the situation that we static fire before the pneumatic valve is ready.

Roughing the adapter

The flanger adapter integrated with the bottom end cap just produced a stunning part. The last task was integrating everything together. To index the 26 holes on the oxidizer tank tube, I used the CNC lathe. The length of the stock was really stretching its abilities, but it was necessary to align the holes.

One change I made was switching from 70D to 90D 254 o-rings for sealing between the end cap and the ox tank tube. The main benefit of this change is improved resilience during installation and higher resistance to extrusion during testing. The final assembled tank and the 2023 oxidizer tank recovered from ballistic impact last June are shown side-by-side.

Before cold flow testing, it was critical to verify that the ox tank was safe. Fortunately, the civil engineering lab, Carleton, had hydrostatic testing equipment that we used. Hydrostatic testing is safe because water is approximately incompressible. This means that it takes relatively little energy to raise water to meaningful pressures, and thus if the pressure vessel failed relatively little energy would be released.

FAR-OUT requires we proof our pressure vessel to 1.5x MEOP, so 1,350 PSI, for 30 minutes. After a couple of iterations of pressure transducer seals, I settled on using copper crush washers since the o-rings were temperamental at high pressures and would begin extruding.

Turning some copper crush washers

The third attempt at hydrostatic testing went smoothly. We pressurized to 1,350 PSI with no signs of plastic deformation of the pressure vessel and no noticeable leaks.

Hydrostatic testing rig! Huge thanks to Will, Liming, and Freddie at Carleton.
Highest proof pressure in CSI Rockets history!

After holding pressure for 30 minutes, it dropped to 1,250 psi. This indicates that there were small leaks, which we confirmed visually at the air release ball valve and pressure transducer crush washers as small drips.

I plan on revisiting the hydrostatic test stand in two weeks for proof testing of the pneumatic valve, and will also have some new solutions for even more robust leakproofing that can hopefully ensure it holds at 1350 for 30 minutes.

This was a huge success and I’m extremely happy with the results. The 2024 ox tank was the most challenging project that I’ve been responsible for thus far, and I’m elated that not only was I able to make design optimizations to reduce mass, but that every hand calc and FEA I ran was validated in hydrostatic testing, while the final product will measurably improve fluids system performance for launch at FAR-OUT.