July 3, 2005 notes:
I was having difficulty finding commercial aluminum pressure
vessels in the size and pressure range we want, so we decided to make our
own. It turns out that our vehicle size
is a very good fit for spherical tanks.
A 3 diameter sphere holds about 100 gallons, so two 3 spheres is about
the right amount for a single-man-to-100km vehicle, and a 3 diameter cabin is
Spherical tanks are nice in that you only have a single weld
bead, and with metal, you get roughly a third higher mass fraction than you would
get in a barrel section due to not having hoop stress at twice the axial
I tried to get pricing information from http://www.northlandstainless.com/products/heads.php
about their hydroformed hemispheres, but they didnt return my email, so I
would up with the metal spinning shop http://www.amsind.com/.
We are initially assuming no post-weld heat treatment, so our
alloy selection was going to be from the 5XXX series of work-hardening, highly
weldable aluminum alloys. The heat of welding
anneals the nearby metal, greatly reducing the strength, but the effects vary
from alloy to alloy.
5052 is the most common aluminum used in rolling and welding
fabrication operations. If you ask for
something formed out of sheet aluminum and dont specify otherwise, you will
probably get 5052. All of our cones and
cylinders for various rocket bodies have been from 5052.
5083 and 5086 are more highly alloyed versions with higher
strength, but somewhat more expensive and less readily available. 5083 is a little higher strength and a
little less formable. The spinning shop
Tensile strengths in ksi (from AMS Metals Handbook, similar
data here: http://www.thermaflo.com/engref_tensile.shtml
5052-O 13 28
5052-H38 37 42
5083-O 21 42
5083-H116 33 46
5086-O 17 38
5086-H116 30 42
The prices were quite reasonable, only $370 per hemisphere
for a 3 diameter, 1/8 wall 5086-H116.
I ordered four for testing.
The order arrived with a sheet of actual metal test results
for that heat of aluminum, showing 45 ksi UTS.
In general, tested values are expected to be slightly higher than book
We had a bit of a surprise on the sphere weight -- strictly
from surface area calculations, a 1/8 wall 3 diameter hemisphere should weigh
25 pounds. The hemispheres we received
only weighed 18 pounds each. I was
under the impression that manual sheet spinning operations, as opposed to
automated shear-forming, didnt produce much thinning of the metal. This was incorrect for something as deep as
a hemisphere. The metal at the polar
boss was the full 0.125 thick, but the metal at the girth edge was thinned
down to 0.100. That thinning alone wouldnt
account for the low weight, so it must have thinned even more higher up the
dome. When we later sectioned a dome,
we found that it got down to 0.060 at the thinnest.
The thinning during spinning would be accompanied by
additional work hardening which could additionally strengthen it, so it might
be a wash higher up the hemisphere, but only having 0.1 at the weld area was
going to reduce the strength for sure.
Getting two 3 hemispheres with weld-beveled edges to line
up for welding is a bit of a trick. Our
first attempt was to put the hemispheres together inside the 3 diameter
tubular section we had fabricated for an inter-tank, and do the initial tack
welds through cutout holes. We then
pulled the sphere out and ran a bar through the entire thing so it could be
rotated on two sawhorses. James was out
of town this weekend, so Russ did all the welding. We used 5356 filler rod instead of the 4043 we use for all of our
We welded fittings on both ends and hung the tank on a scale
under a forklift as we filled it with water for hydrotesting. The tank was under 40 pounds, and it held
over 800 additional pounds of water, for a mass ratio of 21.
The tank burst at only 210 psi, popping open right along the
weld. Inspection showed that it
ruptured at a section where the weld had bridged a small gap between the two
imperfect hemispheres. We also saw that
the back sides of the weld didnt show good closure, with a visible line down
the bead in many areas.
We prepared the other two hemispheres more carefully. Russ tacked about 16 little tabs on the
inside of one hemisphere so that the mating hemisphere was forced to meet up a
lot more tightly, and we spent some time hitting problem spots with a hammer
until we got almost perfect conformance.
We cleaned the weld area more aggressively, and we made a new support
rod that filled the interior of the sphere with helium backing gas during the
weld operation. Russ had one problem
with a weld blowout due to having the helium flow rate too high, but overall
the welding went very smoothly.
This tank ruptured at 340 psi, and it tore in the base metal
right above the weld in the heat-affected zone, exactly where we expected it. The back side weld beads showed good
closure, but there were a couple spots that were still a little less than
perfect. This was still lower than what
we had hoped for (0.1 metal at 38 ksi should give 420 psi), but there are a
lot of possible explanations: increased cross sectional area and forces as the
tank stretched, multi-axial tensile stress may be lower than single direction
test results, the metal may not have been a uniform 0.1 all the way around, and
one other thing that we can easily correct:
The hemispheres came with a 60 grit brushed finish (spinning lines
removed), and the tear seemed to follow the brush marks. We are going to take the next set of
hemispheres as spun and bead blast the area around the weld to remove stress
Based on our results, I am going to go conservative for our
first high performance vehicle. We are
going to use ¼ thick stock for the spun hemispheres, which should give us a
weight of under 80 pounds, a tankage mass ratio of 11, and a burst pressure of
640 psi or greater if the stress relieving helps. Our operating pressure is 300 psi, so that gives us quite a lot
of margin. I ordered six hemispheres,
so we can burst another one, then put a vehicle together. Still very cheap at $650 each.
I am considering the possibility of machining a ring
(probably in several pieces) of thicker metal to go between the two hemisphere
halves. That way, the hemispheres would
only have to meet up with the wider ring, rather than the beveled edge on the
matching hemisphere. The cross section
change would be a stress riser, but guaranteeing a full weld with no thin
bridging sections would be a benefit, and our weld beads are already a stress
When we are ready to increase the performance, we can either
use a heat treatable alloy like 2019 and heat treat the entire tank set after
welding, or try and find a metal spinning shop that can perform CNC machining
on the domes after spinning them, allowing us to start with a ¼ thick plate,
but turn most of the hemisphere down to 1/8 or less except for a gentle
increase to the full thickness in the couple inches near the weld line. AMS cant do this, but I expect there is a
shop somewhere that can (if anyone knows of one, tell me
Another page with lots of aluminum welding information: http://www.weldreality.com/aluminumalloys.htm
Phil fabricated a base heat shield for the vehicle out of
nomex honeycomb. The aluminum legs will
be in ceramic cloth sleeves, but we are still a bit concerned about the engine
plume cooking things during takeoff and landing. We will be finding out soon with a hold-down test firing.
Total vehicle weight is 350 pounds dry, and it can hold 140
pounds of propellant. This is a bit
heavier than we guessed, but not too bad.
When we get all the parts in for the duplicate vehicle we can start
weighing various combinations and seeing how much we could save with various
alternatives, like using aluminum for all the manifolds and hose ends.
We tested the computer control of the roll thrusters with
the vehicle hanging from a hoist, and they worked well. I experimented with a couple different
parameter values, and you can have a fairly generous dead-band between firing
the opposing thrusters. It stops manually
induced rolls pretty quickly and with minimal gas use, but we are still
concerned about any off-axis engine thrust induced roll being difficult to
counter with small thrusters. I will
probably feel better with two engines gimballing to provide roll control.
We also did the final computer / IMU / gimbal actuator alignment
and testing, liquid nitrogen pressure testing of the LOX plumbing, and pressure
testing our high pressure plumbing. The
only thing we still have to bench test before the first hold-down firing is the
Our new engine was a disappointment. Going from 20 x 1/16 holes to 80 x 1/32
holes for each propellant and pulling the fuel injection point in from the side
to the same radius as the lox injection seems to have brought on some degree of
combustion instability (about +/- 20% fluctuation), and the Isp still isnt any
On the bright side, we have made several 30+ second burns
with it so far without a problem. We
got a heavy professional hardcoat applied to the chamber, and we have been
doing most of the tests with 1% by mass ethyl silicate added to the 95% (190
proof) denatured ethanol. The ethyl
silicate leaves a crusty white coating inside the engine, but it is supposed to
significantly reduce the heat transfer.
We did do one run without the ethyl silicate, in case that was causing
the combustion instability, but it didnt change anything (or melt).
The single-billet construction of both the lox and fuel
injectors and manifolds is holding up well, with no signs of localized
flameholding or melting.
One new bit of data we were able to learn: previously with the side injection we couldnt
tell much about the liquid distribution, but now that the fuel flows into a
manifold at the top of the chamber, we were able to water test the chamber
before welding the top fuel manifold on, giving a straight-up fountain of water
from each cooling channel. There was a
quite significant variation in fountain height, with the highest point about 50
degrees forward of the tangent entry point, and steadily decreasing the rest of
the way around. Varying the inlet
pressure moved the high point around a bit, but it retained the steady
decline. We will make a much larger
fuel manifold on the next from-scratch engine.
We expanded the LOX manifold on this engine a bit after seeing that, but
we dont have a convenient way of testing the distribution there.
An older edition of Sutton listed L* for lox-alcohol motors
as ranging from 36 to 120, and our existing motors have been at the very
bottom of that range. A couple people
have told us that we should be able to achieve high efficiency with that volume
if we use a high performance injector, but since we are going to be throttling,
we really dont want twitchy injectors.
We are making a chamber extension that adds 5 more barrel section to
the existing motor, for 36 more L*. We
are probably going to have to make a brand new injector to clear up the
combustion roughness, but well try it with the existing one first. For the next injector, I am going to just
drop the hole count significantly to get more pressure drop across the orifices. Our previous engines ran perfectly smooth
with very little injector pressure drop, but that is a classic recipe for
All of our test stand instrumentation is now running through
the flight computer, making the test stand completely wireless. This has a lot of benefits compared to our
dedicated data acquisition system, and is working well. Keeping the flight computer active and
exercised even when you arent flying a vehicle is a Good Thing..
The hard firebrick ( Louisville dry press high duty brick
) we have been using on the test stand is holding up remarkably well. One of these days we are going to put one of
our Buran tiles (tossed in for free from the guy we bought our Russian space
suit from) on our blast deflector and see how it holds up under the same
circumstances. It wouldnt be the
lightest thing, but it looks like you could make a reentry heat shield out of