April 2, 2006 notes
Our new Haas VF-3 CNC machining center (http://www.haascnc.com/details_VMC_NEW.asp?ID=241#VMCTreeModel )is installed, and we love it! I wouldnât have normally purchased something
this expensive for Armadillo, but my wife got it for me as a birthday present
(isnât that sweet!).
A 20HP enclosed machining center with flood coolant and an
automatic tool changer. Ahhhhh. No more chips strewn all over half the shop.
Machining the manifold side of one of our new injector heads
used to take six hours and many tool changes on our old Sharnoa
mill, but it now gets done in 45 minutes, with the only operator actions being
to help clear chips a couple times and place clamps on the final cutout pass. With more optimal tool selection and pushing
it harder, I could probably cut that in half if we ever needed to make dozens
and dozens of engines, for, oh, say, an OTRAG cluster.
The old mill had a fourth axis rotary table that I used to
mill channel wall chambers, but I got a tilting rotary table on the new machine
to give full five axis control, for the primary use of drilling arbitrary
injector angle patterns. We may have
gone a little weak on this with the little TR110 (http://www.haascnc.com/details_rotary_trunniontables.asp?ID=459#RotaryTreeModel
) model, because we canât push very hard on it while machining, but it works
great for the injector holes made with tiny tools.
One thing I was rather surprised at was how different the
G-code was between the old SDC-850 with a Tiger IV control and the new Haas
control. I expected the modern control
to be a superset of the what I was used to, but
practically everything outside of the half dozen most basic features is
different. The G codes other than G0 /
G1 / G2 / G3 seem to be essentially randomly assigned, and mean completely
different things. I was especially
surprised to find that there were some features on the old CNC that actually arenât
present on the modern one, like general purpose variables with arithmetic, and
some things that were more convenient, like looping without having to create
Iâm still writing all the G-code by hand, but now that I
have a common, modern control, I will probably get some basic CAM
software. The 2D graph mode on the
control is very helpful, but a full rendering with depths and tool diameters
will be better.
1000 inches per minute rapid transit is frightening. I leave it on
25% most of the time because having it chuck around hundreds of pounds of
machinery and look to all the world like it is going to pile drive your tool
right into the bed, only to drop to machining speed 0.1â above the surface is
stressful. After I know everything is
working well I will sometimes put it at full speed, but I still cringe after
tool changes. Something that surprised
me: my old mill did rapids proportionally, and always did Z raises first if the
destination Z was higher, which seemed eminently sensible to me. The new mill (and evidently this is standard
practice in the industry now) moves all the axis at the max speed, so if you
are rapid moving to another point that isnât on an 90 or 45 degree path, the
tool will actually make a 45 degree move, then a straight move. I cut an unintended hole in a part before I
learned this. If I had had the rapid at
100% it would have surely broken the tool.
Iâm starting to do finish passes on some of our machined
parts, and I finally got around to buying a chamfer mill so I can start
automating the edge breaking on the parts instead of just having Tommy do it on
the wire brush wheel.
We have generally had a very frustrating month, because we
kept melting a lot of wiring and various other things on our test stand. A couple times it was due to fuel leaks on
the engine plumbing (we were in good company with that this month â hi Elonâ¦), which we started addressing by making a sealing
plate we can bolt under the injector to allow us to pressure test everything. We tried putting a plumbing plug in the nozzle
so we could also check the graphite and phenolic
seals, but it blew right out at a very low pressure. Even after we got all the leaks fixed, we were
still burning things on the test stand just due to the hot gas mass flow around
it. It looks like 1500+ lbf engines are just too much for our current testing
arrangement, so we only got short runs in between putting out fires. We are going to do some longer runs in a horizontal
orientation, and the really long runs on the actual vehicle, while it is
suspended in the air attached to huge concrete blocks to keep it from going
We also had a lot of trouble with our flow meters, so our
data has not been good lately.
Now that everything is being done with the new mill, I am
using a 1/16â diameter diamond coated carbide end mill to do the injector hole
drilling now. This is happy to plunge
into aluminum at a 45 degree angle at 10 inches / minute / 7500 rpm with no
deflection at all. On the latest injector,
after the hole is drilled I mill a 0.002â path around it to get a very clean
wall and break off the exit chip. I can
also now tailor the holes to be other shapes.
The lox holes are slightly stretched circles to get the 20% larger area
they need to flow the correct amount with an equal number of holes as the fuel
manifold. For unlike impinging, oblong
holes are better than larger holes, because if you have a smaller diameter jet
impinging with a larger diameter jet, the outside edges of the larger jet will
just go past the impinging jet. An
oblong hole can have the same forward area, and meet up exactly.
The experiments with top mounted injectors so far have been:
(all with an L* of 50, 3â throat, 6â ID chamber)
#1: Unlike impinging
45 degree to 45 degree (90 degrees total) head with spot drill cone left on. Fuel on the inside, so any overspray would be
directed towards the chamber wall. The
streams came out messy because of the spot drill marks, and performance wasnât
#2: Unlike impinging
30 degrees to 30 degrees with the spot drill marks faced off. Excellent performance, but it burned through.
#3: Exactly as #2, but with tapered
manifold inserts to keep the propellant velocity higher and improve cooling. Severe erosion after a short run, burn through
#4: Straight down showerhead injector. Extremely poor performance, and even had some
trouble starting due to the lack of atomized propellant. When we did straight shot injectors from the
side they performed much, much better than this, probably because they still
impinged together in the center, and they were shooting crosswise to the
direction of gas flow. Performance could
probably be better if we arranged to have more than just two rings of holes,
but that would complicate the manifolding a lot.
#5: Like impinging (30 on 30) with radial fans impinging
edge on with the opposite propellant. I
couldnât get the fans closer than about a half inch, so the fuel and oxidizer
couldnât mix until at least an inch away from the injector face, and at low
pressure drops they probably didnât directly impinge at all. Performance was moderate, and the injector
didnât show any signs of heat damage.
#6: Unlike impinging 45 degrees on 0 degrees, with the lox
shooting straight down and the fuel shooting inwards from the outer ring, plus
film cooling holes added at a 10 degree angle to the chamber wall. This was done with the new hole drilling
process, with the lox holes oblong for better mixture ratio. Excellent performance, no signs of injector
melting, and the film cooling seems to reduce the chamber erosion rate
We have additional injector designs to try if necessary (a
like impinging 45 on 0 and 0 on 45 for a quadlet
effect was the next on the list), but it looks like #6 will do the job for us.
The latest injectors also have integral gimbal
arms.machined on the flange, which is really neat.
We moved to a phenolic spacer to
replace the machinable ceramic ones that kept
breaking. It has worked fine so far, but
we havenât been able to do any really long runs.
We rebuilt the test stand blast deflector to use graphite
plates, and it is essentially not eroding at all, just pitting a bit so far.
We sent in our experimental permit.application
for this vehicle to AST, but they have already come back with a bunch of things
we need to fix on it.
We found that the gimbal mounting
points we had made by hand werenât very square, so we machined a combination
engine mount / gimbal mount plate completely from a
thick plate of aluminum. Combined with
the integral gimbal arms on the engine flanges, we
are now guaranteed good alignment.
When we are ready to do ground liftoff hover tests at our
remote site, we are going to be using a new tether system. Shock loads are a huge issue with
tethers. If a vehicle can accelerate for
20â before hitting the tether, it will break almost anything hard. Energy absorbing tethers are likely to get
cooked by rocket exhaust. Our solution is
to have a fairly stout wire cable going from one vehicle leg to a pile of
extremely heavy anchor chain. If the
vehicle goes runaway, it will just start picking up 100 pound chain links one
at a time, putting a gradually increasing force on the vehicle. Pulling on a single leg will arc the runaway
vehicle into the ground quickly. It will
be a heck of a crash, but it will happen within 40â of the launch point.
We have started making parts for the 65â vehicle. We are planning on making everything with the
mill, avoiding any hand fabrication, so we can churn out more vehicles quickly
when necessary. We have a new landing
shock / ground sensor design that gives a much stronger side load support and
fully encloses the ground contact spring.