I'll third the Belleville washer; this thing will be seeing a fair bit
of thermal expansion and contraction, so a spring washer to keep the
preload from varying too much is a good idea.
Not that I've really figured out what the idea is with the preload.
Looking at your diagram:
http://www.watzlavick.com/robert/rocket/rocket1/drawings/ball_valve_3_assy-annotated.pdf
preload could come from:
1. Compressing the inner races vertically, via the preload
nut, whereupon they bulge out horizontally (per Poisson's
ratio).
2. Press fitting the bearings onto the shaft and into the
hole, which you've been doing from the start but obviously
wasn't enough (though probably would have been with a really
hard press fit),
3. Forcing a bearing's races in opposite directions (on the
diagram, one up and the other down), which you may be doing
some of, via the flange on the top bearing being opposed to
the valve ball, and via the two bearing spacer sleeves being
slightly different lengths. But to get preload on the bottom
bearing via that last, you'd want the outer sleeve longer,
not the inner one... though that is probably how things end up
once the preload nut is tightened and the whole stack gets
compressed.
You'll also get some of #3 from thermal expansion. As a rough figure,
with the housing being aluminum and the stem stainless, the difference
in thermal expansion coefficients seems to be about 5 parts per
million per degree C; with 200 degrees C temperature change, that's 1
part per thousand.
But as regards preload coming from the flange on the top bearing being
opposed to the valve ball (the flange pulling the outer race up, and
the ball, via the preload nut, pulling the inner race down), that is
the most vulnerable to thermal expansion of these sorts of preload,
and also the least desirable, since it adds to the friction of the
valve ball. Doing a quick search on those part numbers, the bearings
you've specified are "deep groove"; there are also "shallow groove"
bearings which allow more of a vertical float. (Not that deep groove
bearings will necessarily be a problem here -- in fact, I doubt they
will, since the dimensions are relatively small -- but if they do
cause problems, there are alternatives.)
On Sun, May 22, 2016 at 02:24:39PM -0500, Robert Watzlavick wrote:
One thing I forgot to mention is that the inner sleeve is a tiny bit
longer (0.001-0.002) than the outer sleeve so the inner sleeve is in
compression before the outer one makes contact. That ensures the inner
races are always preloaded more than the outer races. The sleeves are
just standard 1/2 and 5/16 tubing so they are pretty quick to cut off
and clean up on the lathe. I had planned on tightening the nut just
enough to keep it from backing off but not too much so it doesn't
overstress the bearing. A Belleville washer would provide a more
accurate preload because I could turn the nut a specific number of
turns, etc. after finger tight. I may have to modify an existing washer
to get the OD and ID that I need. The anti-rotation clip is a good
idea. I seem to be reinventing the standard ball valve one piece at a
time...
For those that haven't followed this for the past couple of years, the
idea is to make a lightweight cryo ball valve that can be actuated with
a small servo motor instead of a pneumatic actuator. The stem in a
traditional cryo valve is too long (and heavy) for a small vehicle and
the majority of the torque needed is due to the stem seal. The idea was
to use an inexpensive off-the-shelf PTFE spring energized seal since the
torque requirements are much lower but it has turned out to be tricky to
get that seal to work with rotary motion.
-Bob
On 05/22/2016 05:05 AM, Ben Brockert wrote:
Interesting approach Bob, thanks for sharing. I agree that a
Belleville washer would be a good addition to maintain preload. You
also may want a dorky little clip that goes over the nut after it's
torqued and keeps it from backing off with rotation of the stem. The
clip being missing on repeatedly rebuilt pneumatic valves is
responsible for about 80% of stem seal leaks on test stands that I've
seen.
Instead of the two sleeves you could use a frusto-conical spacer (or a
thick ring with appropriate lips) inside that loaded the outer race of
the upper bearing into the inner race of the lower bearing to drop the
parts count by 1. It would guarantee that the load on the races was
identical between the bearings, whereas in your setup that's only true
if the sleeves are exactly identical in length. But making that part
might be more complex than making the two tube spacers.
Oh, and it'd be loading between a spinning and non-spinning part. So
that's probably a nonstarter... you'd have to add a little thrust
bearing.
Ben
On Sun, May 22, 2016 at 5:57 AM, Robert Watzlavick
<rocket@xxxxxxxxxxxxxx> wrote:
I was never really satisfied with my semi-custom cryo ball valves (see
aRocket archives around 2014-11-24). These valves use the ball and body
seals from a Swagelok SS-62T6 but with a new aluminum body, stem, and stem
seals. I had some stem leakage due to shaft wobble that I was able to
address by adding a 2nd spring energized PTFE seal and pressurizing the seal
cavity but I had the feeling I was approaching it the wrong way. Several
folks made good suggestions including Peter who provided some sketches of
alternate configurations. I had wanted to get the bearing closer to the
seal so I spent some time designing and fabricating an alternate arrangement
the past few weeks. One thing I discovered is that even with two bearings
on the shaft, there is still an excessive amount of free play until the
bearings have some axial preload. Apparently, bearings (at least standard
ones) have to have a minimum amount of preload on them just to meet their
specifications. I set up a test fixture with two bearings separated by
about 0.5 inches and even with the outer races held tight in a collet, I
could wiggle the shaft up to about 0.005 inches by pushing sideways with my
finger. After some of reading about bearing design, I modified the valve
body and stem to use 0.0005 inch slip fits and then added two sleeves
between the bearings. The preload is applied to the inner races through an
inner sleeve (recommended for maximum stiffness) and the outer sleeve is
used to help keep the bearings aligned with each other and to press them
into the housing without damage. Here are some sketches of the new design:
http://www.watzlavick.com/robert/rocket/rocket1/drawings/ball_valve_3_assy-annotated.pdf
http://www.watzlavick.com/robert/rocket/rocket1/drawings/ball_valve_body_3-nodims.pdf
http://www.watzlavick.com/robert/rocket/rocket1/drawings/ball_valve_stem_3-nodims.pdf
When I assembled it, applying a small amount of preload using the nut
reduced the shaft wobble a factor of 10 (0.005 inches down to 0.0005) with a
fair amount of force from my uncalibrated finger. From that standpoint, the
design is much better than the previous one. One improvement would be to
use a spring washer between the preload nut and the upper bearing. When I
test it with LN2, we'll see how well the single seal holds up at cryo temps.
I was trying to keep the overall dimensions the same as the previous valves
so there is only room for a single seal. The 0.500 bore in the aluminum
valve body will shrink about 0.002 inches while the seal OD will shrink
about 0.009 so the squeeze on the OD of the seal will go from 0.020 to about
0.010 (still some squeeze though). Ideally I would use a seal that has a
heel on the OD to ensure it stays attached to the wall but those are very
expensive in small quantities. The difference in thermal expansion
coefficients between the aluminum body and stainless steel bearing/shaft
will tend to increase the grip on the bearings so hopefully that won't lock
them up. Also, instead of Krytox for lubricating the shaft and seal, I used
Molykote-Z to see if that makes a difference.
After I test it in a few weeks with LN2, I'll report back on how well it
works.
-Bob