To: R. Johnson
From: M. N.
Lovellette, J. Haughton, R. Haynes
Glast
Lat Tracker Tower Comments
Re:
Telecon 6/19/2002
We
agree with these conclusions from the presentation:
- Proceeding
on with the inferior screws in the first test should not have been done -
Once you do this you are testing a configuration that is not flight like
and not definable.
- Proceeding
on with a known crack in the second test should not have been done - Here
again you are testing a configuration that is not flight like and not definable.
Some
things that we don’t agree with:
- “The test
served its purpose” - Very little
was accomplished since the test configuration was wrong from the
beginning.
- “Most of
the tower module design is sound” – This design directs the entire load
from the tower inertia through the bottom tray interface. It’s probably too late now, but a
better approach would be to have an independent load path that does not
involve the tray.
Some
other issues
- The
structural model should be verified and the design should be qualified
(including minimum frequency requirements) without consideration of the
gasket. Then the gasket can be
added and that configuration qualified.
A major effort should be made to find a gasket or alternate
configuration that does not change properties during test.
- Estimated
design loads or the results of coupled loads analysis should be used to
assess the vibration test responses at the CG of the tower to see if
over-testing occurred. Using the
responses measured during a single tray test doesn’t relate to this test
since the load paths and stress through the lower tray are totally
different. On future testing the
input should be assessed in pre low level testing to determine if notching
is warranted to maintain CG responses below predicted design loads.
- Three
sigma responses are generally assumed in predicting peak random vibration
loads for metallic structures due to their ability to redistribute load by
local yielding. Composites,
however, are not as forgiving and can experience ultimate failure due to
the 4 to 5 sigma peaks that are many times seen in random vibration
testing. This should be
considered when setting notch levels.
- In the
thermal cycling of the grid interface, only the bottom tray was used. It seems that the addition of the side
walls and other tray elements could affect the strain distribution
significantly.
- It is
somewhat hard to tell from the pictures, but it seems that the fix that
was performed on the lower tray added additional closeout fasteners pretty
much along a line with the original fasteners. This won’t help much with any heel and toe type loading that
is usually a problem in highly loaded areas near the interface. If you can swing it we should probably
have one of our analysis people take a look at it.