[AR] Re: fatigue life (was Re: Re: SpaceX F9 Launch/Update...)
- From: "Findley, Jeff" <jeff.findley@xxxxxxxxxxx>
- To: "'arocket@xxxxxxxxxxxxx'" <arocket@xxxxxxxxxxxxx>
- Date: Tue, 29 Dec 2015 13:32:22 +0000
On December 28, 2015 5:35 PM, David Weinshenker wrote:
Yeah, I can think of ways to -expose- "dynamic" fatigue sensitivity - e.g,,
instrument the
amplitude as well as the force of the exciter interface to the "system-under-
vibration-test",
and sweep the excitation frequency looking for features in the impedance that
suggest
resonances at certain frequencies - then lean on those frequencies watching
especially for
shifts in the impedance suggesting that some degradation (cracking,
work-hardening, etc.)
is occurring.
With a nasty enough mentality and a powerful enough vibration source you
should be able
to make something crack -somewhere-; the question then becomes the
correlation between
this observation and the expected operating stresses. (I suppose one could
adopt the
philosophy of fixing the fatigue resonances wherever they appear, whether or
not they
would be excited by "expected" flight loads.)
I'd have to pick the brains of our former "test" people here at work for more
details, but this is what I remember from the days when we used to consult with
sometimes very big companies on just such work. Our original company was
founded to perform exactly this sort of work.
We had a "shaker table" in the basement of our former building that was big
enough and powerful enough to do things like simulate the impact of a carrier
landing on a jet engine (not that I can confirm we ever did that exact thing).
At any rate, we typically didn't perform dynamic testing until something
failed. We'd do something more along the lines of what David suggests. Sweep
the frequency range looking for modes of vibration where parts of the structure
were overly excited (think Tacoma Narrows Bridge collapse). You then use that
data to correlate results with a finite element model of the system, tweaking
the FEA model until they match relatively well. Then you can modify the FEA
model in order to "fix" the problem. When the "fix" is agreed upon by all the
stakeholders, it can be implemented in the real world on the test model. Then
you wash, rinse, and repeat as many times as is necessary.
Absent a "shaker table", you can also do much the same with an instrumented
hammer.
Jeff
Other related posts: