[AR] Re: Parabolic Nozzle Approximation Function
- From: Henry Vanderbilt <hvanderbilt@xxxxxxxxxxxxxx>
- To: arocket@xxxxxxxxxxxxx
- Date: Tue, 6 Dec 2016 12:16:21 -0700
On 12/6/2016 1:32 AM, Uwe Klein wrote:
What surprised me in a recent discussion elsewhere is that
number crunching used in connection with trial and error optimisation
is seen as the primary/sole driver for tech progress.
No understanding for the potence of abstraction and
theoretical understanding.
Uwe
Depends on who you're discussing things with. There are parts of NASA
that take the extreme opposite view - they won't give the time of day to
results derived from iterated-best-estimate trial-and-error
optimization. If you can't show them an abstract theoretical basis
sufficient to derive the design in optimized form on the first build,
they aren't interested.
I can sympathize with the viewpoint you mention as a reaction to this
sort of thing. It is an overreaction, yes.
My view is that theory is useful to get you into the ballpark, but
sometimes deriving precise enough results for an optimized-first-time
design is either not possible at all, or requires disproportionate
computational effort.
Put another way, sometimes the most cost-effective computer for a
complex physical process is analog. Building your best estimate of
what's needed then testing it is in fact an analog computational
process. Iterate a few times and you can end up with an optimized
design at a fraction of the time and expense the pure-theory approach
has been known to take.
You can also easily spend six months of test engineering saving yourself
a couple days of boning up on existing theory.
The "right" approach varies with the problem, but in my view is most
often some balance between the two.
Henry
Other related posts: