[AR] Re: Nels Anderson data request

  • From: William Claybaugh <wclaybaugh2@xxxxxxxxx>
  • To: "arocket@xxxxxxxxxxxxx" <arocket@xxxxxxxxxxxxx>
  • Date: Tue, 1 Dec 2015 10:49:27 -0500

Nels:

I cannot say anything more about unregistered launches.

Using only the Atlas, Delta, Shuttle, and Titan data and ignoring the tens
of thousands of launches by the Russians, Chinese, Europeans, Indians and
others...as well as the other American launchers...(which your discussion
suggests you are doing) produces a highly skewed result. That is fine if
you plan on duplicating the engineering discipline and process control used
to achieve the high reliability of those vehicles.

If, however, you are an amateur with ideas about putting something in earth
orbit, these vehicles and their reliability statistics are no model for
your likely experience.

You really need to look at the whole data set; even just adding the other
American launchers will significantly change your conclusions.

Bill

On Tuesday, December 1, 2015, Nels Anderson <nels.anderson@xxxxxxx> wrote:

Thanks very much for the reply. But are there really enough illegal space
launches to significantly affect the statistics? If the illegal launches
have the same statistical properties as those publicly declared, then they
won't affect the statistics much unless they outnumber publicly-declared
launches. If they have dissimilar characteristics, then they might not be
very relevant. For example, if the clandestine launches represent secret,
cost-is-no-object military operations, then they might not be very relevant
for commercially-oriented purposes.

In posing my question, I was hoping that there were additional public
studies that I was not aware of. Otherwise, I'm influenced by Futron's
study of US space-launch vehicles over the 20 years from October 1984
through September 2004 (it seems no longer to be on the web, but I can
provide a copy). The sample included 1255 liquid stages suffering 6
propulsion failures and 1831 solid stages suffering 4 failures. On the
face of it, the liquid failure rate (0.48%) looks distinctly higher than
the solid rate (0.22%). But the solids are comprised of two very different
beasts, namely small Delta II-like solids without TVC, and large Titan
SRM-like solids with TVC. The small solids undeniably have a great track
record, with just one failure (the Delta 7925-10 in 1997) in 1260, i.e.,
0.08%. Large solids, though, fare no better than liquids, with 3 failures
in 571, i.e., 0.53%.

Or consider propulsion-related failures of the heavy Titans.
Conveniently, each had two SRMs and two liquid core stages. In 123
launches, SRMs failed twice (34D-9 in 1986 and K-11 in 1993) and core
stages failed twice (2nd stage of 23C-17 in 1978 and 1st stage of 34D-7 in
1985).

Consider the Shuttle. One out of 270 SRBs failed against five or so SSME
anomalies. But which would you prefer to have avoided: the one SRB failure
or all of the SSME anomalies combined?

About half of the above liquid anomalies were manifest at ignition. In
the Shuttle's case, thanks to hold-down, the consequence was merely a
delay. That's not an option for a solid.

I've considered only a fraction of the publicly-available data, but that
fraction is nonetheless rather voluminous. It indicates that small solids
without TVC are indeed very reliable. But larger solids don't appear to be
any more reliable than liquids, especially for ground-lit stages (which,
ironically, is just where the rocket equation would tell you that solids
are most appropriate). And solids' failures tend to be both less
predictable and quite a bit nastier.

On 11/07/2015 09:54 PM, Anthony Cesaroni wrote:

Forwarded message below at Bill’s request.



---------- Forwarded message ----------
From: *William Claybaugh* <wclaybaugh2@xxxxxxxxx>
Date: Friday, November 6, 2015

My regular email server is under DDOS attack and so I have neither been
receiving or replying to arocket posts.



A review of the archive showed that Nels had asked for the data backing up
the assertion that the 1980 - 2010 space launch data set shows a liquid
failure rate double the solid motor failure rate; I'd like to try and reply
to that request:



The organizations for which I--twice--developed this data set paid a very
great deal of money to get it done; naturally, I did not take a copy when I
left each of those organizations.



Building a accurate data set required access to the private data of three
space insurance companies; after reconciling those, it was necessary to
further reconcile that "master" data set against a classified data set.
This is because, 1) despite treaty obligations, not all launches to earth
orbit are registered with the UN, and, 2) failed attempts are sometimes
not announced--particularly during the first decade of this timeframe.
Further, the binning of failure modes into GN&C, liquid, or solid caused
often requires access to classified analysis for certain foreign systems.



Obviously, it is not possible to reconstruct this data set for public
release; further, any data set based only on publicly available information
is certain to underrepresent the overall failure rate and, particularly,
the cause of failures.



That said, there are certain conclusions that have been publicly drawn
from the comprehensive data: that overall, the failure rate is just under
10%, that after GN&C failures liquid rocket bits account for about 2/3 of
the remainder and solid motors for 1/3; it is also public knowledge that
half of all failures occur in the first 14 launches of new vehicles--the
long run failure rate is less than 5%. This last observation is the
technical basis for NASA requiring 14 consecutive successful launches
before it will contract to fly it's most valuable payloads on a new vehicle.



We might also note the Aerospace Corporation observation that launch
failures are initially design or manufacturing failures but after about a
dozen launches failures tend to be process failures.


Other related posts: