Henry Spencer wrote:

In article ,
Uddo Graaf wrote:

I read that the Russian N-1 moon rocket kept blowing up because the Russians
couldn't control the thrust of 30 rocket engines firing in tandem.


The N-1 had various problems -- it was different each time -- but it
wasn't anything as simple as being unable to control the engines. If
memory serves, the first two failures were actual mechanical failures in
the propulsion system, the third was a deliberate maneuver that wasn't
well thought out, and the fourth was an engine fire (at the very end of an
otherwise-successful first-stage flight).

The hardware failed every time, and there's no doubt that was the
source of the problem. But I did some analysis on KORD, and it did
indeed seem to be rather crudely designed. There's no doubt that similar
throttle-steering systems can and have been made to work with analog or
TTL electronics, it's would have been a lot easier with at least a
simple microprocessor.

One thing is clear - there's no good physical or attitude control
reason to shut off engines in opposing pairs. If you know that one
failed, it's far better to use the control authority of the other
engines "throttle steering" law and simply shut off the propellant to
the failed engine, than to shut off it's opposing number. Not with 25
other engines that have to be throttlable anyway. Shutting them off in
pairs unnecessarily reduces the acceleration.

This is all speculation, but I presume they were smart enough to
have figured this out, but had no way to implement it. The way you would
do it with a processor is to create a pseudo-inverse using the 26 ring
thrusters that maps torque commands from the autopilot to throttle
positions of each engine. If an engine failed, as soon as you knew it,
you would recompute the pseudo-inverse using the 25 remaining engines,
and go about your business.

Doing in analog controls, however, makes this pretty darn tough. I
guess that you would hard-wire the mapping/pseudo-inverse. Trying to
create a logic matrix that is reconfigurable in analog electronics would
maybe be possible, but exceptionally complex. What I expect they decided
to do was to shut them off in pairs, use the same hard-wired
pseudo-inverse, and just "eat" the reduced control gain and whatever
cross-coupling that resulted.

So this may have been one of the rare situations when superior
computing systems WOULD have made a significant difference in the results.

Brett