> ...Something was shaking the
> accelerometer pretty hard at 40 Hz. Very unlikely it was on the whole
> vehicle. We'll take Henry's and Doug's suggestions and see what NASA
> has to say about sensor isolation.
In case it's of interest, I attach (as plain text) my notes on SP-8036.
It's mostly assuming a level of analysis beyond what's probably reasonable
for the KISS vehicles, and it represents what I thought was interesting
several years ago, but it might still be of some use. (I can also supply
a PDF of SP-8036, if it's needed and not on hand. The MSFC tech-report
server reportedly is hors de combat from recent tech-transfer paranoia.)
Henry Spencer
[EMAIL PROTECTED]
SP-8036, Effects of structural flexibility on launch vehicle control
systems, 1970. MIT A&A.
Most controllers apply both gain stabilization (avoid adding energy
to modes and rely on passive damping; easiest at high frequencies)
and phase stabilization (phase the control forces to take energy
out of mode; often necessary at lowest frequencies).
Putting the gyros up top may, because of first bending mode, have
them seeing a deflection to the left at the same time as the engines
are deflected to the right.
SA-203 went through three gimballing cycles before holddown release,
trying to fight wind disturbances.
The first four Thor-Agena A flights had a 5Hz oscillation in yaw
during first-stage flight which could have been fatal had it not
been limited by the hydraulic system. Yaw rate gyro was reporting
first bending mode 180deg out of phase due to local deformation of
the sidewall structure being coupled into the gyro bracket. Pivot
supports at each end of the bracket eliminated sensitivity to local
deformation.
At the "tail-wags-dog" frequency of engine gimballing, engine
inertia can cause a reversal in effective gimballing forces, because
the engine is over to one side but is accelerating towards the
other side. This showed up on early uprated Saturn Is durinig
staging, when thrust decay was rapid but loss of hydraulic pressure
in controls was slow. For a few seconds there was an active control
system operating without thrust, and the first vehicle vibration
mode oscillated strongly. Gains were reduced during staging to
fix it. The same thing happened to an Apollo CSM tested on the
pad, fixed by revising lead compensation. Also, beware cases where
a vehicle mode is below the t-w-d frequency and actuator load
feedback is used.
Engine failure can cause large transient responses because of
rapidly-changing thrust vector. Original Saturn V design could
generate enough bending moment that way to break the Apollo
spacecraft; fixed by canting the outboard engines to move the thrust
vectors closer to the center of mass.
Transient response during switchover from primary to backup control
system must be considered. Needed attention on Gemini.
Stored energy in structure at high acceleration can be increased
by hydrostatic pressures in tanks. If shutdown is rapid enough,
remaining propellants can be thrown forward ("slingshot effect").
One reason to locate sensors forward, rather than in the engine
bay, is reduced levels of vibration and noise at liftoff. Beware
aliasing into sensor output bands; the Saturn IB had trouble with
this in some gyros. Usually the worst noise problem is sensor
saturation.
Pogo can also cause saturation. Coupling between lateral and
longitudinal modes can put Pogo energy into lateral motion at gyros.
Saturn V could do this because of asymmetric payload stiffness,
but fortunately the filter networks in the control systems already
attenuated heavily at the bad frequency.
High thrust in a long, flexible launcher can cause "garden hose"
instability, like a hose whipping back and forth.
Control system design should start by assuming rigid body and
negligible coupling between axes, with coupling, flexibility, and
slosh added later. It will typically be necessary to phase-stabilize
the first mode (although some in-space upper stages, e.g. Centaur,
can get away with gain-stabilizing everything) and might be necessary
to do likewise for the second. A good start is gain margins of
6dB and phase margins of 40deg. Preprogrammed gain changes may be
needed to deal with changes in dynamic characteristics during
flight.
If adequate stabilization is not possible with conventional methods,
consider notch filters to attenuate response at critical frequencies
(limited by the rapid change in response characteristics during
flight), use of multiple feedback sensors, or use of adaptive
control.
Once a linear design is done, examine nonlinear elements. Hard
nonlinearities like saturation, dead zones, and backlash are
particularly relevant because they can cause limit cycle oscillations.
Quantization and sampling effects can do the same.
If possible, put gyros near nodes and accelerometers near antinodes
of phase-stabilized modes, and conversely for gain-stabilized
modes... but the actual position is usually a compromise. Consider
multiple sensor positions; Titan avoided trouble with first vibration
mode by blending output from two separate sets of rate gyros.
Locating sensors away from buffet-prone areas is wise.
Choose important modes by modal gain: modal deflection at gimbal
times modal slope at gyro over generalized mass. Low modal gain
usually indicates that a mode is not significant, but beware that
if it's because the point in question is a node or antinode, small
changes in mode shape can shift those points. Consider effects of
configuration changes. If vibration mode frequencies are near
controlled-rigid-body frequencies, consider coupling; preferably
the control frequency should be at most 1/5 the first bending-mode
frequency. Beware pre-liftoff cases, especially if there's a
holddown period with engines firing and controls active.
If possible, sensor-mount natural frequencies should be at least
twice sensor bandpass. Include sensor mounts in structural modelling,
and predict slopes for sensor locations. Try to put sensors away
from massive or dynamically active components that can cause local
deformation. Consider mounting pitch and yaw gyros separately on
respective structural neutral axes, to further avoid problems of
local deformation.
Assessment of engine-mount flexibility should include the case of
rigid engine attachment, to catch local flexibility at actuator
attach points. Note issues like fluid compressibility, hose
restraint, gimbal friction, and structure flexibility. Choose
actuators to be good enough; don't arbitrarily add large margins
on velocity and force capabilities, because hydraulic-system
saturation can limit moment applied to vehicle during fast
oscillations. If possible, keep engine gimballing resonance above
tail-wags-dog frequency.
Consider thrust transients due to unsynchronized shutdown of engines.
Switchover to redundant control system should consider actuator
rate limits, and switchover circuit frequency must not match a
vibration mode.
