Omar & Jim,
Today looked at why the actuators seem to have less magnification horizontally than vertically.  Discovered that when exciting the actuators laterally from the other horizontal axis, that rotation occurs that robs energy from the pure bow-string.  We mounted an accelerometer on the horizontal center of the actuator in the direction of lateral excitation.  Another accelerometer was mounted on the Moog servo-valve in the same direction.  We find that the 2 accelerometers are 90 degrees out of phase at the bow-string frequency.  Therefore, what is reducing the amount of magnification at bow-string is the rotation of the actuator.  The swivels, of course, have friction and tend to dampen the rotation, robbing energy from the free mass system.  This all happens because the lateral force on the end of the actuator is not in the plane of the center-of-mass.  The servovalve and manifold off-balance the actuator, thus causing rotation to occur when lateral force is applied at the end of the rod.  In the vertical axis, no such center-of-gravity imbalance is present; therefore, the full force is applied to pure bow-string.   Accelerometers again confirmed this fact; there is virtually no rotation of the actuator when excited vertically.  Perhaps a center-of-gravity imbalance in the vertical plane could have positive effect.  Of course, the down side is that the swivels will eventually loosen up and not provide the required damping.

Ito and Fuku and Bami talked to Mr. O today, and he still wants full performance.  He thought that if vertical is the only problem, then a single-axis vertical damper could work.  Only a few % of damping is required to radically lower the magnification.  The current magnification is 8 from table to actuator, but at the actuator itself, the magnification is actually 16 because at the center of the actuator, the low frequency acceleration is exactly 1/2 of the table acceleration.  For a magnification of 16, then,
the damping ratio is a little or 3%, which is quite believable for a mechanical structure like this.  In order to stay safely under 12 g actuator limit, the table to actuator magnification has to be around 3 to 3.5 for 3 g input.  This translates to an actuator magnification of 6 to 7.  This magnification represents about 14-16% damping.

Due to non-linearity of the vibration problem, the sine wave tests show a slightly larger magnification.  For the vertical axis, the random magnification is 8 and the sine magnification is 10.  For the horizontal directions, the magnification only goes up about 10%.

A far bigger problem, I think, is the loose bolts.  We will check bolts again tomorrow, but, I think that they will be loose again after only running random and earthquake records.  The customer does NOT know anything about the loose bolt problem.  Putting accelerometers on various parts of the table near the wings is quite informative.  In the horizontal directions, there is almost no real deflections of note.  However, the vertical direction is quite another matter.  There are significant deflections in this axis (as noted by the 1.33 magnification due to deflection of the table).  After much close measurement, we have confirmed that referenced to the wing acceleration, the actuator itself has a magnification of 6.0 and the wing amplifies by 1.33 to give a total of 8.0.  So, even perfect wing still gives a problem for the actuator itself at
3 g's where the actuator would still get 18 g's.  Another interesting note:  when running the table now, there is a definite low frequency panel vibration resulting in a kind of a low "booming" sound from the table.  This was not noticeable with the old wings, but probably couldn't hear it then for all the clanging and banging that was occurring then.

Final analysis tomorrow, but for now I think:

1. Loose bolt problem MUST be solved somehow.

2. Might be able to deal w/ customer about performance issue, but some consideration will be likely.  A vertical damping system might be best idea.  I think that our past on this job will NOT help us in negotiations w/ customer.  If this was only major problem, then maybe could skate.  My conservations w/ customer leave me w/ impression that he does not want to give on this one after all of the other problems.

Now the good news:  This system really runs earthquakes GREAT!!!!!  I did some training today about how transfer fcn compensation really works.  We were able to get the 8th floor earthquake down to <0.1% rms time history error on the horizontal axes and the vertical to <2% (hard to do cuz very small level).  The other standard earthquakes produced spetacular results w/ errors so small you can just barely see the time history errors in the plots even with different colors for desired and response.  THIS might save our butts w/ the customer in the long run.  MOC has NEVER achieved such good STEX results; but it may be cockpit errors.  I did notice that when Jim Langseth was here last time that he did not know the optimal iteration percentages to do and what is a good transfer fcn measurement and what is not.

All for now, (like this isn't enough)
Rod Larson
