Subject: RE: Servovalve Erratic Behavior
Date: Fri, 9 Apr 1999 16:44:34 -0500
From: "Albright, Joe" <Joe.Albright@mts.com>
To: "Chen, Hans" <Hans.Chen@mts.com>, "Anderson, Harlan"
     <Harlan.Anderson@mts.com>, "Gram, Marty" <Marty.Gram@mts.com>,
     "Saari, Byron" <Byron.Saari@mts.com>, "Larson, Rod" <Rod.Larson@mts.com>,
     "Nelson, Terry" <Terry.Nelson@mts.com>, "Lorenz, Josh"
     <Josh.Lorenz@mts.com>
CC: "Melander, Dan" <Dan.Melander@mts.com>, "Carmein, Judy"
     <Judy.Carmein@mts.com>, "Hessburg, Thomas" <Thomas.Hessburg@mts.com>

Hello all,

Thank you for your inputs.  The problem seems to have been solved.  We
found that the inverse model was calculating large, quick transients in
the region of the problem.  Most of the time the command transient would
be "averaged" out by the valve, yielding very acceptable results.
However, for some combinations of transients it generated an offset
instead of the average.  This was the problem we were observing.  The
solution has been to eliminate the command transient before it is sent
to the model.  For the modelers on the list, it seems that the
compressibility flow needed to position the poppet is the term that is
the largest generator of the noise.  By setting the bulk modulus to a
very high number most of the noise was eliminated.  Use of a 5 point
moving average filter (2 kHz sample rate) looks like it will eliminate
the rest of the noise.  I will know more next week.

One other effect we have noticed is that if the inverse model encounters
a transient in the desired, which is substantially faster than what the
system can follow (i.e. like following a small step), the response
becomes substantially slower than the desired.  My impression is that it
is because in our case the greatest part of the tracking accuracy comes
from the inverse model.  When the system fails to follow a very rapid
transient, its "closed loop response" is quite slow in comparison with
some very strange results.

Again, thanks for all of the input.  FYI, we have done 2 pulses this
week where nearly all or all of the pulse was within spec in 3 - 4
iterations.

Joe

> -----Original Message-----
> From: Albright, Joe
> Sent: Wednesday, April 07, 1999 2:44 PM
> To:   Chen, Hans; Anderson, Harlan
> Cc:   Gram, Marty; Saari, Byron; Larson, Rod; Nelson, Terry; Melander,
> Dan; Lorenz, Josh
> Subject:      Servovalve Erratic Behavior
>
> Hi Hans/Harlan,
>
> While testing the Toyota Crash Simulator, we believe we are seeing
> short periods of time during a pulse where the 256 spool position
> moves in an unexpected direction for a short distance, and then
> catches back up with the expected position.  It may have been doing
> that for an extended period of time.  We are not certain.  It became
> very evident recently during a higher jerk test because the
> unexplained motion occured right at the time of poppet unseating,
> creating a very large acceleration transient.  I am interested in any
> ideas you (or any of the cc's) might have about why it occurs and how
> to remedy it.
>
> Byron and I have started investigating possible control output errors
> from the model, and will probably change pilot valves to see if it is
> pilot specific.  One possible explanation proposed was 252 ball slop.
> If that is indeed the case, an electric feedback pilot valve would
> address that problem.  Along those lines, what is the status of the
> Bosch valve with electric feedback we evaluated a year or so ago.
> Harlan, I was unsuccessful finding it in the back room.  Do we still
> have it, and if so what is the pressure rating on the valve and what
> is required to try it on our system?
>
> Any ideas anyone would care to share would be greatly appreciated.  At
> the moment, this problem is proving to be a bit of a show stopper.
>
> Thanks!
>
> Joe A
>
> joe.albright@mts.com
> http://intranet.mts.com/users/a/albright/index.htm
> X4355
> MIPS Dept (
> http://intranet.mts.com/users/b/bigotB/reference/mips/index.htm )
> Toyota Crash Sled
> (http://intranet.mts.com/users/a/albright/1997/CrashSim/Checkout/Daily
> Log.htm)
>
>
