I would consider the BIAS parameter along the FF0 FF1 and FF2.
Then all the the terms can be considered Feed Forward terms of the
controller, and the part of the control they deal with is conceptually
called "feed forward control".
The feed forward capabilities implemented by those terms can be enough
to implement feed forward compensation of mayor dynamical effects in the
context of Cartesian machines. If the dynamics is nonlinear, accurate
feed forward terms would in general be nonlinear.
The meaning of these parameters are dependent in the meaning of the
output of our controller.
For example if the output of your controller is torque,
then the feed forward terms can be used to compensate (examples):
BIAS: Torque that is needed in an axis to compensate the gravity
effects.
FF0: Torque that is needed to compensate the forces induced by springs
(if present at all), for example, belows covering linear guides can
induce a elastic component (also viscous etc).
FF1: Torque that is proportional to velocity, for example due to viscous
friction
FF2: Torque tat is proportional to acceleration, this is almost always
torque used to compensate the inertia.
A interesting torque that EMC ca not compensate for is Coulomb type
friction, that would be like a BIAS but dependent with the sense of
motion (v/abs(v)).
If the output of the controller is velocity you should judge the meaning
yourself.
Finally when assigning values to these parameters, I think that a good
idea is to be conservative.
Gravity effects, and inertia effects are likely to be constant along the
life of the machine (ignoring variable loads), but this is not the case
of friction (viscous or coulomb), elastic component can be pretty stable
if there is not malfunction or broken elastic elements.
I recommend only to compensate the part of the torque that you know is
going to be there always. I'm not sure but overcompensating a frictions
not a good idea so don't use friction compensation or compensate only
the minimum friction that you are expecting. Also values of friction can
only be estimated after experimentation (identification), this can be
the case of stiffness, but usually gravity and inertial terms can be
estimated from cad data (part weight and -additionally- gravity center
position in the case of gravity compensation).
Feed forward compensation can be in general a good idea, as it
theoretically removes work from the standard PID controller what
theoretically should make things work better/smoother.
In the case of nonlinear dynamics, you can still use those terms to
compensate the first term in the tailor series expansion of the above
mentioned torques.
If you implement your own pid module, and use a dynamic model of your
machine to accurately compute the output of the FF? terms you ar making
essentially what is know as CTC (computed torque control), that is the
generalized version (for non Cartesian architectures) of the
implementation present in EMC, that in my opinion is almost the best
(except for Coulomb terms) implementation without introducing particular
characteristic of the machine, what turns out to be the case.
I hope this helps,
Javier
------------------------------------------------------------------------------
Start uncovering the many advantages of virtual appliances
and start using them to simplify application deployment and
accelerate your shift to cloud computing.
http://p.sf.net/sfu/novell-sfdev2dev
_______________________________________________
Emc-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/emc-users
