friction

Modelling friction is very inexact and fraught with uncertanty and error. It's also the most reliable way I've found to crash Simulink. "The Control Handbook", 1996, William C Levine (editor), CRC Press has an excellent article on modelling friction on page 1369.

Because the friction model is so hard to pin down and complex (and because it's so good at crashing your simulator), your best model isn't going to be my best model, and it's not going to be the best model for the next system you design for.

My favorite models are: (a) just dry coulombic friction, where the friction force is equal to the sign of the relative velocity times a constant (b) a very high gain on the relative velocity plus a saturation -- this keeps Simulink from crashing but lets your system "creep" where it wouldn't in real life, and (c) coulombic friction plus stiction.

thx but ur explanation is for simulation only...but i see so many IEEE papers on friction compensation...but none i am really convinced..most of the stuffs are written for just for the shake of number of publication.......i tried contacting some of the authors but no reply...hmmm thx kristo

I think the reason that I mentioned crashing Simulink is because it's and indication of how mathematically untractible friction is. Since control theory boils down to applied math, mathmatically untractible systems means difficult theoretical solutions.

The number of papers is probably because friction is so hard to model, yet it it so pervasive you just can't ignore it. In motion control systems it's often the limiting factor in performance, and if it isn't it's often because some mechanical engineer has sweated blood to lower it. This is the kind of thing that will generate lots of papers. I haven't been keeping up, but if you did a survey I bet you would find that they fall into two types: 1, failed attempts at a "general" or "universal" solution, and 2, working solutions to specific cases.

In my work I have to deal with friction often. I do it by remembering that mathematics is just a subset of common sense, so applied math is really applied common sense. Therefore it's perfectly OK to apply common sense to the problem of compensating for friction, even if I don't know what the heck I'm doing mathematically. Then you test the heck out of your system, cross your fingers and ship it.

I usually do it either by driving the motor with a fairly high-level PWM signal and varying the pulse width to control the distance moved, or designing substantial deadband into my controller (deadband is usually the only thing that will work to stabilize a plant that has backlash in it's gears, it comes right after friction in the race for ubiquitous problems that crash Simulink, and if you have geartrains you have both backlash and friction anyway).

-- snip --

thx...any info on the types of control used in such systems thanks a lot.