vfd tuning

Hi all,

Basic question. Need to install a vfd on an exhaust blower fan on a piece of machinery. Currently has a motor starter on it. Once vfd ( GE AF-600FP) is connected I need to run the autotune parameter to do a full autotune. My question is do you usually decouple the load fromm the motor ( in this case remove the belts)when performing this. Also, what if you had an application where you couldn't decouple the load from the motor before tuning such as a pump. Thx

Generally...

Typically, it is best to tune ANY control loop at or near actual process conditions . So, loads and upsets should be present to asses response performance to achieve optimum tuning values.

I don't believe the "autotune" runs the motor on most brands. I'm not sure about the GE. Have you checked with your supplier?

See page 34 of http://www.geindustrial.com/publibrary/checkout/DET-610?TNR=Installation and Instruction|DET-610|generic

I will check with the supplier but figured I would ask here first. GE drives you can do an autotune two different ways according to the manual. First one is a basic which the drive just goes out and measures the different values ( resistances etc.) while the motor is not turning and loads them. The ohter is a full autotune. When this is selected the motor actually runs up and down at different speeds to get the values. The manual is kind of vague and doesn't say which is best.

Mitsubishi Drives

This reply doesn't address GE drives but is just a comment on one other brand. At my site we are now installing Mitsubishi A700 drives. These particular drives can be auto-tuned in rotational or non-rotational mode. The manual states that rotational mode is more accurate (with no explanation why). The drive also records if an auto-tune has been done or not, and whether or not rotational mode was used. If my vendor is called for performance support, he typically checks this parameter first.

Edit : The vendor pushes for un-coupled, rotational auto-tune.

Cheers,

Dan.

Part of this is terminology confusion. Most of the time, "autotune" is what I would call a "motor ID run". This is only needed if the drive is going to operate in sensorless vector mode. V/Hz or Scalar operation doesn't need it.

A motor ID run seeks to measure a number of motor parameters and, from that data and the motor nameplate data entered in the drive parameters, the drive builds a motor model in its memory that it can consult for estimated shaft speed, shaft torque, and sometimes other things.

If the ID run is done with the motor stationary, the rotor inertia and the conversion of input amps to shaft torque will not be known. So the model will not be good enough to expect any precision from the motor control system. If, however, the drive does a dynamic rotating ID run, then these motor details and probably several more are known and can be incorporated into the motor model. The result is much better precision both in torque and speed. Operation at or close to zero speed is also greatly improved.

Drives that have external PID loop capabilities will often have "autotune" as part of setting up the P, I, and D gains. That is a true autotune since it is measuring loop characteristics and setting gains accordingly. It has nothing to do with a motor model or the precision of the drive's motor management. PID loop autotune will typically be very conservative and often not "tight" enough. You then need to go in and increase the individual gains manually until you either get the response you want or the loop goes unstable. Manual loop tuning is an art that isn't acquired quickly and can be a huge frustration on some loops. There are mathematical methods for doing loop tuning and maximum precision and response can be acheived that way. But I haven't a clue how to do it and my manual methods have been good enough up to this point. Most pressure, level, or flow loops common to VFD's don't need to be razor-sharp anyway.

If you are doing an auto tune to compute the PID gains then you should do this with the average load.

DickDV, most drives running in velocity mode only need a PI control. Does your auto tuning provide a gain and a time constant or band width. If so then I can provide the formulas for setting the controller gains. It is easy IF you know the actuator gain and time constant or bandwidth.

Yes, Peter, I kind of thought my comments about loop tuning would generate a response from you. And, I'm glad it did. You are correct that the motor velocity loop which is internal to the vfd rarely uses the D gain. I don't believe I've used it more than twice in 20 years. The external loop which often is level, pressure, or flow rate, however, can benefit occasionally from a bit of D gain. At least, using my manual methods, it does!

You have identified precisely why manual tuning is generally used on VFD's. There is no loop data available. So, while you have given me the formulas some years ago, I've never been able to identify the inputs necessary to use them. Shucks, its like pulling teeth to get machine manufacturers to provide WK^2 values for large flywheels.

Fortunately, I've always been able to get the static and dynamic performance needed using manual methods. My method is to run up the P gain until I get some signs of resonance in the torque (it seems to show up in the torque before it shows up in the speed). I then back off about 2% and run the I time down again until I start seeing resonance in the torque. I back off again about 2% and that's it. Note that ABB usually uses Integration time rather than gain. I'm not sure just what the significance of that is except that you tighten the time response by going down instead of up.

I'm glad I don't have to deal with these systems with lots of hysteresis or deadband. About the most challenging tuning I encounter is with pressure loops on certain types of centrifugal pumps where the pump is being operated in a really steep part of the curve. By that, I mean lots of pressure change for very little flow change. The opposite is also a challenge where large changes in flow barely affect the pump pressure output. This last one usually means that the pump was sized wrong for the system flow demand. Of course, I'm just the drive guy so I can't say that and be believed at all.

Thanks for the offer. I think I'll just keep you in mind for that tough job that might just be lurking out there. I've probably got another three years or so and then it won't matter any more. Been doing this since 1989 so, at 67 years old presently, that will probably be long enough. It is my delight to be able to help young people get up to speed on these things and I agree with your recent observation that the BBS is getting smarter. That is very satisfying.

DickDV, the inertia doesn't need to be known. That is what the auto tuning software is for. Our auto tuning software will identify a simple speed control system as a gain and a time constant. The gain is velocity/percent output. The time constant is in seconds. If a system would take 1 second to get to speed then that would be 5 time constants.

Most systems use some kind of pulse but I don't like that method because a pulse can damage a system. If you use a pulse system while loaded with a 90 ton roll of steel you will get asked to leave and not come back. These pulse type auto tuners only look at about 3 points of data. We uses an open loop signal most of the time but we use ramps instead of pulse. The ramps must be fast enough to be faster than the response time of the system but not much more. The ramps allow one to auto tune while loaded which is the only way one can truly tune a system. A system will respond much differently when loaded with a 90 ton roll of steel than with out.

What I would need is a text file of the time, control output and position or velocity.