New VFD is hunting

Hello all, I have been using Hitachi drives for the last few years, have installed about 40 or 50 in all types of machinery, mostly pumps. So when I was asked to retrofit an existing pumping system with a new VFD, I had no worries. The previous VFD was an ABB , the new one is a Hitachi L300P, 480 vac, 75hp. The signal in question is the analog speed reference, 0-10v.

The VFD "hunts" at a frequency of about 2 cycles per second.

I have been in touch with the mfg of the pumping system controller, and he shared info about the PI loop, and tuning the variables. I have tried a variety of combinations, some resulting in less or no "wow", but completely useless as far as following the pressure, correcting for disturbances, etc.

I have set the accel and decel in the drive, bit by bit, with the same result: if I get rid of the "wow", the perfomance is deplorable, or almost non-existant.

The controller mfg told me that they know their system has that tendency, but they just tune it out with the "analog filter" in the ABB, by setting it to average for 2 seconds. So I found the "analog filter" in the Hitachi, which sets from 1 to 30... and set it to the max, 30 (which represents 30 samples taken, then averaged before loading the value as the speed reference), but whatever it is it isn't enough.

The customer SWEARS it never DID that before I put in the new VFD. I have talked to KELE and they say they can make me an "external analog filter" but I have to tell them what "time constant" I want. I'm going to the site tomorrow with an IDEC PLC, (my usual brand) and an analog i/o card.... so I can MAKE an averaging circuit with settable time constant, and I can determine an optimal range.

I KNOW there will be questions... I CAN'T have included absolutely EVERYthing... But I have tried to share what the problem is (the "wow" or "hunting" at about 2 x sec across the entire ranges of pressures and flows) and all the things I've tried thus far.

If there is an easier or more straight forward way.... I'm ALL ears.

ANY comments from people with actual experience in this area will be appreciated....

Thank You,

Stationmaster

What would happen if you took a couple of flashlight batteries and put in a constant DC voltage(s) ie 1.5 3 4.5 etc to the 0 to 10 V input? At least this would confirm your suspicion it is the sensor that is not giving a steady signal.

The other thought is at what speed are you getting the 2 Hz variation - could this be caused by variations in pump load - pump loads down sensor compensates VFD speeds pump up etc etc ?? the battery trick I mentioned may answer this question.

Dan Bentler

If you have the drive in PID mode, then you don't have a 0-10V speed reference. You have a setpoint input and a process feedback input.

From all the talk about filtering, I'm going to assume that you are refering to the process feedback signal---the one that brings the output fluid pressure back to the drive for comparison to the setpoint.

It is common on pumping systems to see 2 seconds of filtering on the process feedback input. But, it sounds like you are talking about lots more filtering with the Hitachi drive which can't be good. If you are getting oscillation or hunting with 2 seconds of input filtering, I would suggest that the P, I, and D gains are not set properly.

Setting these loop gains on a water system can be particularly tricky if there is no accumulator installed in the piping. An accumulator makes the pressure response much lazier and easier to tune. I can't really offer much in the way of advice on strategies for tuning PID loops. Years of experience is about the only bank I can draw on and sometimes it runs very close to dry too.

You may want to enlist the help of someone who has done this often. I don't think smothering the signal with filtering is going to be satisfactory.

Dan, I have aloop calibrator..... I didn't mention it, but I've done one even BETTER.... The pump controller has a "manual VFD" capability which gives the technician the ability to take control of the speed reference from the PLC using up and down arrows on the HMI.... Once I go to "manual" the "wow" is gone....

Perhaps I need to describe the "wow" a little better. It's not 2hz as read on the VFD HZ out... the wow itself is an estimated 2 cycles per second. That is as the motor emits "wow-wow-wow-wow-wow-" each "wow" is about a half second long. The VFD HZ OUT display varies, visibly, but only in tenths of a HZ... it doesn't even vary an entire hz...

The analog/mechanical guage varies about a pound or so, with the needle shaking at a speed that matches the "wow" sound.

The digital pressure display updates in "notches" and is therefore hard to analyze... it seems to jump all around within about 3lbs above and 3lbs below the setpoint, a WIDER range than the mechanical gauge displays....

Dick, yes it is a closed pressure loop.... but it includes a PLC external to the VFD. The PI settings reside in the PLC. The signal in question is between the PLC analog out, and the VFD analog in. The 2 second filter was in the ABB drive, I don't know what time constant is represented by the "30" in the Hitachi drive. The system worked okay with the 2 second filter in the ABB, but not with the Hitachi.

Thank you,

Stationmaster

Stationmaster said:

Dan, I have aloop calibrator..... I didn't mention it, but I've done one even BETTER.... The pump controller has a "manual VFD" capability which gives the technician the ability to take control of the speed reference from the PLC using up and down arrows on the HMI.... Once I go to "manual" the "wow" is gone....

Perhaps I need to describe the "wow" a little better. It's not 2hz as read on the VFD HZ out... the wow itself is an estimated 2 cycles per second. That is as the motor emits "wow-wow-wow-wow-wow-" each "wow" is about a half second long. The VFD HZ OUT display varies, visibly, but only in tenths of a HZ... it doesn't even vary an entire hz...

The analog/mechanical guage varies about a pound or so, with the needle shaking at a speed that matches the "wow" sound.

The digital pressure display updates in "notches" and is therefore hard to analyze... it seems to jump all around within about 3lbs above and 3lbs below the setpoint, a WIDER range than the mechanical gauge displays....

Click to expand...

Well it is good you did the same thing I was thinking of.

Ive heard MG set that sounded like canary was inside ie CHEEP CHEEP, but for sure I never heard a WOW. Kinda crossed my mind that maybe the pump shaft or impeller has a wow in it but with VFD in "manual" and no noise that eliminates that.

Where on the pump curve are you running and are you positive you have NPSH at this speed including top speed at the wow??

Either way the VFD should hold freq and RPM at a steady level and you have seen it do that in "manual".

Most mysterious re the wow sound. Let me know how you make out with it.

Dan Bentler

What are the ramp settings on the drive? It could just be chasing the signal from the external PI loop.

If the VFD is following a speed reference from the PLC, and the PID system is in the PLC, then the filtering should not be on the speed signal to the drive. It should be on the process signal going into the PLC.

On water systems with no accumulator, it is almost always necessary to smooth the feedback signal a little to get close to stability. With a "hard" piping system, there will always be a little "dance" in the pressure gage but I wouldn't expect it to be so big that you can hear it in the motor/pump.

I'd put the 2 second or so of filtering on the process feedback signal and then retune.

RD Rast,

Yes, clearly the VFD is chasing the signal from the external PI loop. I have talked to the developer of this external PI loop, and he says their package has the tendency to do that, they just tune it out in the analog filter of the VFD they use.

I have tried Accel and Decel settings from 8 seconds to 30 seconds. At thirty seconds, the hunt/wow goes away, but performance is awful. The presssure undershoots and overshoots are unacceptable.

All,

I made a time based averaging filter last night, from a PLC and an analog I/O card. I am sampling the data in the "analog in" register at a rate of once per millisecond, and can adjust the number of miliseconds/samples that are averaged, before loading the average into the "analog out" register, giving me the "time based" filter control I want. It works great on the test bench, as VISUALLY displayed with touch screen line graphs that seem to represent exactly what I'm trying to accomplish. I am heading to the site this morning to experiment with my "filter". If I can find a happy "averaging time", that is a decent comprimise between "eliminating the hunting" and "acceptable performance" I will then order a signal conditioner with that "averaging time" built in. The signal conditioner has been quoted at about 1/4 the cost of the PLC-i/o card combo. (I'll just put the PLC and analog card back on my test bench.)

Wish me luck. I'll post the results this evening.

Stationmaster

DickDV said:

If the VFD is following a speed reference from the PLC, and the PID system is in the PLC, then the filtering should not be on the speed signal to the drive. It should be on the process signal going into the PLC.

On water systems with no accumulator, it is almost always necessary to smooth the feedback signal a little to get close to stability. With a "hard" piping system, there will always be a little "dance" in the pressure gage but I wouldn't expect it to be so big that you can hear it in the motor/pump.

I'd put the 2 second or so of filtering on the process feedback signal and then retune.

Click to expand...

Dick, VERY good point. I was focused on the input to the VFD because that's where the original developer said his company "tuned it out". But I HAVE seen the "dance" of which you speak coming from the xducer and "hard" piping configurations... I will try that too. Thank you.

Stationmaster.

Stationmaster said:

RD Rast,

Yes, clearly the VFD is chasing the signal from the external PI loop. I have talked to the developer of this external PI loop, and he says their package has the tendency to do that, they just tune it out in the analog filter of the VFD they use.

Click to expand...

Well... Hrm... their 'package' needs some work, but I agree with Dick above, filter the feedback going to the PI loop.

Stationmaster said:

I have tried Accel and Decel settings from 8 seconds to 30 seconds. At thirty seconds, the hunt/wow goes away, but performance is awful. The presssure undershoots and overshoots are unacceptable.

Click to expand...

I make extensive use of PID loops in PLCs, and always set the accel and decel ramps for the absolute minimum the drives will take. Typically 0 seconds for DC drives, and 0 or 0.1 seconds for AC Drives.

rdrast said:

Well... Hrm... their 'package' needs some work, but I agree with Dick above, filter the feedback going to the PI loop.

Click to expand...

YOU sir, are ABSOLUTELY correct ! ..... Their 'package' is the only one that has EVER given me this headache.

rdrast said:

I make extensive use of PID loops in PLCs, and always set the accel and decel ramps for the absolute minimum the drives will take. Typically 0 seconds for DC drives, and 0 or 0.1 seconds for AC Drives.

Click to expand...

I usually set the Accel and Decel relative to the mass I am moving. These are 75hp vertical turbine pumps, so there is a lot of mass. 5 sec on the Accel works on MOST systems, with 8 seconds on the decel. The slower decel is necessary to prevent regeneration. Those settings are useless on the system i'm struggling with now.

Thanx

Stationmaster

On your PLC how many counts are there on the Analog in and out ( 10 bit to 15 bit ) are common and what range in PS, speed, HZ, do they cover when dealing with 1 -10 volt or 4 to 20 mA ranges

What I am wondering is are the counts stepping in course steps as against fine steps, have they made a universal package that has a wide pressure range to serve multiple customers.

Generally, I am in agreement with rdrast concerning setting accel/decel times very short. The P, I, and D gains should regulate the slew rate.

However, setting the decel rate up a little to prevent regen probably can't hurt as long as it is still down around 1-2 seconds. Anything longer than that will make the loop tuning less likely to work right.

Just to be clear, I'm talking about the Accel and Decel rates in the VFD, with the PI loop acting from OUTSIDE the VFD, as opposed to an accel and decel that may be INCLUDED in some PID function blocks.

Another engineer pointed out that I can make an RC network to "filter" the signal.... some quick calculations have me looking for some off-the-shelf componenets at Radio Shack.

I was unable to try ANY of the good ideas we've discussed, because the CUSTOMER didn't want to deal with it on a Friday....