Is a PIDE instruction enough?

Hi:

I have a process where I need to control flow with a bomb with Powerflex700. I have a periodic PIDE instruction, Flowmeter is the PV and CV is PowerFlex Speed. A very simple an standard configuration.

The problem is that the flow is very unstable, even with the pump at fixed speed (working with PIDE in Manual mode).

With fixed speed I have fluctuations of up to 30 m3/hr in a span of 0-500 m3/hr of the flowmeter.

For example the meter is reading 341, 310, 340, 360, 355, 350, 330, 320, 325, 315, etc, etc every one second update rate.

I have tried to tune the PIDE with a lot of methods (Logix AutoTune, RSTune, test and error, Ziegler-Nichols, using just P, just PI, etc) without good results.

Something that I am missing or definitely PIDE isntruction never will work here?

thanks

Do you have a Digital Filter setup on the AI card?

Noisy analog input? Is the wire shielded?

You could apply a filter. See txt.

The 4-20 mA wire is shielded and when I see the flowmeter display, the signal also reads unstable, with similar values that the indicated in the 4-20 mA signal (scaled to Eng units).

Thanks

Some flow meters need to have some conductivity in the fluid to function properly. This can make commissioning / water batches difficult if the water is too pure, unless one throws a handful of salt into the tank first, to simulate more realistic conditions.

Then again, is the goal is to get the tank clean, and those are the types of flow meters your stuck with, then no amount of filtering or tuning is going to help, and you just have to set constant conditions and ignore the meter.

I have a process where I need to control flow with a bomb

Click to expand...

???

If you see variations at fixed speed the PID is not the problem. Unless you later try to control a fast disturbance which will cause oscillations.

What liquid do you have?
Flow meter model?
Anything that changes inside the liquid? Both temperature and conductivity will upset a mag meter.
Is it installed correctly?

yeah, I wondered about that too, Mickey ... but I was afraid to ask ...

Might be a translation error - "bomba" = pump in spanish.

HoboWhisperer said:

Might be a translation error - "bomba" = pump in spanish.

Click to expand...

2 years of anxiety, I feel better now.

If the signal from the flow meter is unstable it could be a number of things first the flow meter needs to be full of liquid all the time, second do not terminate the screen of the cable both ends this causes reflections in the transmission line only terminate at one end especially if there is also a grounding wire as well as the screen.

HoboWhisperer said:

Might be a translation error - "bomba" = pump in spanish.

Click to expand...

yes, that is correct, he means pump, pump translated to Spanish is bomba

I have had similar experiences with flow control loops being hard to stabilise. From your description of the flow-rate varying on the flowmeter display, it sounds as though you have turbulent flow through the meter.

There are several things you can try to get things under control....

1) Ensure that your flowmeter is installed correctly, and to the manufacturers recommendations. In many cases these include - a) installation in upwards flowing pipework if possible, b) installation a certain number of pipe diameters away from bends or valves that will cause turbulent flow.

2) Remove any "ramp" functions from your VSD driving the bomba, it was the first thing I did on any flow control commissioning jobs I was involved in.

3) Try "filtering" your analog input signal to damp out irregular readings. This can most often be done on the analog input module itself, or it can be done in software, usually by averaging over several readings. There are numerous ways of doing averaging. EDIT : Most flowmeters also allow filtering to be applied.

One method I have used with great success on an input signal from a yeast slurry capacitance sensor (it gives a measure of the live yeast cell count, but is prone to irregular readings), goes as follows...

Capture the input signal into a shift register of readings at a regular rate.

Copy the shift register array to another, then Sort that array. abnormally low readings will go to one end, abnormally high to the other.

Average the middle n readings, effectively ignoring the spurios data.

For example capture 12 readings, copy, sort, then average the middle 10, or 8 etc. readings. The lower the number you average, the more aggressive the filtering.

A possibly simpler alternative to filtering is to use the PIDE deadband; with a noisy signal and no deadband, a PID controller can add to the process noise by chasing the sensor noise.

If there are +/-X units of noise in the measurement with the output held at a constant position, then with a 1.02X deadband the PIDE will only control (change the output) when the PV is more then 1.02X above or below the setpoint.

Straight from the Powerflex700 manual, default accel and decel rates for all speed changes - 10.0 S !!

You are not going to get good PID control with those ramps set in the drive.

EDIT : Another thing that comes to mind is to make sure the pump is sized correctly for the flow-rate you want. Most centrifugal pumps (bomba's !) achieve maximum efficiency (and hence maximum flowrate) running at 70 - 90% of their rated speed, not 100%. Also most centrifugal pumps do not pump anything at a speed less than about 20 - 30% of their rated speed. The pump data should come with efficiency curves that you can use to determine your min and max output scaling for the flowrates you require. To most people a pump data sheet is just a minefield of numbers and diagrams, so perhaps you can talk to the technical team of the pump manufacturer to see if it is suited to the application.

ANOTHER EDIT : Where I used to work (in the process industry, beer and milk), it was rare to see a VFD pump to control flow-rate, unless it was a Positive Displacement pump, in which case the VFD controlled pump was essential. The vast majority of flow control systems used fixed speed DOL pumps, and variable control valves.