Butterfly valve in PID loop

Hi all,

I'm new to process automation. I have a loop that contains (one butterfly valve + pneumatic actuator + positioner) and (flow meter). I was asked to calibrate the positioner and the valve. The point is that the butterfly valve as far as I could see is not linear. I mean that flow isn't linear with valve position, am I right?? I don't have valve chart so, I'm trying to obtain the curve by some ways. In all cases, I'm asking do I have to teach positioner the curve (input signal 4-20 mA / valve position)?? Or I could rely on PID controller only to handle this non-linearity?

You are right, the butterfly valve is probably not very linear at all. However, what you say is that you have been asked to calibrate the valve positioner, not tune the loop. Do you mean that the PID loop controller is in the box you are calling the positioner? If so, then is it really a controller, not a positioner, and the "positioner" is actually contained in the pneumatic actuator?

Most likely, if it is a closed loop, your description below is missing the controller. Just hooking a flowmeter straight into the positioner is going to give extremely rough control, if it gives any control at all.

If you will calibrate the "positioner + actuator" portion to stroke the valve from 0% to 100 % with a corresponding 0% to 100% input signal, whether it's 3-15 psi, 4-20ma, or some other standard, then *IF* the PID is properly tuned and the flow transmitter is also properly calibrated, the integral in the controller will bring the valve to the proper position after several reset periods.

Now, if the task is to "calibrate the loop" rather than "calibrate the positioner" then you need to calibrate the positioner+valve assembly, calibrate the flow meter, and calibrate the controller. Then you can tune the controller - but if it is an existing loop that has worked before you can probably just set it up the way it was before and it will probably be pretty close.

Can you shed some light by telling us a little more about the components?

Michael
--
Michael Batchelor
www.IndustrialInformatics.com

With this being a flow loop the PID controller will probably handle it. If you still want to linearise the loop though, you need to step the signal to the valve (Process Demand) from 0% to 100% in 5% or 10% steps and note the flow reading. You would then plot the flow against Process Demand in Excel and note the shape of the curve. You need to then plot a characterisation curve that when applied to the flow response curve gives a linear response (difference of the two responses).

To get a characterisation automatically in Excel, you should normalise the flow response numbers (i.e. change them to % of span), then plot the Process Demand against normalised flow and you get the characterisation curve required.

The points of this curve are the points used in the characterisation table inside the positioner or DCS.

The characterisation can be applied to either the PID output as a furmula or look-up table or inside the positioner as a lookup table.

Mr. Hassona

The valve+positioner calibration is usually only the 0 and 100% opening setting (and a 50% check).

I agree that the butterfly valve nonlinearity impacts very negatively in the loop performance. But "flow calibration" is a difficult task. The resulting flow is influenced by several factors:

. Upstream and downstream pressures

. Pipe friction loss

. Other valves losses

. Pumps, etc.

You may use a flowmeter to evaluate the actual relationship between valve opening and flow, and implement a linearization curve. But some of these conditions may change after some time (months, weeks, days, hours�), making your linearization curve give wrong results. The most common problem is positioner calibration drift.

Therefore, and following the KISS principle, I suggest you to try first the PID only, if possible. How many maintenance people (at your site) will understand the linearization, and make a new curve when needed? Think about the long run, and not only immediate performance.

Regards,

Marcos T. Bergamo

I'm really very thankful for those who posted replies and I'm sorry as my description missed some info.

The job required was to calibrate the assembly ( valve + actuator + positioner) to make sure that 4mA will give no stem travel and 20 mA will give 100% of stem travel.

I did that but I noticed that between 0 and 100% the relation between input signal and estimated flow by such valve positions was not linear.
These assemblies will work in closed loops with flow meters and PLC with software PID function will take control.

From the beginning I was intending to make a look-up table after PID's output as I thought that this will improve loop performance.
Now, I don't know whether to keep it simple with no curve and let the PID handle this non linearity or to make a look-up table after PID's output.

One more question I have, what are the factors that determine valve's ramp up time and ramp down time?
Once again, I'm really thankful for your replies.

Don't worry about the non-linear flow, the PID will take care of it. The nice thing about flow loops is they're fast so you don't need much gain and the reset rate can be reasonably high to overcome any offset.

By the ramp time I assume you mean from closed to open with a step change in signal, thats usually a function of the positioner orifice size (assuming it is pneumatic) A typical 6" valve would stroke in less than 5 seconds, much faster than the control loop tuning requires.
I always set my positioner up so that it doesn't start to move before about 4.5 mA that way you can be sure it's closed when the controller is at zero. If the valve is oversized you can also reduce the 20 mA setting to say 60% of full travel that will give a more linear response.

Hope this helps
Roy

Non-linearity in the flow control valve will not matter as long as the dynamic performance of the control loop is not critical. By dynamic performance I mean the speed with which the loop will correct an error in the process variable. You can make almost any PID control loop stable by reducing the performance sufficiently. If the time the loop takes to settle after a disturbance is not critial to you, then don't worry about. Just tune the PID loop for stability at the worst-case operating point (probably at the almost-closed end of the valve travel).

On the other hand, if you need to push the performance to its limit for some reason, then it may help to insert some non-linearity into the loop. Hopefully the non-linearity that you insert will just counteract the non-linearity in the control valve, making the total loop linearity better. This can be done in the form of a lookup table that defines a piece-wise linear transformation of the commanded valve position, for example.

One case where I can see the need for such linearization is if the dynamic range of your process variable is very large - like trying to control the flow over a 100 : 1 ratio. In that case tuning the control loop for one end of the range will probably result in unacceptable performance at the other end of the range.

Robert Scott
Real-Time Specialties