Need help on flow control tuning

In a recent plant turnaround, our instrument people replaced an old flow control globe valve. It was originally installed in early 80's
where the actuator functions through an I/P converter. Apparently, the new globe valve was installed and fitted with the old actuator together with the I/P converter. When it comes to start up, it was found that the valve is not able to close despite sending direct output signal from the DCS. This is because the new control valve high DP pushes it upward against the supply pressure from instrument air. (Note that it is a failed open control valve).
The instrument later fitted the control valve with a positioner which then enable the valve to travel from open to fully close position.
However, the flow control was quite poor where it oscillates (approx 1 m3/hr) around the setpoint. Initially, I placed the old tuning value to this controller and it resulted in heavy oscillation which indicates that the valve performance is no more the same as before. After all, it is a new valve and it's characteristic is already different though they both are an equal-percentage valve.
Despite executing close loop response tuning, I still could not bring the performance equal to the old valve. In fact, with Proportional Band of 330% and Integral time of 500, it is impossible to put the controller in auto as it oscillate the flow heavily. Note that the DCS is Yokogawa Centum.
I did an open loop step response (1%) and to my amazement, the flow responded rapidly with an overshoot. In stepping up (open), there was 2 to 3 overshoots while 1 to 2 overshoots when stepping down before the flow settle to a new value.
What can we say about this valve?
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload

Just a guess: the valve's tight new seals cause enough hysteresis for the valve positioner to have problems?
Regards, Andrey Romanenko
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
TheRomanov wrote:

A positioner is a closed loop controller that may be unstable, all by itself. Before you try to tune the flow control loop, you have to make sure the position control loop is stable. Some positioners have a gain or speed adjustment that can be set wrong enough to cause position instability.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload

Whatever you do, your end objective should be for an open loop valve response that is overdamped, ie. clearly no overshoot. A bit overslow is better than wobbly in the control world on most occasions. You should also check for stiction, that's a common cause of flow loop instability.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
Bruce,
The valve responded quite resonably fast(as indicated by change in PV) when I did an open loop test. Therefore, I don't think it is sticky.
bruce varley wrote:

Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
I don't understand how you can relate speed and stiction so readily. They are different things.
Stiction presents as a form of hysteresis. A better open loop test for stiction is perhaps a sinusoidal or bang-bang control signal using a variety of amplitudes. If there's stiction it will be evident in the response. At low amplitudes you may get no response at all.
On the other hand, the response to a giant step will probably overcome the stiction rather immediately and all you will see is the expected speed of response with not much indication of what stiction is doing.
Fred

Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
Fred Marshall wrote:

Valve stiction is sticky valve.

1% open loop step caused a change in 8m3/hr. Sinusoidal or bang-bang control on a furnace pass flow with low flow trip setting will get you a boot out from the control room by the panel operator. This is not a university work but real plant.

In open loop test with PV responding more than a minute or even half a minute is a good indication of sticky valve. This test indicates a PV change in 3 sec after a step on the controller output.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
I guess my response didn't make any sense to you.
I remain of the belief that stiction represents a form of hysteresis no matter what the mechanism. Now, it may be that your sticky valve presents another type of behavior that isn't of that sort. But, it's not something that I can well imagine.
I still don't understand how you can relate speed to stiction. I surely do understand how stiction can affect another kind of speed - as below.
Here is the model that I had in mind:
The valve will turn at some speed when it is already moving. The speed may be proportional to a control or it may be constant if the control is bang-bang or on-off-on ... well, after an initial transient due to its dynamics (e.g. mass, etc.). So, this "speed" is a parameter of the valve and the control and is independent of stiction.
The valve also has stiction that is characterized by the amount of control force necessary to get it to start moving. The force necessary to overcome static friction / stickiness is indepedent of the speed at which the valve will move once the stiction is overcome.
Depending on how the control signal is done, the stiction can affect response times - which may be called "speed" but not the "speed" mentioned above which is an inherent property of the valve. Surely if the control is varying (shall we say in a sinusoidal manner just for discussion?) then the stiction can well affect the valve position.
Perhaps this model is too simple for the situation. But, separating out critical device parameters seems useful to me! Both parameters can be measured if one has the luxury to do so and both parameters can surely be estimated.
If "1% open loop step caused a change in 8m3/hr" then 1% change in the control is sufficient to overcome the stiction - is that right? The question is, what is the maximum step size that is inadequate to cause any change at all? That's the stiction "force" in my simple-minded model here.
Perhaps a simple stiction model is too simple. Perhaps the globe valve has rotating friction that is a function of speed - so it's not a classical static/dynamic friction situation - perhaps it's a gradual changing friction force that goes from static to ever-decreasing dynamic values as a function of speed.
Only you know the limitations imposed by the environment you're working in. Some things can be done on site and other things cannot. A suggestion for how to deal with things may not fit. That's not a reason to label the suggestion as "academic".
Fred

Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
wrote:

Positioners are not recommended on flow-control valves(although they are often fitted without problem). Clearly, if your valve without positioner cannot close against the process pressure at full control signal, this is because the actuator is not providing enough driving force. It was probably a mistake to re-fit the old actuator. An actuator has to be matched to the valve, and your new valve has, presumably because of its design, a different balance of forces then the previous valve.
Kelvin B. Hales Kelvin Hales Associates Limited Consulting Process Control Engineers Web: www.khace.com
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
Correction......
When we change the new valve, we also fitted it with new actuator but maintaining old I/P converter. After we failed to close the valve, we fitted with new I/P converter.
TheRomanov wrote:

Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload

Polytechforum.com is a website by engineers for engineers. It is not affiliated with any of manufacturers or vendors discussed here. All logos and trade names are the property of their respective owners.