Control Loop Case History 109

“I no longer have any valve problems, as we have fitted smart positioners to all our valves!”
This proud statement was made by a senior instrument technician who attended one of my course series a few years ago, but is it true? Are smart positioners the answer to all our valve problems? Are they much better than the old pneumatic-mechanical types that have been around for nearly a century?
The purpose of a positioner is to make sure that the valve opens exactly to the position as demanded by the controller’s output (PD or process demand). The controller itself solves a relatively complicated equation based on various factors including control error which is the difference between setpoint and measurement (PV or process variable). The controller’s output is then demanding that a certain amount of product is inputted into the process. However it can only put out a signal value (normally in a 4-20 mA range), and cannot physically modulate the flow itself of product into the process. This is the purpose of the valve, which must “translate” the PD signal into a physical thing. There is no doubt at all, that the better and more accurately that the valve can achieve this, the better will be the control, and the smaller the control variance.
As mentioned many times in these articles and in the Loop Signature series, at least 80 -85% of all loop problems limiting good control occur in the valve portion of the control loop, largely due to mechanical limitations, and process difficulties such as abrasion, high pressures, fluctuating pressures, etc. Therefore as a general rule, a positioner is normally an essential piece of equipment to try and overcome such problems. However it is a controller in its own right trying to control the valve position, and it is in fact a cascade secondary control in relation to the main controller which is controlling the process.
The advantages and problems associated with cascade controls were discussed in Loop Signature LS1-17 (available on CD for readers outside Southern Africa). Although in general the advantages of cascade control are huge, one of the main disadvantages is that cascade controls are inherently interactive, and normally it is essential that the secondary controller be at least 6 times faster than the primary. In cases of fast processes like flows and small volume pressures, this may not be the case, and instability can easily occur if the positioner is not set-up properly to prevent this.
In the old type of positioners, the manufacturers of the valves usually were very careful to match the positioners to the valves, to ensure stability on fast processes.
The modern “smart “positioner, so named because it contains a computer and employing modern measuring and computing technologies, should certainly work a lot better than the old pneumatic-mechanical type. However it will obviously depend on what the manufacturer puts into it, and how well it is made. Some of the leading makes have some fabulous features available, though many of them are options which cost a lot more money. However basically a good smart positioner should contain at the very least a good valve position measuring transducer, and a reasonable controller. These are of the essence for a positioner.
The old type positioner had a relatively crude measuring system, generally consisting of a position “feedback” arm sensing the actual valve stem’s position, which was then inputted into a pneumatic force balance proportional only controller. Although relatively crude by modern standards, it worked reasonably well and has been used for many decades.
Good makes of smart positioners, will generally firstly have a reasonably accurate position sensing transducers. However in most makes that I have come across, these generally are connected to the actuator’s stem position, as opposed to the actual valve stem’s position. Then secondly, with the power of the computer, they have the possibility of a much more sophisticate controller than the old type of positioner. Again this is dependent on the manufacturer. Some have used a P only controller and others a P + I controller (which is preferable by far).
Now the important points here are:
1.The position signal which may be accurate in itself, may not be truly accurate as far as the actual valve position is concerned, as there may be hysteresis, and/or mechanical deadband (play) in the mechanisms between where the position transducer is actually attached to, and the actual position of the valve’s modulating element, such as a plug or butterfly’s vane.
2.The position controller needs tuning correctly, which as mentioned earlier must also take into account the fact that this is a cascade secondary controller.
These two points seem to be often ignored by users, and where I find that smart positioners often give huge problems. An excellent example of this is illustrated in by the performance of a flow control loop on a submerged arc furnace, which I worked on recently.

Fig. 1

Figure 1 shows a closed loop “as found” test where the loop was tested in automatic with the existing tuning. A setpoint change had been made about a minute after the start of the test. It can be seen from the test that firstly the loop was in a continuous cycle, and secondly the response to the setpoint change the tuning was very slow indeed.

Fig. 2

The open loop test making steps on the controller output shown in Figure 2 is extremely interesting. Three signals were recorded:
a)The controller output (PD).
b)The process variable flow signal (PV).
c)The valve position feedback signal coming from the smart positioner (VP).
It is quite rare that a VP signal is available, but it can be very useful as will be seen below.
One of the first things to note is that the process gain is huge (?PV/?PD), which means that the valve is very oversized. In this case the valve appears to be about 4 or 5 times oversized. Various other things can be also be seen on the test at various points which have been numbered in the figure:
At Point 1: The PD is constant, but the VP suddenly moved, and there was no change in PV. This would mean that there is mechanical play between the true valve position, and the VP as given by the positioner.
At Point 2: The PD was stepped, but neither the flow nor the VP changed immediately. The PD and the VP were quite apart, but the positioner made no effort to reduce the error, which probably means that the controller in the positioner has no integral term. However after about a minute, the VP suddenly moved as did the PV. This could also mean that the valve is likely to be quite sticky..
At Point 3: The PD has been stepped in the opposite direction, and has been followed by the flow and the VP. However although the VP has come back to the position where it was originally (just before the movement at Point 1), which is correct, the flow has dramatically overshot the mark, and gone way down too far. This behaviour is called “negative hysteresis” (refer Loop Signature LS1-7), and could be due to either an undersized actuator or a positioner problem, which is the case here. Negative hysteresis is a very serious phenomenon and may, as in this case, result in uncontrollable cycling, as the valve overshoots every time its is reversed. Another observation that can be made comparing the differences between PD, VP and PV at points 2 and 3 is that there are huge discrepancies between the true valve position as reflected by the flow PV and the VP being measured in the positioner. It really illustrates very well that there are very serious problems with the positioner.
At Point 4 and at subsequent steps in the downward direction: The valve appears to be sticking on each step, and the positioner is trying to get it to move to the correct position, resulting in downward ramps on each step.
At Point 5: The valve is reversed again, and once again we see a huge overshoot in the opposite direction. However on this occasion the flow overshoots and then comes back down a long way, resulting in a big “spike”. This is very bad.
Basically the responses are very non –repeatable with different things occurring every time the PD moves. It can be concluded from this test that no real control can really be achieved until the positioner/valve combination is sorted out. This is borne out by the “final” closed loop test shown in Figure 3 where even with good but relatively slow tuning parameters, it can be seen that the loop is in a continuous cycle caused by the valve overshooting on every reversal.

Fig. 3

A very interesting “stick-slip” type cycle can be seen on the 2 traces of the actual PD (controller output) and VP (valve position feedback signal). The PD is exhibiting a “saw-tooth” wave format caused by the flow controller’s integral term trying to eliminate the error between PV and SP. The VP has a square wave format showing that the positioner is jumping up and down, overshooting, and is unable to get exactly to the position as required by the flow controller. Again it is a good illustration that the positioner is not set-up properly, as it should track the PD.
The amplitude of the cycle on the actual PV is about 5 times bigger than that of the PD which is due to the huge valve oversize. If the process gain had been about unity (with the valve properly sized), then the PV cycle’s amplitude would have been about a fifth of the size.
This is a wonderful example of the fact that smart positioners are not a “be-all to end-all” solution to valve problems, but may actually magnify them and cause other even worse problems if they are not set-up properly.
The figure does however show that the actual control response to the setpoint change has been dramatically improved, due to vastly better tuning. However nothing can be done to stop the cycle, except to fix the equipment.

Fig. 4

I have put in Figure 4 for interest sake only. It shows a closed loop test with reasonable tuning of the furnace roof pressure in a typical submerged arc furnace. The pressure transmitter is ranged ±500 kPa with the setpoint at zero. It shows the typically huge random disturbances going on in the furnace, causing pressure spikes, some of which are bigger that the span of the transmitter. These not only make control very difficult, but can, and often do, cause trips on high or low pressure. All one can do is to try and get as fast a control as possible to keep the average pressure at setpoint, and at the same time to be able to correct for sudden load changes which can occur quite quickly. One cannot filter the signal without killing the control as the spikes are at too low a frequency, Also if you use too much gain in the controller, then apart from the possibility of instability, the output of the controller will jump around excessively, and this can dramatically shorten the valve life.


Index to articles

Michael Brown is a specialist in control loop optimisation, with many years of experience in process control instrumentation. His main activities are consulting, and teaching practical control loop analysis and optimisation. He gives training courses which can be held in clients' plants, where students can have the added benefit of practising on live loops. His work takes him to plants all over South Africa, and also to other countries. He can be contacted at:
Tel (011) 486-0567
Fax (011) 646-2385
E-Mail:  michael.brown@mweb.co.za