The elements of a control loop normally consist of the process itself, the measuring transmitter, the controller, and the final control element. The latter in the vast majority of loops consists of an I/P (current to pneumatic) converter, and a pneumatically actuated valve which is generally fitted with a positioner.

Good modern transmitters are engineering masterpieces based on decades of scientific research and knowledge, and enhanced with the latest in microcomputer technology. They are generally reasonably, if not highly, accurate, and usually very reliable.

Most PID controllers today are simply blocks of software inside a modern computer which operates extremely accurately and efficiently, and the controller software is based on fantastic feedback control theory perfected by 3 of the world’s greatest mathematicians about 80 years ago. The controller compares the desired operating setpoint (SP) as set by the Operator, or by a higher level advance controller, against the actual process variable (PV), and performs a mathematical calculation, the result of which sets the value or amount of product that should be fed into the control loop to bring the PV to setpoint, and result in zero error.

Unfortunately the controller can only output this result (PD) as an arithmetic value normally given in percentage terms between 0 and 100%. This by itself can obviously not control a physical thing like a flow of a product. Therefore the final item in the control loop, sometimes called the final control element, is used to translate the PD value into the control of a physical phenomenon, such as a flow of the product. The final control element, which will be referred to as the valve hereafter, should be thought of as a ‘translator’ or slave element which must do the controller’s bidding. This leads to the maxim: “The more accurately that the valve carries out the controller’s demand, the better will be the control”.

Herein lies the rub! Whreas the transmitter and the controller are items operating close to perfection, the valve is normally a mechanical device. Mechanical devices can be a host to all sorts of problems including friction, excessive stick-slip, hysteresis, non-linear installed characteristics, corrosion, insufficient actuation power, limited rangeability, and many more things. Even on more sophisticated control elements, such as variable speed pumps for example; there are a whole bunch of physical and inherent problems which limit their performance. As a result about 80% of problems associated with control loops are found to fall within the valve area of the loop. Typically a full third of my 3 day control course is devoted to problems with final control elements, and the diagnosis thereof.

Many of these problems can be dealt with fairly simply by techniques which can be incorporated into the control philosophy and sometimes into the tuning. In fact in most cases, it is not always necessary to shut-down the loop in order to have to physically work on the valve. These techniques can generally allow the loop to carry on operating, until a more convenient time arises to make proper repairs to the actual valve.

Unfortunately in certain cases there are problems where the valve is so bad, that good control becomes impossible. In these cases it is necessary to fix or replace the valve. The two worst cases are firstly when the valve is non-repeatable, and basically does “its own thing” irrespective of the controller’s dictates; and secondly, if it causes cycling. Two actual cases of really bad valves like this are described in this article.

Fig.1

The first example shows one of the worst valves I have encountered in a plant. Figure 1 is a recording of a closed loop test performed on a flow control loop. The flow is at setpoint. Then a step change of setpoint is made. There is a relatively long period where the PV doesn’t change, because the valve is stuck, and the graph of the PD shows how the controller’s integral action is ramping up the output. Eventually the valve breaks away, and the flow comes up to fairly close to setpoint. It stays there for a while, looking quite good, then suddenly, and for no apparent reason the valve jumps down, and the flow drops. The controller tries to correct and you can see the PD ramping up again. The valve sticks for a while, and then breaks free and jumps too much, so the flow goes up way past setpoint. The controller tries to bring it back, but then the valve goes unstable, jumping up and down.

Fig. 2

An open loop test was then performed and is shown in Figure 2. It shows some remarkable things:

a) Steps of the same size on the PD result in steps on the PV the size of which differ frequently, being particularly different when the valve is opening as to when the valve is closing.

b) On many steps, but not on all, there is a huge overshoot and recovery (sometimes much bigger than the actual step size) when the valve moves.

c) The overall changes in PV compared with the overall changes made on the PD are much larger than unity, indicating that the valve is probably largely oversized.

It was later found that the positioner on the valve, which was a modern ‘smart’ positioner, had never been set-up properly. A representative from the manufacturer was called in to do this, and good control could then be attained.

This is an interesting point, for I have come across many people who fully believe that all their valve problems will be eliminated if they use smart positioners. Whilst I fully agree that the a good modern smart positioner should operate much better and more accurately than the old type of mechanical-pneumatic type of positioners, which have been around for many decades, the modern units need proper commissioning. One must always remember that they contain their own feedback (usually P+I) positioning controllers in them which act as a cascade secondary to the main loop controller, and these must be properly tuned, taking the rules of tuning cascade secondary controllers into account.

Fig. 3

The second example is also of flow loop where the valve was very bad. The closed loop test displayed in Figure 3 shows the operation with original “as-found” tuning with a step change of setpoint being made. The first thing that can be seen is that the tuning is woefully slow for a flow control, with the integral setting being too big (i.e. too slow). However when watching the response of a fast process like a flow loop, one of the advantages of too slow an integral is that the slow ramp allows one to get a good idea of how well the valve follows the PD. In this case one can see that the valve is very sticky and then it slips and sticks again. In this case it is very bad.

Fig. 4

Figure 4 shows the open loop test on the same loop. Various steps of the same size had been made on the PD. It can be seen that the PV did not respond consistently, with some steps being of different sizes, and in a couple of cases there was some overshoot. This again shows non-repeatability and non-consistency of valve response, and can only cause problems with the control. However it is not nearly as bad as the valve in the first example, so it might be possible to get more or less acceptable control using better tuning with a small proportional gain and relatively fast integral set as close as possible to the dominant time constant of the process. The idea behind this is to get the valve moving smoothly and as ‘bumplessly’ as possible when control deviations exist.

Fig. 5

Figure 5 shows a final closed loop test with the setpoint being held constant. The PV is pretty close to setpoint with a strange little “stick-stick” cycle occurring on it. This is quite interesting. What is happening is that the valve sticks close to setpoint. One can see on the PD that the controller output is ramping under the integral action to get the valve away, but once it moves it overshoots, and the controller moves the valve back down again. This is repeated ad-nauseam. Provided that there is no interaction with another loop, this little cycle is not serious, and will not do any harm.

In both of the above cases nobody in the plants had diagnosed the problems as valve related, and people were trying to sort out the problems by playing with controller tunings.

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