Q: I am an instrumentation engineer working for a EPC company in Dubai, UAE. I follow your articles closely. I have a query about one of your replies to the query on the subject of direct and reverse controller action. My understanding is that the controller action should be decided based on the control valve failure position, plus the process requirements, while the control valve failure action position is decided independently, considering only the safety requirements. I am still skeptical about the controller action. Can you please review my paper and help me? I tried to get this concept from books and the Internet, but nowhere is it complete.
China Petroleum Engineering Co. Ltd
Q: I always enjoy your articles. You are so well-versed in process control. I enjoyed your last article concerning controllers, but am somewhat confused on one aspect and seek further information. You mentioned that you don't like to use positioners on flow control because the positioners can be slower than the process flow. When I don't use a positioner and just use an I/P converter, the I/P is much slower acting than a positioner in terms of how the two devices function on a spring-and-diaphragm (S&D) actuator. Typically a piston actuator will take much more air pressure than an S&D, and there doesn't seem to be an issue there. I would enjoy your follow-up in further explanation.
Through the years, you have been my primary source for process control answers. Thank you for sharing your knowledge.
A: Loop Action: The needs of the process determines if the control loop should be direct- or reverse-acting. For example, in the case of a steam-heated water heater, a rise in hot water temperature requires a reduction in steam flow, hence, the loop has to be reverse-acting. In the case of a cooler, a rise in the temperature of the process fluid means that the cooling water flow must also rise, so the loop has to be direct-acting. The loop action can be achieved by any of the loop components (transmitter, converter, controller, etc.). For sake of simplicity, I like to keep all loop components direct-acting and let the controller action alone determine the loop action.
Valve Failure Position: The failure position of the valve (closed, open or last position) has nothing to do with loop action. It has to do with safety. In other words, it makes no differernce if the controller action is direct or reverse, the valve failure position is that which makes the process safe when air supply (or other energy supply) fails. For example, in case of a cooling system, safety is served by the valve failing open. In critical applications, such as the cooling water feed valve to a nuclar reactor, the valve must fail open not only if the air supply is lost, but also in response to all failures that otherwise could cause overheating.
Speed and Accuracy: For good (stable) control. the valve has to take the position that the controller demands and do it faster than the measurement of the controller can change. To achieve these dual goals (stability and speed), two things are needed: First, an actuator that is strong enough to overcome valve sticking, nonlinearity and changes in valve pressure drops; and second, a positioner that can put the valve in the required position quickly. So let's look at the actuator and the positioner.
The Actuator: The force generated by the actuator (pneumatic, hydraulic, etc.) must be greater than all forces that are resisting stem movement (stem sticking, process forces on the plug, etc). All actuators are both "velocity and force limited." Pneumatic ones are "velocity limited" due to the limitation in the air flow and "force limited" because they can only increase the actuating force by the size of the actuator and the air supply pressure, which varies from 2 to 7 bars (30 to 105 psig).
The Positioner: The positioner is a "cascade slave controller." It measures the stem position of the valve and throttles the air flow to the valve's actuator. The setpoint of the positioner is the output of the controller (cascade master). The positioner's job is to change the valve opening quickly and accurately. All slave controllers must be about five times faster than the speed at which the measurement of their cascade master can change. This is essential so that the valve will not cycle and will not be late in responding to the cascade master or to close or to open. Therefore, when the loop is controlling liquid flow or liquid pressure, the pneumatic positioners can be too slow. In this case, the old remedy often was to remove the positioner and replace it with a large-capacity air volume booster. In other words, the solution was to sacrifice control quality for loop stability.
Digital Positioner: Today, most vendors say that digital positioners are faster and, therefore, can be used even on liquid flow applications. Unfortunately, this is not necessarily the case. It is true that digital positioners are faster than pneumatic in calculating the required pressure to be sent to the actuator, but the speed of moving the inner valve is still a function of the air flow, and the force generated is still a function of actuator area and air pressure.
Typically, the maximum air supply pressure to a digital positioner (Figure 1) is 105 PSIG, and the maximum air flow (Fa) is 12 SCFM, which at 105 PSIG is 1.7 ACFM. Therefore, the time (t) required to fully stroke the valve (0% to 100%) is a function of the actuator volume (V) divided by (Fa), which in the above example was 1.7 ACFM:
t (in minutes) = V/Fa
and the valve speed (Sv) in % stroke/minutes is
Sv = t/100
Therefore, if the air flow can't be increased, the maximum speed at which the valve stem can move Sv (in % per minutes) is:
Now, because in a cascade loop the valve speed (Sv) must be at least five times faster than the speed at which the measurement can move (Sm):
Sv << Sm
No positioner can be used (digital or not) if one wants stability.
Added to the time the positioner takes to move the valve is the time which it takes for the DCS to execute the PID algorithm, which in case of analog electronic controllers was 0.03 seconds, while in most DCS systems it is 0.5 seconds. So what do we do to properly control such fast processes? We simply replace the pneumatic valve actuators with electro-hydraulic or variable-frequency drives, and in the most demanding cases, we might even bypass the DCS.
A: It's not that complicated. Fail-safe position prevails. Then the positioner and DCS are configured, so what the operator sees with the process and the effect of the final control element matches what is happening in the field.
A: A traditional IP controller sends position instructions to the valve as to what to do. The problem is, via traditional feedback control, we don't learn how the valve responds until the controller receives the PV response.
Now consider the positioner. The controller sends out an instruction to the positioner. The positioner sends an instruction to the valve, then receives feedback from the valve that the instruction has been received. This is classical to feedback control, but this is information only to the positioner. How does this relate to the primary controller?
The answer is that the positioner acts as a cascade inner loop to the primary controller. The benefits of this solution are that the positioner verifies that the valve position is correct, and the positioner acts to correct any errors in valve performance.
To optimize performance, it is important that the positioner, as the inner loop, react at least three to five times faster that the controller that is issuing the control solution. For optimal performance, you need to tune the inner loop (the positioner) prior to tuning the outer loop, the controller.
A: In the days of pneumatic positioners and electronic controllers the cascade problem of the secondary inner loop (valve position) being too slow compared to the outer loop was a concern.
Today, the ability to tune a digital valve positioner aggressively and use volume boosters on the positioner output and the slower PID execution rate (e.g. 0.5 sec) in a distributed control system (DCS) compared to an electronic analog controller with an equivalent execution rate of 0.03 sec, and the primary use of integral rather than proportional action in flow loops means that the cascade rule is much less of a problem than the theoretical studies in the 1970s concluded. In fact the proposed solution at that time of using a booster instead of a positioner on fast loops has led to unsafe conditions where surge valves and vent valves have slammed shut due to the positive feedback (from combination of extreme booster output sensitivity and change in actuator volume with pressure due to diaphragm flexure), and the lack of negative feedback by a positioner. I have personally experienced this instability from the omission of a positioner.
If a valve is too slow for cascade control, you should be using a variable-frequency drive (VFD) with high-resolution cards, pulse width modulation, no dead band, minimal speed rate of change limiting, and fast speed control in the field (speed control in the DCS makes this secondary loop too slow). See my blog on how to get the most out of your variable frequency drive via link below:
Liquid pressure loops and some furnace pressure loops require the fast response of a properly designed a VFD.
The stick-slip from control valve stem and sealing friction is an order of magnitude larger for valves without positioners. The response of the pressure in a diaphragm actuator is integrating. Putting a positioner on the valve adds sufficient negative feedback action to make the actuator response self-regulating. Without a positioner the valve position could differ from the valve signal by 5% to 25% depending on friction and the accuracy of bench settings. Correspondingly, the valve may not be fully closed at 0% signal and may not open until the valve signal is 5% to 25%. Finally, diagnostics in the smart digital valve positioner (e.g., digital valve controller) enable the user to know what type of maintenance is needed. In my book, all valves should have positioners, and the valve should be originally designed for throttling service. There are a lot of on-off valves posing as control valves with excessive backlash and sealing/seating friction and non-representative positioner feedback. For more on this subject see my article and blog on the Control Global site via links below:
If your valve is an imposter, putting a positioner on the valve won't solve the problem. If you have any questions, send me an email.
Greg K McMillan
A: Thee positioner should be the same type of action as the controller, and if the valve is fail-closed, and the actuator is direct-acting, then the positioner should be direct-acting to promote the fail-safe condition, not the reverse.
The level of a tank is measured and controlled to maintain a correct level and guarantee that the tank does not overflow. It may require a valve with a fail-open to drain the tank, but have the positioner and controller working on a direct action so that as the level increases, the level decreases in the tank.
The instrument engineer has to select the safest valve failure position. It is also desirable to configure the system such that the up and downstream process pressures help to keep the valve in its safe failure position. It is at this point that the direct or reverse acting action of the valve itself is chosen.
Place a positioner with direct- or reverse-acting actions on it. Now then remember that the true function of a positioner is to have a better control of the valve action by offsetting higher pressures on the valve plug and getting the valve to the desired control point and maintaining it there more constantly. Also a positioner allows you to linearize the action of a valve.
You correctly indicate that you can have fail-open and closed conditions where the mechanism that guarantees the fail-safe condition may be direct- or reverse-acting. Here you generally want to have the action of the spring or other mechanism working with gravity and line pressure to go to the fail position and guarantee that it's maintained there. Also you correctly indicate that a valve may have a positioner with direct or reverse acting actions on them. Now then, remember that the true function of a positioner is to have a better control of the valve action by offsetting higher pressures on the valve plug and getting the valve to the desired control point and maintaining it there more constantly. Also a positioner allows you to linearize the action of a valve, allowing for better control, an action which many years ago was performed by cam operated positioners.
If a positioner is needed, then the choice of reverse- or direct-acting will depend on the valve fail action. For example if the valve is fail-closed, and the actuator is direct-acting, then the positioner should be direct-acting to promote the fail-safe condition, not the reverse.
The controller is generally defined by the process group and is set up so that the control action will cause the valve to open and close, guaranteeing the proper process actions.
In a control loop where the level of a tank is measured and controlled to maintain a correct level and guarantee that the tank does not overflow, you may require a valve with a fail-open to drain the tank, but have the positioner and controller working on a direct action so that as the level increases, the level decreases in the tank. When a failure occurs the tank will drain, yet during normal operation, the tank will maintain the level inside the tank at the desired level. If on the other hand, the positioner is required for strict critical control of the valve position, the positioner should be the same type of action as the controller to guarantee that the control philosophy is maintained.
A: You are making this too far complicated. Control direction for modern systems does not depend on the fail open/close direction of the valve. This was something that had to be done in the past when using pneumatic controllers.
Control action is based on the whether the valve has to open or close (i.e., go towards 100% or 0% if the process variable increases). For example, flow control with a flow transmitter and valve in series is always reverse-acting. In other words, you just need to consider whether the valve needs to open or close and not what the signal needs to do.
The analog output block of a modern control system has configuration selection for air to systems usually provide means to select the fail-closed or fail-open configuration for the valve or provide means to select if the failure position is to be taken on the failure of the air supply or the failure of the control signal. This could not be done this way with pneumatic controllers (i.e., there was no output block).
It is very bad practice to set up a valve positioner as reverse-acting. You may think of circumstances to do this, but there would have to be some very, very strong reasons for doing so. There were strange things done in the days of pneumatic control, but this did cause some real safety issues.
Example: flow control with flow transmitter and valve in series is always reverse-acting. In other words, you just need to consider whether the valve needs to open or close and not what the signal needs to do.
A: Most modern distributed control systems are designed to allow the PID direct/reverse action to be configured independent of whether the valve is fail-open or fail-closed. This is because in the output block, the ability is provided to address reversal of the current output if the valve is fail-open. Thus, the operator can always think of the PID output in terms of implied valve position. Also, in defining direct/reverse action, you only need to think in terms of implied valve position. Below is some information on this taken from Chapter 11 of the book Control Loop Foundation – Batch and Continuous Processes, available as an e-book in Chinese for both the Kindle and iPad. The website that goes with the book contains workshops that you may find of help if you work with control systems – see http://www.controlloopfoundation.com/.
The valve manufacturers' guidelines on when to use a positioner versus an I-to-P transducer have changed dramatically over the last 40 years. In the early 1970s, many manufacturers of valves recommended that valve positioners be used with only a limited number of applications. Plants constructed in that time frame may still have valves that do not have a positioner. However, based on field experience, the recommendation today is a positioner should always be used with a valve. There are very few installations that would not benefit from all valves being installed with a positioner.
The I-to-P has no means of automatically adjusting actuator pressure to maintain the position requested by the current signal from the control system, since it does not use a measurement of stem position. If the valve is sticking, then it may not move as requested in response to a change in the control system input. Therefore, a positioner should be installed if the valve is to be used in a control application.
When a valve positioner has not been installed on a valve, the control system may interface to the actuator through an I/P transducer. In the case of a sticky valve, the pressure change provided by the I/P may not initially be enough to cause the valve stem position to change, and multiple changes in the control system output and to the pressure delivered to the actuator may be required to move the stem. Once the stem begins to move, less force will be required, and the actuator pressure may now exceed the pressure that is needed to achieve the desired position. As a result, a larger than intended change in flow may be observed. This nonlinear response associated with a valve actuator or damper drive, ( that is, the controlled parameter does not reflect changes in the input to the manipulated parameter), can significantly degrade control performance and make it difficult to commission the control system.
When a valve positioner is installed on the actuator, as long as it is functioning properly, it will automatically increase the actuator pressure until the valve moves to a position that matches that requested by the control system. Feedback of the stem position is used by the positioner to sense if the valve has responded. Thus, it is possible for the valve position to eliminate overshoot or offset.
When a control loop is placed in automatic control, it is easy to detect if a valve or damper is not responding to the control system by observing the response of the controlled parameter to control system changes in the PID output. In the absence of a valve positioner, a significant change in the setpoint may cause a change in the control system output, as shown by the PID output that is large enough to generate the actuator pressure needed to move the valve stem, but result in the valve overshooting the target position and introduce cycling in the control. Small output changes made by the control system for small deviations in the controlled parameter from setpoint must accumulate before enough force is developed to move the valve. When the valve moves in response to accumulated changes in PID output, then a large change in the controlled parameter may be observed. The change in the control system output in response to this large change in the controlled parameter may be enough to again cause the valve to move. Thus, a sustained, characteristic cyclic pattern will develop in the controlled parameter and the manipulated parameter. The amplitude and period of the cyclic pattern will vary depending on the controller tuning and the force needed to cause the valve to move. The cyclic behavior caused by a sticky valve cannot be eliminated through tuning. Changes in tuning will only impact the period of the cycle that develops. The only way to eliminate this type of behavior is to install a valve positioner.
A: I have received a copy of your letter from Mr. Béla Lipták and was impressed immediately by the extent of your work. You have put together, mostly correct, a lot of important issues. Well done!
There are few points where, as you have mentioned yourself, that you are still confused and I shall try to lighten these dark topics.
One more personal remark before plunging into the subject: you have put most of the things correct and if I repeat some of your sayings it is more for me then trying to correct your (excellent) work.
In the preparation of my reply to you, I started with the definition of direct/indirect control action and the related matters as set by Gregory K. McMillan in Good Tuning: A Pocket Guide, 2nd edition, , ISA, p.5.
First of all, I tend to approach this matter (as well as any other in our field) from the process point of view. We have to make the process act and behave the way it has to and the instrumentation is there to help us achieve this goal.
We have to look at the whole system of which the controller, the control valve, the positioner and whatever other hardware, are only a part.
Let's follow the process of setting up a control loop.
The process exists, having its own characteristics and requirements.
Based on process and safety consideration we chose the control valve to be fail-open/fail-close (or none!), which also indicates what is the right actuator and positioner for the job.
The very first settings that must be right are the controller action mode and the valve actions. If these settings are not right, nothing else matters. The control loop will not function properly!
Based on the process behavior, the controller action should behave in the opposite way, to be able to provide feedback correction. A direct-acting process is one in which the direction of the change in the process variable is the same as the direction of the change in the manipulated variable and vice versa. The manipulated variable, the output of the controller, is translated to be valve position that generates flow, the setpoint of a slave loop in a cascade control system, etc.
The controller action should be the opposite of the process action unless there is an increase-to-close (fail-open) control valve for which there is no reversal of the valve signal.
The valve action determines whether a 100% output signal corresponds to a wide open or a fully closed valve. It also determines the direction of a change in the actual signal to the control valve when there is a change in the controller's output.
From the instrumentation point of view, the signal can be reversed in several locations along its route from the controller (output) to the control valve, and you should decide what best suits your needs and availabilities.
Table 1 (from Good Tuning) summarizes how the controller action depends upon both the process and valve actions and on the signal reversal.
This is a device to ensure the control valve executes exactly what it it is told to do. This is a controller (usualy a PID) whose PV is the valve position; its SP is the controller (ours) output, and its manipulated variable is the air pressure within the controller. The positioner will be direct acting if the control valve is "air to open" (fail-close) and reverse acting one if the valve is "air to close" (fail-open).
I hope I was able to set some order into your mind (and thanks to you – into mine as well).
Please fill free to contact me direct for any of these matters or any other within our mutual area of activities.
A: In our experience, that's true that positioner has characteristic like a simple controller; inside the positioner there is button for damping and gain. The buttons are adjustable. We must set the gain and damping so that the positioner respond will satisfy with our process.
When we do not use any positioner, (depend on I/P converter), there is limitation on its (I/P converter's) flow rate of instrument air to actuator. On the other side, without a positioner, maybe we will need bigger size of actuator due to limitation on an I/P converter output pressure.
So in our opinion, it's applicable to use positioner on the flow control application. With right setting on the positioner, the actuator size physically will be smaller because we can use a higher pressure of instrument air (in our plant up to 5 barg), and the valve responce will be faster because I/A flow rate from positioner to the actuator will be higher than I/A flow rate from I/P converter. When the setting of positioner is not correct, and the positioner becomes slower than the master (flow system), the valve will start hunting.
A: We use positioners for flow everywhere. Back in the days of pneumatics, our specifications used to say "Use I/P only or volume booster for flow applications." This is because the all-pneumatic positioner was slow and sloppy. The time constant of the positioner could be greater than the time constant of the process (flow) so cycling/instability was common.
With digital positioners ("smart" positioners) the limiting factor becomes the actuator and valve. How fast can it move? The "dominant time constant" of the final element becomes that of the diaphragm, stem,and valve trim, which is the same regardless of whether a positioner is used. So I can't think of an instance where you wouldn't be better off with a positioner, aside from the all-pneumatic case.
Our plant has a flow application where operations wants to purge a very small amoun—a fraction of a GPM—of a polymerizing material from a distillation column. The valve has to be very wide-ranging so the "chunks" can be washed out periodically, but able to control at very low rates so we don't throw away much useful product. A characterized-seat ball valve would be a better solution, but this application runs today with a rising stem control valve, which normally is throttling in 0 to 1% of its roughly ¾-in. travel. The flow loop is in "automatic" and controlling ±0.05 GPM. You shouldn't be able to do this, and my valve supplier would rather I bought a new valve, but it meets the requirement thanks to the precision of the digital (i.e., microprocessor-based) positioner.
Digital positioners aren't the cheapest option, and you will certainly find applications where perhaps there's no easy access to reliable power, and you have to use pneumatics—in these cases you are better off without a positioner for flow loops.