Greg: I became sensitized to the need for control valves to respond to small signal changes during my many decades of doing pH applications, as described in my ISA book, Advanced pH Measurement and Control, Third Edition. A resolution limit of 0.1% for even the best control valves could translate to a 2 to 12 pH oscillation in a strong acid and strong base neutralization. Since resolution limits are in percent of valve capacity, this translated to an increasing need for multiple stages of neutralization, with the first stage setpoint on the flattest part of the titration curve, and the largest control valve and last stage on the steepest part with the smallest control valve.
Threatening this capability was the realization that the actuator size stocked was marginal, increasing the resolution limit to 0.4%. Further, the response time of the positioners would increase by more than an order of magnitude for changes of less than 0.4%, and the resolution near the seat more than doubled the resolution limit for even the best throttling valves, causing a corresponding loss of rangeability.
My work over many decades on compressor surge control made me keenly aware of the need for control valves to quickly respond to large signal changes and the first-hand-experienced dangers of replacing a positioner with a volume booster, instead of putting the booster on the positioner output, as documented in my Momentum Press book, Centrifugal and Axial Compressor Control. Large valves with consequentially large actuators need an 86% response time of less than 2 seconds for a 50% change in signal for surge control.
Also, I have repeatedly seen on-off valves posing as throttling valves, where the backlash and resolution exceed 8% as revealed in the Control article, “Is your control valve an imposter?” The error between the closure member position and actuator feedback position (readback) often approached 8%, leading even smart positioners to think everything was fine.
Control valve specifications have an entry for capacity and leakage, but nothing so far about precision and speed of response. Often the imposters are less expensive. The thought that bigger is better, the bid process, tightening of project budgets, and lying positioners have put us in dire straits for the past 50 years. This led me to join the committee for the ANSI/ISA-S75.25.01 standard and ANSI/ISA-TR75.25.02 technical report on how to do control-valve response testing. However, I didn't try to address the many control issues, so that we don’t keep making the same mistakes.
James Beall is the chair of the committee and is going to reopen the reports for possible improvements. James has a history of identifying valve response and its effect on process performance. While at Eastman Chemical, he discovered the tremendous impact of small step sizes on valve response due sensitivity limitations in positioners. He went to work for Emerson in the then newly acquired EnTech Control organization that had a mission of helping all realize that more than 30% of the variability in control loops could be traced back to deficiencies in valve response. EnTech worked with Fisher Controls’ research and development unit.
An EnTech valve response report was the starting point for ISA 75.25. We're also fortunate to have Cullen Langford, original chairman and current vice-chairman of ISA 75.25, joining us with some concluding remarks on how we can more specifically help users get the best responding valve for an application.
I realize that in a technical report we can't tell people what to do, but I was hoping we can alert them to the many problems and possible solutions, leaving it up to them to make the final decisions on final control elements as shown in my Control article, “How to specify valves and positioners that don’t compromise control.”
James, can we do this as an annex to the existing technical report? I've prepared a “Valve Response and Control Loop Performance, TR75.25.02 annex proposal," which considers valve nonlinearities such as backlash, shaft windup, slew rate, dead time and valve flow gain from the installed flow characteristic. This is in addition to the existing nonlinearities noted and associated with dead band resolution and changes in 86% response time and travel gain.
James: As you mention, the ISA 75.25 standard and technical report don't make any recommendations on specific performance metrics per ISA guidelines. However, I think it would be quite appropriate to elaborate about the impact of valve nonlinearities on process/loop performance, and help customers translate process performance requirements into valve-performance requirements.
Greg: We should explain the distinguishing characteristics of dead band originating from resolution, backlash, shaft windup and positioner settings. We should also note that the proper definition of hysteresis (hysteretic error) doesn't include dead band. The existing reports only briefly discuss hunting. What can the committee do to provide an understanding of limit cycles from backlash, shaft windup and resolution, as well as the burst of oscillations for small and large signal changes due to positioner limitations?
James: My plans are to add a section to the technical report (TR) on the impact of nonlinearities. For example, as you pointed out, a non-zero “lost motion” nonlinearity such as backlash in the valve system will produce a limit cycle (aka hunting) when at least two integrating elements are in the loop. A non-zero resolution nonlinearity will produce a limit cycle when there is at least one integrating element in the loop. The integrating elements could be in the PID controller(s), the positioner or the process.
Greg: How can we increase understanding of how integral action in a positioner may increase limit cycles and response time, and that the benefit in gain ratio is not as important?
James: I've seen the integral mode used two different ways in positioners. The first way is where the integral action isn't used as a primary positioning function. It's a slow integral with an error dead band (e.g., 1% travel span) that's used to keep the valve position close to the nominal input signal value. For this case, I've rarely seen it cause limit cycles. In the second case, the integral function is used as a major part of the positioning algorithm in the positioner. For this case, I've seen many cases where it resulted in limit cycles and extended response time, since the positioner gain must be lowered to maintain stability of the positioning algorithm.
Greg: Can we introduce the idea of external-reset feedback, secondary flow loops, and tuning to reduce oscillations and stop some limit cycles?
James: If the smart positioner and control system provide fast and reliable valve position feedback, then external reset feedback and the fast, precise readback of actual valve position can be used to reduce the occurrence of oscillations and limit cycles. If the primary nonlinearity is resolution, this technique will result in an offset between process variable (PV) and setpoint. So, if the offset is less detrimental to the process than the oscillation, this technique will provide benefit. If the primary nonlinearity causing the limit cycle is a slow response time for small signal changes due to positioner limitations or due to lost motion, this technique will actually stop the limit cycle without a PV-SP offset.
Greg: Can we provide an alert about the types of actuators and valves that tend to have the worse resolution, backlash and shaft windup?
James: By design, rotary-type valves and actuators have a greater propensity for these nonlinearities. We discovered many years ago that a valve shaft could actually twist due to the torque required to move the closure member. This twist, or shaft windup, could be as much as 8% of travel span. The loose fit of the closure member to the shaft can add significant backlash to the response.
Valves with high packing or internal-component friction typically have more of these nonlinearities. The larger the difference between the static and sliding friction coefficients, the larger the resolution will be. Valves designed for on-off service and “converted” to throttling service often have features designed for good on-off service but create significant nonlinearities for throttling applications.
With pneumatic actuation, actuators with a smaller force capacity tend to have higher nonlinearities. On the other hand, a larger volume in the actuator might produce a slower step-response time, another valve-system nonlinearity. Motor-driven or hydraulically-operated throttling valves could have an intentional dead band setting that's too high (to reduce motor operation). Oversized valves or using incorrect trim characteristics can result in a high loop gain that amplifies the impact of the valve nonlinearity on the process.
Greg: Can we note sources and consequences of excessive error between the actuator feedback (readback) and closure member?
James: Depending on the valve, actuator, connection, linkage and positioner design, there can be a simple offset or quite a variable difference due to backlash, friction forces and shaft windup, particularly for rotary valves designed for tight shutoff with pinned or keylock connections. This can cause poor performance since the positioner response is based on the position feedback!
Greg: Rotary valves with an integral cast ball or disc stem—together with splined actuator shaft-to-stem connections with minimal friction from seals and packing—greatly reduce shaft windup, leaving the problem to being mostly one of actuator design and method of reading actuator-shaft position.
The most precise pneumatic actuators are diaphragm. The piston actuators with the worst backlash or resolution are link-arm, rack-and-pinion, and scotch yoke. Higher pressure diaphragm actuators are now available, enabling their use on large valves and higher pressure drops. Can we include in the technical report calculations of allowable resolution and lost motion (backlash and shaft windup) as a function of allowable control error, process gain and PID-tuning settings?
James: I recommend that you start with your process performance criteria and use the valve gain and process gain and dynamics to calculate the control valve performance specifications of lost motion, resolution and step-response time in the step-response regions.
Greg: Can we provide calculations to estimate installed flow characteristic from inherent valve characteristic and valve-to-system pressure drop ratio? We should alert users to the need for testing near the closed and open position to identify nonlinearities that are accentuated at these operating points.
James: I have seen some excellent presentations on calculating installed flow characteristics using fairly rigorous techniques including the one by Robert Armstrong documented in your Control Talk column, “Why and how to establish installed valve flow characteristics.” However, you can often get an estimate of the installed characteristics, or at least some guidance on selecting the best trim characteristic, by using the information that you mention.
An example I use to teach control engineers about this topic is the discharge-pressure control on a positive-displacement compressor. In this case, the discharge and inlet pressure are nearly constant, and the pressure drop across the recycle valve is much greater than in the recycle piping.
In this case, the installed flow characteristic is very close to the inherent valve characteristic and a linear trim is the correct choice.
Greg: Can we provide calculations to estimate installed rangeability as a function of installed flow characteristic, oversizing, resolution and backlash?
James: The process impact of valve travel nonlinearities such as resolution and lost motion are amplified by the valve-flow gain. Thus, a high installed flow gain due to the installed flow characteristic and/or an oversized valve increase process variability and reduce rangeability.
Greg: Presently, valve specifications don't have anything to help minimize valve nonlinearities. Existing fields inadvertently lead the user to select a valve with maximum capacity, minimum leakage and minimum pressure drop, often resulting in a rotary valve with tight shutoff, a piston actuator and low valve-to-system pressure drop ratio.
These all lead to valves with the greatest nonlinearities. Chapter 7 on the effect of valve and variable-frequency-drive dynamics in Tuning and Control Loop Performance, Fourth Edition (Momentum Press 2014), offers equations on the consequences of poor choices and some best practices. What can we do, realizing that past inquiries about changing the existing forms for valve specification didn't progress?
Cullen: James and I have helped Greg improve his proposed Annex A. I'm working on an Annex B to demonstrate how valve response affects control loop performance.
10. These are the same valves in the piping spec for isolation
9. Tight shutoff is so great; I dare you to try and open them
8. The valves have so much future capacity capability, you barely need to open them
7. We went with a quick opening flow characteristic to speed up the response
6. To make sure our loops are really fast; we went with boosters instead of positioners
5. We standardized on graphoil packing
4. To eliminate any possible emissions, all packings are extremely tight
3. To save energy, the valve pressure drop is always less than 5% of the system pressure drop
2. The actuators are sized to minimize cost
1. The valves are so cheap; we can afford to spend more on IIoT