Features Research & Innovation
VALVES: Control valve performance update — The Last 10 Years

Nearly 10 years ago, EnTech process variability audits in pulp and paper mills identified that sub-standard pneumatic control valves were undoubtedly the single biggest cause of process variability in the industry. At that time, many control valves were unable to track their input signals to anything better than 2% or worse still, 5% or 10%. When a controller moves its output, it expects the flow through the control valve to change by a predictable amount in a predictable time. This expected behaviour is reflected in the controller tuning parameters. If the flow does not change by the expected amount on time, the controller will make additional adjustments in the hope that these will induce the needed flow change. Due to its mechanical nature, the control valve can neither respond instantaneously nor in a completely predictable way. It seems to move when it “feels like it” and, in turn, this delay causes the controller to make unnecessary additional adjustments which eventually result in overreaction and cycling, known as a “limit cycling”. Figure 1 shows typical limit cycles as they appeared in 1991 (left) and now in 2001 (right). The good news is that the cycles are smaller by a factor of five or more since 1991. This remarkable improvement starts with the EnTech Control Valve Dynamic Specification of 1992.

Control Valve Dynamic Specification

In 1992, EnTech wrote its first Control Valve Dynamic Specification and issued this freely to the industry. The specification required that the combined non-linear response of the control valve be inside a 1% envelope or plus/minus 0.5%. Equally important, the speed of response of the control valve should be within specified limits. The intention was to weed out the valves with really poor performance. The valve specification was readily accepted by the industry and paper companies used it to judge in situ valve performance as well as for a guide for valve purchasing. The specification created a challenge and a common language to address the issues. The major valve suppliers took the valve specification seriously and initiated intensive testing programs of their valves against compliance to specification. They built extensive lab facilities to carry out this testing. EnTech was frequently retained to certify that these labs were technically capable of testing control valves to within the tolerances needed by the specification. In the mid-1990s, a number of valve suppliers made marketing claims that their valves were “EnTech Compliant”. In the meantime, the testing programs started to pay off, as design deficiencies were uncovered. This led to design changes that reduced friction, backlash, shaft twist and other performance problems. It became evident that the problems were inherent to the materials and manufacturing methods and at best these various effects could be reduced in amplitude but not eliminated. To further optimize performance, digital valve positioners were developed by several valve manufacturers in an attempt to further improve performance. By 1997, word of the valve specification had spread to other industries. In response, ISA (originally the Instrument Society of America, and now The Instrumentation, Systems, and Automation Society) convened the S75.25 sub-committee to draft a new ISA standard for control valve response to small signal changes based on the EnTech specification. The new S75.25 standard was issued by ISA in 2000 for use by all of the process industries.

Why the limit cycle?

The 1991 example of Fig. 1 shows a flow loop limit cycle caused by a pneumatically actuated ball valve with a diaphragm actuator, spring return and a pneumatic two-stage positioner. The controller output limit cycle is about 2% and represents the tracking ability of this control valve. For 1991, this was a pretty good valve!

To understand these problems, the control valve has to be seen as a “system” — a feedback loop in its own right, consisting of control valve, actuator and (if installed) positioner. To cause the fluid flow to actually change, the valve plug or trim must move and to make this happen the valve drive train must first overcome many opposing forces, all of which are highly non-linear. For instance, the valve packing causes static friction resulting in “stiction” — a tendency to stick and then slip. This causes the valve to make sudden jumps when moving in the same direction. In addition, static friction causes an extended dead time, which is another significant cause of limit cycling.

In rotary valves, there is possible lost motion at each joint connection, as well as twist in the shaft connecting the actuator to the rotary plug. The result is backlash, or deadband, that requires additional motion when changing direction of travel. Combined, all of these effects easily exceeded 2% or even 5% 10 years ago in many valves.

Valve sizing and over sized valves

Even though the 1991 valve limit cycle was 2% in Fig. 1, the resulting flow limit cycle was about 4.5%. The ratio of these two numbers is about 2.2 and represents the sensitivity of the control valve — its “process gain”, which is a direct result of valve sizing and trim characteristic selection. For the same flow range, the bigger the valve the greater the process gain.

Valve trim selection is also critical, and an inappropriate trim can change the process gain by a factor of 10 or more. Valve sizing is the domain of the capital project design engineer. There is a tendency to make absolutely certain that the valve is “big enough” to handle any and all future flow requirements. There is an old saying about valve sizing: “No one ever got fired for making the valve bigger than needed!” And here is the point of conflict between this design practice and the process variability sensitivity that EnTech process audits bring. It is entirely possible that the extra big valve with a high process gain will create so much variability that it will be impossible to manufacture salable product at the present production rate let alone to allow a speed-up! The EnTech valve specification sets a value of 1.0 as the ideal value for the process gain. Bigger values increasingly magnify the control valve’s inherent tracking problems.

Today’s performance

The right hand side of Fig. 1 shows typical results for 2001 for the same style of valve as in the 1991 example, but with the improvements of the last 10 years. The control valve limit cycle has been reduced to a mere 0.3% — a factor of six reduction. This is a direct result of better mechanical design, as well as improvements stemming from the use of a digital valve positioner. The flow limit cycle is now about 0.8%. Even though this is five times smaller than the 1991 result, the ratio of 0.8 to 0.3 reveals a process gain of 2.7. The interpretation is that, while the valve manufacturers have done a superb job of improving the control valve’s tracking ability, the valve sizing engineers are still sizing control valves the same old way by making sure that the valves are plenty big enough; evidently no one has been fired! If the process gain on the right side of Fig. 1 could have been reduced to 1.0 from 2.7 by selecting a smaller valve or a more appropriate valve trim characteristic, then the flow limit cycle would have been only 0.3% — an improvement over 1991 by a factor of 15!

Digital valve positioner

The advent of the digital valve positioner has been significant in allowing on-board intelligence and diagnostics to be available. It has also allowed tuning parameters in the valve positioner to directly affect the valve response. This has been a good thing but it also provides opportunities for problems. The digital positioner must be properly tuned to achieve the desired response. To minimize process variability the objective should be to optimize the small amplitude response, which requires a trade-off between achieving a small dead band, a short dead time while at the same time minimizing overshoot. In some cases the person installing the positioner can overlook these issues and inadvertently cause the digital positioner to become a new so
urce of limit cycling and process variability.

Conclusion

Control valve dynamic performance on average has improved by a factor of at least five in the last 10 years. The EnTech Control Valve Dynamic Specification was the stimulus to drive this change. The improvements came about as a result of much better mechanical design and potentially better performing digital valve positioners. Even better performance could be achieved if over-sized control valves could be avoided, proper trim characteristics were selected and if digital valve positioners were optimally tuned.P&PC

Bill Bialkowski, Mark Coughran and James Beall are with the EnTech Division of Emerson Process Management. Bialkowski is the founder of EnTech and the author of the EnTech Control Valve Specification. Coughran came to EnTech from Fisher Controls where he was instrumental in dynamic testing of control valves. Beall recently joined EnTech from Eastman Chemical. He chairs the ISA S75.25 committee and was instrumental in generating the ISA S75.25 standard for control valve dynamic response.