Effect of Valve and VSD Dynamics on Loop Performance Tips
The control valve and variable speed drive are the final control elements that directly affect the process by manipulating a flow. The expectation is that these elements do their job and do not adversely affect the tuning and performance of the loop. Most control text books do not include details of the final control elements in the analysis or solution. Until recently, most users were not aware of the potential problems. Here is a perspective, overview, and recommendations.
Prior to the 1980s, nearly all final control elements were control valves supplied by throttling valve manufacturers. Since these valves were originally designed for throttling service, backlash and stick-slip were minimal. Actuators and positioners with excellent threshold sensitivity were generally used. Since energy conservation and the possible use of variable speed drives were less of a motivation, the valve to system pressure drop ratio was large enough to provide a good installed flow characteristic and rangeability. The downside was that the positioners being pneumatic devices were out of calibration typically resulting in an offset between actual position and implied valve position based on the valve signal. Since there was no readback, the user was typically unaware of positioner problems. Fortunately, the main problem being simply an offset was automatically corrected by the control loop manipulating the valve to reach setpoint.
These control valves with a throttling valve heritage typically had a deadband less than 0.4%, stick-slip less than 0.2%, and an actuator-positioner resolution or threshold sensitivity of better than 0.1%. Unfortunately, this capability was taken for granted and never put on the valve specification as a requirement. The lack of understanding of valve dynamics and the missing link in terms of readback set the scenario for a disaster in valve performance for decades except where company and plant standards required the time proven valve solution despite a higher price tag.
The typical control valve specification had sizing information and a leakage specification. There was no information on installed characteristic or valve dynamics. Since the main goal was making sure the valve could pass the required flow and stop the flow as completely as possible, valves designed by on-off valve (e.g. piping and isolation valve) manufacturers met the specification with a significantly lower price tag. In many cases the valves were already in the piping spec and being used manually by operators or automatically by batch sequences and safety instrumentation systems (SIS).
The on-off valves had excessive deadband (e.g. 8%) from backlash associated with links and connections between shafts, stems, and internal flow elements (e.g. balls and discs). These valves also had excessive stick-slip (e.g. 4%) from the high seating-sealing friction and stem friction from packing designed for high temperatures and lower emissions. Poor resolution and threshold sensitivity (e.g. 2%) also originated in the accessories due to the use of piston actuators with O-rings or gear teeth instead of diaphragm actuators, high volume spool positioners and volume boosters instead of high gain relay positioners. Rotary valves were used with a nearly quick opening flow characteristic because these were a very low cost solution high capacity solution particularly attractive in larger line sizes. The steep slope of the flow characteristic near the closed position where the stick-slip was the highest from high seating-sealing friction associated with tight shutoff, caused larger amplitude limit cycles. The term "high performance" was often used for these valves because performance was judged in terms of tightness of shutoff and capacity. Who wouldn't want a "high performance" valve at a much lower cost?
Some applications tried to take advantage of the control valve having low leakage by making it serve the dual purpose of isolation and throttling. The lack of position readback and the standard testing procedure of making 25% or 50% changes and eyeballing the stem position did not show the problem created by the use of on-off valves. The limit cycles in the process were often attributed to some other problem.
While awareness has improved by the ISA standards Test Procedure for Control Valve Response Measurement from Step Inputs (ISA-75.25.01) and Control Valve Response Measurement from Step Inputs (ISA-75.25.02), many users still can get into trouble because there is very little response data published. What data does exist is from tests at mid position with hand tight packing to eliminate sealing and seating friction and minimize stem friction. Furthermore, original throttling valve manufacturers seeing the profits generated by on-off valve manufacturers have bought on-off valve companies. To become more cost competitive, throttling valve manufacturers may quote a control valve with an on-off heritage or use the smallest possible actuator size making the valve more susceptible to stick-slip particularly near the closed position or as packing is tightened.
To make matters more confusing, the readback of actual valve position for on-off valves is from the shaft position instead of actual internal closure element position. Tests by the author in separate applications revealed smart positioners saw and reported only a 0.6% backlash in the test results when the use of a travel gage indicated 8% backlash in the ball or disk position. This deception caused by the backlash in the connections between the shaft, stem, and internal closure element is widespread.
Dampers on large gas flows tend to have even more backlash than on-off valves because of greater play (slop) in the linkages. On-off actuators and positioners are often used having a poor resolution and threshold sensitivity.
Variable speed drives do not have backlash or stick-slip. However, an excessive deadband and rate limiting is often introduced due to an over concern about hunting and motor load. Also, the standard input card receiving the PID output signal had a resolution of 0.35%. A special input card had to be requested from the VSD manufacturer to get the resolution as good as the resolution of the modern day DCS output card.
Before a valve can move, enough air must move into or out of the actuator to create a sufficient change in force. This time interval is called a pre-stroke dead time. Once the valve starts to move there is a stroking time (time for 100% stroke) with a lag. For small actuator volumes this dead time and lag is less than 0.2 seconds and the stroking time is less than 2 seconds. Note that the slewing rate is 100% divided by the stroking time. The ISA standards use 86% response time (T86) as a test criterion, which is the time to reach 86% of the final response. For extremely large actuators, these times can be much larger unless volume boosters are added to the positioner output. The ISA standard lump stick-slip and resolution and threshold sensitivity limits together simply as a resolution limit. Note that the response time includes the effect of threshold sensitivity limits.
Deadband, stick-slip, resolution and threshold sensitivity limits cause limit cycles and dead time in addition to the pre-stroke dead time. The limit cycles (sustained equal amplitude oscillations) develop from the integrating action in the process or control system. For deadband, two or more integrators are required to create a limit cycle. The additional dead time can be approximated as the largest of the deadband, stick-slip, and threshold sensitivity or resolution limit, divided by the rate of change of PID output.
The variable speed drive (VSD) is often thought to provide a change in flow that is linear with speed. For this case the rangeability and linearity is impressive. Often not recognized is that when the discharge head approaches the static head as the speed is lowered, the change in flow with speed becomes larger and erratic. If the discharge head becomes less than the static head a dangerous reversal in flow can occur. A check valve is advisable in this case if the check valve is reliable at operating conditions. As a result the installed characteristic becomes steep and a low speed limit is added both reducing the controllability and rangeability of the VSD.
While the inherent dead time in the VSD response is negligible, additional dead time is created by introduction of deadband in the drive setup or a resolution limit in the signal input card. The dead time created is the deadband or resolution limit, whichever is largest, divided by the rate of change of the PID output.
If the speed control is slowly tuned or moved from the field into the DCS, the speed loop is not sufficiently faster than the process loop. This violation of the cascade rule where the secondary loop is not 4 times faster than the primary loop causes poor performance. If external reset feedback is not enabled on the primary loop, an insidious burst of oscillations will occur for large disturbances or setpoint changes. This same problem can occur for valves when positioners are sluggishly tuned or volume boosters are not used on large actuators. See chapter 5 for more details on the use of external reset feedback to prevent the primary PID output from changing faster than the secondary controller process variable (speed or stroke) can respond.
If your company or plant has strong standards as to the use and testing and response of throttling control valves, you probably don't need to pay much attention to valve dynamics except for applications where a fast valve response is needed such as surge control or pressure control or where there are exceptionally large line sizes or high pressure drops. If you don't have the protection of these standards, you are vulnerable to undersized actuators or the use of on-off valves posing as throttling valves. You need to become educated on all the gotchas.
Since the focus of variable speed drive manufacturers is on energy savings, there is very little understanding and guidance offered by the supplier on dynamic problems created by cascade control, deadband, speed rate limiting, and static head. Design and maintenance engineers must develop the knowledge and assume the responsibility for making sure the response of the VSD is fast and the installed flow characteristic is as linear as possible.
•1. For surge control valves and for systems where the valve to system pressure drop ratio is greater than 0.8 use a linear inherent characteristic, otherwise use an inherent equal percentage characteristic.
•2. Ensure the valve to system pressure drop ratio is greater than 0.2 and large enough to meet rangeability requirements.
•3. Select valve type and size to avoid operation on the upper flat portion of the installed flow characteristic.
•4. Keep the largest of threshold sensitivity and resolution limit, stick-slip, and ½ the deadband less than 0.2% by valve, positioner, and actuator design.
•5. Add a separate on-off valve for isolation and to minimize reagent injection delays (volume between tight shutoff valve and process pipe connection) that is coordinated with the opening and closing of the control valve.
•6. Tune the positioner for a fast but non oscillatory response. Only use integral action in the positioner as a last resort to create a fast limit cycle completely attenuated out by a slow process response (e.g. temperature).
•7. If a volume booster is needed for large actuators, put the booster on the positioner output and open the booster bypass just enough to stop the very fast oscillation.
•8. Keep the speed control for a VSD in the field and tune the speed controller for a fast non-oscillatory response.
•9. If the position or speed control is not fast enough, add a fast external reset feedback of actual valve position or VSD speed to the process PID.
•10. Make sure the VSD deadband is less than 0.4% and the resolution limit of the input card is less than 0.2% and the speed rate limit is fast enough.
•11. Make sure the lower speed limit will always prevent the pump or fan curve discharge head above the static head.
•12. Ensure the static head is sufficiently less than the system pressure drop to meet rangeability requirements (add system resistance and pump head if necessary).
For more details, see the May 2012 Control Talk Blogs "Checklist for Best Variable Speed Drive Performance" and "Checklist for Best Control Valve Performance"