«

»

Dec
15

How to Succeed – Part 6

The increased emphasis on minimizing leakage, energy use, and low cost bid has lead to increased process variability. The valve with the lowest leakage, pressure drop, and price tag often has deadband, rangeability, resolution, and sensitivity deficiencies leading to poor control creating continual oscillations even when there are no process upsets. The sustained equal amplitude oscillations can not be eliminated by tuning. Just putting the loop in auto is enough to trigger the oscillations.         

Control valve specification sheets have 50 or more entries but typically do not have a field for deadband, rangeability, resolution, and sensitivity. In fact there is no requirement that the control valve actually respond to a change in signal. The supplier is not held accountable for whether the valve internal closure member (plug, disc, or ball) actually moves and most importantly the flow really changes for a change in signal. No wonder the wrong valve wins the low cost bid. A rotary piping valve with an actuator originally designed for on-off service can generally meet all of the specification requirements. Slapping on a smart digital positioner just intensifies the deception giving a false sense of confidence because the positioner feedback is actuator shaft and not closure member position. The actuator shaft may move even though the closure member does not move due to deadband in linkages and connections, packing fraction, shaft windup, sealing and seating friction. Even if the closure member does move, the installed flow characteristic may be so flat the change in flow is trivial.

Resolution (stick-slip) should be better than ¼ the allowable process variable error divided by the open loop gain over the whole throttle range (min to max flow). The open loop gain is the product of the valve gain, process gain, and measurement gain. The valve gain is the slope of the installed characteristic taking into account any split ranging (change in valve flow per % change in controller output). The process gain is the change in the process variable in engineering units for a change in control valve flow. The measurement gain is 100% divided by the measurement span. The deadband from backlash should not be more than twice the resolution. The threshold sensitivity the actuator and positioner should be better than half the valve resolution. The sensitivity (slope of the installed characteristic) should be greater than the resolution in % divided by the product of the process gain and measurement gain. The sensitivity should not change more than by a factor of 4 unless the nonlinearity is compensated for by signal characterization, gain scheduling, or adaptive tuning. Figures 7-48a-c on pages 411-414 of  the “Essential’ book, Essentials of Modern Measurements and Final Elements in the Process Industry show the acceptable throttle ranges to keep the slope changes within the 4:1 range. There are other things more associated with valve selection and sizing not apparent from the control valve specification sheet. I recommend you use software not only to size the valve but to compute the installed characteristic using details of the mechanical and piping design.

Statements that you can allocate 5% of the system drop to save energy assumes you don’t need any rangeability. Statements on rangeability do not generally take into account the installed characteristic, deadband, and resolution. Equations 7-19a-d on page 418 of the “Essential” book offer a more realistic view of rangeability. Valve drops more than 25% of the system drop are advisable.

Finally, here is my check list with suggested values assuming one does not have the time or info to do detailed calculations of deadband, rangeability, resolution, sensitivity, or nonlinearity requirements:

 Use smart software to size the valve
 Select location and valve type to eliminate or reduce damage from flashing
 Select location and valve type to eliminate or reduce damage from erosion
 Include swage effect from piping reducer
 Use smart software to compute and plot installed valve characteristic
 Size actuator to deliver twice the max torque or thrust required
 Specify actuator threshold sensitivity better than 0.1%
 Specify smart positioner threshold sensitivity better than 0.1%
 Tune smart positioner for application (otherwise you have a dumb positioner)
 Specify deadband less than 0.4% over the entire throttle range
 Specify resolution better than 0.2% over the entire throttle range

 Use step sizes of 0.1% and flow measurement or travel gage to test response

 Ensure valve gain > 0.5% max flow per % signal over the entire throttle range
 Ensure valve gain < 2.0% max flow per % signal over the entire throttle range

Note that for valve gain also known as flow sensitivity, the max flow is valve capacity and the split range amplification effect must be included. For 50% split range point of 2 control valves the amplification is a factor of 2. For small and large valves, the effect of size on valve gain can be mitigated by an intelligent selection of a split range point. For a large valve with 4 times the capacity of a small valve, the split range would be 0-20% for the small valve and 20-100% for the big valve. For valves on different process streams, the process gain needs to be included in the calculation of the intelligent split range point. Since operations are accustomed to a 50% split range, graphics & training are needed.

Threshold sensitivity and resolution are the smallest input change that will cause the automation device to respond. For resolution the response is a step the size of the resolution limit (stair-case or quantized response). For threshold sensitivity, once the response occurs the output change matches the input change. The response of actuators and positioners is commonly characterized by a threshold sensitivity whereas the response of a valve with friction (stiction) is characterized by a resolution assuming the slip equals the stick. Threshold sensitivity and resolution will cause a limit cycle for a controller in automatic if there is integrating action in the process or in the controller via the integral mode. Deadband will cause a limit cycle if there are more than two occurrences of integrating action (e.g. integrating process such as level and integrating action in a level controller or cascade control with integrating action in both the secondary and primary PID).

If you get too much flak, tell them this all came from the author formerly known as Greg.