By Greg McMillan and Stan Weiner, PE
Greg: Do you think the reason readers didn’t reply as to what was wrong with the November Control Talk trend was because the following trend chart (See Figure 1 below) was left out of the column?
Stan: I didn't even get a November issue. When I got your email, I read the article on the computer. Henree gets his copy from me. He doesn't get his own magazine. Maybe we should put him on the mailing list. Henree would have commented directly to you or Control but his pop-up preventer program has been giving him a great deal of problems; so I am forwarding his reply.
Henree: Those sure are nice pictures that you had in November's story. My two-year-old likes to color in between the lines. Her only complaint seems to be that you need to use more colors cause the coloring box has 101 crayons.
When I first looked at the picture, it seemed like you did a pretty good job getting the process variable to sit right on top of the set point. We don't see much of that in our control rooms. Most of the time, the operators keep changing the set point and the process variables just keep swinging back and forth.
Anyway, after I looked at the picture for a while, it seems like something is stuck. What we like to do in these cases is to have the field operator hit the control valve with a hammer. They tell me that in the old days they would hit the transmitters and controllers also, but it doesn't seem to do much good to beat a DCS or PLC. Anyway, the old boys still feel that it helps to hit the valves.
FIGURE 1: CONTROL TALK TREND
November 2005 Trend Graphic.
Greg: Well-placed tools on control valves is a time-tested practice to jar a shaft, flapper, or linkage free, but hitting a digital valve positioner is not advisable. Also, in this case the hesitation in the approach to the set point was caused by a controller reset time that was too large. The contribution from the proportional mode was greater than the contribution from the reset mode. Reset action will keep increasing the controller output as long as the process variable (PV) is less than the set point (SP). However, gain action will decrease the controller output as the PV increases before the PV crosses the SP for this reverse acting controller. The result is a hesitation that is a common problem particularly when the process time delay (dead time) is larger than the time lag (time constant) and the reset time was set as a ratio of the loop period typical for closed loop oscillation methods.
For dead time dominant processes, the ratio of reset time to loop period decreases dramatically. Lambda tuning, where you set the reset time basically equal to the process time constant, eliminates this problem. However, for processes where the process dead time is small compared to a very large process time constant, setting the reset time equal to the process time constant can also cause a faltering and reversal of the controller output from gain overwhelming reset action. Here, a batch controller developed in the 1970s for analog controllers but lost in the 1980s and forgotten in many a distributed control system (DCS) can reduce the time to reach set point. Of course, the DCS allows you to come up with your own strategy to hold an optimum valve position until the PV is within one dead time of reaching the set point. Just remember to release the controller in time to do its job and that while your customization is job security, you may have to maintain someone else’s creation someday.
Stay tuned for a future CONTROL article on how to generically optimize the set point response for batch, startup, and transition sequences.
Stan: To encourage readers to reply even if the puzzler or trend is missing, we are offering to send an autographed free copy of “Control Talk – The Early Years” if the reply includes a mailing address. For January, we are looking for the readers to send in their own problem and solution. Trends are nice if you remove the process engineering units and stream names.
Greg: Now for something to help you more than your control courses as you enter the control room. The number “five” seems to come up over and over again in Control.
- To keep coupled or cascade loops from fighting among themselves, these loops should be tuned to provide closed loop responses that differ by more than a factor of five. For cascade control, the closed loop response of the secondary loop should be at least five times faster than the response of the primary loop.
- When step testing, the step size should be at least five times greater than the valve dead band, or measurement noise band.
- The resolution of the measurement and valve should be better than one-fifth of the control band of the process variable and controller output for upsets.
- The total scan or execution time of asynchronous digital devices or controllers should be less than one-fifth of the process dead time or time constant, whichever is largest. To prevent A/D chatter, set the scan or execution time just large enough to see a true process variable change greater than the A/D resolution limit.
- The process variable filter time or transmitter dampening time should be less than one-fifth of the process dead time or time constant, whichever is largest. To prevent valve dither, set the filter time just large enough keep the controller output fluctuations from noise exceeding the valve’s resolution limit.
Greg: The number “ten” keeps coming up in lists as we conclude with a timely list from Mark Sowell on reasons to migrate.
Top 10 Reasons to Migrate to a New DCS
- The Function Sequence Table (FST) critical to your operation is named TOMMY.FST
- You can’t open a valve because your last COOP student is the only one that knows how to bypass the interlock code
- The application requires an 8 loop override control scheme implemented across 4 controllers and you are still emotionally scarred from the past virtual communications problems
- Your technician can replace a controller power converter card blindfolded with one hand tied behind his back
- The person you are talking to on the Support hotline wasn’t born when your system was installed
- The color of your console furniture looks more like Mississippi mud than Seafoam gray
- Your project is debottlenecking a unit configured with indirect addressing and you have a life outside of work
- The Control Diagrams in your unit have not been updated since the Reagan administration
- You have to go out to the field to see if an instrument still exists before you can delete a point from your overloaded controller
- And finally… You need to update your collection of hats and jackets.
|About the Authors|