Resolutions are made to be broken

Stan: In Trend 1 below, which is the December puzzler, the process lag was increased from the 5 seconds to 50 seconds and the controller reset time was decreased from 20 to the 5 seconds that eliminated the falter in the November trend discussed in the January issue. If you are having trouble keeping the months straight, you are not alone.

Greg: The reset time that is correct for the dead time dominant process is too short for the process where the process time constant (lag) was increased by a factor of 10. If the reset time is increased from 5 to 50 seconds, the slow oscillations and overshoot characteristic of too much reset action (too small a reset time) are eliminated. The only problem is that now the time to reach set point (rise time) is several times longer. This can be cut by one third by tripling the controller gain. This increase in controller gain is made possible by the time constant now being much slower than the dead time. In the Trend 2 for the same process dynamics, the controller gain was increased from 0.25 to 0.75 and the reset time was increased from 5 to 50 seconds. Note that the controller output goes immediately to the final resting value. If we had added in a feedforward signal that is the controller set point after it has passed through a lead-lag, we could have overdriven the output and significantly reduced the rise time. The feedforward gain is set so that the combination of the feedforward and feedback gain gives the final resting value after the lead lag transient of the controller output for the new set point. It is important to remember that the set point feedforward action is direct for a reverse acting controller and vice versa.

Stan: Finally, in Trend 3 we changed a third tuning parameter. We were going to pose it as a puzzler but the time delay between posing the puzzler and discussing the solutions is too long. So the answer is we added a rate setting of 150 seconds, which is about 10 times larger than it should be. The result is high frequency oscillations in the controller output and a staircase in the set point response of the PV. It is also actually working against the set point change because the rate action is on PV. Even if rate on error, a set point filter, and the proper amount of rate were used, it wouldn’t improve the set point response much for this single lag plus dead time process.

Greg: In real life there is hardly ever just one lag and rate can help if it is set equal to the second lag. For integrating processes, we can view the integrator gain as a big first lag and set rate equal to the additional lag. The problem, of course, with derivative is that measurement noise is amplified. The best applications for rate are back mixed temperature and pH loops. A narrow range transmitter to minimize A/D resolution noise for temperature and a PV filter to reduce sensor noise for pH should be used.

Stan: While we are going to suspend the use of trend charts as puzzlers, please send in your favorite trend along with the analysis and your mailing address and we will put it in the column and send you a free copy of “Control Talk – The Early Years.” 

Greg: For you “Top Ten List” fans, we intend to keep the lists going since they are timeless.

Top Ten Broken New Years Resolutions

1. Stop rooting against Texas teams now that my home is in Austin. Why did I enjoy the massacre of the Cowboys by the Redskins?
2. Listen intently to my wife’s instructions. Why does my mind still jump to weighty matters like what is next for dinner?
3. Stop making cheap control valve jokes. Could the next final element reputation I hurt be my own?
4. Help make smart diagnostics smarter. Do I need to de-fussify my fuzzy logic?
5. Stop lusting in my heart for more computing power. Is it the PC or me that is the constraint?
6. Turndown the volume on my headphones. What did you say?
7. Stop drinking cheap wine. Does good wine ever come in a size large enough?
8. Read a college text on control theory. Can I watch Star Trek without setting up the state space equations?
9. Stop answering a question with a question. Why should a consultant do this?
10. Get Henree a job. Who is Henree anyway and why does he know so much about me and Stan?

CAPTION: ..."I thought you were going to start spending more time with me, and less time with that magazine!"