«

»

Sep
08

Review of Deminar #9 – Process Control Improvement Primer

Greg McMillan Deminar Series

To view the recording of Deminar #9, click on the above picture. If you want to just view the slides click on Deminar #9 – Process Control Improvement Primer

A lot of ground was covered in an hour to address the major aspects of basic control, the foundation of higher levels of control. You can sum up the first ten concepts as keys to how you can improve a loop to quickly and consistently see and enact intentional changes (setpoint changes) and correct for unintentional changes (load disturbances).

The main job of a loop is dealing with ramping disturbances as inputs to a process with variable delays, speeds, and gains. In control theory we are accustomed to frequency response and setpoint response metrics for systems with fixed and rather small delays. When a disturbance is introduced in control theory, it is often downstream of the process. If applications were so straight forward in industrial processes we would be home free or at least at happy hour instead of working on how to understand what is really going on.

One of the common mistakes is to not realize the critical role controller tuning plays in any analysis of control system performance. You can prove almost any point by how you tune the controller. The tuning relative to the fastest practical settings for load disturbance rejection determines the opportunity for reducing peak and integrated errors. Simple equations are provided to estimate errors for a given set of tuning settings and to estimate the fastest practical settings. The ultimate limit to performance depends upon dynamics throughout the loop. Here again simple equations are offered. The deeper understanding offered by these equations is probably more valuable than any number crunching.

One interesting point not covered is that the sensitivity-resolution limits and backlash-deadband in every control valve inherently reduces the transfer of measurement noise to the process by preventing the valve from moving in response to fluctuations in the controller output. Since variable speed drives (VSD) have potentially negligible resolution limits (provided the A/D and speed pickup are properly selected) and essentially no backlash, a deadband is normally introduced to prevent reaction to noise. Often this setting is too large because the user doesn’t realize that deadband creates deadtime for a ramping disturbance. A velocity limit is also commonly introduced that makes the (VSD) response much slower than a control valve slewing rate. The result is a burst of oscillations for large upsets when the process controller output tries to change faster than the VSD can respond (subject of Deminar #3). Thus, the lack of a fundamental concept of speed and delay results in VSD settings that deteriorate loop performance.