The maximum level controller gain for stability is one or two orders of magnitude higher than expected. The main limit to how high you can go in controller gain is often measurement noise and control objective. The common practice of using much less than the permissible controller gain causes slow rolling oscillations. There is a low as well as a high controller gain limit for level control. The controller gain used to prevent vessels from being overfilled or running dry or smoothing out feeds to downstream equipment is typically below the low gain limit. Even when tight level control is needed for reactor residence time control or distillation column overhead receiver reflux control, the controller gain is typically still far below the high gain limit for stability.
In order to prevent slow rolling oscillations, the product of the controller gain (Kc) and controller reset time (Ti) should not be less than twice the inverse of the integrating process gain (Ki) that for level loops is incredibly small (e.g., Ki < 0.0001 % PV per sec per % OUT).
Kc* Ti> 2 / Ki
For surge tank level control, gain scheduling is used to reduce the changes in manipulated flow for small changes in level. This tuning helps eliminate the reaction to noise and short term transients but does not stop the eventual excursion from the level ramping out of the low gain region due to the non-self-regulating nature of integrating processes. Also, the minimum controller gain limit is violated in the low gain region. The solution is to schedule the reset time to be much larger besides the gain setting to be much lower for permissible level changes.
Note that the addition of filter is sometimes necessary but will greatly deteriorate the maximum gain for stability. Since level changes are normally so slow, a velocity limit on the measurement signal is effective in screening out noise as being unrealistically too fast. Also, the level controller execution time should be set large enough so the actual level change is significant. Most level controller execution times are an order of magnitude too small.
An adapted velocity limited feedforward can provide maximum smoothing that is especially useful for the transition from batch to continuous operations. The velocity limit is adapted so the unbalance in flows into and out of the tank is spread over the available operating inventory as described in Tip #94 in the new ISA book 101 Tips for a Successful Automation Career.
Level loops are susceptible to limit cycles from deadband due control valve backlash or variable frequency drive (VFD) settings. Integral action must be suspended in all controllers including the positioner to stop the limit cycles. Where ever integral action is used in the PID controllers or positioner, the integral deadband must be set larger than the PV limit cycle amplitude caused by the valve or VFD deadband.
For a better understanding of the importance of level control see the InTech article "Advances in level and flow measurement enhance process knowledge, control". See the Control Talk Blog "Checklist for Maximizing Radar Level Measurement Performance" to achieve the greatest sensitivity and least noise.
The following checklist is not intended to cover all the application requirements but some of the major considerations to be addressed.
•1. Is the level sensor sensitivity high enough and noise low enough to meet control objectives?
•2. Is the level sensor adversely affected by temperature and density changes?
•3. Is the level PID execution time large enough to provide a good signal to noise ratio?
•4. Is a signal filter or velocity limit needed to reduce measurement noise?
•5. For surge tank level control is the controller gain decreased and reset time increased for permissible level changes to smooth out feed changes to downstream equipment?
•6. Would an adapted velocity limited feedforward be useful for maximum the smoothing of the feed to downstream operations for extremely discontinuous upstream operations?
•7. Is the product of the controller gain and reset time greater than or equal to twice the inverse of the integrating process gain to prevent slow rolling oscillations?
•8. For reactor level control and distillation distillate receiver level control, is the controller gain maximized to provide tight residence time and material balance control, respectively?
•9. For minimum inventory control is the level control tight enough to prevent running dry?
•10. Is integral action turned off or integral deadband enabled in the level controller, secondary flow controller, and positioner to prevent limit cycles from valve backlash and VFD deadband?
