Just as split personalities cause mental instability, split ranged loops cause process control instability. Actually, the situation is worse for split ranged loops because valve stiction is particularly bad near the closed position, and the transition between fluids (e.g., between water and steam) may involve changes in phase besides changes in dynamics.
If you have a split ranged loop, I bet you 25 cents that the loop oscillates at times across the split range point. If oscillations never occur when operating at the split range point, please mail me a copy of the trend chart, along with a self-addressed, stamped envelope, and I will mail you the quarter.
Stiction causes a limit cycle in a loop with integrating action at one or more points in the process or in the controller. The introduction of a deadband into the split range point in an attempt to stop unnecessary crossing back and forth of the split range point also causes a limit cycle if there is integration in two or more points in the process or control system. Batch temperature, pH and composition control, vessel pressure control, and vessel level control all have integrating action in the process. Thus, the addition of one PID loop causes limit cycles from deadband. The use of cascade control, where both the primary and the secondary loops have integral action, will cause a limit cycle regardless of process type. Deadband also appears in control valve backlash and in variable speed drive setup.
We think about the difference in process gain for a change in phase in the fluid being manipulated. We may not think of the difference in process deadtime and time constant for the same change. When there is a change in phase, there is also a significant change in process dynamics. For example, as a reactor temperature controller switches from steam to cooling water, or a bioreactor pH controller switches from carbon dioxide (gas) to sodium bicarbonate (solid), there is a big change in the process time constant. In addition, there is an undetermined residual effect of one phase. For reactor jackets, switching from cooling water to steam can result in pockets of water and droplets in the steam. Switching from steam to water can result in pockets of vapor and bubbles in the water. The result is cold and hot spots and an erratic response in the transition.
In chemical reactors, there is often a time when there is no significant demand for heating or cooling. In bioreactors there is often no significant demand for an acid or base. The best thing for the control system to do is nothing but because of integrating action, limit cycling will occur.
Needless to say, most loops oscillate at the split ranged point, decreasing process efficiency and in some cases decreasing product quality. For bioreactors, the unnecessary addition of sodium bicarbonate increases cell osmolality (internal cell pressure), causing cell rupture and death. A batch worth several million dollars can be lost due to high osmolality from oscillations at the split range point.
Concept: Most split ranged loops oscillate across the split range point. Stopping these oscillations requires a comprehensive approach, addressing changes needed in the process, control system, and final control elements. Missing just one of the many sources will result in continuing (but usually less severe) oscillations.
Details: To avoid the discontinuities and unpredictable behavior from the changes in phase between using steam and cooling water, use steam injection so the transition between heating and cooling is a transition between hot water and cold water. To reduce the effects of stiction and backlash at the split range point, add a small sliding stem valve with a diaphragm actuator and digital positioner for fine adjustment at the split range point. Even if this does not stop the oscillations, the consequences in terms of process efficiency and product quality will be less severe. The fine adjustment valve must be kept from going wide open by the manipulation of a large valve by a valve position controller (Tip #97). This strategy can provide greater precision at all operating points. Do not add deadband in the split range control configuration or in a variable speed drive setup. Use an auto-tuner or adaptive tuner to identify the different process dynamics of split ranged valves or secondary loops and schedule the tuning based on the primary controller output. Use external-reset feedback and directional setpoint rate limits on an analog output block for split range valves and on PID setpoints for secondary loops to suppress movement into the split range point. To stop limit cycles use an enhanced PID developed for wireless with a threshold sensitivity setting to ignore noise and insignificant changes.
Watch-outs: The split ranging of a small valve and a large valve will create another point of oscillations. The valve position controller eliminates the addition of a split range point but will interact with the process controller unless an integral-only structure or an enhanced PID is used. For exothermic reactors, transitions from heating to cooling should not be suppressed and valve position controllers must not slow down the need for more cooling to prevent a runaway. Valves with very low leakage specs tend to be on-off valves. Do not use on-off valves for any throttling service (Tip #84). For tight shutoff and positive isolation use an on-off valve that is coordinated with the opening and closing of the small valve.
Exceptions: For valves that have difficulty opening after being closed for a long time due to temperature or phases, the periodic opening of the valves caused by split range oscillations may help prevent them from seizing.
Insight: Split range oscillations occur due to process nonlinearities and discontinuities, control valve deadband and stiction, and controller integral action and overreaction.
Rule of Thumb: Use precise valves, scheduled tuning, external-reset feedback, directional rate setpoint limits, and an enhanced PID to stop split range oscillations.
Look for another tip next Friday.