Here is the third part of a point blank decisive comprehensive list of what we really need to know in a detailed attempt to reduce the disparity between theory and practice. Please read, think and take to heart the opportunities to increase the performance and recognized value of our profession. The list is necessarily concise in detail. If you want more information on these opportunities, please join the ISA Mentor Program and ask the questions whose answers can be shared via Mentor Q&A Posts.
You can also get a comprehensive resource focused on what you really need to know for a successful automaton project including nearly a thousand best practices in the 98% new 2019 Process/Industrial Instruments and Controls Handbook Sixth Edition capturing the expertise of 50 leaders in industry.
The following list reveals common misconceptions that need to be understood to seek real solutions that actually address the opportunities.
- Dead time dominant loops need Model Predictive Control or a Smith Predictor. There are many reasons for Model Predictive Control but dead time dominance is not really one of them. Dead time compensation can be simply done by inserting a dead time block in the external-reset feedback path making a conventional PID an enhanced PID and then tuning the PID much more aggressively. This enhanced PID is much easier to implement than a Smith Predictor because there is no need to identify and update an open loop gain or primary open loop time constant and there is no loss of the controlled variable seen on the PID faceplate. An additional pervasive misconception is that dead time dominant processes benefit the most from dead time compensation. It turns out that reduction in integrated error for an unmeasured process input load disturbance is much greater for lag dominant processes, especially near-integrating processes. While the improvement is significant, the performance of a lag dominant process is often already impressive provided the PID is tuned aggressively (e.g., integrating process tuning rules with minimum arrest time). For more details see the ISA Mentor Program Q&A Post "How to Improve Loop Performance for Dead Time Dominant Systems"
- The model dead time should not be smaller than the actual loop dead time. For Model Predictive Control, Smith Predictors and an enhanced PID, a model dead time larger than the actual dead time by just 40% can lead to fast oscillations. These controllers are less sensitive to a model dead time smaller than the actual dead time. For a conventional PID, tuning based on a model dead time larger than the actual dead time just causes a sluggish response so in general conventional PID tuning is based on largest possible loop dead time.
- Cascade control loops will oscillate if cascade rule is violated. For small or slow setpoint changes or unmeasured load disturbances, the loop may not break out into oscillations. While it is not a good idea to violate the rule that the secondary loop be at least five times faster than the primary loop, there are simple fixes. The simplest and easiest fix is to turn on external-reset feedback that will prevent the primary loop integral mode from changing faster than the secondary loop can respond. It is important that external-reset feedback signal be the actual process variable of secondary loop. There is no need to slow down the tuning of the primary loop, which is the most common quick fix if the secondary loop tuning cannot be made faster.
- Limit cycles are inevitable from resolution limits. While one or more integrators anywhere in the system can cause a limit cycle from a resolution limit, turning on external-reset feedback can stop the limit cycle. The feedback for the external reset feedback must be a fast readback of the actual manipulated valve position or speed. Often readback signals are slow and changes or lack of changes in the readback of actuator shaft position are not representative of the actual ball or disk movement for on-off valves posing as control valves. While external-reset feedback can stop the limit cycle, there is an offset from the desired valve position. For some exothermic reactors, it may be better to have a fast limit cycle in the manipulated coolant temperature than an offset because tight temperature control is imperative and the oscillation is attenuated (averaged out) by the well-mixed reactor volume.
- Fast opening and slow closing surge valves will cause oscillations unless PID is tuned for slower valve response. It is desirable that surge valve be fast in terms of increasing and slow in terms of decreasing vent or recycle flow for compressor surge control. Generally, this was done in the field by restricting the actuator fill rate or enhancing exhaust rate by a quick exhaust valve since the surge valves are fail open. The controller had to be tuned to deal with the crude unknown rate limiting. Using different setpoint up and down rate limits on the analog output block and turning on external-reset feedback via a fast readback of the actual valve position make the adjustment much more exact and visible. The controller does not need to be tuned for the slow closing rate because the integral mode will not outrun the response of the valve.
- Prevention of oscillations at split range point requires a deadband in the split range block. A dead band anywhere in the loop adds a dead time that is the dead band divided by the rate of change of signal. Dead band will cause a limit cycle if there are two or more integrators anywhere in the control loop including the positioner, process, and cascade control. The best solution is a precise properly sized control valves with minimal backlash and stiction and a linear installed flow characteristic. External-reset feedback with setpoint rate limits can be added in the direction of opening or closing a valve at the split range point to instill patience and eliminate unnecessary crossings of the split range point. For small and large valves, the better solution is a valve position controller that gradually and smoothly moves the big valve to ensure the small valve manipulated by the process controller is in a good throttle position.
- Valve position controller integral action must be ten times slower than process controller integral action to prevent oscillations. External-reset feedback in the valve position controller with fast readback of actual big valve position and up and down setpoint limits on analog output block for large valve can provide slow gradual optimization but a fast getaway for abnormal operation to prevent running out of the small valve. This is called directional move suppression and is generally beneficial when valve position controllers are used to maximize feed or minimize compressor pressure or maximize cooling tower or refrigeration unit temperature setpoints. One of the advantages of Model Predictive Control is move suppression to slow down changes in the manipulated variable that would be disruptive. Here we have the additional benefit of the move suppression being directional with no need to retune.
- High PID gains causing fast large changes in PID output upset operators and other loops. The peak and integrated errors for unmeasured load disturbances are inversely proportional to the PID gain. A high PID gain is necessary to minimize these errors and to get to setpoint faster for setpoint changes. Too low of a PID gain is unsafe for exothermic reactor temperature control and can cause slow large amplitude oscillations in near-integrating and true integrating processes. Higher PID gains can be used to increase loop performance without upsetting operators or other loops by turning on external-reset feedback, putting setpoint rate limits on the analog output block or secondary loop and providing an accurate fast feedback of manipulated valve position or process variable.
- Large control valve actuators and VFD rate limiting to prevent motor overload requires slowing down the PID tuning to prevent oscillations. Turning on external-reset feedback and using a fast accurate readback of valve position or VFD speed enables faster tuning to be used that makes the response to small changes in PID output much faster. Of course, the better solution is a faster valve or larger motor. Since there is always a slewing rate or speed rate limit in VFD setup using external-reset feedback with fast readback is good idea in general.
- Large analyzer cycle times require PID detuning to prevent oscillations. While the additional dead time that is 1.5 times the cycle time is excessive in terms of ability of loop to deal with unmeasured load disturbances, when this additional dead time is greater than the 63% process response time, an intelligent computation of integral action using external-reset feedback can enable the resulting enhanced PID gain to be as large as the inverse of the open loop gain for self-regulating processes even if the cycle time increases. This means the enhanced PID could be used with an offline analyzer with a very large and variable time between results reported. While load disturbances are not corrected until an analytical result is available, the enhanced PID does not become unstable. The intelligent calculation of the proportional (P), integral (I) and derivative (D) mode is only done when there is a change in the measurement. The time interval between the current and last result is used in I and D mode computations. The input to I mode computation is the external-reset feedback signal. If there is no response of the manipulated variable, I mode contribution does not change. An analyzer failure will not cause a PID response since there is no change in P or D mode contribution unless there is a new result or setpoint. The same benefits apply to wireless loops (additional dead time is ½ update rate). For more details see the Control Talk Blog “Batch and continuous control with at-line and offline analyzer tips”