All Search Results matching “deadtime”

To view the recording of Deminar #8, click on the above picture. If you want to just view the slides click on Deminar #8 - PID Control of Runaway Processes

Self-regulating processes are the easiest to control given similar dynamics (e.g. delays, lags, and gains), nonlinearities, and upsets. In manual, the process variable will eventually reach a steady state for a self-regulating process. Integrating processes are the next most difficult to control because in manual the process variable will always be ramping even if there are no disturbances. Runway processes are the most challenging and potentially the most dangerous because in manual the process variable is always moving and can accelerate in its divergence even if there are no disturbances. Runaway processes are termed "open loop unstable." The acceleration is characterized by a positive feedback time constant. Both integrating and runaway processes have a low gain limit that causes slow rolling oscillations and a divergence off-scale, respectively. Integrating processes are more sensitive to integral action and secondary lags than self-regulating processes and runaway processes are more sensitive to integral action and secondary lags than integrating processes. The most common problem with integrating and runway processes is too much integral action (too small of a reset time) and the omission of derivative action for secondary lags (rate time should be set equal to largest secondary lag). Some highly exothermic polymerization reactors have proportional plus derivative control to avoid the potentially unsafe situation of someone adding too much reset action. I have been in the control room when an exothermic reactor has reached a point of no return where the temperature acceleration was so high despite full cooling, the only thing the operators could do was prepare for the rupture discs to burst and the reactor contents blow over to the flare stack tank. Highly reactive chemicals lead to rapid and complete reactions but can also lead to an uncontrollable temperature rise since the reaction rate and hence heat release doubles for every 6 degree increase in temperature. Runaway processes can look like integrating processes unless the temperature controller is left in manual long enough for the temperature change to be large enough.

Deminar #8 shows the dramatic correction needed for the tuning settings. The factors used in the short cut tuning method for near-integrators in Deminar #6 and the classic Ziegler Nichols ultimate oscillation method are detailed and demoed. Equations are offered to predict the ultimate gain and ultimate period showing the dramatic effect of a secondary process or thermowell lag and loop deadtime. If a secondary lag or the loop deadtime approaches the positive feedback time constant, the window of allowable controller gains closes and the loop is unstable for all tuning settings. The virtual plant is where you want to learn about runaway processes. You can't experiment much or have the loop in manual for more than a few deadtimes with a true runaway process.

To view the recording of Deminar #7, click on the above picture. If you want to just view the slides click on Deminar #7 - PID Control of True Integrating Processes

Time is money. If you can get to optimum setpoints faster during fed-batch operations and for startup and product transitions of continuous operations, the increase in production revenue can be significant. For continuous operations there may also be an appreciable decrease in the processing, recycle, and waste treatment costs of off-spec material.

For cascade control, the speed of the secondary PID setpoint response largely determines the ability of the primary PID to get to its setpoint quickly and reject disturbances in the primary loop. A slow secondary PID setpoint response may require detuning of the primary PID to prevent interactions between the secondary and primary loops.

In Deminar #7 we explored how we could use PID structure options, setpoint feedforward, and bang-bang control to improve the setpoint response for integrating (e.g. batch) processes. The concepts are also applicable to the continuous process startup and transitions. The demos showed a big reduction in rise time (time to reach setpoint) by the use of "PID on Error" instead of "I on Error, PD on PV." The benefit of the additional bump from derivative action on error is rather marginal for the small rate setting used. In other words most of the speedup in the setpoint response could be achieved by "PI on Error, D on PV" unless there is a large secondary lag and hence a large rate time setting. The use of setpoint feedforward helped reduce overshoot, rise time, and settling time by about 25%. For deadtime dominant self-regulating processes, the improvement would have been more impressive. The most dramatic improvement occurred for full throttle bang-bang control. With some adjustment of logic and resting value as noted on slide 6, the bang-bang logic can also be effectively used for self-regulating processes. You can try out setpoint feedforward and bang-bang control on the virtual plant website starting August 20.

The PID enhancement for wireless (PIDPlus) offers an improvement wherever there is an update time in the loop. In the broadest sense, an update time can range from seconds (e.g. wireless updates and valve or measurement sensitivity limits) to hours (e.g. failures in communication, valve, or measurement). Some of the sources of update time are:

(1) Wireless measurement default update rate for periodic reporting (refresh time)
(2) Wireless measurement trigger level for exception reporting (sensitivity limit)
(3) Wireless communication failure
(4) Broken pH electrode glass or lead wires (failure point is about 7 pH)
(5) Large valve operating on upper part of installed characteristic (low sensitivity)
(6) Valve with backlash (deadband) and stick-slip (resolution and sensitivity limit)
(7) Valve with solids, high temperature, or sticky fluid that causes plugging or seizing
(8) Plugged impulse lines
(9) Analyzer sample processing delay and analysis or multiplex cycle time
(10) Analyzer resolution and sensitivity limit

The PIDPlus waits for an update in the measurement whereas a traditional PID continually ramps the output acting on old information. When there is an update, the PIDPlus considers the changes to have occurred over the elapsed time from the last update whereas the traditional PID thinks the entire change occurred in the PID module execution time. The result is a spike from derivative action by a traditional PID that is particularly large when a measurement recovers or a valve trim or solids break free.

The improvement in control by the PIDPlus is most noticeable as the update time becomes much larger than the 63% process response time (defined in the white paper as the sum of the process deadtime and time constant). When the update time becomes 4 times larger than the 63% process response time that roughly corresponds to the 98% response time frequently cited in the literature, the controller gain can be set equal to the inverse of the process gain. This controller gain can provide an exact correction for changes in the measurement and setpoint.

The PIDPlus execution is kept fast so that the PID immediately responds to changes in setpoint, feedforward, mode, tuning, detail display parameters, and remote output. We have the interesting result that when the update is much larger than the 63% process response time so we can set the controller gain equal to the inverse of the process gain, the controller output goes immediately to the value needed to achieve the setpoint. An increase in update time to prolong battery life can actually translate to a faster setpoint response. However, if the process gain changes with time or operating point, the PID will require several updates to home in on the proper correction. An increase in update time will increase the settling time for unrecognized changes in the process gain. The use of an adaptive tuner such as DeltaV Insight that automatically identifies the process gain and schedules the tuning setting accordingly can sustain a fast setpoint response despite nonlinearities and a large update time.

The Emerson White Paper DeltaV-v11-PID-Enhancements-for-Wireless.pdf discusses these opportunities in more detail. Later this month, an entry on this site will show and discuss the trend plots that compare the enhanced PIDPlus with the traditional PID for the applications tested including valves with stick-slip and backlash.

It is important to distinguish between an update time and process deadtime. The update time is the time interval between successive updates by the final control element (initiated changes to the process input) and successive updates by the measurement (reported changes in the process output). The process deadtime is a continuous train of values delayed by the deadtime. The most common source of a pure process deadtime is a transportation delay of temperature and composition changes in a conveyor, extruder, dip tube, heat exchanger, pipeline, sheet line, or any volume where there is plug flow (no back mixing). Small time constants such as thermal lags, sensor lags, signal filter times, transmitter damping settings, effectively become additional deadtime in terms of a first order plus deadtime approximation (single time constant plus deadtime). The PIDPlus algorithm does not correct for process deadtime. As the process deadtime increases and approaches the update time, the opportunity to increase the PIDPlus gain decreases. For compensation of deadtime, a standard deadtime block can be inserted between the BKCAL_OUT of the AO block and the BKCAL_IN of the PID block if the DCS uses the positive feedback method for the integral mode (external reset) as reported in Advanced Application Note 3 "Compensation of Deadtime in PID Controllers".

In a future Deminar we will look in greater detail at the effect of updates time of discontinuous measurements and process deadtimes on the ultimate period and ultimate gain and if there is an improvement in loop performance offered by a combination of PIDPlus and deadtime compensation.

You can click on the above to view and hear the recording of the Deminar.

Would you like to find tuning settings and develop a real time simulator for the more important loops in your plant in less than 10% of the time normally required? If this is of interest, check out Deminar #6. The test, triggered by a setpoint or output change, only needs to last about 3 deadtimes. Since the process time constant for the composition, pH, pressure, and temperature response of vessels and columns is 6 to 100 times the observed deadtime and the time to steady state is 4 time constants plus the deadtime, the time savings varies from 90% to 98%. The reduction in test time also minimizes the possibility of the test being disrupted by a disturbance. One of the problems we have with testing large columns to identify the dynamics for tuning or model predictive control is that the time to steady state is a day or more. Day to night temperature changes, feed changes, and shift changes usually disrupt the test of these columns. With the near-integrator approach the test time is a matter of hours and if there is a disruption, the test can be readily repeated. Also, the upset to the process from the test is significantly less because the excursion during the shorter test is much smaller.

The near-integrator gain parameter used to dramatically shorten the test time leads to a simpler expression for the controller gain that is just a function of the near-integrator gain and the observed deadtime. All of the tuning methods reduce to this same expression for maximum disturbance rejection as shown in "Appendix C - The Unification of Controller Tuning Relationships" in the ISA bookNew Directions in Bioprocess Modeling and Control. The controller gains differ by a factor that varies from about 0.5 for a Lambda tuning with a closed loop time constant equal to the process deadtime to 1.0 for the Ziegler Nichols Reaction Curve (ZNRC) method (not to be confused with the widely remembered and unpopular Ziegler Nichols ultimate oscillation method). Note that the ZNRC method requires an open loop test (change in manual output of the controller) and waits for the process to reach steady state to construct a tangent to the inflection point and find its intersection with the final value. The near-integrator method finds the maximum ramp rate for a step change in the controller output regardless of PID mode (e.g. triggered by a setpoint change or a remote output change for batch control).

What about the secondary time constants? If these time constants are much less than the primary process time constant, these secondary time constants result in an increase in the observed deadtime. Keying on a multiple of the observed deadtime self-compensates for this situation. For non-interacting secondary process time constants that approach the primary time constant (an interesting but relatively rare case), the search for the maximum ramp rate would need to be extended for several more deadtime intervals. The search can stop if the ramp rate is not increasing. For equal interacting time constants, the secondary process time constant is about 1/6 of the primary time constant. This methodology can be readily automated to identify the dynamics whenever there is a step change in controller output significantly larger than the final control element (e.g. valve) resolution limit.

For a simple real time process simulation that uses standard function blocks, the controller output and process variables from a scan or snapshot of the actual process for a representative relatively quiet operating point can be used to create deviation variables and provide a correction of the model.

The Deminar focuses on self-regulating processes that look like integrating processes because the process response ramps in the control region. The appearance can be caused by a time to steady state that is beyond the practical time range for observation or by a steady state that is beyond the operating limits of the equipment. For example, an increase in vessel pressure can force more flow out the vent valve but the vessel pressure required for the vent flow to balance the incoming or generated gas flows can be beyond the pressure relief valve setting. The time constant or ramp rate for gas pressure is generally order(s) of magnitude faster than for liquid temperature but the pressure loop deadtime is even faster. For example, the deadtime and time constant for a column pressure response might be 5 and 100 seconds, respectively whereas the deadtime and time constant for column temperature might be 5000 and 30,000 sec, respectively.

The near-integrator method can also be applied to true integrating processes which means level loops and composition, pH, pressure, and temperature loops in batch besides continuous processes can be rapidly tuned and simulated. Loops not suitable for this method are liquid pressure and flow loops and inline (pipeline) blending, pH, and temperature loops because the observed deadtime is comparable or even larger than the process time constant. However, the time to steady state for these loops is a matter of 2 to 20 seconds so that the test time is already fast and conventional methods can be employed.

Higher value added products are generally produced by batch operations. Often these products are sold out and extra batches translate to significant increases in revenue. Prime opportunities are specialty chemicals and drugs, especially new biopharmaceuticals where optimization took a backseat to time to market in the initial plant and automation system design.

I looked over my past experience with Monsanto, Solutia, and Emerson and have come up with myriad of methods to reduce batch cycle time. I have divided them up into opportunities to help feedback loops to get to setpoint faster that are important for Fed-Batch operations and for startups and transitions of continuous operations (Part 1) and opportunities to shorten phases and holds that are important for Pure-Batch operations (Part 2). These techniques like all new configurations and strategies should be thoroughly tested by simulation and closely monitored and adjusted for safe and efficient operation. Today's blog is a preview of Deminar #7 on July 14.

Fed-Batch Opportunities

1) PID on Error Structure - This structure maximizes the kick of the controller output for a setpoint change. The overdrive (driving of output past resting point) is essential for getting slow loops, such as temperature, to the optimum setpoint as fast as possible.

2) SP Track PV - With this control option the setpoint is changed to its optimum with the controller in automatic providing the kick from the PID structure (1). For batch operations this option is commonly used. For continuous operations with few setpoint changes (no grade transitions) and extremely long run time (e.g. years), the setpoint is held at its last value. However, even here loops with slow reset action (large reset times), such as level, the use of the SP Track PV option can prove useful when putting these loops back in service after maintenance.

3) SP Feedforward - For low controller gains (controller gain less than inverse of process gain), a setpoint feedforward is useful. The setpoint feedforward gain is the inverse of the dimensionless process gain minus the controller gain on a percent basis. If the setpoint and controller output are in engineering units the feedforward gain must be adjusted accordingly. The feedforward action is the process action, which is the opposite of the control action, taking into account valve action. In other words for a reverse control action, the feedforward action is direct provided the valve action is inc-open or the analog output block, I/P, or positioner reverses the signal for a inc-close (fail open) valve.

4) Output Lead-Lag - A lead-lag on the controller output or in the digital positioner can kick the signal though the valve deadband and sticktion, get past split range points, and make faster transitions from heating to cooling and vice versa. When combined with the enhanced PID algorithm described in Deminar #1, the lead-lag can potentially provide faster control when online analyzers are used for closed loop control of the integrating response associated with batch operations.

5) Deadtime Compensation - The simple addition of a delay block with the deadtime set equal to the total loop deadtime to the external reset signal for the positive feedback implementation of integral action (see Deminar #3). The controller reset time can be significantly reduced and the controller gain increased if the delay block deadtime is equal or slightly less than the process deadtime (see Advanced Application Note 3 entry March 25, 2009 on this website).

6) Full Throttle Batch - The controller output is put at its output limit to maximize the rate of approach to setpoint. When the projected PV equals the setpoint less a bias, the controller output is repositioned to the final resting value captured from the last batch. The output is held at the resting value for one deadtime. For more details, check out the Control magazine article "Full Throttle Batch and Startup Response."

7) Feed Maximization - Valve position control, Model Predictive Control (see Advanced Application Notes 1 and 2 entries March 25, 2009 on this website), or override control is used to maximize feeds to limits of operating constraints (e.g. maximum vent, overhead condenser, or jacket valve position with sufficient sensitivity). Alternatively, the limiting valve can be set wide open and the feeds throttled for temperature or pressure control. For pressure control of gaseous reactants, this strategy can be quite effective. For temperature control of liquid reactants, the user needs to confirm that the inverse response from the addition of cold reactants to an exothermic reactor and the lag from the concentration response does not cause temperature control problems. All of these methods require tuning and may not be particularly adept at dealing with fast disturbances unless some feedforward is added. Fortunately the prevalent disturbance is a feed concentration change that is often slow enough due to raw material storage volume to be corrected by feedback control.

8) Profile Control - If you have a have batch measurement that should increase to a maximum at the batch end point (e.g. maximum reaction temperature or product concentration), the slope of the batch profile of this measurement can be maximized to reduce batch cycle time. For application examples checkout "Direct Temperature Rate of Change Control Improves Reactor Yield" in a Funny Thing Happened on the Way to the Control Room E-book April 3, 2009 entry on this website and the Control magazine article "Unlocking the Secret Profiles of Batch Reactors"

This blog was kind of fun to write "With A Little Help from My Friends" (beer and music). By the way, the album "Joe Cocker" was apparently only produced on vinyl and 8-track tape. I fondly remember riding in my roadster with the top down listening to "Saint James Infirmary Blues" on the way to Monsanto's New Orleans plant. There are great songs on this album that carried me through startups that never made it onto a CD.

I got an excellent question during Deminar #3. An attendee asked how fast does the readback of actual valve position need to be as a secondary variable from a smart positioner. I said it depended on the speed of the valve. For flow loops, I thought once per second would be fast enough. However, since the communication of the actual valve position is not synchronized with PID module execution, there needs to be more than one communication per module execution time. Also, for very fast valves, the valve response time could be much less than the module execution time. The dynamic reset limit needs to know the valve is actually moving or it will slow down the change in controller output. For wireless communication of position measurement, exception reporting could be used where the deadband for updating the position readback is the resolution limit of the valve.

A guideline for the conventional PID could be:

When the controller output changes by an amount greater than resolution of the valve, the communication of the valve position for the dynamic reset limit of a conventional PID should be less than ½ the module execution time and less than ¼ the valve response time.

For an enhanced PID as described in Deminar 1, it is possible that valve position only needs to be communicated when a new measurement value is communicated.

The response time per the ISA-75.25.01-2000 (R2009) standard Test Procedure for Control Valve Response Measurement from Step Inputs is the time the valve takes to reach 86% of the final stroke. As noted in slides 12 & 13 in Deminar 3, the response time for small signals and small actuators is a second order exponential response (response time is approximately twice the sum of the time constants) whereas the response time for large signal and large actuators is a ramp (e.g. response time is 86% of the step change in signal (%) divided by the slewing rate (%/sec)). For valves with hydraulic or digital actuators or small valves with a negligible deadtime from backlash and stiction and with a high sensitivity actuator and positioner (e.g. sliding stem valve diaphragm actuator and digital positioner), the response time could be less than a second. For extremely large valves with excessive deadtime from backlash and stiction and with a low sensitivity actuator and positioner (e.g. piping valve with scotch yoke actuator and pinned shaft connections) the response time could be more than 100 seconds. Thus, we have the ironic situation, where if we have a poor valve choice, the resolution and update rate of actuator position communication can be decreased and the filtering of noise can be decreased to keep fluctuations in controller output from measurement noise less than valve dead-band and resolution. If you don't do small step tests or have no communication of actual valve position, the poor loop performance from a piping valve posing as a control valve may be attributed to disturbances or noise.

The accuracy of the valve position communicated is not as important as precision since it is the change in valve position rather than the value of valve position that is important. The bias and span errors in valve position are corrected by feedback control of the process loop. Since even the best valves with pneumatic actuators do not respond to changes in signal less than 0.1%, the greater resolution of digital values of valve position communication is unnecessary. Consequently, to get faster communication for fast valves and small signal changes, analog signals of valve position should be used for the dynamic reset limit even though they may not be as accurate as digital signals.

The precision of the valve position communication should be better than resolution limit of the control valve (e.g. 0.1% for sliding stem valves with diaphragm actuators and digital positioners).

All of what I have presupposed here needs to be tested and investigated. There is no shortage of interesting scenarios to investigate via dynamic simulation.

The first Deminar is history. The seminar-demo showed how an enhanced PID controller can reduce cycling caused by sampled measurements. The benefits are not only the obvious one of less process variability but includes extending valve packing life by reducing the accumulated valve travel and battery life of wireless measurements by reducing the number of communications. The name of this series of live meetings was the result of me mistakenly saying "Deminar" when I meant to say "Seminar-Demo."

To keep the demo fast enough the process dynamics were in seconds instead of minutes. In other words, the 1 second deadtime and 10 sec time constant of the primary process were chosen to be indicative of a well mixed vessel with a mixing delay of 1 minute and a residence time of 10 minutes. Setpoint changes were made to show the response of a standard PID and an enhanced PID (DeltaV PIDPLUS). In future labs, the testing and importance of dealing with load disturbances will be discussed and demoed. Even though the process dynamics were relatively fast, I did not want to waste precious viewer time or risk viewer boredom staring at a trend chart waiting for the response to develop. Consequently, I shuffled back and forth between the demo and the seminar presentation WebSeminarDemoLab01.pdf and user screens to discuss the concept of the enhanced PID and flexibility of the lab and virtual plant to explore, test, and quantify process control improvements. I could have presented comparison trend charts of a traditional versus enhanced PID as typically seen in most presentations but choose to make the demo more interactive and show the dynamic transition when the enhancement was turned on.

The demo started out with a controller tuned for composition control of a self-regulating process with an online analyzer providing a continuous measurement of vessel composition by means of a probe (e.g. NIR probe in a circulation line). The setpoint response of the standard PID for the continuous measurement was fast and non-oscillatory with almost no perceptible overshoot.

I then set the sample time to be twice the primary process time constant and made another setpoint change. If the time scale was minutes instead of seconds, the 20 minutes sample time would be typical for a chromatograph. Now the setpoint response exhibited a significant overshoot and oscillation. I then cut the reset time in half, a common scenario because of tuning misconceptions or change in process dynamics. The setpoint response developed severe and persistent oscillations . When I switched on the PID enhancement, the oscillations quickly died out. A subsequent setpoint change showed that the enhanced PID response had no overshoot or oscillation.

The last test involved the removal of the sample time and the addition of a 2% sensitivity limit to show the result of an analyzer or wireless measurement with a detection or reporting threshold (called deadband for wireless measurements). The sensitivity limit was purposely chosen to be larger than typically expected to show a clearly recognizable oscillation. I had intended to switch back right away to the traditional PID but instead made the setpoint change to the enhanced PID. I wondered why the response did not show the expected cycling until I realized I had forgotten to switch back to the traditional PID. When I did make the switch to the traditional PID, the cycling started but we ran out of time to show the subsequent limit cycle (perpetual constant amplitude square wave cycle in the process variable and saw tooth cycle in controller output).

For your viewing pleasure, checkout the ScreenCast courtesy of Jim Cahill.

We expect to have the audio glitches worked out for the next Deminar on "PID Control of Valve Sticktion and Backlash" set for April 21 at 1:00 Central Daylight Time - my personal apologies to Europe about the time.

Let's pull together this series on errors and conclude with a check list. The idea was prompted by perusing a popular book written on just the value of check lists. I didn't think you could write a book on just one concept but the result of saving lives for surgical procedures is impressive. I know as I have gotten older, check lists are essential to just remember what I am suppose to be doing. I have found checklists to be helpful for me from both a practical and psychological viewpoint when rushed or overwhelmed with details, tasks, and objectives.

In the following list, increases in on-stream time can increase efficiency besides capacity by eliminating the time and off-spec and waste associated with abnormal operations, startup, and shutdown. An increase in yield or decrease in recycle can be taken as a decrease in raw material costs (same production rate for lower feed rate) or an increase in production rate (higher production rate for the same feed rate). The order of the list is in order of things to check and somewhat in the order of priorities.

Check List to Improve Process On-stream Time, Production Rate, and Efficiency
(composition measurements include conductivity, dissolved oxygen, pH, and turbidity)

1. Use smart transmitters with the best sensor technology and integration of process and ambient conditions compensation.

a. Avoid older technologies particularly ones with mechanical elements

b. Seek sensor and transmitter with the best sensitivity and repeatability

2. Pick sensor location and installation method to provide the most representative measurement in process with no stagnation, best velocity, fastest response, and least noise.

a. For DP and pressure transmitters, avoid impulse lines (sensing lines) by direct mounting transmitters or using diaphragm seals and filled systems

b. For DP and vortex flow meters insure uniform velocity profile

c. For thermowells and electrodes increase velocity to reduce response time and coatings but not so high to cause abrasion, static charge, or vibration

d. For thermowells and electrodes pick location with good mixing, minimal transportation delay, and least bubbles, slime, and solids

3. Use real throttle valves with smart positioners.

a. Avoid on-off and isolation valves posing as throttling valves. Go to a control valve manufacturer instead of a piping valve manufacturer

b. Seek actuator, positioner, and valve type with best sensitivity of installed flow characteristic and signal response with least stick-slip and backlash

c. Verify positioner feedback measurement is representative of internal closure member (e.g. ball, disk, or plug) and not just actuator position

4. Tune control loop with on-demand auto tuner or adaptive controller to meet loop objectives. Tuning speed is chosen to:

a. Insure an exceptionally smooth PV and output response by decreasing transfer of variability from PV to output (increasing Lambda) for:

i. level loops on surge tanks to minimize feed upsets
ii. deadtime dominant loops (deadtime >> process time constant)
iii. interacting loops (e.g. headers)
iv. loops on piping or equipment with no back mixing (e.g. blenders, heat exchangers, extruders, static mixers, sheets, webs, and yarns)

b. Provide good load rejection of moderately fast disturbances by increasing transfer of variability from PV to output (decreasing Lambda) for:

i. Fed-batch and continuous agitated vessel and column composition, level, pressure, and temperature loops

c. Provide good load rejection of extremely fast disturbances by setting the gain and reset as a factor of deadtime rather than the time constant for:

i. Continuous agitated vessel and column composition, pressure, level, and temperature loops

d. Provide minimal overshoot of setpoints of slow lag dominant loops (process time constant >> loop deadtime and slower than 10 minutes) by tuning the loops as near-integrating processes for:

i. Fed-batch and continuous agitated vessels and column composition, pressure, and temperature loops (setpoint changes occur at startup or for changes in batch phase and product grade)

e. Provide minimal peak error by maximizing controller gain even if it requires increasing reset time to maintain robustness for:

i. Prevention of SIS activation
ii. Prevention of pressure relief
iii. Prevention of environmental violation
iv. Prevention of equipment damage

5. Add DCS signal filter or damping adjustment to keep loop output fluctuations from noise less than the valve deadband to prevent excessive valve packing wear and inflicting disturbances on loop. For wireless transmitters use damping adjustment to reduce keep transmitter output fluctuations from noise less than wireless deadband to eliminate unnecessary communication and extend battery life.

6. Eliminate on-off actions

a. Replace on-off control by switches with loops.

b. Eliminate manual actions by adding loops, keeping loops in highest design mode, adding feedforward, and automating and tuning loops to handle startup and abnormal operating conditions

c. Replace pure batch with fed-batch automation by replacing discrete sequential actions (e.g. stepping feeds) with loops (e.g. throttling feeds)

7. Tune loops that create feed disturbances (e.g. surge level loops) to provide a smooth slow transition in feed rate.

8. Add cascade control to compensate for nonlinearities and pressure disturbances (e.g. secondary flow loop and secondary coolant temperature loop).

9. Add feedforward control of measurable fast disturbances not compensated by secondary loop.

10. Optimize setpoints by operating closer to constraints for production rate or product quality spec.

a. Eliminate operating margin imposed by shift's perceived sweet spot or operating margin caused by process variability from not doing check list items 1-9

b. Find more efficient operating points based on R&D reports and virtual plant exploration - confirm with process tests

b. Add model predictive control to optimize setpoints as process conditions and market requirements change.

At my recent presentation to the ISA Saint Louis section meeting on "pH measurement", I had several people around my age say how nice it was to see me still involved in advancing our profession. Maybe it was the beer and the top ten lists but just maybe it was also that I represent a generation of expertise rapidly disappearing via retirement. The ability to still learn and share keeps me going but I realize time is running out so I intend to take this blog to the next level by coupling it with a web lab series to provide an interactive self-learning experience for exploring process control improvements (PCI). I intend to start the web lab series on April 7. Recordings of the PCI topics and demos along with instructions on using the associated labs will be viewable anywhere anytime.

In the meantime, we need to finish up this series so let's see what we can do as automation engineers to minimize loop errors.

The first thing is to make sure the measurement is fast and precise enough. So far as loop performance is concerned, precision is more critical than accuracy. The bias or offset in a measurement and control valve position can be corrected by feedback control. The offset in valve position is eliminated by the process loop. Similarly, the offset in a process loop is eliminated when the loop is in cascade or remote cascade modes. For loops operated in the auto mode, operations have often compensated for the measurement offset by tweaking the set point. This is not to say that measurement accuracy is not important.

Improving the loop's speed of response often comes down to keeping sensors clean (e.g. electrodes and thermowells), minimizing signal damping and filtering, selecting sensor locations that eliminate transportation and stagnation delays, using boosters for big valves, maximizing positioner sensitivity, minimizing deadband, and maximizing the controller gain (last week's blog).

Control loops have a difficult time dealing with the poor precision experienced as excessive stick-slip and backlash (deadband) in control valves and insufficient resolution, repeatability, and sensitivity in measurements (older measurements technologies, such as floats and rotameters can also exhibit stick-slip and backlash). Fortunately, an increase in A/D input card bits have greatly improved the resolution of transmitted signals so that sensitivity and repeatability is the remaining focus. This is unfortunately not the case for variable frequency drives manufacturers whose standard input cards have only 8 bits. A resolution limit is more degrading than a sensitivity limit. For example for a 1% resolution and 1% sensitivity and a change in the true process variable of 1.5%, the changes in measurement would be 1% and 1.5%, respectively. The deadband setting in wireless transmitters is really a sensitivity setting. When the change in a wireless transmitter measurement exceeds this setting regardless of the direction, the full change in the process variable is communicated.

Pages 12 through 15 of EffectsLoopTuning&Dynamics-KPI.pdf show the relative effect of measurement accuracy and resolution on variability. For control valves, process variability is introduced when excessive slip-stick and deadband causes an appreciable limit cycle in loops that have single and two or more integrators, respectively (pages 19 and 20).

The total loop deadtime can be approximated as the sum of all the delays and small lags in the loop whether they are in the DCS, valve, process, or measurement. For flow, pressure, level, and inline temperature and pH loops, most of the loop deadtime comes from the automation system. If you consider that the remaining loops that have significant process deadtime, such as vessel or column temperature, have seriously detuned controllers that create an effective deadtime per Advanced Application Note 5, you realize you have the opportunity as a process control engineer to make big reductions in loop deadtime that are also low cost and quick compared to changing process piping or equipment to reduce transportation or mixing delays.

Fast disturbance originate from manual operation, on-off actions, sequences, or setpoint changes. The elimination of operator actions, on-off control (e.g. level switches), and the use of set point rate of change limits and fed-batch rather than pure sequential batch, can dramatically slow down disturbances since throttling control by intention is smooth. If we keep all loops in their highest design mode and limit on-off valves to SIS actions and isolation, we could eliminate step disturbances. Page 22 shows how slowing down the disturbance dramatically reduces the peak and integrated errors for an integrating process. Not shown here is the fact that slowing down disturbances can also reduce interaction between loops. This phenomenon explains why it is difficult to get pharmaceutical companies excited about doing a better job of bioreactor control after reaching setpoint. The disturbances from cells are incredibly slow (e.g. process time constants of days).

Maybe we should not slow down disturbances because all of our control texts are based on step disturbances. Slowing down the upsets relegates us to improving the set point response in the startup of a continuous process or for changes in phase in a batch process. Whoops, even here we could use strategies such as "Full Throttle Batch and Setpoint Response" to eliminate most of the job of the loop.

There are always opportunities to make us more appreciated even when we are not improving loops. Since spouses were at the ISA Section Saint Louis Meeting, I interjected the following list. The spouses laughed although a second opinion was suggested for some of the items. See what your spouse or significant other thinks.

Top Ten Reasons Why an Automation Engineer Makes a Great Spouse or at Least a Wedding Gift

(10) Reliable from day one
(9) Always on the job
(8) Low maintenance - minimal grooming, clothing, and entertainment costs
(7) Many programmable features
(6) Stable
(5) Short settling time
(4) No frills or extraneous features
(3) Relies on feedback
(2) Good response to commands and amenable to real time optimization
(1) Readily tuned

How does controller tuning affect on-stream time and environmental costs?

The basic process control system (BPCS) forms the inner protective layer for safety instrumentation systems (SIS) as shown on page 5 of EffectsLoopTuning&Dynamics-KPI.pdf. The performance of the BPCS loops must limit excursions to be well within the operating limits that correspond to the trip points of the SIS. Specifically, the peak error for the largest and fastest disturbance should not cause a trip. The SIS should only be activated for failures or extremely abnormal conditions. The trip of a process unit not only causes downtime but can cause off-spec and additional waste during the shutdown and startup of the unit. The start-up of the process unit is often the most operator intensive and hazardous time. The importance of minimizing peak errors to prevent shutdowns can involve all types of loops (e.g. flow, level, pressure, and temperature). We normally think this is important only for continuous loops but I have been able to increase a fed-batch reactor capacity by 25% by eliminating level, pressure, and temperature trips by a series of override controllers tuned to minimize peak errors.

The peak error from the closure of a downstream valves (e.g. trip of reactor feed valves) on the discharge of a compressor controller must not cause an excursion of the operating point of the compressor to the left of the surge curve. If the operating point reaches the negative slope of the characteristic curve, it is like the compressor is falling off a cliff. The operating point jumps to a negative flow operating point in 0.03 seconds. This precipitous drop rivals water hammer in disturbance speed (both phenomena involve momentum balances that are orders of magnitude faster than material balances). Once a compressor gets into surge, the feedback controller is helpless and needs an open loop back-up (e.g. kicker) to get out of trouble as detailed on pages 6 - 8. Surge cycles can cause a decrease in compressor efficiency and damage by excessive vibration. I have also seen where surge caused a runaway speed response.

RCRA environmental regulations may classify a pond as hazardous waste if the pH of an effluent stream going into the volume momentarily ventures outside the permissible 2 to 12 pH range. Even though a short term excursion can not possibly change the pH in the volume and is effectively filtered by the volume where the change in pH is not detectable, the volume may still be classified as hazardous. For these systems, peak errors are incredibly important and kickers are used as shown on page 10 to prevent RCRA violations that not only can cause excessive fines but necessitate the process unit to apply for a new permit that might not be approved. A violation could result in the permanent shutdown of a unit because operation is no longer economically feasible or even allowed under new permit requirements.

Many process units have relief devises (e.g. relief valves and rupture discs) to prevent the over pressurization of piping and equipment. Often, pressure letdown and vent loops are the first line of defense. The peak error for the largest and fastest disturbance should be sufficiently away from the relief device setting to prevent fatigue and activation of the relief device taking into account setting tolerances and fatigue that cause a premature relief. The activation of a relief device is hazardous and causes downtime and waste burned in a flare stack or at best in a waste heat boiler.

So how do we minimize peak error? Given a set of dynamics and disturbances, the solution is to maximize controller gain even if it means increasing the reset time. This is seen in the first equation on page 1 but also intuitively from the realization that gain provides an immediate response whereas reset provides a gradual response. In the fed-batch reactor example cited above, the override controllers were proportional-only with their gains set high enough to cut back the reactant feeds immediately when the reactor pressure and level from the gas released as a byproduct or the temperature from the exothermic reaction approached settings that would cause a trip and the associated delay and disruptive restart of the feeds.

We can reduce the peak error per the first equation on page 2 by increasing the process time constant and decreasing the loop deadtime which increases the maximum allowable controller gain. We can also decrease the open loop error in the time frame of the controller's response by increasing the disturbance time constant. The fastest possible tuning should be able to stop the excursion from a disturbance after the loop deadtime. Thus, slowing down the disturbance slows down the excursion and reduces the peak reached in one loop deadtime (more on this next week). The process time constant is typically set by process equipment size and design, but we as automation engineers can greatly affect the disturbance time constant and the loop deadtime and sensitivity. We can iimprove the degree of automation, interaction, speed, reliability, and precision in automation systems. The opportunity may be larger than we realize. Up to 50% of downtime is attributable to instrumentation problems as noted in the March 2010 Control magazine article "Look to Valves for More Uptime"

Next week we will look at how the dynamics and precision of measurements, valves, and disturbances affect peak and integrated errors. We conclude this series with a check list for improving loop performance (Part 4).

If you increase the controller gain by the same factor that you increase reset time (e.g. double the gain and the reset time), how does it affect key performance indicators such as quality, yield, on-stream time, and environmental costs? If you make the valve and measurement faster, how does it affect these same KPI? If you want to improve a KPI, what is the priority of solutions?

The equations for the peak (Ex) and integrated error (Ei) in terms of controller settings, shown on slide 1 of EffectsLoopTuning&Dynamics-KPI.pdf, provide an answer to many of these questions if you embrace your inner geekness as advocated in the Control Talk Jan 2010 issue "The Future is Now"

Both equations were derived in Appendix A and B of Tuning and Control Loop Performance (scheduled to be back in print by Momentum Press, 2010). The derivation of the equation for the integrated error was included in Appendix C of New Directions in Bioprocess Measurement and Control (ISA, 2007) along with a unification of controller tuning rules. This unification, which showed how all the major tuning rules give basically the same result for a controller gain to minimize peak error, was personally satisfying but possibly not for people who are adamant about the relative merits of personal favorite tuning rules.

Since the integrated error is inversely proportional to the controller gain and proportional to the reset time, doubling the controller gain and reset time cancel each other out. However, doubling the controller gain halves the peak error since reset time doesn't appear in the equation of the peak error. Reset time has an effect on peak error but it is negligible unless the reset time is decreased to the point where it approaches the loop deadtime. This can happen for deadtime dominant systems, but the peak error here is basically the open loop (error with the controller in manual) as evident from the equations on slide 2 of EffectsLoopTuning&Dynamics-KPI.pdf.

Nearly all the process control literature focuses on integrated absolute error (IAE) as the measure of loop performance. The IAE is a good measure of product that is off-spec that can lead to reduce yield and the raw material or recycle processing to product cost ratio (euros per kg and dollars per lb). If the off-spec cannot be recycled or the feed rate cannot not be increased to compensate, there is also a loss in production rate. If the off-spec is not recoverable, there is an additional waste treatment cost.

What we usually don't take into account is the filtering effect of back mixed volumes as indicated by the equation on slide 3 of EffectsLoopTuning&Dynamics-KPI.pdf. For chemical and pharmaceutical plants and refineries, there are large volumes that provide significant attenuation of oscillations. However, in other process industries, various pathways of variability do not have significant filtering and culminate in the final product. These processes are also more vulnerable to interactions because there is no smoothing of effect of one loop's control valve movement on another loop's process variable. This changes the whole view on how you tune controllers. For systems with little back-mixing, controllers are tuned to limit the transfer of variability from the controlled variable (controller PV) to the manipulated variable (controller output) to prevent interactions and to provide a smooth response. The controllers are also tuned for coordination by enforcing a closed loop time constant (Lambda). For pulp and paper plants, nearly all of the variability expressed by the IAE ends up in the sheet since most of the processing is done in pipes and inline or unagitated equipment. Lambda tuning has been exceptionally successful in optimizing the transfer of variability and the coordination of loops. The same requirements could occur for plastics and textiles, since the IAE in the polymer lines and extruders shows up in the yarns and webs. However, these plants may have extensive blend tanks that average out the plus and minus fluctuations in product quality.

I ran into a process control improvement (PCI) study, where after an hour of discussion and investigation it became obvious a reduction in the considerable variability observed in each textile line had no value because the product coming out of the huge blend tank was always in spec and the variable speed pumps were maxed out. My decision to move on to better opportunities was not well received, so we stayed for 2 days to confirm there were no PCI opportunities (reducing the size or inventory in the existing tank or replacing the pumps were considered accounting or process design improvements).

When loops are oscillating across the split range point (common case due to valve stick-slip and installed valve characteristics), there can be a cross neutralization of acids and bases or a cross compensation of hot and cold heat transfer fluids that increases reagent and energy costs. Here the IAE is important but an integration of individual reagent and heat transfer fluids is a better indication.

If there are appreciable back mixed volumes whose residence time is much larger than the control loop period, the integrated error (Ei) where the plus and minus errors cancel out for a disturbance can be a better indication of the effect on product quality. Taking into account that the integrated error is also the IAE for an over-damped or critically damped response, we realize the simplification of the relationship of off-spec to an integrated error offers considerable understanding as to the effect of tuning settings.

This topic will roam on for 4 parts. In part 2, I discuss the effect of the peak error on onstream time and environmental costs. In part 3, I cover how measurement and valve dynamics impacts both types of errors and hence KPI. In part 4, I conclude with some rules of thumb on the priority of PCI solutions for various scenarios.

It is well known that measurement noise reduces or eliminates the use of derivative action. Since rate is not popular (another story), the exclusion of rate is not seen as a significant disadvantage even though temperature loops could benefit from rate since it can compensate for thermowell and heat transfer surface lags and reduce overshoot. In the 1980s and 1990s many temperature loops suffered from the prevalent use of 12 bit I/O and wide range thermocouple input cards that caused a resolution error of 0.25 degrees in a signal whose true rate of change of temperature was usually much slower than 0.25 per minute. The result was a poor signal to noise ratio. We tried to filter the heck out of the signal so we could use rate but this added another lag. Fortunately, today we have 16 bit I/O systems and smart transmitters so that signal resolution is better than the sensitivity of the sensor - just one of the many reasons to get your automation system into the 21st century.

A wider consequence of measurement noise not so readily recognized is the reduction in permissible controller gain. For loops with a true integrating or "near integrating" response where the process variable ramps when the controller is put in manual, the high limit for controller gain is way above the normal range of consideration. For example, level and batch temperature loops normally have a ramp rate so slow (0.000001 %/sec), that the controller gain could be higher than 50 if there was no measurement noise and the reset time was not too small (a big "if"). Since the peak and integrated errors are inversely proportional to the controller gain, these and other loops could significantly benefit from a smoother signal and better tuning.

What is measurement noise and where does it come from? In my book, measurement noise is any fluctuation in the measurement signal that should be ignored by the controller. If the controller reacts to a fluctuation it really cannot correct, the loop inflicts a disturbance upon itself. If resolution problems are behind us, the biggest sources of measurement noise are inadequate axial (back) mixing, bubbles and foam in liquids, liquid droplets in steam or gas, inconsistent profiles, lqiuid and pressure waves, and insufficient measurement rangeability. Measurement noise is amplified by high process gains (e.g. steep titration curve for pH control) and sensitive measurement ranges (e.g. - 0.25 to 0.25 inches of water column for draft pressure control). The Table in MeasurementNoiseSourcesControlBandAmplitude.pdf provides a summary of my assessment of noise sources, control bands (allowable control error), and noise amplitude (peak to peak) for common loops. The noise amplitude should be less than ¼ the allowable control band for fast disturbances. A reduction in noise amplitude is ideally achieved by eliminating the source of the problem. If the correction is not practical or is not yet implemented, a signal filter is often used to attenuate the noise. The ratio of the amplitude of the filtered signal to raw signal is roughly proportional to the ratio of the period to the filter time when the filter time is greater than the period (simplification of the Bode plot attenuation equation). The filter time becomes effectively additional deadtime in a loop when it is less than the process time constant. If the filter time is considerably greater than the process time constant, the measured process variable amplitude may look better but the real amplitude is worse because you are seeing a very attenuated version of the real world. I have seen where an ISA conference speaker said he almost did not get permission to give his presentation because the improvement was so great it was considered proprietary. He had increased the measurement filter so much he was drawing a straight line no matter what was happening in the process. I have seen where a biochemist withdrew a temperature sensor halfway in its thermowell and proudly said this was the way to run the bioreactor because the temperature reading was so much smoother. Then there were the cases of sand in thermowells and the mounting of extruder temperature sensors in massive blocks of metal giving the illusion of smooth temperature. These are all old stories but I am sure people are being fooled today especially since one can so easily add a filter via the damping setting in the transmitter, the analog input block, and the PID block. Provided the filter setting is not so large it eliminates any recognition of process variability, the key symptom of too large of a filter setting is a long control loop period or recovery time if the controller gain is not so detuned you can't see the effect of more loop dead time (see Advanced Application Note 5 for estimation of how the detuning of a controller is equivalent to additional deadtime in the loop). To prevent the loop from inflicting disturbances upon itself by reacting to noise, the filter time should be set just large enough to keep the fluctuations in the controller output smaller than the resolution (stick-slip) of the final control element (e.g. control valve). A less desirable but widely used way of keeping the fluctuations in the controller output small enough is to reduce the controller gain.

Simulation was such an integral part of my job it is difficult for me to visualize a process control career without models. I was asked to join Engineering Technology (ET) at Monsanto in 1976 because I had developed a dynamic compressor model as the lead Instrument and Electrical engineer for what was the largest Acrylonitrile plant in the world. I developed the model in order to understand more about the incredible surge phenomena where reversals of flow could occur in less than 0.01 seconds leading as a minimum to a loss in efficiency and in some cases to the damage of shafts and seals of large and expensive compressors from the extreme momentum swings and vibration. In most plants the ability to initiate and explore abnormal situations is severely limited or not allowed. A dynamic model allows you to readily and quickly try out "What if Scenarios" whose only limit is your imagination.

ET developed FLOWTRAN, a process simulator that was directed by the government to be sold to Aspen institute. Several key specialists left with the FLOWTRAN to develop the process modeling software that eventually was the state of the art process design modeling software by AspenTech. In the ET process control groups, we used FLOWTRAN to get the process gains and then used IBM's Continuous System Modeling Programs (CSMP) followed by Raytheon's Advanced Continuous Simulation Language (ACSL), and ultimately HYSYS Plant for dynamic simulations. After retirement from my career in ET, I focused on using the DCS as a Virtual Plant for simulation and control. The graphical configuration environment where function blocks are equipment and wires are streams (e.g. DeltaV Control Studio and MiMiC) allows the development of dynamic process models in the same familiar way as the configuration of control strategies.

My vision of a virtual plant has a simple first principle model that starts with one component (e.g. water and air) that is corrected by an experimental model automatically generated by a simple test that takes less 10 minutes to execute for most loops. The result is a plant wide simulator. As more information is available and desired, the process knowledge embedded in the model grows but the fundamental basis is the same. No re-write is required. The opportunities and associated fidelity needed are as follows:

1. Control system set point optimization - Fidelity 5

2. Control strategy analysis and R&D - Fidelity 4

3. Root cause analysis and data analytics R&D - Fidelity 4

4. Operator training for abnormal situation management - Fidelity 4

5. Controller tuning and PID structure and options analysis - Fidelity 3

6. Batch configuration checkout and operator training for system familiarization - Fidelity 2

7. Loop configuration checkout - Fidelity 1

Fidelity 1: loop process variables respond in the proper direction to their loop output

Fidelity 2: measurements respond in the proper direction when control and block valves open and close and prime movers (e.g. pumps, fans, and compressors) start and stop.

Fidelity 3: loop dynamics (e.g. process gain, time constant, and deadtime) are sufficiently accurate (e.g. 50%) to tune loops and see process interactions

Fidelity 4: measurement dynamics (response to valves, prime movers, and disturbances) are sufficiently accurate (e.g. 25%) to track down and analyze disturbances

Fidelity 5: process metrics (e.g. yield, raw material costs, energy costs, product quality, production rate, production revenue) are sufficiently accurate (e.g. 5%) to find optimums

In the ISA New Orleans section short course I am teaching on March 3 and 4 titled: "Exceptional Process Control Opportunities - An Interactive Exploration of Process Control Improvements", I will use a virtual plant suitable for process control research, development, and education. I will demonstrate how a user can perform a 10 minute test of a manipulated process flow to provide a fidelity level 3 and 4 model. The contact for the course is Robert Deeb (ISA New Orleans section education chairman).

In the InTech Jan-Feb 2010 Web Exclusive "Advances in Flow and Level Measurements Enable Dramatic Improvements in Process Knowledge and Control", the following perspective was offered on the importance of flows for many types of process models including the following:

• Projection to Latent Structure or Partial Least Squares (PLS)
• Model Predictive Control (MPC)
• PID Adaptive Controller Tuning
• Neural Network
• First Principle

Flows determine what is going on in a process. If you don't get the flows right, not much else matters. Because of valve backlash, stick-slip, nonlinearities, and variable pressure drop, all types of process models have suffered from the use of valve positions rather than flow measurements. PLS, MPC, and PID performance assumes dynamics that are linear and independent of direction and size, all bad assumptions when valve positions rather than flows are used as inputs. Additionally, the valve nonlinearity from the installed characteristic varies with pressures at the inlet and outlet of the valve.

Pioneering advances in dynamic modeling by Alex Muravyev offer a next generation of pressure-flow solvers that will be robust and flexible enough to provide flows from valve positions. The solver is expected to handle complex piping networks and the discontinuities from batch and startup sequences (AdvancedSimulationPressureFlowSolver.pdf). The ability to consistently and comprehensively compute flows for all streams will enable dynamic models to reach the highest levels of fidelity required for research, development, and design of automation systems for nearly all applications. Presently, models can only move up in fidelity when flow control loops are installed on the key streams so that feedback action removes the nonlinearity and unknowns of the valve and piping system. New pressure-flow solvers can eliminate this precondition. A side benefit will be the demonstration by these models of the improvement in process performance that can be gained from cascade, feedforward, and ratio control. The quantifiable benefits from demonstrable test cases can justify new flow devices to provide missing flow measurements or improve the accuracy of existing flow measurements.

For comic relief for the New Year I offer here some of my broken resolutions published in my Control Talk column plus a new number one.

(20) Listen intently to my wife's instructions. Why does my mind still jump to weighty matters like what is next for dinner?

(19) Stop making cheap control valve jokes. Could the next final element reputation I hurt be my own?

(18) Help make smart diagnostics smarter. Do I need to de-fussify my fuzzy logic?

(17) Stop lusting in my heart for more computing power. Is it the PC or me that is the constraint?

(16) Turndown the volume on my headphones. What did you say?

(15) Stop drinking cheap wine. Does good wine ever come in a size large enough?

(14) Read a college text on control theory. Can I watch Star Trek without setting up the state space equations?

(13) Stop answering a question with a question. Why should a consultant do this?

(12) Spend more time with my wife than with Control magazine. Whatever happened to my January issue?

(11) More bark than bite. Can I at least growl? Will I be forced to wear an anti-bark collar? Is this better than a muzzle?

(10) Stop making fun of seniors. Who else do I know? We certainly aren't an endangered species with the influx of baby boomers.

(9) Stop focusing on deadtime. What else is there at Sun City? Whoops, I am already breaking my last resolution.

(8) Final element resolution resolution. Why should I get unstuck when valves are stuck and it gives me a chance to repeat words?

(7) Get into hybrids. No issue here with fashion models. Can a hybrid face up to a 2 ton high lift truck with a cattle guard? Can I drive under a cow?

(6) Show my more sensitive side. Wait, will I confuse people including myself? Do I have to start watching "Brothers and Sisters?"

(5) Stop drinking cheap booze. I will give this another shot.

(4) Listen to hip-hop. What if I am not hip and can't hop?

(3) Become rich and famous. How about poor and infamous?

(2) Lean how to sell. How can I sell a product when I can't sell myself?

(1) Stop pushing the Essential Book even though the royalties are donated for wireless research at the University of Texas and the book is like a fine wine with a lush blend of technology with rich overtones, a balanced feel, and a lingering finish. What if readers like cheap wine? What if readers are not UT fans? What if they are rooting for Alabama? What if they think wireless research will be used in the BCS game?

The figures in the attached VariableSpeedDriveRangeability.pdf and the following discussion is an excerpt from the ISA book The Essentials of Modern Measurements and Final Elements - A Guide to Design, Configuration, Installation, and Maintenance.

The 4 main practical reasons that variable speed drives (VSD) drives are not used as extensively as one might think for pump control are as follows [35].

1. Drives are generally not built just for pumps. They handle conveyors, extruders, etc. There are a lot of VSD menu choices and options not pertinent to pumping applications.

2. Users don't like the complexity of the VSD. The user must address setup, maintenance, and design issues. Special practices are needed to prevent EMI in instrument signals and from getting harmonics back into the power supply.

3. Someone needs to do the right calculations on dollars saved. Typically calculations don't take into account the drop in drive efficiency at low speeds. The duty cycle (amount of time speed is really turned down) is not known in advance. If there is a high static head, the energy savings of a drive disappear.

4. It is rare to compare a VSD and valve. There are generally no decision points in the project for this comparison.

Is a Valve or VSD Faster?

Exceptionally fast loops (e.g. furnace pressure, liquid pressure, and surge control) can ramp off-scale in milliseconds. These loops have essentially a zero process deadtime and may have a high process gain due to a narrow control range (e.g. fractional inches of water column for furnace pressures). These loops require DCS scan times of 0.05 to 0.1 seconds. Special fast scan rate digital controllers or analog controllers are needed. DCS scan time requirements of 0.2 seconds or less signify a VSD opportunity. A properly designed VSD has no measureable dead time while control valves and dampers take anywhere from 0.2 to 2.0 seconds to start to move. For example, an incinerator pressure and polymer pressure loop that could get into trouble in less than 0.1 second required a VSD and analog controller to stay within the desired control band [20][23][35].

The VSD has a negligible time delay unless a deadband or dead zone is introduced in the drive electronics to reduce reaction to process measurement noise or a low resolution input card is used. A control valve or damper has a deadtime that is proportional to the resolution limit (sticktion) or deadband (backlash) divided by the rate of change of the process controller output. For large or fast changes in signal this deadtime disappears.

A pneumatic actuator has a pre-stroke deadtime that is the time it takes for the actuator pressure to change enough to move the actuator shaft. For large actuators, the pre-stroke deadtime can be several seconds unless a booster is added.

The inertial time constant of liquid flow response is inversely proportional to flow. Consequently, the process lag at low flow rates and at the initial start of flow can be quite slow (e.g. 5 seconds) compared to the process lag at normal flows (e.g. 0.5 seconds). The comparison between VSD and control valve response should be at normal flows.

In a published comparison of the dynamic response of a control valve and a pump for flow control for a system with negligible static head, the integral times were about the same for the VSD and valve loops. However, the controller gain could be increased by over a factor of 6 for the VSD loop. As a result, the set point response was faster [38]. In this test the valve deadband was about 8% and there was no static head. In unpublished lab test results of control valves with low sticktion, low backlash, and a digital positioner and a VSD with a volts/hertz PWM drive for liquid flow control, the speed of response of the valve and VSD were similar.

Variable speed drives, control valves, and dampers have a velocity limited exponential response. The velocity limiting in a drive depends upon the available motor torque and the inertia of the motor rotor, the pump shaft, and the pump impeller. The exponential term is generally much smaller for a VSD than for a control valve or damper. On the other hand, the velocity limiting is slower for a VSD unless the actuator size is large and boosters are not used. Consequently, for small changes in signal, a well designed VSD is faster. Conversely, for large changes in signal, a small control valve is faster (see section on dynamics). This leads to the conflicting statements about whether a VSD or control valve is faster. Which final element is faster often depends upon the size of the change in signal.

VSD Best Practices

To summarize, a VSD is most likely to offer energy savings or better loop performance as a final element for the following types of applications:

• Loops that require 0.2 seconds or faster scan time
• Valves and dampers with 0.5% or more sticktion or backlash
• Large utility flows
• Integrating and runaway processes without a secondary flow loop
• Low static head processes requiring frequent turndown

A tachometer or inferential speed feedback signal should be sent to the process controller in the DCS that is sending the signal to the drive. The speed feedback should be used in a similar way to the position feedback from a digital positioner to prevent the process controller output from changing faster than the final element can respond. The use of the dynamic reset limit option for the loops in the DCS can automatically prevent the process controller from outrunning the final element response (see section on dynamics).

For best performance users should consider the following during the specification and implementation of variable speed drive systems:

• High resolution input cards
• Pump head well above static head
• On-off valves for isolation
• Design B TEFC motors with class F insulation and 1.15 service factor
• Larger motor frame size
• XPLE jacketed foil/braided or armored shielded cables
• Separate trays for instrumentation and VFD cables
• Inverter chokes and isolation transformers
• Ceramic bearing insulation
• Pulse width modulated inverters
• Properly set deadband and velocity limiting in the drive electronics
• Drive control strategy to meet rangeability and regulation requirements
• Dynamic reset limiting using inferential speed or tachometer feedback

VSD Response

The response of variable speed drives more closely resembles a pure ramp with no rounding or time delay provided a filter or deadband has not been added in the drive electronics to attenuate process noise in the process controller output signal. The ramp time in the VSD depends upon the size of the load compared to the available torque from the motor. In general, the ramp time of a VSD is longer than the stroking time of a control valve but is shorter than the stroking time of a large damper. Longer than necessary VSD ramp times may inadvertently be imposed in the drive electronics.

There is essentially no sticktion or backlash in variable speed drives for axial and centrifugal blowers, fans, and pumps but this does not necessarily mean there is no resolution limit or deadband in the VSD response.

Controller outputs invariably have fluctuations that originate from process or sensor noise and transmitter resolution limits. These fluctuations are not representative of the actual value of the process variable and are best ignored. These fluctuations are particularly large and fast for flow and pressure loops. A deadband is sometimes introduced in the VSD electronics to prevent changing the speed. The effect may be a true deadband where the desired speed does not change upon a change in direction until the change in signal is larger than the deadband setting. The effect here is similar to backlash in a control valve. In other cases, it may be a deadzone setting, in that the desired speed does not change until the accumulated change in signal since the last change in speed is larger than the deadzone setting. Here the effect is similar to a resolution limit.

If there is no deadzone setting, the resolution limit in a VSD is largely determined by the input card. Assuming there is no sign bit, the VSD resolution limit is simply 100% divided by 2 raised to the number of bits (n) of the input card. Unfortunately, VSD manufacturers did not understand the limit cycle that would result from the resolution limit and offered an 8 bit input card (0.4%) as the standard card. Higher resolution input cards (e.g. 12 bit and 16 bit) should be specified to make the VSD I/O resolution comparable to the DCS I/O resolution.

VSD Installed Gain

In a variable speed drive for liquid flow, the pump characteristic curve shifts with pump speed. Since there is no control valve, there is no valve drop and the flow is at the intersection of the pump curve and the system frictional loss curve.

For a negligible static head and an idealized pump, motor, and VSD, the change in flow with speed is linear. If the static head is negligible, the loss in pump efficiency and the increase in slip at low speed, cause a decrease in gain (sensitivity) at low speed. This loss of sensitivity is seen as a flattening at low speed in the plot of flow versus speed.

If we ignore the loss in pump efficiency and increase in slip, a pump curve that approaches the static head will show a sharp bend downward to zero flow at low speed. The plummet of the speed at low speed causes a significant increase in gain and a nosier flow at low speeds [46].

A flat pump curve will cause almost a quick open type of flow characteristic. The high gain (sensitivity) at low speed can cause cycling [46]. Operation on a relatively flat pump curve can occur from improper pump selection or over-sizing.

VSD Rangeability

For variable speed drives, estimating rangeability gets tricky. The decrease in process gain from speed slip offsets the increase in process gain as the pump discharge head approaches the static head. If there are no overheating or cogging problems as suggested is the case for a pump and valve system with a well designed open loop (volts/hertz) PWM drive, high resolution input card, and negligible static head, the rangeability is normally 40:1. When the pump head is operating near the static head, the minimum controllable flow is set by rapid changes in the static head and frictional loss. These rapid changes could be due to noise and sudden or large disturbances. The speed can not be turned down below the amplitude of these fast fluctuations.

The rangeability of a VSD could drop to 4:1 for the following systems:

(1) Older VSD technologies such as 6-step voltage (excessive slip at low speed)

(2) Systems with a high static head (flow plummets to zero at a low speed)

(3) Operation on the flat portion of the prime mover curve (cycling at low speed)

(4) Hot gases (motor overheats at a low speed)

In the process industry, what a control loop eventually manipulates in nearly all applications is a flow via a final control element such as a control valve, damper, or variable speed prime mover (pump, fan, or compressor). Dampers and variable speed prime movers are commonly found in utility systems. Peristaltic pumps are used in labs and positive displacement pumps are used for extremely low additive flows in plants. In instances, mass flow controllers (thermal mass flow meters with an integrated PI controller and valve) and remotely set pressure regulators are used. However, in production units, control valves are used as the final element in 95% or more of the loops.

Do we know for a change in controller output, did the valve actually move and if so when? Do we know when the control valve is the source of process variability? Do we know what makes a valve "Good" or Bad" in terms of its ability to do its job?

In valve selection and specification, a lot of effort is put into making sure the valve passes the required flows, has minimal leakage, no plugging, and has materials of construction and packing that withstands process composition and conditions. The dynamic response is often neglected possibly because response criteria and requirements are not well understood. Since most loops are digital, the question comes down to whether the change in controller output in a given scan results in a change in position of the internal trim (closure component such as a plug, ball, or disc). Of course most valves will eventually re-position, but the internal trim may not move until the total accumulated change in the controller output is large enough to

(1) Exceed the sensitivity of the positioner and actuator
(2) Change the pressure in the actuator enough to move its shaft
(3) Work through the play in shaft/stem linkages or connections (backlash)
(4) Break free the internal trim from packing, seating, and sealing friction (sticktion).

The result is a delay and a jump followed by a slow transition to a new position. The jump from sticktion causes a limit cycle in any PI or PID control loop. The deadband from backlash causes a limit cycle in any PI or PID control loop on an integrating process (e.g. level or batch temperature). The delayed and slow response adds pure and effective deadtime, respectively, to the loop.

The ultimate question is what should a user specify in terms of valve response? The table ControlValveResponseCriteria.pdf provides a summary of the parameters that makes a valve rated "Great", "Good", "Fair", "Poor", and "Bad". For most loops where process variable deviations of 0.5% are tolerable, a "Fair" valve will suffice. For loops where tighter control is needed (e.g. column, crystallizer, evaporator, or reactor temperature), a "Good" valve is needed. For loops with high process gains (e.g. pH), a "Great" valve is required to prevent self-inflicted oscillations from limit cycles being larger than the allowable deviation around set point (pHControlValveSizeandResolution.pdf). For tight control in loops with extremely fast dynamics (e.g. polymer pressure and incinerator pressure) a "Great" valve or a special variable speed drive may be needed (see "Analog Control Holdouts" on this website).

The ISA-75.25.01-2000 (R2006) draft standard "The Test Procedure for Control Valve Response Measurement from Step Inputs" and ISA-TR75.25.02-2000 (R2006) draft technical report "Control Valve Response Measurement from Step Inputs", use the time to reach 86% of the final response as a major criteria. This assumes the step input size is larger than the valve resolution and deadband for steps in the same direction and reverse direction, respectively. This 86% response time for small steps can be estimated as the sum of the pre-stroke deadtime and secondary lag time plus twice the primary lag time. For example, the 86% response time of a "Good" valve would be about 1.3 seconds for a 0.5% step (0.2 sec + 0.1 sec + 2*0.5 sec). For large step sizes encountered in surge and vessel pressure control systems, the 86% response time can be estimated as the sum of the pre-stroke deadtime and secondary lag time plus the stroking time to reach 86% of the step size. For example, the 86% response time of a "Good" valve would be about 2.45 seconds for a 50% step (0.2 sec + 0.1 sec + 0.86*0.5*5 sec). Note that the actuator size, pneumatic connections, and accessory (e.g. booster, positioner, and solenoid valve) flow coefficient determines the pres-stroke deadtime and stroking time, The pre-stroke and stroking values are based solely on actuator shaft movement and are determined by the manufacturer for tests of an actuator not connected to a valve. The sensitivity of the actuator and positioner is the minimum change in signal that causes a change in shaft position within a reasonable time frame (e.g. 10 seconds). Diaphragm actuators and digital positioners have the best sensitivity. Rack and pinion actuators and spool positioners have the worst sensitivity. Pneumatic positioners and scotch-yoke actuators are also bad news. The deadband from backlash in stem and shaft connections and the resolution from friction in packing, seats, and seals are determined after the actuator shaft moves. For practical purposes, the sensitivity of the actuator and positioner can be combined with the resolution limit of the valve for a total resolution of the package.

I have been particularly sensitized to valve response due to working on pH, furnace pressure, and compressor control. To add insult to injury, a proliferation of piping valves with piston actuators and spool positioners developed as a result of the emphasis on tight shutoff and low cost rather than response. These on-off valves posing as throttling valves created a problem for all types of loops. The idea was if the on-off valve worked well for sequencing and safety systems and was already in the piping spec, why not slap on a positioner and make it a throttling valve. Often the process variability from valve limit cycles was attributed to unknown process disturbances since there was no readback of actual closure component position.

This blog is getting long so I will just close with some figures on valve dynamics (ControlValveDynamics.pdf) from my new book The Essentials of Modern Measurements and Final Elements - A Guide to Design, Configuration, Installation, and Maintenance.

In upcoming entries we will seek to sort fact from fiction and hopefully provide some insight on valve rangeability and variable speed drive dynamics and rangeability.

After all is said and done, articles and books have been the main method of advancing and sharing the technology for industrial process control.

I don't know of an undergraduate degree in process automation. Chemical, electrical, mechanical, and systems engineering programs offer an undergraduate course or two on process control. However, the typical university control course needs to spend most of the time on Laplace transforms, frequency response, and state-space to provide a theoretical understanding and groundwork for graduate courses. Outside of chemical engineering the focus is more on set point response and signal noise for servo mechanism and aerospace control. Consequently, the student doesn't learn about the critical characteristics of control for the process industry where nonlinearities, deadtime, valve stick-slip, unmeasured load disturbances, and incredibly long time frames are the cause of most tuning and control loop performance problems. Throw into the mix the unknown features of proprietary PID algorithms, and you have a script for islands of expertise. I personally like tropical islands so maybe this is OK. I could retire to one and conduct web based courses instead of doing cross word puzzles.

Courses may not be the whole answer considering that more than 80% of the details presented are forgotten. The PowerPoint slides often don't tell the real story. In my days, professors used the chalk board with only passing references to a book so my only record of knowledge is in notes long gone. Maybe the best way to make courses have a greater long term value is by providing labs for hands-on learning and refresher exercises, key memorable concepts, and resources for reference and further investigation. Audio should be combined with the presentation as exemplified by the slidecast of my Boston ISA presentation Exceptional Process Control Opportunities.

Considering that people don't have time to read books maybe courses and seminars and the structure of books themselves could provide better direction to areas of specific interest to solve problems. This is an argument for electronic books with interactive queries and demos.

For process automation, the articles and books written by practitioners are our best way of capturing and advancing the technology. Unfortunately users are not given the time or priority to write and most companies are reluctant to disclose information that could be considered to provide a competitive advantage for manufacturing. Consequently, suppliers of automation systems and services write most of the magazine articles and books on the practical application of process control. University professors write most of the journal articles and technical conference papers on the theoretical advancements in process control. The two groups don't talk much to each other. The use of industrial control systems for labs is one glimmering area of hope for the meeting of minds from universities and industry (see my last entry on "Exceptional Opportunities in Process Control - Expertise Development" and the June 1, 2009 entry "What I have Learned? - Bridging the Gap between Universities and Industry").

For me writing books was a way of organizing and expanding knowledge gained on the job. I found it allowed me to put technologies to bed (at least temporarily) so I could clear my head for the next area of expertise. My serious technical books in order of oldest to most recent publication date are: Axial and Centrifugal Compressor Control, Biochemical Measurement and Control, Continuous Control Techniques for Distributed Control Systems, Tuning and Control Loop Performance, Advanced Temperature Measurement and Control, Process/Industrial Instruments and Controls Handbook, Good Tuning - A Pocket Guide, Advanced pH Measurement and Control, Advanced Control Unleashed, Models Unleashed, New Directions in Bioprocess Modeling and Control, and The Essentials of Modern Measurements and Final Elements. My favorite book, which is a mostly serious collection of case histories written in a humorous way, is A Funny Thing Happened on the Way to the Control Room. My mostly humorous books in order of oldest to most recent publication date are: How to Become an Instrument Engineer - The Making of a Prima Donna, Logical Thoughts at 4:00 am, How to Become an Instrument Engineer - Part 1.523, Dispersing Heat Through Conviction, The Life and Times of an Automation Professional - an Illustrated Guide, and The Funnier Side of Retirement for Engineers and People of the Technical Persuasion. The last two books were written solely for comic relief.

While I had to largely write the books on my own time (except for the last serious one), the companies I worked for were supportive in terms of approval and recognition. In the end I expect books helped me along with my heroes Shinskey and Liptak to be the first group of inductees into Control magazine's Process Control Hall of Fame.

I think the following message titled "Why Books" from Ted Stillwell who is of the same vintage as me concisely offers "memories of the way we were."

Because I learned process control on the job books provided the only formal learning environment. Starting with the first treatment plant, with a control panel that would not fit through the door, I began my knowledge quest about instruments and process control. Chemical Engineering published Process Automation a 14-Part Series. My first book purchase was Liptaks' Instrument Engineers' Handbook that I read commuting back and forth to the office. The process control companies offered a great training ground for young engineers. Highly experienced application specialists at these companies wrote most of the articles and books on process control. I have five books by Shinskey, the most recent being Feedback Controllers for the Process Industries (McGraw-Hill 1994).

I hesitated at first to include sample time as one of the exceptional opportunities in process control because in most loops it is not issue. Then I realized I should give my perspective on the effect of sample time for the following reasons:

(1) Since we live in a digital world, sampled data is the norm. Just from the volume of applications, the opportunity is large

(2) There are no clear guidelines for various types of process control applications

(3) In some applications conventional sample times can cause severe safety and performance issues

(4) In most cases the tuning of the controller dictates that sample times could be significantly slower. If DCS module execution times and wireless communication time intervals could be increased, controller loading is reduced and wireless battery life is prolonged, respectively

(5) If we want more at-line analyzers to provide measurements of stream compositions that tell us what is really going on in the process and offer the opportunity for a more advanced level of control, we need to understand and address sample processing and analyzer cycle times

(6) If we want to move to more wireless measurement that give us the flexibility and intelligence for process control improvement, we need to understand and address wireless communication intervals

I am considering sample time as the time between updates in sampled data in the broadest sense. The following discussion should be useful for determining whether DCS scan or module execution times, wireless communication time intervals, model predictive control execution time, and at-line analyzer cycle time will affect control system performance.

If you are pressed for time you can skip the discussion below and just check out ProcessControlSampleTimes.pdf

There is considerable confusion as to when sample times affect the ability of a control system to compensate for unmeasured disturbances. The following is my quick attempt to provide some concepts to sort out fact from fiction and provide some guidance.

The performance of a control loop depends upon the tuning. Specifically, the peak and integrated errors are inversely proportional to the controller gain. The peak error is not affected much by the integral time setting. However the integrated error is proportional to the integral time. Thus, a loop with good dynamics can be made to perform as poorly as a process with bad dynamics by sluggish tuning. The effect of slow sample times is hidden by large integral times or small controller gains. Thus, it is critical for any comparison, that tuning criteria be specified. In fact there is an implied deadtime as a result of the tuning of the loop as derived and discussed in Advanced Application Note 5. The tuning of the controller puts a practical limit on how fast the sample time must be for the effect to be negligible.

If a controller is tuned for maximum performance, the peak error is proportional to the loop deadtime to process time constant ratio. The integrated error is proportional to the deadtime squared. These statements are strictly true only when the process time constant is large compared to the loop deadtime. The loop dead is the sum of final element deadtime (e.g. valve pre-stroke time delay, deadband, and sticktion), process deadtime (e.g. mixing, thermal, and transportation), automation deadtime (e.g. sensor lag, transmitter damping, and sample times), and small process time constants. All of the time constants smaller than the largest time constant become effectively deadtime in the first order plus deadtime approximation used in industry. Process and automation system dynamics places an ultimate limit on loop performance. There is a corresponding ultimate limit on the sample time.

The relationships between process dynamics (e.g. total loop deadtime), controller tuning, and loop performance is detailed in the Theory section in Chapter 2 of Advanced Control Unleashed, and Appendix C in New Directions in Bioprocess Modeling and Control. All of my books and many of my articles take advantage of the fundamental understanding gained from these relationships.

The effect of sample times can be accessed in terms of practical and ultimate limits on performance. Critical loops where peak errors can cause destruction or environmental releases such as compressor surge control, furnace pressure control, exothermic reactor temperature control, and RCRA pH control, the tuning is necessarily aggressive. As a result the practical limit is much closer to the ultimate limit. For a discussion of cases where exceptionally fast sample times are needed, checkout the April 2, 2007 entry "Analog Control Holdouts."

For excellent final elements, clean sensors, and transmitter damping settings of 0.2 sec, we can suggest practical and ultimate sample times for different types of processes with typical dynamics. The ultimate limit (fastest conceivable sample time requirement) is set to be less than 1/10th of the sum of the minimum loop deadtime and minimum process time constant with some consideration as to maximum practical controller gains to reduce valve cycling and noise amplification. For any loop with a a large control valve, the minimum loop deadtime is about 1 second for an unmeasured disturbance (unless volume boosters have been added to the output of the positioner) so the ultimate limit on sample time is about 0.1 second. The practical limit reflects current tuning practices (much slower tuning to insure a smooth gradual response despite unknowns and nonlinearities). For integrating processes, the process time constant shown is the inverse of the integrating process gain (denoted by single exclamation point). The double exclamation point denotes a runaway (positive feedback) process time constant. Consultants says it is impossible to generalize but I think some guidance is helpful to the user with the realization there are always exceptions and the actual process dynamic and tuning should be identified by automated online tuners and adaptive controllers (e.g. DeltaV Insight). I didn't consider ultimate sample times slower than 60 sec. Note that slower sample times will affect the deadtime identified. A Rough Guide to DCS and Measurement (e.g. Wireless) Sample Times is offered in ProcessControlSampleTimes.pdf

For many digital devices the update is available near the beginning of the sample time (latency is negligible), which means the average deadtime from the sample time is about half the sample period. For at-line analyzers (field analyzers with automated sample systems), the result is not available until the end of the sample processing and analyzer cycle time, which translates to an average effective deadtime that is about 1.5 times the time interval between updates in the analyzer output signal. Theses deadtimes determine the minimum peak error for an unmeasured step disturbance at the input to the process.

The detrimental effect of sample time is greater than deadtime in that for continuous sources of dead time such as process transportation and mixing time delays and small process time constants, there is a continuous train of updates. For sampled data there are no intervening values. Consequently, the effects can be worse. For example, there is aliasing of oscillations where the indicated amplitude is smaller and the period is larger than actual. There can be jitter due to variations in latency and lack of synchronization of digital data that introduce variable time delays and noise for rapidly changing signals.

The PIDPLUS modification of the traditional PID developed for wireless applications helps the PID deal with the sample time from digital devices and communication, and at-line analyzers. The improvement is most dramatic for self-regulating processes but is also significant for integrating processes as seen in the tests documented in ControlStudiesPIDPLUS1.pdf. The PID-Plus algorithm also breaks the limit cycle from the resolution limit from the deadband setting for exception reporting of wireless devices because integral action is only done when there is a measurement update.

My publications are notorious as "no-fluff" zones. My articles "Life's Batch" and "Maximizing PAT Benefits from Bioprocess Modeling and Control" should have been a 5 part series. After 120 blogs, 84 Control Talk columns, and 14 articles since I retired from my full time job, you might think I might be running out of ideas. I wonder myself when I sit down to write but once I feel a flow with the music, the main constraint is time. There is always something to say even if it is just shedding more light on an old subject. It is kind of surreal since I am a quiet guy. As I get older I am going to have to make sure I don't repeat myself, repeat myself, repeat myself.

Here are the key points for my 2005 - 2006 articles

"Life's a Batch", Control, May, 2005
(Click "Download Now" button at end to get Equations and Figures)

1. The key to good batch temperature control is the secondary loop setup and tuning

2. An inlet or outlet secondary temperature loop linearizes the process gain of the primary batch temperature loop and makes the primary loop dynamics faster

3. An inlet jacket or coil temperature can correct for coolant disturbances before they appreciably affect the batch temperature

4. An outlet jacket or coil temperature can correct for heat transfer surface disturbances before they appreciably affect the batch temperature

5. The use of a heat exchanger in a recirculation loop instead of a jacket or coil creates a delayed integrating response in the secondary temperature loop that is problematic if much integral action is used (not discussed in this article)

6. The difference between an inlet and outlet jacket or coil temperature multiplied by coolant flow provides a measurement of heat release and hence reaction rate. The inlet temperature should be delayed by the transport time through the coils or jacket (Volume/flow) to match up the inlet time wise with the outlet temperature

7. If the jacket or coil flow rather than a makeup flow is throttled, the increase in the process gain and process delay of the secondary loop can causes oscillations

8. The secondary loop should be tuned with mostly gain action for a fast response otherwise disturbances start to affect the batch temperature and an exothermic reactor can develop a runaway response

9. Coolant valves should be judiciously sized sliding stem (globe) valves with digital positioners to reduce the limit cycles from stick-slip and deadband

10. Most batch temperatures will oscillate across the split range point because of the dramatic difference between the installed valve characteristic curves and the increase in sticktion near the closed position

11. Trim coolant valves should be considered to reduce oscillations around the split range point and provide fine adjustments (see the March 16 and March 24 entries on this site on the "Manipulation of Multiple Flows")

12. The integrating response of batch temperature will cause a limit cycle from deadband even if the secondary temperature loop has no integral action

13. A highly exothermic reactor can runaway if the secondary temperature measurement or heat transfer rate is too slow

14. To reduce the batch cycle time for to reach a batch temperature end point, the jacket and coil valve can be set wide open and a control strategy such as the following used where appropriate:

a. A temperature rate of change calculation multiplied by the deadtime triggers the shutoff or positioning of the coil or jacket valves. If the feeds are to continue or there is some residual heat generation, the batch temperature should be put in automatic (see 2006 article "Full Throttle Batch and Startup Response" for details)

b. A reactor temperature controller can throttle the reactant feed rates nut there may be an appreciable inverse response from the dilution and cooling effects of increasing a reactant feed rate

15. Model predictive control is more effective approach where there are multiple constraints for batch reactors being pushed beyond their nameplate capacity

16. Coriolis mass flow meters can correct of reactant concentration and provide a model of reaction product concentrations

17. Equations can estimate the ultimate gain of self-regulating, integrating, and runaway process for process gains, lags, and dead times and provide a deeper understanding of what affects performance and why batch reactor temperature loops require higher controller gains and lower integral times

18. The primary temperature controller integral time setting should be scheduled based on totalized feeds and the secondary temperature controller gain and integral time setting scheduled based on the position of split ranged valves

"What If? Virtual Plant Reality", Control, Aug, 2005
(Pages 3 and 4 of "How to Survive the Oncoming Train of Technology")

1. Process flow diagram (process design) simulations circa 2005 that are made dynamic

a. Can provide a reasonably accurate steady state process gain and the residence time based process lag time if the physical properties are well known

b. Generally do not model mixing lags, transportation delays, installed valve characteristics, valve backlash or sticktion, mixing or sensor noise, and sensor lags, or bubble or particle distribution and size

c. Have trouble simulating batch operations, startups, and shutdowns because equipment instantaneously go to equilibrium conditions and the program can develop numerical instabilities for extreme conditions and zero flows

d. Cannot possibly emulate all of the batch and loop control capability in a DCS and thus must relay upon being interfaced to a DCS which is problematic in terms of running faster than real time (synchronization and acceleration issues)

2. Dynamic simulations that focus on the dynamics of interest can focus on the details important for process control

"Model Predictive Control can Solve Valve Problem", Control, Nov, 2005

Advanced Application Note 002

I don't need to say anything here since it is covered in the application note and the March 16 and March 24 entries on this site on the "Manipulation of Multiple Flows." Dare I repeat myself?

"Maximizing PAT Benefits from Bioprocess Modeling and Control", Pharmaceutical Technology, IT Supplement, Nov, 2006

There are so many uses of a virtual plant it is mind boggling. Just search for Virtual Plant on this website. In particular, check out the Oct 8, 2008 entry "High Fidelity"

"Full Throttle Batch and Startup Response", Control, May 2006

This article shows a simple calculation when the reactor temperature will reach set point based on rate of change and deadtime can minimize the time to reach set point. The calculation is particularly appropriate for the integrating response encountered in a batch operation or in the startup of a continuous piece of equipment where the discharge flow has not started. It is important to remember for integrating processes, the controller output must be driven past the balance point (resting valve position) to make the process variable move. With self-regulating processes, you can go to the balance point directly but even here you get there faster if the output is initially drive past the balance point.

I really like blogging. The only reason the blogs are fewer these days is that my time is consumed with finishing up the "Essential Book" so it is available in time for ISA Expo. What free time I have is spent taking advantage of Austin being the "Live Music" capital.

As I reflected on my career, I reaffirmed that what drives me is gaining a deeper understanding and sharing what I have learned, hopefully with a few laughs along the way. Throughout my career I sought with an open mind the knowledge and insights of the leaders in process modeling and control. I then used simulations to rapidly explore process relationships and to prototype control improvements that incorporate process understanding. The knowledge prepared me to solve tough plant control problems.

During my career at Monsanto I wrote a bunch of articles in the 1980s for InTech on my time in the plants with some humor introduced to help make the material more accessible and memorable. These articles were compiled and published in the book A Funny Thing Happened on the Way to the Control Room available for viewing as an E-book in the April 3, 2009 list of my books on this website. This is my favorite book, I didn't write much in the way of articles or books in the 1990s. I was on the road most of the time.

When I retired from Monsanto-Solutia in 2001 (sans package), I taught at Washington University. The students were great but after the course and lab was developed, it became routine. Also, I felt isolated.

I tell people I flunked retirement. I moved to Austin in September 2004 and started a second career as a part time consultant at Emerson Process Management. This gave me a chance to keep up to date with the latest new tools besides continue my exploration of process control opportunities. Plus it felt like home since Monsanto and Fisher Controls were one for most of my career.

I have been blessed with access to the best minds. In Monsanto's Engineering Technology I got to work with the leaders in process modeling and control. Some went on to distinguished chairs at prestigious universities, several were inducted into the Process Control Hall of Fame, some served as presidents of ISA and AIChE, and others left to become the principal technical resources for leading simulation companies. Here in Austin in Applied Research I get to work with the brains behind DeltaV. Plus my second career is more balanced. Except for the spike in work this year, I take a total of 4 months off each year to travel to see relatives, friends, and neat places and to write books.

Key points of my articles written in my post retirement years provide a quick overview of what I have been doing. The entries on this website in July will focus on the dozen articles I have written since retiring from my full time job. Here are the articles from 2003-2004.

"Has Your Valve Responded Lately", Control, May, 2003
"What is Your Flow Control Valve Telling You", Control Design, May 2004

Putman publications decided to do an encore publication in a second magazine. Some nomenclature typos were corrected in the reissue of the article in Control Design.

1. Deadband originates from backlash in the linkage and connections between the actuator and the plug, disc, or ball. Stick-slip comes from friction in stem packing and seals around the sealing of the plug, disc, or ball for process isolation

2. Deadband from linkage and connection backlash and stick-slip from trim and packing friction create deadtime for slowly changing controller outputs

3. Deadband will create a limit cycle in any control system where there are two integrators in series, such as a PI controller on an integrating process (e.g. level)

4. For deadband, the limit cycle amplitude is the ratio of deadband to controller gain

5. For stick-slip, the limit cycle amplitude is the product of the open loop gain and the stick-slip

6. For both deadband and stick-slip, the limit cycle period is proportional to the controller integral time and inversely related to the controller gain

7. Large actuators can have a large stroking time for a large change in signal

8. The size of the changes signal typically used to checkout control valves will not reveal the deadband or stick-slip and make all but the largest valves look good

9. A volume booster can reduce the stroking time of big actuators but has a large deadband. The booster should be put on the positioner output to quickly drive through this deadband. The booster bypass must be opened enough to prevent fast cycling from the positioner output looking into the booster's small inlet volume

10. Unstable oscillations can break out for large disturbances when the integral action in process loop becomes faster than the valve response. The integral time must be greater than the product of the valve slewing rate, disturbance size, and controller gain. (Not mentioned in the article but frequently discussed on the this website is that position read back from digital positioners and the PID dynamic reset limit option can automatically prevent the controller output from outrunning the valve)

11. Limit cycles are attenuated (filtered or washed out) by vessels or columns. The ratio of the attenuated to original amplitude is proportional to the period of the oscillation and inversely proportional to the residence time (volume/flow)

12. The control valve with the best response is a sliding stem valve with a digital positioner. If one must use a rotary valve, avoid tight shutoff and high friction packing and use a diaphragm actuator with a short shaft and splined connections between the actuator shaft and the stem of ball, disc, or plug. Make sure the stem is cast with the ball, disc, or plug to avoid another connection with backlash

Postscript: Rotary valves designed by piping manufacturers have a lot of deadband and stick-slip as discussed in the July 2009 Control Talk column "Downturn Turndown" in Control magazine.

"The Next Generation - Adaptive Control Takes a leap Forward", Chemical Processing, September, 2004

1. Nearly all controllers are detuned (backed off from maximum performance) to some degree to provide a smooth response and to deal with the inevitable changes in the process dynamics

2. Older technology adaptive controllers had these undesirable features
a. The process had to be disturbed or oscillated (e.g. patter recognition)
b. The dynamics were embedded in tuning settings
c. No real insight as to where the process has been or where it is going
d. Tuning method was fixed
e. Always playing catch up even if same situation was seen a thousand times

3. The next generation adaptive controller can
a. Normal changes in a controller's set point or manual output are used
b. The process dynamics are displayed and historized
c. From changes in the process dynamics, plant problems can be diagnosed
d. Several tuning methods are available
e. Tuning settings identified can be scheduled for preemptive action

4. "The information on changes in the process model may be directly used to monitor loop performance and to provide more intelligent diagnostics. The models can provide the dynamics for simulations and identify candidates for feedforward control and advanced control techniques. For example, loops dominated by a dead time or exhibiting disturbance models for multiple variables, are prime candidates for model predictive control. The dynamic process models in general can be used to create or adapt real time simulations for prototyping new control strategies, exploring "what if" scenarios, and training operators. Process gains that decrease or time constants that increase with feed totals are ripe for real time optimization of the run time between defrosting or cleaning and catalyst reactivation or replacement. The beauty of this route is the models and tuning settings are available from the adaptive controller for a higher level of control by a better knowledge of the topology"

"Advanced Control Smorgasbord - A Lot of Tasty Choices", Control, May, 2004

The online version is missing the following introductory sentences at the beginning of the first paragraph.

"By the time I was assigned to my first electronic control room project, some very smart engineers had already developed most of the techniques to exploit PID controllers.
Relative gain arrays and simple decoupling of the controller output were used to analyze and deal with interaction on a steady state gain basis. The outputs from PID controllers, whose process variable was a constraint variable, were sent to a signal selector to form an override control scheme to maximize or minimize a manipulated variable."

1. Previously, advanced process control (APC) required software packages at $100K a clip, separate computers, special interfaces, and consultants to do the studies and implementation. The total bill could easily approach or exceed a million dollars for a medium project, the biggest chunk being the consultant's time charges. Even a greater consideration was that the process knowledge to exploit or to just maintain the system disappeared when the consultants left the site

2. At the turn of the century, APC technologies were integrated into the basic process control system. License fees were minimal and whole cost of implementation decreased by a factor of twenty or more by the automation of the configuration, displays, testing, simulation, and tuning

3. In the time it takes to read this article, a model predictive controller or neural network could have been configured

4. Perhaps the biggest opportunity for driving the application of APC is the development of online process performance indicators

5. The key variable for process performance monitoring is the ratio of the manipulated flow to the feed flow

6. The controlled variable is best expressed and plotted as a function of the flow ratio (e.g. pH versus reagent to feed ratio, column temperature versus reflux to feed ratio, exchanger temperature versus coolant to feed ratio, and stack oxygen is versus air to fuel ratio)

7. The process efficiency is seen in difference between the actual and optimum ratio rather than in the gap between the actual and optimum controlled variable

8. A novel method has been developed to use model predictive control (MPC) to simultaneously adapt multiple first principle process model parameters

9. For closed loop process control, consider
a. PID for tight control of integrating or runaway processes
b. MPC for multivariable control, interactions, and optimization

10. For online property estimators for continuous processes, consider
a. ANN for highly nonlinear predictions with uncorrelated inputs
b. LDE for lag dominated linear predictions with uncorrelated inputs
c. PLS for steady state predictions from large number of correlated inputs

ANN is an artificial neural network, LDE is a linear dynamic estimator, and PLS is a projection to latent structures or partial least squares prediction discussed in Chapter 8 of Advanced Control Unleashed

I need to minimize the time delay to dinner so I will minimize this discussion of how to minimize the oscillation from analyzer sample time delay. So many minimums and so little time.

Composition measurements with sample systems and cycle times termed "at-line analyzers" offer incredible opportunities for understanding and controlling what affects what you ultimately want to know for a process output stream - the composition. The sample time delay from the cyclic results from an at-line analyzer is more problematic than the transportation delay for a continuous measurement via a probe in a sample line termed "in-line analyzers". The "at-line analyzer" has a stepped response and sometimes spikes from bad readings with no intermediate values. The result is a propensity for oscillations when used for feedback control.

One might think a deadtime compensator would help the traditional PID deal with the deadtime from a cyclic time delay. However, these deadtime compensators are notoriously sensitive to a mismatch between the actual process deadtime and the estimated deadtime used in the compensator. The loop deadtime from unsynchronized digital devices and at-line analyzers is extremely variable and can at best be estimated after the fact.

It is interesting that the solution for suppressing oscillations from at-line analyzers resulted from improvements to the PID developed for variable updates from wireless devices (see February 9, entry on "Unexpected Wireless Benefits"). The control solution for WirelessHART requires no estimate of deadtime and is more robust than a traditional PID. The PID execution is kept relatively fast (once per second). The contribution of the proportional mode is computed every execution. The proportional action every scan provides a good set point response for a PID structure with proportional action on error. The contribution of the integral and derivative mode is only computed when the measurement has changed per the resolution setting of wireless device. Furthermore, the time used in the integral and derivative mode calculations is not the scan time but the elapsed time from the last measurement update.

The use of the elapsed time in the integral calculation and a reset time the same as the process time constant provides an integral correction that is equal to and opposite to the process response in the elapsed time. Even if the process time constant changes, making an integral correction only when there is update eliminates the extraneous ramp of the integral mode in the traditional PID acting on old information. The suspension of integral action until there is new information also helps the PID deal with a valve that is momentarily stuck provided position read back is used for dynamic reset limiting.

The use of elapsed time instead of PID execution time in the derivative calculation spreads the change in the process over the elapsed time rather than taking it to all occur in the single execution time. This more intelligent rate action eliminates spikes in the controller output that would occur in a traditional PID when there is an update.

The wireless PID greatly stabilized the glucose control of a bioreactor which had at-line analyzer sample time delays that varied from 6 to 12 hours. The improvement is greatest for self-regulating processes and controllers tuned for maximum performance. The suppression of oscillations can be seen on slides 29 - 33 of the Interphex 2009 Presentation "Advances in Bioreactor Modeling and Control."
Interphex2009_Advances_In_Bioreactor_Modeling_and_Control.pdf

If you want to know how to minimize oscillations from final elements and don't have time to read the supporting information you can use the following rules of thumb and move on to more important tasks like reading email. The final elements considered here are throttling control valves and variable speed drives (VSD) on pumps or fans.

• Use a sliding stem throttling valve with a properly tuned digital positioner (position feedback) or a VSD with a properly tuned speed controller (tachometer feedback) to minimize the amplitude of the limit cycle from a final element
• Make sure the DCS and final element I/O cards have at least 12 bits
• Enable "Dynamic Reset Limit" in PID block and use position or speed feedback as PV for BKCAL_OUT of AO block to prevent a burst of unstable oscillations when PID reset action is faster than valve or VSD response
• Set IDEADAND in the PID block equal to the limit cycle amplitude from the final element to kill the limit cycle during quiet periods of operation (e.g. periods when there are no disturbances or set point changes) for a self-regulating loop

Resolution is the minimum change in the element's output. Changes in the output smaller than the resolution cannot be made. For a control valve, the resolution limit is the result of friction in the packing, seat, and seal. For a VSD, the resolution limit is the result of an artificially imposed deadband, which is really a dead zone or from a speed sensing element resolution limit. Resolution can also result from a quantize limit from the number of bits in a microprocessor or I/O card. The number of bits in A/D and D/A cards for most DCS has increased from 12 bits to 16 bits. In both cases, the resolution limit from these I/O cards is negligible. However the standard input card of some VSD manufacturers is only 8 bit causing a significant resolution limit. The resolution in the stroke of a control valve or in the speed a variable speed drive will cause a limit cycle in any loop with integral (reset) action.

The term deadband is often used in automation systems to specify a dead zone (a bandwidth around a reference value where there is no response). Examples are deadband (dead zone) specifications in VSD configuration for noise rejection and in a PID configuration for integral action suspension.

For final elements, deadband has a significantly different definition. Here deadband is the change in signal required upon a reversal of direction to get a change in the element's output. Once the output reverse direction, deadband places no limit on how small a change can be made in the same direction. In reality, valve deadband is usually accompanied by a resolution limit. In the stroke of a control valve, deadband is the result of backlash from gaps or play in linkages and shaft or stem connections. Deadband normally doesn't exist in a VSD. Deadband will cause a limit cycle if there are two integrators in series in the control system. Multiple integrators in series can occur from a PID with integral action on a process with an integrating response such as level. Alternately, the limit cycle can occur if there is a cascade control loop where there is integral action in more than one controller. If both the temperature and flow PID blocks have integral (reset) action in a temperature to flow cascade control system, then deadband can cause a limit cycle. Most people forget that a positioner or digital valve controller creates a cascade loop where the positioner controller is the secondary loop. Positioners until recently were proportional only controllers.

The amplitude of the limit cycle is the smallest change in flow associated with the smallest possible change in valve position or speed multiplied by the process gain (change in process variable in engineering units divided by the change in flow). To get the smallest possible change in flow of a control valve, multiply the valve's resolution limit in % of stroke by the installed characteristic curve for the valve at its operating point. Note that valve stick-slip and the resolution gets worse near the seating or sealing surface. The manufacturer's quoted numbers are at a 50% throttle position. To get the smallest possible change in flow of a VSD multiply the resolution limit of the input card resolution of the tachometer sensing element, or noise deadband, whichever is largest, and convert to flow based on the interpolated shift in the installed characteristic curves with speed for the pump or fan. Be careful, many VSD have an adjustable deadband (dead zone) to prevent the VSD from responding to noise. This adjustment is often set with no regard to the effect on loop performance.

Resolution limits and deadband add dead time to the control loop for slow disturbances because it takes time for the PID output work through the zone of no final element response. The dead time is the resolution limit or deadband divided by the rate of change of the controller output. This additional deadtime increases the peak and integrated error for the upset. Note that step changes in the controller output larger than the resolution limit or deadband will not reveal the deadtime.

Control valves have an inherent velocity limit from the limitations imposed by actuator fill and exhaust rates. VSD have an application set velocity limit from the motor load limitations imposed by the impeller inertia. Make sure the valve actuator and VSD motor have enough muscle for the valve sticktion and pump inertia, respectively or you can get into poor valve position or speed control and hence even bigger loop problems.

Use the "dynamic reset limit" option of a PID block in a DCS, such as DeltaV, where the PID uses a positive feedback network for its integral action. The BKCAL_OUT for the AO block which in connected to the BKCAL_IN of the PID block should be actual valve position or VSD speed. Select the PV (position or speed) option in the AO block for the BKCAL_OUT. This feedback of actual position or speed to the PID enables the PID algorithm to curtail its integral contribution to the PID output so that the PID output from reset action does not change faster than the valve or drive can respond. If this protection is not in place, everything may look OK until the loop gets a disturbance large enough PID to cause the PID output to change faster than the final element. The mysterious bursts of instability for big load upsets often go unresolved.

Set the IDEADBAND option in the PID block to a value about equal to the limit cycle amplitude. IDEADBAND will suspend the integral action when the PID error is less the IDEADBAND. This suspension will stop limit cycles from a resolution limit or deadband for a self-regulating process at a steady state. It will not stop the limit cycle on a process with an integrating response because the process has no steady state and will continue to ramp until the process variable exceeds the IDEADBAND.

For more info on final element response, check out the "Deal or No Deal" Control Talk column in Control magazine, the article "What is your Valve Trying to Tell You" in Control Design magazine, and "Improve Control Loop Performance" in Chemical Processing magazine.

If you have manipulated flows with counteracting effects (application 5), such as steam and coolant or acid and base reagents, your most straightforward solution is split range control because split ranged control prevents a loss in efficiency from both streams flowing at the same time if there is no overlap at the split range point and no low limits in the manipulated flows.

What about applications to increase plant turndown and capacity (application 1), reduce process variability (application 2), and improve plant efficiency (applications 3 and 4)?

(1) Extend rangeability
(2) Improve resolution
(3) Enable preferential use of flows based on cost
(4) Send flows to multiple destinations possibly based on priorities
(5) Provide counteracting effects

If the manipulated flows had perfect valves and no discontinuity at the split range point, we could use split range control for applications 1-4 if we addressed the tuning considerations for the different dynamics of the manipulated flows. If the manipulated flows had the same time constant and deadtime, compensation would reduce to setting the split range point to compensate for the different process gains for each manipulated flow as mentioned in Part 1. When the speed of response is different, a more effective technique may be to schedule controller tuning settings based on which flow is being manipulated. Scheduling of the gain, reset, and rate time will take into account the changes in the process time constant and deadtime as well as process gain. For example, if a loop is manipulating waste bark feed and natural gas flow to a boiler, the response of steam generation to waste bark flow will be much slower than to natural gas flow. Often the less expensive manipulated flow is the one with the slowest and most variable response. An adaptive controller, such as DeltaV Insight, can continuously update the scheduling of the tuning of the settings for a manipulated flow with variability, such as the heating value of waste fuels, the acid and/or base concentrations of waste reagents, the composition of recycle flows, and the temperature of heat recovery streams.

What are the options for dealing with the specific problem of a single critical process controlled variable and two manipulated flows with different costs, dynamics, stick-slip, and backlash? Can we mitigate the consequences of non-ideal valves? Can we avoid the nasty discontinuity of the split range point and limit the need to schedule PID settings to the effect of just one manipulated flow on the critical process variable?

A solution in the regulatory control world is to continuously manipulate the flow with the faster and fixed dynamics (FFD) for tight control of the critical process variable and only move the flow with the slower and variable dynamics (SVD) when absolutely necessary.

This strategy uses a PID to tightly control the critical process variable by directly and rapidly manipulating the FFD flow. A valve position controller (VPC) keeps the FFD flow from getting too high or low by slowly manipulating the SVD flow. The valve position control (VPC) is an integral-only controller that is optimizing the FFD flow. Proportional and rate action are not used in the optimizing VPC because fast and abrupt changes create interaction and disruption. A description of VPC starts on slide 25 in ControlUsingTwoManipulatedVariables.pdf

Control valves, particularly rotary valves, lose their sensitivity at high positions (installed valve characteristic flattens). Consequently, there is a maximum throttle position for good control. At the other end, it is undesirable to ride the seat of any control valve. Many develop more stick-slip and backlash as you approach the closed position (< 20%). As a result, there is a minimum throttle position for good control.

The VPC set point is the optimum desired FFD flow. If the FFD flow is more costly, the VPC set point is a minimum FFD flow that still enables good control. A minimum FFD flow may also exist for stability, such as a minimum gas natural flow for flame stability. If the FFD flow is less costly (less common case), the VPC set point is a maximum FFD flow that still enables good control. If there is no cost difference between the FFD and SVD, the VPC set point is the mid throttle range of the FFD (e.g. 50%). Whenever, small and large valves are used on the same stream to increase rangeability and resolution, the small valve is considered the FFD flow because the smaller valve generally has a faster response and a finer resolution in terms of total flow. In this case, the critical PID directly throttles the small valve (fine adjustment) and the VPC throttles the big valve (coarse adjustment). The VPC set point is the best mid throttle position of the small valve. The best mid throttle position is a function of the room to roam on the best part of the installed valve characteristic and keeping away from the seat.

The VPC process variable is the FFD flow. Typically, the critical PID controller output is used. Since the VPC response is intentionally slow and the optimum VPC set point knowledge is rarely better than 1%, the use of actual flow or valve position read back is unnecessary as the PV of the VPC. There might be some advantage in using actual flow in terms of linearization, but there are bigger issues like what is the ball park for tuning? The good news is we have only one VPC tuning setting, integral time. The bad news is this integral time tuning is not defined for applications. We know the VPC should be slow enough to prevent interaction with the PID but fast enough to allow the PID to do its job. The best paper I have seen on VPC tuning is "Analysis of Valve-Position Control for Dual-Input Processes" by Cheng-Ching Yu and William L. Luyben published in the American Chemical Society journal in 1986 (0196-4313/86/1025-0344$01.50/0). The conservative tuning in this paper appears to me to be the best and simplifies to the integral time setting being approximately the ratio of the SVD process time constant to the FFD process time constant for stable (self-regulating) processes. For unstable (runaway) processes, a satisfactory integral time is about half the ratio. For the exothermic reactor example cited, the integral time is about half of the ratio of SVD heat removal time constant to the FFD heat removal time constant. This article implies an independence of the VPC integral time from other process dynamics. This independence should be confirmed through more analysis and testing. The VPC integral time might also be a function of the ratios of process gains and dead times in the response of the critical process variable to the manipulated flows.

It is important that the critical PID be tuned first for tight control. For unstable processes this PID must have enough gain and rate action to prevent a runaway. The VPC is then tuned next and any fighting between the loops or oscillations created in the PID loop for a set point change in the VPC loop must be prevented by increasing the VPC reset time. For large and fast disturbances that drive the FFD flow out of the good control range, it is important to add feedforward control to put the valves in the right position without having to wait for the slow VPC loop to respond. If we are doing the small valve PID and big valve VPC control deal, it may be useful to turn off integral action in the VPC when the fine valve is within an acceptable throttle range (e.g. 40-60%) so the big valve ("Mr. Big") with its big problems is only asked to move for a big disturbance. This eliminates a big limit cycle from the big stick-slip and big backlash of "Mr. Big."

Stay safe. Always monitor and test any new strategy or tuning for worst case scenarios.

When there is more than one manipulated flow, split range control is commonly used, the subject of Part 1. In succeeding parts to this series we will see how regulatory control connoisseurs use valve position control and advanced control enthusiasts have gotten increasingly innovative in the use of model predictive control. The manipulated flows are set points of secondary flow controllers or the signals to final elements (control valves and variable speed drives). The primary conceptual objectives in the manipulation of multiple flows to control a single process variable are as follows:

(1) Extend rangeability
(2) Improve resolution
(3) Enable preferential use of flows based on cost
(4) Send flows to multiple destinations possibly based on priorities
(5) Provide counteracting effects

Some applications require more rangeability than can be achieved by a single valve or flow meter. The standard definition of valve rangeability that uses a low flow limit based on the uniformity of the inherent valve characteristic at low positions is too large and misleading. The lowest flow that can be reasonably throttled is the minimum flow of a repeatable installed characteristic multiplied by the stick-slip or backlash, whichever is greater. Per the standard definition, linear valve maybe cited as having the best rangeability even though the installed characteristic is quick opening and it is missing a positioner. Alternatively, a rotary valve may be stated as best because of its high flow capacity even though it may have high sealing friction and breakaway torque and excessive backlash in its linkage. The best rangeability is achieved by a sliding stem valve with an equal percentage characteristic, low friction packing, a generously sized actuator, and a tuned digital valve controller, which also provides the best resolution as discussed in the article "Improve Control Loop Performance"

The differential head flow meter rangeability of 4:1 is the lowest although it can be extended to 16:1 in cases where the differential pressure is large enough and the noise is low enough by the use of extended range d/p transmitters. The vortex meter, magmeter, and coriolis flow meter have an approximate rangeability of 16:1, 100:1, and 250:1, respectively if the maximum velocity of the meter size matches the maximum velocity of the process.

When the valve or flow meter rangeability (turndown) is not good enough, split range control is used where the controller output is split between a small and large flow meter or valve. The split range point between small and large flow control is conventionally chosen as 50%. However, to keep the control loop more linear, the split range point should be based on the capacity of the meter and control valve. For example if the big meter or big valve has a capacity 9 times larger than the small meter or small valve, the split range point should be 10%. In terms of control valves, this means the small control valve should be stroked 0 to 100% as the controller output goes from 0 to 10% and the big control valve should be stroked 0 to 100% as the controller output goes from 10% to 100%. In the old days, the split ranging of valves was done in the positioner. Today it is widely recognized that split ranging should be done in the control system for better maintainability, flexibility, and visibility.

The most common loop where you see small and large valves is the pH loop because of the exceptional rangeability and sensitivity of the pH measurement. The 0-14 pH scale covers fourteen orders of magnitude of hydrogen ion concentration and control at neutrality (e.6. 7 pH) involves controlling concentration to 7 or more significant figures. Often the small and large reagent valves are split ranged. However, for split ranged control, the resolution needed to meet the sensitivity requirement only occurs at low reagent demands before the transition from the small to large valve.

Split range creates a severe discontinuity not only from the change in gain but also because one valve is often opening and/or another control valve is closing. The stick-slip and the uniformity of the valve characteristic are worst as the valve goes into or trys to come off the seat. Many loops oscillate across the split range point. A hysteresis or deadband setting at the split range point can prevent both valves from being open but this deadband adds deadtime and the controller output will never settle out within the deadband but will always be traversing back and forth across the deadband for manipulated flows needed that are close to the split range point. The deadband provides some useful purpose in preventing noise or a short term disturbance from causing a transition from one valve to another. A deadband is commonly used in pH control of bioreactors in an attempt to decrease the number of excursions from carbon dioxide (acid) and sodium bicarbonate (base) to reduce the rise in internal pressure of the cells from an increase sodium ion concentration.

In some cases, cost dictates the preferential use of a flow. For example a waste reagent stream would be maximized compared to a purchased reagent stream for pH control, a waste fuel would be maximized compared to a purchased fuel for combustion control, air flow would maximized compared to a oxygen flow for bioreactor dissolved oxygen control, suction flow would be minimized via speed control compared to a flow vented for air compressor discharge pressure control, and a letdown valve position would be maximized compared to a flow vented, for header pressure control. Sometimes the allocation is made based on the priority of the user. For example, the more profitable reactors would loaded up with feed first for reactant compressor pressure control and the more efficient heat exchangers would be loaded up with coolant flow first for chiller pressure control.

In Parts 2 and 3 we will see if there are better solutions than split range to achieve objectives (1) through (4). Split range control is admittedly the best solution for loops that manipulate flows with competing effects, objective (5). Primary examples are acid and base reagent valve manipulation for pH control and heating and cooling valve manipulation for temperature control. The use of split range readily prevents the acid and base valve from both being open and the heating and cooling valves from both being open wasting reagent and energy, respectively.

The Emerson 2005 Exchange Presentation titled ControlUsingTwoManipulatedVariables.pdf offers implementation details and examples of split range control, valve position control, and model predictive control for the manipulation of multiple flows to control a single process variable.

In the February 9 entry on "Unexpected Wireless Benefits" we saw how the positive feedback implementation of the integral mode enabled an enhancement of integral action had benefits that extended beyond wireless devices to any loop with appreciable measurement delay. The positive feedback network sends the controller output or an external reset feedback back through a filter and adds the result to the controller output from the proportional mode as shown in PIDPLUS_Results.pdf. The positive feedback network also offers a convenient method of deadtime compensation by just inserting a deadtime block in front of the filter as shown in Advanced Application Note 3.

Additionally the positive feedback network provides a significant improvement for batch control, override control, and cascade control as described in the article "The Power of External Reset Feedback." In cascade control the use of the secondary process variable (PV) as the input to the filter (external reset feedback) enables the primary controller to deal with a poor (e.g. slow) response in the secondary controller. To set up external reset feedback of the secondary loop PV, the "Dynamic Reset Limit" must be enabled in the primary (outer) loop PID and the secondary (inner) loop PV must be selected for the BKCAL_OUT of the secondary PID as shown in Cascade_Control.pdf. The use of external reset feedback prevents the primary loop from acting faster than the secondary loop can respond, which could occur if the secondary loop has a slow reset setting or has a slow valve. To include the effect of the valve response, the valve position read back must be selected as the PV for the BKCAL_OUT of the AO block. Without this configuration, the primary controller does not know a valve has excessive stroking time, stick-slip, or deadband, a loss of signal, solenoid valve failure, or unexpected de-energization of the solenoid due to a discrete process action, sequence, or interlock. The loop may seem OK for small changes load if it is just a slow valve or slow variable speed drive without external reset feedback, because the velocity limiting (rate limiting) in the final element has little effect for small changes in the controller output. However, a large disturbance or set point change will trigger oscillations when the controller output outruns the response of the final element.

Did you know the peak error for cascade control decreases for an unmeasured disturbance as the size of the secondary process time constant (lag) increases? In the single loop, the secondary process time constant is detrimental because it creates dead time whereas putting the secondary time constant as the largest time constant in the secondary loop is beneficial because it allows a high secondary controller gain and slows down process disturbances entering the secondary loop. The ratio of cascade to single loop peak error goes from about 0.25 to 0.1 as the ratio of the secondary (inner) to primary (outer) loop time constant increases from 0.2 to 1.0 for a 0.6 inner to outer loop dead time ratio as shown in Tuning_and_Control_Loop_Performance_Figure_11.2.pdf. The figure also shows how a smaller deadtime in the secondary loop compared to the deadtime in the primary loop decreases this same peak error ratio. The improvement is even more dramatic when the primary loop has an integrating or a runaway process.

The idea of a lag in the control loop just sounds bad but are lags always bad news? The popular consensus is yes. Could a lag could be your best friend despite its bad rep?

I was instructed in a graduate class on distillation column modeling and control decades ago that the big problem with columns is the big process lag. This didn't sit well with me but I didn't stand up and object. I concluded there seems to be a lag in the understanding of lags.

If the lag is a process time constant in the input path of the disturbance into the process, it is actually beneficial. This process time constant slows down the effect of a disturbance and gives a chance for controller to catch up. The controller gain for the ultimate performance of most tuning methods is proportional to the ratio of the largest time constant to the total loop deadtime as seen in Appendix C of New Directions in Bioprocess Modeling and Control. BioprocessModelingControlBookAppendixC

Furthermore, the peak error from a disturbance for this tuning is inversely proportional to this ratio of time constant to total loop deadtime as shown in Equation 2-40 in Chapter 2 of Advanced Control Unleashed. This ratio is about 5:1 for distillation column temperature due to interactive process lags. In other words, you can anticipate a 5:1 reduction in error by closed loop control for a step disturbance. For well mixed crystallizers, evaporators, and reactors this ratio could be 50:1 or more. This leads to permissible controller gains much larger than we are accustomed to using.

Often there is a similar type of process lag on the path of the manipulated flow into the process used for correction of the disturbance. When reflux rate is adjusted directly or indirectly to compensate for a feed disturbance and the temperature used for control is about half way between the feed tray and the top of the column, the process lags could be about the same. If the tray for temperature control is closer to the feed tray, the feed upset would be seen before the correction can arrive, not a good deal.

For continuous composition and temperature control of well mixed volumes, the process lag is approximately the residence time (volume/flow). This process lag is in the path of both the disturbance and the manipulated flows and temperatures. The process delay (process deadtime) is usually quite small relative to the process lag except for neutralizers where small reagent flows cause incredibly large injection delays.

When the process lag also exists in the path of the manipulated variable, it is important to use a high controller gain to overdrive the controller output so the loop can catch up to the disturbance.

Large process lags from large process volumes smooth out oscillatory disturbances from poor control or limit cycles and are the principal reason why we don't see as much variability in storage tanks as we might expect. Equation 3-4 in Chapter 3 of Advanced Control Unleashed can be used to predict this attenuation of oscillations by process volumes.

If there is a final element lag (e.g. slow valve or positioner) or there are volumes in the path of the manipulated flow that don't exist in the path of the disturbance, then the controller can't react to a disturbance fast enough. If there is a measurement lag due to a sensor lag (e.g. thermowell lag) or DCS filter (e.g. AI or PID PV filter time), then the controller can't see the disturbance fast enough. What is hideous and not well recognized is that the time constant in the tuning equations is for the largest time constant in the loop and doesn't matter where it is located in the loop. If the largest time constant is in the measurement, the user is seeing an attenuated version of the real process variable. An increase in the measurement lag allows the user to increase the controller gain. The oscillation amplitude may also look smaller due to filtering. The key indicator is an increase in the oscillation period. Not all measurement lags are bad. A small judiciously set PV filter to keep measurement noise from causing fluctuations in the controller output greater than the valve resolution can prevent self-inflicted disturbances from reaction to noise.

While control textbooks show step disturbances, most process disturbances have a process lag because they are the result of control loop reset action and valve throttling and are smoothed by intervening volumes. The worse case disturbance is a manual action by an operator, a discrete action by a batch sequence or interlock, and an on-off flow from level switches or an overly aggressive level controller that directly feeds into an important unit operation. The best bet is to slow down the disturbance, and then use a properly tuned PID and MPC and add feedforward control.

Lags slow down the set point response and make tuning a test of patience. However, if you tune the controller with a gain close to the maximum permitted by the use of small Lambda factors or simplified internal model control tuning per Appendix C, the closed loop time constant can be made much less than the process time constant by overdriving the output past its resting point. This is only true if the PID structure chosen has proportional action on error so the loop kicks the output from the set point change. If the setpoint change is large enough to saturate the loop output, you don't see the full boost in the response from the higher controller gain. If rapid changes in controller output upset the operator or another loop, set point velocity limits or filters can be employed but these limits or filters should not be used on the secondary loop of a cascade control system.

To summarize; a lag in the disturbance path on the input to the process should be maximized and lags anywhere else should be minimized for disturbance rejection. If the largest lag in the loop is much larger than the total loop dead time, the Lambda factor should be set less than one to give higher controller gains for a faster response. Tuning tests take a long time for long process lags and people get frustrated but if the higher controller gains that are permitted are used, the results can be great.

With all of the advanced algorithms and smart instrumentation available today, we can sometime lose sight of what are the real limits to loop performance. While it doesn't take an Einstein to figure this out, as a former physicist, I found an interesting analogy.

Einstein's reasoning that nothing can travel faster than the speed of light lead to incredible insights and revolutionary equations. For example if you substitute the speed of light for velocity into the equation for kinetic energy, you now have the famous equation that relates mass and energy (energy is equal to mass multiplied by speed of light squared). You also end up with a unification of space and time and warping by gravitational fields.

The absolute limit to feedback control system performance is the total dead time in the loop, which is the summation of all the final element, process, measurement, I/O, and controller execution time delays. A feedback control system cannot correct for something it hasn't seen yet and hasn't been able to change yet in the process (see "Funny you should Ask a Process Control Engineer" in the Funny Thing E-book). http://www.modelingandcontrol.com/FunnyThing/page-123.asp

The fastest closed loop time constant (Lambda) possible is the deadtime. If you substitute deadtime for Lambda into the controller gain equations for Lambda tuning, you end up with the Simplified Internal Model Control and factored Ziegler Nichols equations for the highest controller gain with a relatively smooth response. This unification of equations for controller gain was documented in Appendix C of New Directions in Bioprocess Modeling and Control. This Appendix also provides the derivation that the performance achieved in terms of integrated absolute error (IAE) for an unmeasured load upset is proportional to reset time and inversely proportional to controller gain. BioprocessModelingControlBookAppendixC

The implications of this for sample delays in terms of there being an additional implied dead time for detuned controllers is explored in Advanced Application Note 5.
http://www.modelingandcontrol.com/repository/AdvancedApplicationNote005.pdf

The hype of some advanced process control (APC) algorithms may lead one to believe this limit can be violated. Many of the early APC algorithms significantly increased the loop deadtime (See "Advanced Control Algorithms- Beware of False Prophecies in the Funny Thing E-book). While model predictive control (MPC) can potentially help dead time dominant systems, the original execution time (e.g. 1 minute) of separate MPC software packages was so large their applicability was restricted to slow processes. With the advent of the MPC embedded in the DCS, the execution time can be as fast as 1 second which means MPC can be applicable to all but the fastest processes (e.g. liquid pressure control and furnace pressure loops).
http://www.modelingandcontrol.com/FunnyThing/
http://www.modelingandcontrol.com/2008/08/tipsntechniques_tnt_tuning_fur_1.html

Deadtime compensators such as the Smith Predictor can make the PID algorithm think there is no deadtime in the loop. You can get fooled as well if the PID faceplate shows the compensated PV that has the deadtime removed from the consequences of its own actions instead of the original PV. Deadtime compensators allow the user to increase the controller gain. If the deadtime compensation is perfect, the increase in controller gain can be huge. However, many sources of deadtime are variable and unknown.
http://www.modelingandcontrol.com/2007/06/deadtimes_secret_identity_part_1.html
http://www.modelingandcontrol.com/2007/06/deadtimes_secret_identity_part_2.html

For PID controllers an underestimate of deadtime can lead to instability if one goes for the gusto of ultimate performance and pushes the limit beyond the original unfactored Ziegler Nichols equation for controller gain. For deadtime compensators and model predictive control, you can also get into some oscillations for overestimates of deadtime.
http://www.modelingandcontrol.com/repository/AdvancedApplicationNote003.pdf

Finally, there is another limit to control loop performance, signal resolution. You can't control to a degree finer than the resolution of the final element, measurement, or I/O. The resolution limit of digital devices today is nearly negligible (e.g. 16 bit A/D) but some older DCS (e.g. 12 bit A/D) could cause noticeable stair-steps in the temperature response from wide range thermocouple input cards and the standard input card. It is strange to me that the standard input card for many variable speed drives still uses an 8 bit A/D that significantly restricts the resolution of the final element. Today most of the resolution limit seen in control loops is from control valves. The principal cause is stick-slip and is usually lost in the smoothing afforded by process volumes and the compression of data (except for pH and other high gain processes) unless you have on-off or isolation valves posing as control valves or choose to save money by not buying digital positioners.
www.ControlDesign.com/articles/2003/164.html
http://www.ChemicalProcessing.com/articles/2007/200.html
http://www.controlglobal.com/articles/2008/063.html

Not well recognized is that for PID control of integrating processes, valve deadband causes an implied resolution limit in the PV for a reversal of direction that is the deadband multiplied by the integrating process gain and size of the correction needed in the controller output to balance out the disturbance. Real control valves with digital positioners have a deadband that is less than twice the resolution limit.

I think the future is advanced process control (APC). My definition of APC is any technology that puts process knowledge on the line online. Feedforward control is APC when the feedforward gain and dynamic compensation are based on process knowledge. On-demand and adaptive auto tuners, such as DeltaV Insight, are APC tools because these tuners identify the process dynamics that are useful for process diagnostics and training besides model based tuning. For example, the process deadtime can be monitored as an indicator of heat transfer surface fouling in temperature loops and the dynamics can be inserted in simulations for operator training and scenario testing and prototyping of PID enhancements (e.g. set point filtering and structure) or Model Predictive Control (MPC). There are many higher level technologies. In a recent presentation I made to a major chemical company I showed these technologies, the results from a benchmarking study of the top ten companies in the use of process control, and practical tips on how to conduct an opportunity assessment. The presentation can be seen at:

http://www.emersonprocessxperts.com/archives/2008/11/assessing_oppor.html

Slide 8 shows the pyramid of technologies that includes process performance monitoring (data analytics and process metrics), abnormal situation prevention, property estimators (inferential composition or quality measurements), model predictive control (MPC), rampers and pushers to maximize or minimize a controlled variable (e.g. feed rate), linear programs (LP) for optimization given defined constraints and economics, and real time optimization (RTO) for variable constraints and economics. The importance of process knowledge in all of these technologies is obvious. Slide 9 gives a straightforward "easy to remember" relationship between controller tuning for loop performance. The equation indicates before, during, and after APC implementation, the controllers should be tuned.

The amount of effort and the performance of the upper level technologies rest upon the strength, breadth, and integrity of the foundation of basic control. As you improve the number, type, and sensitivity of the measurements and control valves, the performance of these systems improve by reducing the number of unidentifiable disturbances and enabling more first principle calculations and inferential measurements, such as frosting rate, fouling rate, crystallization rate, and reaction rate important for diagnostics and batch profile control as discussed in a recent article in Control magazine.

http://www.controlglobal.com/articles/2008/230.html

Decades ago, field pressure and temperature gages were installed. These were not very accurate. prone to be broken, and obviously were not visible in the control room or historized. With wireless, we can afford to get many more measurements into the control system. Wireless measurements offer the opportunity to provide many of these missing measurements at a reasonable cost. However, the choice of measurements for data analytics (principal component analysis and projection to latent structures) must be judicious. Randy Reiss, the developer of online data analytic algorithms for Emerson, says "more measurements for analytics means more correlations. However, it introduces the possibility of dominate correlations that do not relate to product quality. That would skew the model for the worse. So there is a double edge sword there."

For portable bioreactors, laboratory analyzers, and sterilization systems, wireless adds flexibility and utility. Wireless access to process and loop performance monitoring systems in the field makes troubleshooting much smarter. Wireless access anywhere to virtual plants with process performance scores for university courses on process control makes learning almost like a video game. There are many more applications for wireless than the monitoring of remote tanks and pipelines. The following Control Talk column slated for the December issue of Control magazine discusses the role of wireless in APC.

WirelessControlTalkColumn

Randy Reiss's list of the "Top Ten Reasons You Will Go Wireless Next Year" in the above column provides a reality check in case we are thinking of making everything wireless. This list has the insight, bite, and humor typical of the lists Randy has contributed to my column in recent months. Upon reading the draft of the column, Randy said "it's the best argument I have heard for wireless." Randy agreed to the post of this quote after checking with his PR agent.

Scott Broadley, the president of Broadley-James, is participating in a beta test with Emerson on the use of wireless transmitters on portable single use bioreactors (SUB) whose size is steadily growing from pilot plant (100 liter) to production (1000 liters) scale. Scott is also looking forward to the elimination of ground loops and noise by wireless pH transmitters particularly where the solution ground is not used or where AC noise gets through the power supply. Scott says, tongue-in-check, "we could hook the pH and DO transmitter up wirelessly to a Twitter account so your cell phone is getting constant text updates on how your bioreactor is "feeling" . Scott offers the following additions to the top ten list for going wireless.....(11) Each bioreactor can have its own Face book page where operators from different shifts can leave their comments......(12) Each transmitter can be on Twitter and send you instant text messages on your phone when it is moody..."

If the pressure set point is a fraction of an inch of water column, you have a high integrating process gain. The response is often a high speed ramp in the control region. For a waste incinerator and a phosphorous furnace I worked on decades ago, the pressure could ramp off scale in 0.2 and 5 seconds, respectively. Trying to control the incinerator and furnace pressure was reported to be like trying to control an explosion when there was a shutdown or a slag slide, respectively, and a corresponding burst of vapors and gases. Needless to say these pressure loops could never go to manual and open loop tuning methods were down right dangerous. In the old days I used a modified ultimate oscillation method and a high speed recorder. As with runaway reactors, the reset time (e.g. sec/repeat) was increased by 100x to make reset action negligible and the controller gain was increased until there was the start of an oscillation. The reset time was set equal to the period of the damped oscillation and the controller gain was halved. A set point change was then made and if the response was more oscillatory than dictated by valve limit cycles from stick-slip or deadband, the controller gain was decreased. If damped oscillations persisted and got worse or slower, then the reset time was increased until the oscillations period and decay rate were faster. This test was repeated and the gain decreased or the reset time increased until the response was sufficiently smooth.

Before we go further, one should realize that the original ultimate oscillation method asked the user to increase the controller gain until there were equal amplitude oscillations. This was too exciting and gave controller tuning settings that were too oscillatory especially if there was an increase in the loop dead time or process gain or a decrease in the process time constant. The damped oscillations mentioned here are rapidly decaying where each succeeding peak is less than ¼ the previous peak.

The damped oscillation period is larger than the ultimate oscillation period and the damped oscillation controller gain is smaller than the ultimate oscillation gain and the factors of 1.0 for period and 0.5 for controller gain are not per the textbook definitions of the Ziegler-Nichols ultimate oscillation method. Using the text version of the closed loop (ultimate oscillation) or open loop (reaction curve) Ziegler-Nichols tuning method and thinking that tuning settings with more than one significant digit are practical, is a great way to reject the pioneering work of Ziegler and Nichols and to glorify new tuning methods. What I found early in my career is a simple change of using damped oscillations instead of ultimate oscillations and using easy to remember rounded off factors, gave me the proportional mode action needed for these loops that lack self-regulation and can be headed for a trip point. I also quickly realized that the nonlinear and non-stationary nature of chemical processes and valve stick-slip and backlash meant that the long term tuning setting accuracy of better than 50% was wishful thinking.

Today, integrated online adaptive tuning tools that look at set point changes, such as DeltaV Insight, should be able to automatically identify tuning settings of most fast integrating processes. However, some pressures can be so fast (e.g. the cited incinerator) digital delays must be eliminated and tuning tools that directly connect to the I/O, such as those used by EnTech, are needed. It is important that the module execution time, the tuning tool, and the trend chart update time not cause aliasing or an extra observed dead time. The controller, final element, and pressure sensor must also be extremely fast. Finally, it is particularly critical to test and observe new tuning settings for these and other types of loops that require aggressive feedback control.

If you want to get more details on the importance of making the loop fast enough, check out the chapter "Pressure Control: Without Deadtime I Might be Out of a Job" in the free E-book A Funny Thing Happened on the Way to the Control Room on pages 31-41: http://www.modelingandcontrol.com/FunnyThing/.

This week I completed a model with the help of Roger Reedy that allowed me to confirm some concepts besides detail how the design of the control system can cut a project cost almost in half by the use of 10,000 gallon instead of 40,000 gallon neutralization tanks. It wasn't an easy pH control application but not many of them are. The titration curve slope and the hence process gain changed by a factor of 1,000:1 from the extremes of the pH scale range to the neutral point. The influent pH could swing from 12 to 2 pH during the regeneration of a demineralization system or an area pump out. The disturbances could be fast because of plug flow, batch sequences, manual operations, and the stick-slip action of control valves. If pH control is not your thing and it is "High Time We Went" per the Joe Cocker song I am listening to, here is the escape clause.

"Besides embedded process models saving projects a chunk of money, improving plant performance, and justifying better controls and valves by studying the dynamics and integrated functionality of the process and automation system design, you can learn neat stuff like:"

(1) Speed besides size is important
(2) Feedforward signals can do more harm than good
(3) Feedforward head starts based on deltas can help
(4) Linearization of the process variable can be robust and useful
(5) Valve stick-slip can be the upset that keeps on giving

If you are caught within the gravitational pull of this study, I can't guarantee it is not a black hole that sucks you into another dimension.

A process model constructed and embedded in the DCS was used to study a conventional pH and a reagent demand control system with and without feedforward control. In all cases the control loop was in the recirculation line of a vessel to provide a fast feedback correction of abrupt and large disturbances. The feed and reagents were injected at the inlet of a static mixer just before the recirculation stream reentered the vessel. Middle signal selection of 3 pH electrodes was used to inherently ignore a single sensor failure of any type, reduce measurement noise, ignore spikes and slow sensors, and facilitate online diagnostics and calibration. The inline control loop was extremely fast. The transportation delay was only about 2 seconds. The largest potential source of deadtime was injection delay associated with opening and closing of the reagent control valves but this was minimized by coordinated action of close coupled isolation valves at the injection point. Insuring model fidelity for a pH system simply came down to matching the slopes of the model's titration curve with the slopes of the plant's lab titration curve. The following file shows the model and lab titration curves on slides 1 and 2 and the control system on slide 3. Not readable is the slope of 0.015 at 2 and 12 pH.
pH System02 Study Results

First you need to get good lab curves by taking samples of the influent at key times such as steps in a batch sequence when acids or bases are used or during unusual operations such as the pump out of containment areas. The samples should be at the process temperature and titrated with the same reagents used in the automation system. The sample time, temperature, and volume and reagent type and strength must be noted and reagent addition volumes and pH must be tabularized. The typical graphical plots of titration curves showing a vertical line between 3 and 11 pH are next to useless.

The feedforward signal and linearized process variable for reagent demand control were created by use of the same signal characterizer block where the input array was pH values and the output array were corresponding X-axis values per the titration curve. The X-axis was scaled 0 to 100% for the Y-axis and the pH measurement scale of 2 to 12 pH. The first input to the signal characterizer for feedforward control was influent pH. The first input to the signal characterizer for reagent demand (feedback control) was static mixer outlet pH. The second input to both signal characterizers was the pH set point.

Since influent pH measurement errors as small as 0.04 at 2 and 12 pH can cause feedforward errors of 20% or more per the titration curve, it was decided that continuous adjustment by means of a pH feedforward signal could be making large incorrect changes in the reagent flow. It was reasoned that large changes computed in feedforward signal due to large changes in influent flow or pH could be useful as a delta head start to pre-position the valves for the start of a large upset and then let the feedback controller do its thing. This proved to be the case although the feedforward was complicated by the blend of the recirculation stream with the influent at the inlet to the static mixer. Unfortunately, the accuracy of the feedforward curve depended on the accuracy of the titration curve.

Reagent demand control does not deteriorate significantly for changes in the titration curve because only relative changes in the slope are important for linearization and any information is usually better than no information about the shape of the curve. Reagent demand control uses the X-axis of the titration curve scaled as a 0-100% process variable and set point. This control ignores the pH fluctuations near neutrality because these correspond to very small changes in reagent demand due to the steep slope. Reagent demand control also recognizes the true distance of the influent from the set point, which is important for startup and well as disturbances.

Results of the auto tuner showed that the pH controller gain needed to be very low (e.g. 0.02) because of the high process gain from the steep slope of the titration curve at the 7 pH set point. The reagent demand controller gain could be 10 times larger (e.g. 0.25) - see slides 5 and 6 for screen prints of auto tuner results.

A comparison of the conventional pH and reagent demand control is shown on slide 7. The spikes in the static mixer pH are caused by 0.4% stick-slip of the water valves. If the resolution of the water valves was improved from 0.4% to 0.1%, the spikes went away. If the resolution of the acid and base valves upstream or at the static mixer deteriorated from the specified 0.1% to 0.4%, there were many more spikes from the limit cycles of these valves. Normally, a 0.5% resolution control is consider good. This is not so for high process gains. Neutralization systems with pH set points near neutrality are excellent indicators of actual valve resolution and a perpetual stick-slip limit cycle. If you want to know more, check out "Improving pH System Design and Performance" at the Emerson Global Users Exchange this September and the Chemical Processing article on control valves last October.
http://www.chemicalprocessing.com/articles/2007/200.html

The process of creating a model is in itself a knowledge building activity because it makes you think through first principal and dynamic relationships. Often insight into the problem is gained before the model is completed but then again there are the surprises from the test runs of the models due to the interactions and multivariable nature of most processes. Last week I gave my first model as an example. Here I jump forward 30 years to my most recent model, which explores glucose control of a mammalian cell culture.

Even if you are not into bioreactors, you might want to read on because there are insights here for concentration control of batch reactors in general. Just substitute your favorite reactant for glucose and reaction rate for consumption rate.

Most bioreactors to date have a glucose feed rate scheduled as part of batch sequence. The feed rate changes are usually developed during research and development and fixed for the commercial process for the industrial plant. The chances that the glucose feed rate exactly matches the glucose consumption rate is next to none.

The advent of analyzers to measure glucose at-line and NIR probes to measure glucose online opens the opportunity for glucose concentration control and consequently finding and maintaining the optimum glucose concentration for cell growth and product formation as discussed in the article titled "Unlocking the Secret Profiles of Batch Reactors" http://www.controlglobal.com/articles/2008/230.html

For a step change in glucose feed rate, there should be an integrating process response. However, test results from a bioreactor model show that for step increases in glucose feed rate the response started to ramp but then leveled off and decreased for steps on day 2 and flat lined for a +5% step and accelerated upscale for a +10% step on day 8. This odd behavior is the result of a glucose consumption rate that parallels the exponential, stationary, and death phases of the batch. These phases can be seen in the dissolved oxygen controller output. In Glucose Test Results, slide 1 shows a batch with automatic glucose control and slides 2 and 3 show batches with the glucose controller in manual with steps at day 2 and 8 of +5% and +10%, respectively in the controller output. Not shown is the ramp down and eventual depletion of glucose for decreases in feed via steps of -5% and -10%. In each case, the glucose feed and consumption rate were in balance because the controller was in automatic prior to the first step. This may not be the case for new or improperly tuned controllers.

Slide 4 shows the glucose control test results described in the aforementioned article for an online probe (no delay) and an at-line analyzer (11 hr sample delay). These test results show that determining the integrator gain and arrest time is essential and that the use of a feedforward signal can provide remarkable improvement especially for at-line analyzers.

Obviously the size and duration of the steps and their time in the batch determines whether you see a self-regulating, integrating, or runaway response and even an eventual reversal of the process gain. So how does one tune such an animal? The short cut method as described on pages 53-57 of the Good Tuning: A Pocket Guide 2005 second edition published by ISA, which uses the initial change in the ramp rates, may be your best bet. The method identifies a "pseudo integrator" ("near integrator" ) gain and does not require the controller be in auto or the process be lined out at the start of the tests. The referenced pages are in the book excerpt Good Tuning Short Cut Method.

The relay oscillation auto tuner can provide successful results if the step size is large enough to overcome the changes in the consumption rate. However, for at-line analyzers, the ultimate period and consequently the test time may be too long. For the default setting of 3 cycles and assuming an average period of 4 deadtimes, the auto tuner would typically take 12 deadtimes. The user may find it adequate to use the results available after just one cycle (4 deadtimes). The short cut method can provide an estimate of the tuning settings in about 4 dead times assuming you make 2 steps and wait at least 2 deadtimes to see the change in ramp rate. Regardless of method, the tests for an at-line analyzer with a 4 hour or longer sample time will cover several shifts. New adaptive online tuning tools such as DeltaV Insight offer the opportunity to non-intrusively find better tuning settings from the initial response of the loop at the start of the batch and the response to normal set point changes during the course of the batch. However, the auto tuner and the short cut method might be still be useful for getting a new loop in the ball park to enable basic closed loop control from the get go.

The glucose consumption rate depends upon cell growth and to a lesser extent on product formation rates. The oxygen uptake rate can be estimated from the dissolved oxygen controller output (more specifically the secondary loop flows for air and/or oxygen sparge). If this inferred oxygen uptake rate is then corrected for maintenance and yield factors, it is a good candidate for a feedforward signal if the dissolved oxygen control is fast and tight so that changes in process are rapidly transferred to the controller output and the mass balance of dissolved oxygen in the broth is maintained.

What steps can be taken to make the real loop deadtime step forward? Last week we found that a step made in the controller setpoint rather than its output for a controller gain less than one increases the deadtime because of the time it takes for the controller output to work through the valve deadband and resolution. Slower tuning makes the deadtime larger. Subsequent increases in Lambda factors for additional robustness can get the user into a downward spiral in terms of loop performance (slower tuning -> larger deadtime -> slower tuning -> larger deadtime).

Additionally, small steps in the signal to control valves, particularly those with pneumatic positioners, have a dramatic effect on valve response time and hence loop deadtime. The following tests show that the response time of positioner can increase from 1 second to 100 seconds when the step size is decreased from 10% to 0.2%. While you may not be making such a small change in controller output, consider that a 1% change in setpoint to a controller with a 0.2 gain translates to a 0.2% step in the signal to the valve.

Effect of Step Size on Positioner Response

For slow loops like tank level and temperature, the time it takes for a change in the process variable to work through the resolution limit or noise band of the measurement creates another increase in deadtime. For a 1980s vintage DCS with 12 bit A/D (one sign bit) wide range thermocouple cards, the resolution limit of about 0.25 degrees adds significant deadtime besides loop A/D noise. The additional deadtime can be estimated as the measurement resolution divided by the rate of change of the process variable. For a temperature loop changing 0.05 degrees per minute from a step change in controller output, a resolution limit of 0.25 degrees can add 5 minutes of loop deadtime.

I remember trying to use an auto tuner on level loops on large tanks and waiting what seemed like forever for the measurement to get out of the noiseband. I quickly realized that I needed to take larger steps to drive the level faster before the auto tuner or my brain timed out.

Where tight control is needed for slow level and temperature loops, the controller is normally tuned with a controller gain much larger than one. This is a tip that the step changes in the controller output should be large so you are not waiting till the cows come home to see the process variable stir. For more Texas talk, see my Control Talk Column "Puzzler Roundup" in the July issue of Control magazine.

For a constant flow and set of process operating conditions, is the observed total loop deadtime relatively constant? We know from last week's blog, the deadtime also depends upon the sensor time constant and hence it's fouling. Less recognized is that it depends upon whether a step change is made in the controller output versus its setpoint.

The closed loop deadtime (e.g. deadtime in automatic mode) is generally greater than the open loop deadtime (e.g. deadtime in manual mode).

The deadtime from control valve stick-slip and backlash is the valve resolution and deadband, respectively divided by the rate of change of the controller output. For small step changes (particularly for pneumatic positioners), the response time also gets incredibly slow. For a large step change in controller output, the dead time from stick-slip and backlash is zero and the response time is minimal (except for large actuators). Next week, we will discuss some other ramifications of step size.

For a step change in controller setpoint, there is a kick from proportional action (for a PID structure with proportional action on error) and a ramp from reset action. If the kick is not enough to get the valve to move then the loop has to wait on reset action and the chosen closed loop time constant. Thus the deadtime identified for a setpoint change depends upon the controller tuning. Equations 2-47 through 2-50 in the book Advanced Control Unleashed show the development of an equation to estimate the increase in the deadtime from a control valve based on the open loop deadtime. While, these equations are for deadband, they can be used for stick-slip if you consider that half of a deadband is roughly equal to a resolution limit, which is often the case for the best throttle valves (e.g. sliding stem valves with diaphragm actuators). Note the presence of a detuning factor Kx that is approximately the inverse of the Lambda factor (the ratio of closed loop to open loop time constant).

For adaptive controllers or on-demand tuning software that rely upon setpoint changes, very sluggish initial tuning or an unnecessarily large closed loop time constant specified will lead to a larger identified deadtime and overly conservative settings that tends to keep the loop deadtime larger and hence the controller detuned.

Dead time is bad news because the controller has no effect on the process during this time interval. The minimum peak error for a disturbance is basically how far the process is driven away from set point during the total loop deadtime by the process upset. The minimum peak error from a load upset can be estimated as the average rate of change of the process variable multiplied by the dead time. The minimum integrated error is proportional to the deadtime squared. These relationships for peak and integrated error are developed in Equations 2-38 through 2-44 of Advanced Control Unleashed. If this is not enough to get you to rush out and buy a copy, I am offering for a limited time a $0.25 rebate (generous considering the royalties are donated to a university). Just send me your receipt in a self-addressed and with enough postage to get to my secret island hideaway.

When we first started Emerson's advanced control program in the early 90's, one of the initial objectives of the program was to develop an adaptive control capability that could be used in our control products. However, we realize that adaptive control is one of the most challenging advanced control areas to address from a technical standpoint. Thus, most of the programs resources were initially focused on other areas e.g. on-demand tuning, property estimation using neural networks, simulation, fuzzy logic control and model predictive control. Adaptive control was kept on the backburner for many years with work in this area restricted to technical evaluation of different technologies. Gradually, starting in the late 90's, a more focused effort was put into addressing adaptive control. As a result of this work, the first release of our adaptive control technology was recently introduced as part of the DeltaV Insight product in the v9.3 release. The things that we learned in researching and developing this technology greatly influence the final design of DeltaV Insight.

In the early 90's, one of the first adaptive control technique that we investigated was one developed by Professor Karl Astrom, Lund University. This technique allows the controller gain to be automatically adapted through on-line assessment of process gain. As part of this investigation, we worked with Professor W. K. Ho from the National University of Singapore in researching this technique. Even though the approach proposed by Astrom is technically very sound and is utilized in some commercial products, its application is limited to feedback control and adaptation of controller gain. Since our ultimate goal was to find a technique that could be used to adapt all components of PID feedback control (Gain, Reset, and Rate) and feedforward control (gain, Lead/Lag Time constant, and deadtime), we did not pursue this approach past this initial investigation.

At one point we were offered the rights to an adaptive control technique that had been developed by the engineering department of a major chemical company. To avoid polluting the Emerson development team, we hired an outside consultant to evaluate this technology. It turns out that the technique was based on pattern recognition and the application of rules to establish tuning. Even though this approach is used by some major process control companies, the feedback from customers who had tried this technology was not encouraging. There were reports of erroneous adjusted of controller tuning base on cyclic upstream disturbances that were interpreted as a sign of too much controller gain. Thus, we decided to avoid this approach.

In the late 1990's, Willy Wojsznis came across a very interesting paper on model free adaptive control. This paper helped sparked work that lead to a unique design and implementation of model free adaptive control that we later patented. In the summer of 2000, we sponsored a graduate student under the guidance of Professor Dale Seborg, University of California at Santa Barbara, UCSC, to test and further investigate this technique using process simulations. The basic approach provided to be a reliable method for directly establishing feedback tuning. However, only through inference from the controller tuning was it possible to gain any insight into the process gain and dynamics. Also, the method could only be used for the adaptation of feedback tuning. Therefore, we continue to evaluate other approaches that better met our requirements and would give direct insight into the process gain and dynamics.

In the mid-90's, a number of papers on the application of controller switching appeared in some of the major control conferences as a technique for evaluating best tuning. Also, a few papers were published on the use of model switching to identify process gain and dynamics. The concept as proposed was not practical to implement. However, these techniques offered the promise of allowing process models to be identified for both the feedback and feedforward path. After some consideration, Willy and I developed a new approach which we labeled model switching with interpolation and re-centering. This new approach to model switching required the evaluation of only a limited number of models at any given time. Testing of this technique by UCSB from 2001-2003 showed the method to converge very quickly for a variety of self-regulating and integrating processes.

An alpha version of adaptive control based on model switching with interpolation and re-centering was installed at two chemical plants in early 2004. The results from one of these sites, Solutia, were published in September 2004 issue of Chemical Processing. Based on the positive results of these installations, beta testing was conducted at four sites from 2005-2006 on approximately 1000 loops. As part of this beta testing, a special emphasis was place on quantifying the benefits of adaptive control for the batch industry. We created a video of the Lubrizol installation in which the customer discusses the benefits they realized from adaptive control on their batch process. The things we learned from these beta installations had a great impact on the final product design. In particular, the beta test proved the value of maintaining a record of the models that are identified over time from each loop. Also, the capability to automatically provide tuning recommendation using this technology was seen as a major benefit in improving plant operations independent of whether closed loop adaptive control was applied in the plant.

If you have an interest in learning more about the adaptive modeling technique used in Delta Insight, then the following presentation that Willy Wojsznis and I gave at Emerson Exchange provides information on the technical details on this technology.

Adaptive Technology

Also, additional detail can be found in the two patents that we have on the basic technology and its use with non-linear applications.

How can we get rid of dead time in our loops so we can be rich and famous by Friday? PID controllers with dead time compensation are reported to eliminate dead time in terms of a controller seeing the effect of changes in its controller output. For set point changes where all the controller needs to be concerned with is how its output responds to a new set point, the results are impressive for an exact knowledge of the process dead time. However, for unmeasured load disturbances at the process input, the only way to eliminate dead time other than an improvement in the plant or control system design is to accelerate the control system to the speed of light. So unless you have Scotty and Warp Drive on the Starship Enterprise, you are stuck with the dead time from the process equipment, piping, control valves, instrumentation, and digital devices. A dead time compensator can offer some improvement in load rejection by facilitating more aggressive tuning of the PID but with a considerable risk of oscillations from an inaccurate dead time.

If you don't have time for the details or just want to cut to the chase, here are the recommendations

(1) First improve the PID controller tuning before even considering dead time compensation. Setting Lambda equal to the maximum dead time (Lambda factor equal to the maximum dead time to time constant ratio) is effective for load disturbances at the process input if there are no extenuating circumstances.

(2) Add feedforward control whenever it is possible to measure or infer load disturbances at the process input.

(3) If there is economic justification for further improvement and the dead time can be updated within 25% accuracy for varying operating conditions, trial test and closely monitor a PID with delayed external reset for low dead time to time constant ratios.

(4) For loops with high dead time to time constant ratios, multiple manipulated variables, interactions, or constraints, consider model predictive control.

The ultimate performance achievable in terms of load disturbance rejection depends upon the dead time. In the "Theory" section of Chapter 2 of Advanced Control Unleashed equations are developed that show the minimum peak error is proportional to the dead time and the minimum integrated error is proportional to the dead time squared for unmeasured load upsets. How close the actual performance of a control loop comes to this ultimate performance depends upon PID structure, tuning, and enhancements. This blog focuses on the effect of variations in dead time on the performance and robustness of dead time compensation as an enhancement and Lambda as a tuning rule for disturbance rejection. The two predominant methods of dead time compensation studied here are the Smith Predictor PID and the PID with a delayed external reset.

The Smith Predictor was extensively documented in the 1970s. It provides a new controlled variable that is the response of the process variable to its controller output without dead time. It requires entry of three parameters commonly known as process gain, dead time, and time constant. The Smith Predictor uses these parameters to create models of the process from the controller output. In its most documented form, the Smith predictor subtracts a model of the process with dead time from a model of the process without dead time and adds the net result to the measured process variable to create a new controlled variable. If the model is perfect, the new controlled variable has zero dead time in terms of the controller seeing the effect of its own controller output. Since the maximum allowable controller gain is inversely proportional to dead time, the controller gain can theoretically be increased without limit for a perfect model provided you ignore extenuating circumstances, such as loop interaction, measurement noise, and final element dead band and resolution. One of the practical issues with the Smith Predictor is that the new controlled variable of the PID is no longer the actual process variable. The original process variable must be restored for the operator interface to the PID. Also, performance monitoring or trending must look at the original process variable rather than the new controlled variable used by the PID. Terry Blevins proposed in the 1979 ISA paper "Modifying the Smith Predictor for an Application Software Package" a multiplicative and additive correction of the process variable to deal with changes in the slope (gain) and intercept (bias), respectively in the process model.

The PID with a delayed external reset was informally presented in the 1980s and published in the early 1990s. It simply consists of putting a dead time (DT) block in the external reset. This method only requires that a single parameter commonly known as process dead time be entered as the dead time in the DT block. Terry Blevins documented in the early 1990s how the Smith Predictor for a particular Lambda tuning reduces to this PID with a delayed external reset.

The results presented here show that for a perfect model and the same controller tuning the PID with a delayed external reset performed better for processes with a small dead time to time constant ratio (time constant dominant), whereas the Smith Predictor performed better for processes with a large dead time to time constant ratio (dead time dominant). The Smith Predictor did not do as well for small dead time to time constant ratios because the control error seen in the controlled variable by the PID is much smaller than the actual control error in the process variable. In both cases, the improvement was not as impressive as the improvement gained from setting Lambda equal to the dead time rather than the time constant. Surprisingly the improvement in load disturbance rejection from dead time compensation was greater for processes with small dead time to time constant ratios. This goes against the conventional wisdom that the best opportunity for dead time compensation is for dead time dominant loops. The results can be explained in terms of the ultimate limit for performance of dead time dominant loops being lower. The reduction in the peak excursion from more aggressive tuning settings is negligible for dead time dominant processes because the peak error is essentially the open loop error.

Another startling result was how quickly a Smith Predictor erupted into rapidly growing oscillations in the controller output when the model dead time was more than twice the actual process dead time. The fast full scale oscillations in the controller output resembled on-off control. While it is relatively well known that dead time compensators are sensitive to model mismatch, the effect was expected to be gradual and thought to be more in terms of a model dead time being too small. The concern for rapid deterioration for a model dead time being too large was raised in Good Tuning - a Pocket Guide and was documented for model predictive control in Models Unleashed. While a PID with delayed external reset is also adversely affected by a dead time mismatch in both directions, this PID develops a small amplitude high frequency dither rather than a full scale oscillation in controller output for an excessively high model dead time. The consequence is less severe and may be adequately handled by the addition of a small dither filter inserted in the PID controller output, but this was not tested.

PID controller tuning for self-regulating processes without extenuating circumstances can develop oscillations when the identified (model) process dead time is too small. PID controllers with dead time compensation and model predictive controllers can develop oscillations when the identified (model) dead time is too large as well as too small.

In order to get the performance benefit from dead time compensation, the PID must be tuned more aggressively. In other words, a PID with dead time compensation will perform the same as a PID without dead time compensation if they are tuned the same. While the improvement in integrated absolute error (IAE) for load upsets from more aggressive tuning (higher controller gain and lower reset time) can be accurately estimated for a regular PID, the equation does not work well for a dead time compensator. Furthermore, a dead time compensator soon reaches a point of diminishing returns. For example, the improvement in load rejection of a Smith Predictor from a controller gain that is quadrupled may not be noticeable whereas for a regular PID, it normally results in a four fold reduction in IAE. It is important to remember there is a tradeoff between performance and robustness for any feedback controller in that as you make controller tuning more aggressive to improve load rejection you make the controller more sensitive to changes in the process gain, dead time, or time constant.

A nonlinear gain from the installed characteristic of a control valve has been widely discussed. However, the nonlinearity of the process gain of the temperature or composition response is the inverse and consequently the combined effect is less than documented when these loops directly manipulate a control valve. The variability of dead time is often larger than the variability of the process gain or time constant because the dead time is inversely proportional to a rate (e.g. flow rate or pumping rate or rate of change of a signal) and has many different sources (e.g. valve deadband or resolution, piping transportation delay, mixing delay, process lags in series, sensor lags, signal filters, and discrete communication or scan intervals). Thus, it is problematic to compute the dead time accurately enough to get the benefit of a dead time compensator.

In all of the following test results AC1 is always an uncompensated PID with Lambda equal to the process time constant (lag), which is equivalent to a Lambda factor of one.

The first set of test results illustrates the effect of different tuning. Here AC2 is an uncompensated PID with Lambda equal to the process dead time (delay), which is equivalent to a Lambda factor set equal to the dead time to time constant ratio.

Tuning Rule Test 1

The second set of test results shows how well a Smith Predictor can do. Here AC2 is a Smith Predictor PID with the gain doubled and the reset time halved after Lambda has again been set equal to the process dead time. In other words, this AC2 has twice the proportional and integral action of the uncompensated AC2 in the first set of test results.

Smith Predictor Test 2

The third set of test results shows how well a PID with a delayed external reset can do. Here AC2 is a PID with delayed external reset with the gain doubled and the reset time halved after Lambda has again been set equal to the process dead time. In other words, this AC2 has twice the proportional and integral action of the uncompensated AC2 in the first set of test results.

Delay Comp Test 3

For discussion of the test results and configuration, request from me a copy of the Advanced Application Note 003 titled "Compensation of Dead Time in PID Controllers."

My first exposure to model predictive control, MPC, was in late 1979 when I attended a meeting called by Bob Otto, ISA Fellow. Bob had just returned from the AIChE 86th Annual National Meeting where he sat in on Charlie Cutler and Ramaker's presentation of their paper Dynamic matrix control-a computer control algorithm. This landmark work by Shell was the fore runner of modern day model predictive control, MPC. Bob's assessment was that this technology represented one of the most important developments he had seen in process control. The power of MPC technology comes from the fact that the controller is generated based on a process step response or impulse response model and is designed to minimize the control error over a prediction horizon. Control performance is determined by parameters that specify penalty on error and penalty on move. Soon after Shell's public announcement of their work on dynamic matrix control, Charlie went on to form the DMC Corporation. Since that time, major suppliers of MPC technology have successful addressed a variety of applications. The wide spread acceptance of MPC technology is well documented in the paper by Professors Joe Qin and Tom Badgwell, A survey of industrial model predictive control technology.

In the early-80's, Bob Otto lead an initiative within Emerson to explore the feasibility of embedding MPC technology within a distributed control system. This research focused primarily on single loop applications as documented in the paper Development of a Multivariable forward modeling controller by Bob Otto and Kelvin Erickson. Field trails were conducted using a prototype of single loop MPC. One of the technical challenges that prevented general deployment of this technology at that time was the need to provide a robust means of process identification. Also, it was not feasible at that time to embed general MPC in the controller because of the associated CPU and memory requirements.

By the later-90's, the availability of low cost memory and vastly improved processor performance made it feasible to fully embed MPC technology within the control system. By embedding MPC in the control system, a control system supplier can provide an environment that makes it easier and quicker to engineer and commission MPC applications. Also, by embedding MPC in the controller, it is possible to address applications that require faster control execution e.g. 1sec period of execution. In many cases, embedded MPC control is a valid alternative to the traditional PID based strategies for deadtime compensation, feedforward and override control. If you have no experience with MPC, then some examples of how MPC may be effectively used to replace traditional PID based strategies are contained in the following:

MPC for smaller applications

These examples are based on the DeltaV MPC capability introduced in 2000, DeltaV Predict. This initial capability was targeted at smaller applications (no larger in size than 8x8). The DeltaV advanced control team later developed DeltaV PredictPro to address larger applications (as large as 40x80 in size).

We could talk about how important communication is for our society and even more importantly our marriage but let's stick to something we are more interested in as automation engineers particularly since we essentially have no control over politicians and spouses. So let's talk about communication intervals, control execution intervals, analyzer cycle times, and input scan times.

We tend to think that faster is better but this is not always the case. For example, a bioprocess control engineer recently suggested model predictive control of growth rate in a fermentor would not work because the changes in growth rate were too small. If you consider it is just a matter of time frame, you see a resolution (pun intended). If an analysis was made every hour, the true change in biomass concentration would be small compared to the repeatability of the analysis. The signal to noise ratio for the rate of change of biomass concentration (biomass growth rate) would be poor. However, process control is still possible if the time interval between analysis data points is increased and the result fed to a rate of change calculation described in the article "Full Throttle Batch and Startup Response" in the May 2006 issue of Control. Note that even though this calculation uses a dead time and velocity (rate) limit block, the proper setup of these blocks does not introduce additional dead time. Further details on the configuration and the proper filtering and rate limiting of the process variable before it goes into the dead time block for the rate of change calculation is offered in the following screen print of a module.

Rate of Change Module

The use of a rate of change as the controlled variable is described for PID control of an exothermic reactor in the book A Funny Thing Happened on the Way to the Control Room and for model predictive control of a bioreactor in the book New Directions in Bioprocess Modeling and Control.

Whether we are talking about analyzers, or any sort of digital communication, control, and processing, a dead time is created for unmeasured disturbances from the time interval. The actual dead time to detecting and reacting to an upset depends upon the relative timing of the read (input), write (output), and the upset. If the output is done right after the input, the dead time varies from nearly zero to one time interval for an upset that arrives just before and after the input, respectively. On the average, we can say the upset arrives in the middle of the interval so the average dead time is 1/2 of the time interval. For unsynchronized digital devices, the worst case dead time could be the summation of the time intervals. If the output is done at the end of the time interval, the dead time varies from one to two time intervals for an upset that arrives just before and after the input, respectively. This is the case for chromatographs and other analyzers where the sample is processed and the analysis is ready at the end of the cycle time. Here the average is 1.5 times the time interval (cycle time). The following slide illustrates the concept.

DeadTime from Discrete Devices and Analyzers

Even when dead time is introduced, it has minimal effect on performance for controllers that were detuned since the integrated absolute error for the upset depends on the controller tuning settings. In my Control Talk column in the November 2006 issue of Control magazine, we discussed how an increase in digital time intervals did not have an affect on a controller tuned with a Lambda factor of one until the total dead time exceeded half of the process time constant. Thus, tests on the effect of intervals and cycle times should use different relative timings of the unmeasured disturbance and various tuning settings.

(The above is an excerpt from my Control Talk column in the upcoming December 2006 issue of Control Magazine. Please see the column for a more complete discussion and the latest "Top Ten List").

A small fraction of the control loops in industry are characterized by the process deadtime being dominant i.e. greater than the process time constant. In most cases the source of the process deadtime is associated with transport delay or analyzer sample time for the process measurement. In many cases the loop directly impacts final product and thus can have a significant influence on the process efficiency and product quality. For such a process, the loop response to load disturbances and setpoint changes may be slower than desired since the dominant deadtime limits the amount of reset and gain that may be applied in the loop tuning. One approach that may be taken to improve the control of a deadtime dominant process is to utilize deadtime compensation with the PID. The Smith predictor is one of the best known techniques for deadtime compensation. Also, the Dahlin algorithm has been successfully applied by the pulp and paper industry in the control of deadtime dominant processes such as the paper machine. Having confronted some difficult applications in which process deadtime was a limiting factor in the loop performance, I took some time to look into the different implementations of the Smith predictor and to compare these with the Dahlin algorithm.

Interesting enough, it turns out that mathematically the Dahlin algorithm is identical to a Smith predictor applied to a PI controller if the PI tuning is set in a specific manner. This specific tuning of the PI controller is based on the loop period of execution, process gain, time constant, deadtime and the desired closed loop time constant. Through a sight modification of the Smith predictor, it is possible to extend the use of the Smith predictor to address processes that are characterized by unmeasured disturbances that modify the process gain. Also, it is possible to structure the Smith predictor to allow control to be done using sampled process measurements e.g. composition from a gas chromatograph. These modifications of the Smith predictor are the basis of the Provox deadtime compensation PCA and the DeltaV PID_DEADTIME module template. If you have an interest in this area of control, then the mathematical analysis and derivation of the tuning to provide the response of the Dahlin algorithm using a Smith Predictor and details on the modifications used to extend the applicability of the Smith predictor can be found in the paper Modifying the Smith Predictor for an Application Software Package, T.L. Blevins, ISA National Conference, 1979.