Modeling and Control

At a luncheon meeting of Automation Xchange in Park City on August 22, many key users from the biopharmaceutical industry were interested in doing a better job of pH measurement. Many bioprocesses are adversely affected by a change as small as 0.2 pH. During the course of a fermenter batch, which can take up to two weeks, the drift of the pH measurement can exceed 0.2 pH. It is suspected that coatings may be cause of many problems. The removal of electrodes is problematic because of concerns about contamination. There are no easy solutions but the following general techniques in the design and maintenance of the pH systems may improve electrode performance.

The following practices are offered to improve the performance of pH electrodes where an accuracy of 0.2 pH or better is needed based on field experience and literature as documented in my ISA book; Advanced pH Measurement and Control – 3rd Edition. It is assumed the user has already selected the best features, such as electrolyte and glass type.

Practices Offered to Improve the Performance of pH Electrodes

1. Use a flowing junction reference and spherical glass bulb measurement electrode for accuracies better than 0.1 pH
2. Use smart digital transmitters with built-in diagnostics
3. Use middle signal selection of three pH measurements
4. Allocate time for equilibration of the reference electrode
5. Use ‘in place” standardization based on a sample with the same temperature and composition as the process. If this is not practical, the middle value of three measurements can be used instead as a reference. The fraction and frequency of the correction should be chosen to avoid chasing previous calibration adjustments.
6. Use a fixed process fluid velocity at the highest practical value to help keep the electrodes clean and responsive

The following file provides some background information on these practices.

Info on pH Measurement Practices

The problems in pH control systems for chemical processes and environmental discharge can be more dramatic. Even though the accuracy and control band required may not be any where near as tight as for fermentors (with the exception of chemical reactors and crystallizers), the titration curve can be exceptionally steep and variable and the process fluid extremely harsh. Chemical attack of the glass and poisoning of the reference can be major concerns. Solidified references, special glass formulations, thicker glass, and flat glass electrodes may be important. Still the life expectancy can be so short for some environments (e.g. a few days) that the use of three electrodes is not feasible. In these cases a piston actuated retractable injector assembly may be needed to reduce the time exposure to the process and to provide automated periodic cleaning, rejuvenation, and hydration of the electrodes. A word of caution for low water content streams; variations in the water concentration affect the pH reading even if the gel layer remains hydrated.

Top Ten Signs of a Rough pH Startup

10. Food is burning in the operators’ kitchen
9. The only loop mode configured is manual
8. An operator puts his fist through the screen
7. You trip over a pile of used pH electrodes
6. The technicians ask: “what is a positioner?”
5. The technicians stick electrodes up your nose
4. The environmental engineer is wearing a mask
3. The plant manager leaves the country
2. Lawyers pull the plugs on the consoles
1. The president is on the phone holding for you

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").

If the total loop dead time was zero, you could set the controller gain as large and the reset time as small as desired. If there were no disturbances, you could simply sequence the controller outputs for startup, transitions, and shutdown. Process dynamics, controller tuning, and loop performance would be a non issue.

I once had a loop with zero dead time. I was studying the performance of my new algorithm for adaptive pH control in an Advanced Control Simulation Language (ACSL) program for my Master’s Thesis. The larger I set the controller gain, the tighter the control I got. I was ecstatic. I was going to become “way famous”. Then the let down - I had inadvertently turned off the dead time function. All I had left for process dynamics was a single time constant. The operating point nonlinearity of pH had no effect because I could stay incredibly close to set point. Since then I have seen tuning studies for a single time constant that beat to death a scenario where all the normal concerns are non existent. I decided to become sensitive to dead time especially since I could reduce my time on a pH startup by reducing dead time.

Control textbooks and studies tend to focus on set point responses ignoring unmeasured disturbances at the process input (e.g. load upsets). Special algorithms can be designed and tuned to prove a point. This may work well in simulations, aerospace, and hydraulic systems where dead time is either negligible or predicted/compensated and the servo response rules, but the real world of industrial process control isn’t so kind.

The variety and variability of the sources of dead time and disturbances in process control is quite impressive. The following lists are just some major sources that come to mind.

Sources of Disturbances

1) Limit cycles (split ranged point discontinuity, resolution, and cascade dead band)
2) Interaction between loops
3) Slow secondary loops (cascade control)
4) Design limits (equipment operating limits)
5) Low residence times (e.g. undersized feed, recycle, surge, and waste tanks)
6) Manual procedures and manual valves
7) Field switches (e.g. on-off level control)
8) Activity (catalytic and metabolic)
9) Ambient conditions
10) Interlocks and sequences
11) Raw materials
12) Recycle streams
13) Startups, shutdowns, and product transitions
14) Fouling (e.g. process coatings) and frosting (e.g. crystal accumulations)
15) Parallel trains
16) Undersized cooling towers
17) Bored board operators
18) Shift change
19) Initiatives
20) Goal reviews

My worst experiences have been with undersized recycle, surge, and waste tanks. The residence time (volume divided by throughput rate), which is the process time constant, is so low there is not enough filtering of the changes in stream composition. Also, the level control on these tanks is forced to jockey the feeds to downstream operations to keep the tank from overflowing or running dry. Plants tend to avoid putting in the bigger tank to save money and reduce inventories when they need to debottleneck or push a process.

Sources of Dead Time

1) Discrete execution and communication interval
2) Analyzer cycle time (e.g. chromatograph)
3) Transportation delay (e.g. sample line)
4) Mixing delay (e.g. agitator, eductor, and sparger)
5) Injection delay (e.g. back filled dip tube)
6) Resolution limit (e.g. VSD, control valve)
7) Dead band (e.g. VSD, control valve)
8) Instrument time constants in series (e.g. sensor and signal filter lag)
9) Process time constants in series (e.g. thermal lags and residence times)
10) Lab samples (e.g. sample hold, processing, and analysis time)

Dead time is often inversely proportional to a rate and therefore a function of test conditions. The dead time from transportation delays, sample lines, sensor lags, and residence times in series is inversely proportional to flow rate. Mixing dead time is inversely proportional to agitator pumping rate or eductor flow rate. The dead time from dead band and resolution limits is inversely proportional to the rate of change of the signal (e.g. rate of change of process variable for measurement resolution limits and rate of change of controller output for valve dead band and stick-slip). The time it takes a measurement to get out of its resolution limit or noise band can be significant for level or temperature and depends upon how fast the process is driven to change and hence the step size in the controller output or set point. The dead time for control valves becomes just the summation of the pre-stroke dead time, discrete processing, and communication interval (all usually small) if the step in controller output is larger than the valve dead band or resolution limit. The dead time effect of dead band and resolution limits unfortunately does show up for unmeasured load upsets at the process input.

My intention is now to avoid any further dead time or disturbances to an evaluation of dead time compensators and model predictive control so check here next week for more fun than control engineers should be allowed to have with advanced control.

A control valve isn’t doing much to help a control loop deal with the minute by minute onslaught of disturbances if it does not respond to the controller’s output. Yet there is normally nothing in a control valve’s specification form to insure the control valve actually moves. A step forward has been the ANSI/ISA standard 75.25.01 for a control valve step response testing procedure but I wonder if any where near as much effort is put on making sure the valve movement is smooth and sensitive as is spent on the valve size and leakage spec?

I was sensitized to the sensitivity of the control valve because my first area of expertise was pH when I moved from E & I Construction to Engineering Technology. The high process gain for strong acids and bases makes pH loops ideal for identifying valve response limitations. A jump in valve position of just 0.1% can cause a several pH swing. Putting a pH loop in automatic may initiate large amplitude oscillations even though there are no load upsets. In the end I realized great control valve sensitivity could reduce the number of stages of neutralization and save big bucks in process equipment required (see the 3rd edition of the ISA book titled Advanced pH Measurement and Control).

There is a growing awareness that a resolution limit from stick-slip in a control valve can cause a limit cycle in a control loop because the valve position is never exactly were it needs to be. Even if there are no disturbances, integral action in the controller drives the output until it moves, but then it steps right past the right valve position. Besides the limit cycle, there is also a dead time that is the resolution limit divided by the rate of change of the valve signal (controller output). To make things worse a slower rate of change of the controller output increases the resolution limit in some positioner designs. Consequently as the controller tuning is slowed down (Lambda is increased), the dead time and possibly the resolution limit is increased.

Deadband can be just as deadly. Whenever the controller output has to reverse direction, the change has to be greater than the deadband before the valve moves. The result is a dead time that is proportional to the deadband divided by the rate of change of the valve signal (controller output). If the are two integrators in the loop, deadband also creates a limit cycle. The two integrators can be the result of a controller with the integral action on an integrating process (e.g. level) or a cascade loop where the secondary and primary loops both have the integral mode (e.g. PI or PID controller) as discussed in the article “Life is a Batch” in the June 2005 issue of Control magazine.

Stick-slip normally originates from friction in stem packing or from sealing surfaces on the trim. Excessive tightening of the packing, high temperature packing (e.g. graphoil), older types of environmental packing, tight shutoff ball and disc seals, and low gain or spool positioner designs create more stick-slip. The friction is generally worse near the closure position, so most tests results are cited at higher valve positions (e.g. > 20%).

Ever since I started my career almost 40 years ago, inexpensive actuators and positioners have been added to tight shutoff rotary valves original designed for on-off or isolation service. The package is attractively priced and pitched as a control valve that meets or more unfortunately exceeds the valve’s capacity and leakage spec. If the process, mechanical, and instrument design engineer each add extra capacity in the piping, pump, and valve, the result is the extreme sport of a control valve riding the seat. If engineers attempt to make the control valve serve the additional purpose of isolation besides throttling, the problem of popping on and off the seat is magnified. In general, an isolation valve does not make a good throttling valve and vice versa.

In rotary valves, shaft windup can occur, where the actuator shaft twists but the ball or disc does not move because of high friction of the sealing surfaces. Eventually, the ball or disc breaks free and jumps to a new position. If the positioner, no matter how smart it think it is, measures actuator shaft position rather than ball or disc travel, it may report everything is relatively OK. I have seen a whole series of fancy plots from a smart digital positioner with vertical travel actuator shaft position feedback consistently show the stick-slip was less than 0.5% for a butterfly valve designed for tight shutoff (not too bad for the particular application). A travel gage added to the disc in the shop test setup gave the reality check that the stick-slip was actually 9% (lousy for any application).

Deadband is also known as backlash and is often larger in rotary valves because of rotary actuator and shaft coupling design or the need to translate from vertical to rotary motion. Be careful about the use of the term deadband. Purists will argue that deadband is the offset in the plot between an increasing and decreasing valve position for a full scale change in valve signal. In practical terms we think of deadband as the reversal in valve signal necessary to reverse valve position anywhere in the signal range. In the following plot of actual ball travel versus controller output, the stick-slip is evident for changes in the same direction and the deadband shows up for a change in direction of the valve signal. This plot is for the controller in automatic and shows that with a bit of understanding and practice, the dead band and resolution limit can be identified from trend charts. For rotary valves, this presumes there is a measurement of the actual ball or disc position or flow thorough the valve. For sliding stem valves, actuator shaft position read back is normally sufficient because there is a more direct connection of the shaft to the trim stem and no translation of motion.

Valve Deadband and Resolution

For the use of a model predictive control to achieve better valve sensitivity and rangeability see the article “A Fine Time to Break Away from Old Valve Problems” in the October 2005 issue of Control magazine. For equations on how to estimate the amplitude and period of limit cycles from a resolution limit or deadband see the article “What is Your Flow Control Valve Telling?” in the May 2004 issue of Control Design magazine.

To end on a lighter note, here is list to identify with:

Top Ten Exceptional Valves

(10) A measurement with 0.1% repeatability
(9 A control valve with 0.1% dead band
(8) A control valve with 0.1% resolution
(7) A controller that is tuned
(6) A process that is simulated
(5) Any computer picked out by your son
(4) Any canceled all week team building exercise
(3) Any afternoon meeting at the Oasis in Austin
(2) Any conference in Park City
(1) Any writing expedition in Naples

Next week’s blog discusses the merits of a block added to the PID controller output to compensate for valve resolution and deadband.

Last week, we discussed the deadly and sticky situations of some control valves. I once thought a variable speed drive (VSD) was the solution until Bob Heider pointed out that some VSD packages may not be as sensitive as a well designed throttling control valve. A VSD resolution may be 0.4% whereas a sliding stem control valve with low friction packing and digital positioner can be 0.1%. Jae Park looked at the specs for a particular model VSD and found out the following:

I Speed control

a. Speed regulation without the feedback (without the encoder): 0.1% of base speed across 120:1 speed range

b. Speed regulation with the feedback (with the encoder): 0.001% of base speed across 120:1 speed range

II Command signal accuracy

a. Typically 9 to 12 bits for analog in VFD from different manufactures. If one bit is a sign bit this corresponds to a resolution of 0.05% to 0.4%.

b. If using digital input, accuracy is 0.01% of set output frequency.

Obviously, here the limitation is the analog command signal and a low resolution A/D.

If you refer your project manager to this website and you are still not set free to buy a fine final element and are stuck with a sloppy control valve, where the backlash plus sticktion can range from 0.5% to 10%, you may need to resort to desperate measures. This resorting can be retirement to Stan’s country (Naples) or adding a resolution and deadband compensator from a library of composite templates to the controller output.

A deadband (backlash) compensator can be as simple as an addition or subtraction of a half deadband offset to the controller output when it reverses direction. The following screen prints show the configuration of the composite block.

Deadband Compensator

The compensation of resolution (stick-slip) is a bit dicey. One implementation uses the relay auto tuner method of a single step (kick) of the output in the direction to return the process variable (PV) to its set point (SP) when the PV gets out of the noise band. However, in this case the kick is equal to the stick and is not necessarily large enough to cause the PV to cross back over the SP. This method requires that the control action and valve action be correctly provided as inputs. The block also shows an optional dither of the loop to help reduce the sticktion since some valves have trouble breaking free when stuck in one position for a long time (often called freezing in position even when it caused by hot temperatures). The worst case is if the valve is normally closed and designed for tight shutoff. The whole motionless gig is kind of like me in a lazy boy chair getting commands from my spouse. I can be as sensitive as the finest example of my gender but as I get older my joints get stiffer the longer I sit.

Since constant dither can wear out valve or body parts, dither amplitude and frequency is important from both a maintenance and variability view point. The dither is more suitable for composition, gas pressure, or temperature loops on a large well mixed volume, because these have a slower natural period (less frequent dither) and larger process time constant (more effective filtering). The following file shows the reduction in oscillation amplitude in the primary temperature loop of a cascade temperature control system by the addition of a resolution compensator on the secondary (coolant) temperature controller’s output. Also shown are screen prints of the configuration of the composite block.

Resolution Compensator

These compensators need to be tested and adjusted carefully because of many practical issues. The deadband and stick-slip are never constant. For example, the value depends upon the throttle position, magnitude and direction of the change in the controller output, and the time in service of the valve. The slip can also be greater than stick, particularly if the actuator is undersized or the valve is coming off the seat. Fine valves do not age like fine wines. Crud can build up on the stems and trim (another reason why dither may help). A kick or offset that is too large will create additional slip and do more harm than good so underestimates of the deadband and resolution limit are wise. Some software packages can identify the deadband and resolution automatically online as documented in the paper “Valve Diagnostics in an Adaptive Control Loop”.

The following file, which is Appendix A of the paper, describes the use of a composite block that has a concise code for the simulation of valve dead band and resolution. The block also models the rate limited second order second order response of the actuator. The user can set the pre-stroke dead time, the slewing rate for increasing and decreasing directions, and second order lags. For large valves, the actuator dynamics are significant because it takes time to move enough air in and out of the actuator to move its shaft. These dynamics are particularly important for compressor anti-surge control valves. Until recently, when a supplier provided the dynamic response of a control valve, it was for an actuator not connected to a control valve even though not too many of these were sold this way. The dynamic response did not include the effects of backlash and sticktion.

Valve Diagnostics Appendix A

Control systems assume linearity. Unfortunately the world is basically nonlinear. For the next few weeks we are going to explore how gains, process time constant, and dead times change with plant design and operating condition. This week we start out looking at valve gains.

A plot of the flow versus valve position (installed characteristic) of most control valves is nonlinear. Here the slope is the valve gain. If we were to plot a process variable versus this flow, such as temperature or composition, it would also be nonlinear. Here the slope is the process gain. These are called operating point nonlinearities. If the process variable stays close to its set point, the slope doesn’t change much. Thus, for a constant set point, minimal dead time, and good tuning, the process nonlinearity is not much of an issue. On the other hand, the control valve may have to move a lot to achieve tight control. The loop is more likely to see the nonlinearity of the control valve. Generally the slope of the installed characteristic gets too flat at low and high positions. Entech published a gain specification that the % flow divided by % signal should be between 0.5 and 2.0 (a gain change of 4:1). The following examples of installed characteristics show that the throttle range is shortest for a butterfly valve and longest for a sliding stem valve. For a detailed discussion of these figures see Chapter 2 of Advanced Control Unleashed.

Valve Gains

The rangeability statements by valve manufacturers are defined in terms the uniformity of the inherent characteristic. These statements do not take into account a gain specification, an installed characteristic, or the increased stick-slip at low valve positions from friction of the seating and sealing surfaces, particularly for tight shutoff valves.

A signal characterizer block can be inserted between the controller output and analog output block to compensate for the nonlinearity of the control valve gain. The characterizer is set up to calculate the % flow from % position (the Y axis from the X axis of the installed characteristic). The input signal to the control valve is now % desired flow rather than % desired position. This can confuse operations and maintenance if not adequately documented and displayed. The accuracy of this gain compensation depends upon the knowledge of the system pressures and friction losses that affect the pressures at the inlet and outlet of the control valve. Software can predict the installed characteristic but this is done typically offline with manual entry of data. There is an opportunity for pressure measurements upstream and downstream to provide better compensation of the valve nonlinearity besides facilitate the monitoring and trouble shooting of disturbances. Many times I wished more pressure transmitters were installed to figure out why a loop just got clobbered, but this is another story.

Another practical issue relates to valve stick-slip and backlash, whose effect and compensation we alerted readers to in our Dec 4 and 11 blogs. For operation on the steeper portion of the installed characteristic, the characterizer makes the change in signal to the control valve smaller. Thus it takes longer for the signal to work its way through the resolution limit and dead band. However, for operation on the flatter portion of the installed characteristic, the change in the control valve signal is larger reducing the dead time from the resolution limit and dead band. If you ever waited for the controller output to work its way along the upper flat portion of a butterfly valve characteristic for a process unit operating at or beyond its design limit, you can appreciate the acceleration offered by the signal characterizer. Of course, at some point you just run out of valve and need to take a look at the pump and piping system design besides the valve size.

First principal relationships can define process cause and effects that can lead to improved controller tuning and performance by the selection of better tuning rules and process variables for scheduling of tuning settings. It also affects the choice of control valve trim and the feedforward design. The understanding of these relationships does not require a degree in chemical engineering but presumes just some understanding of common terms (e.g. heat transfer coefficient and area), relationships (e.g. ideal gas law), and physical concepts (e.g. conservation of mass and energy).

Equations have been developed from first principal relationships for the process gains, dead times, and time constants of volumes with various degrees of mixing. The results show that for well mixed volumes with negligible injection delays, the effect of flow cancels out for the controller gain if one of the following methods is used: Lambda self-regulating rule where Lambda is set equal to the dead time, or the reaction curve method. The effect of flow also cancels out for the reset time besides the controller gain if the process is treated as a “near integrator” and the Lambda integrating tuning rule is used. This is because the flow rate cancels out in the computation of the ratio of process gain to time constant that is the “near integrator” gain. This ratio and “near integrator gain” are inversely proportional to the process holdup mass (e.g. liquid mass). However, for temperature control the effect of changes in liquid mass cancels out because a change in level increases the heat transfer surface area covered. Several authors have mistakenly tried to schedule controller tuning based on liquid level for reactor temperature control. One author has reported being bewildered by its failure. This is not the case for gas pressure control. The equations show that liquid level has a profound effect on the process integrating gain for vessel pressure control because it changes the vapor space volume without any competing effect. To summarize, the integrator gain for composition and gas pressure is inversely proportional to liquid level (liquid mass). For temperature, the effect of level cancels out unless the level is above or below the heat transfer surface area, which is unusual but can occur at the beginning or end of a batch when coils instead of a jacket is used for heat transfer. For temperature, the integrator gain is nearly always proportional to the overall heat transfer coefficient that is a function of mixing, process composition, and fouling or frosting.

The equations also show that if the transport delay for flow injection is large compared to the time constant, which does occur for reagent injection in dip tubes for pH control), then the controller gain will be proportional to flow. Note that pH control is a class of concentration control.

For the control of temperature and concentration in a pipe, the process dead time and process gain are both inversely proportional to flow and the process time constant is essentially zero, which makes the actuator, sensor, transmitter, or signal filter time lag the largest time constant in the loop. Thus, the largest automation system lag determines the dead time to time constant ratio. For a static mixer, there is some mixing, and the process time constant is inversely proportional to flow but is usually quite small compared to other lags in the loop. The controller gain is generally proportional to flow for both cases.

Finally, the above has implications so far as whether a flow feedforward multiplier or summer and whether a linear or equal percentage trim should be used. A flow feedforward multiplier and equal percentage trim, which both have a gain proportional to flow, can help compensate for a process gain that is inversely proportional to flow provided the process time constant is not also inversely proportional to flow. This is generally the case for temperature and concentration control of essentially plug flow volumes (pipelines, static mixers, and heat exchangers). For well mixed volumes, feedforward summers and an installed linear characteristic for valves is generally best. For control valves this corresponds to a linear trim when the available pressure drop that is much larger than the system pressure drop or critical pressure drop so the installed flow characteristic is close to the inherent flow characteristic.

The results are also useful for determining the dead time to time constant ratio, which has a profound effect on the tuning factors used and the performance of dead time compensation, which has been discussed in the category of controller tuning and control performance on this blog site. A copy of Advanced Application Note 4 that summaries and derives these equations is available from me.

(10) Stop talking about process feedback. I promise to do this right after this blog.
(9) Listen intently to my wife’s instructions. Why does my mind still jump to weighty matters like what is next for dinner?
(8) Stop making cheap control valve jokes. Could the next final element reputation I hurt be my own?
(7) Help make smart diagnostics smarter. Do I need to de-fussify my fuzzy logic?
(6) Stop lusting in my heart for more computing power. Is it the PC or me that is the constraint?
(5) Turndown the volume on my headphones. What did you say?
(4) Stop drinking cheap wine. Does good wine ever come in a size large enough?
(3) Read a college text on control theory. Can I watch Star Trek without setting up the state space equations?
(2) Stop answering a question with a question. Why should a consultant do this?
(1) Spend more time with my wife than with Control magazine. Whatever happened to my January issue?

Last Chance to Learn About Process Feedback

Processes can have negative feed back, no feedback, or positive feedback. A negative feedback process is a self-regulating process and will decelerate to a new steady state. A process with no feedback is an integrating process and will continually ramp. A process with positive feedback is a runaway process and will accelerate until hitting a relief device or safety interlock setting. As in circuits, positive feedback is problematic. Control systems use negative feedback for regulation and provide a correction to the process in the opposite direction of the excursion from set point to compensate for load changes. The control system has to be progressively more aggressive for integrating and ultimately positive feedback processes. The following file shows the response to of these three types of processes to a change in controller output with the controller in manual (open loop responses of self-regulating, integrating, and runaway processes).

Types of Process Feedback

Over 90% of the processes are self-regulating. However, many of the continuous and fed-batch processes involving large back mixed volumes with the greatest direct economic benefits behave in the time frame and control region of interest like they have an integrating response and can be best treated as “near integrating” processes. The classic integrating process is a pure batch or level process. Less than 1% of the processes are runaway. When these exist, understanding the runaway response is critical in terms of safety and control because of the propensity to accelerate and reach a point of no return. Runaway responses are almost exclusively associated with highly exothermic batch reactors used in plastics and specialty chemical production. I was in a control room when a batch reactor reached the point of no return and was going to blow over to a flare stack tank. There was nothing the operators could do except call for the evacuation of the reactor area and make sure the flare stack was ready.

I haven’t gotten any feedback on process feedback, which is really fortunate because it allows me to keep my new years resolution, so I will slog the blog along to the question that has been keeping people awake at night; “When is a batch a batch?”

Some say all processes like life are batches because there is beginning and an end.

Traditionally labeled “continuous processes” can significantly benefit in terms of process efficiency, time, and safety from the application of batch sequence technology for the automation of startups, grade transitions, and shutdowns. Conversely, batch processes can benefit from what has been classified as continuous control techniques.

There are some important advantages in terms of the application of PID and model predictive control by being more definitive in the distinction of batch and continuous operation. A process is best classified as “continuous” when there is both a feed flow and a discharge flow. Thus, the startup and shutdown of continuous processes, where the discharge and feed flows are zero, respectively, is better controlled if recognized as effectively a batch process. Also, fed-batch processes, which have a feed flow but no discharge flow, while termed by some as “semi-continuous” is better treated as batch.

Batch processes have an integrating (ramping) response as described in last week’s blog. Such responses have different tuning rules and settings. If there is a zero load (e.g. no heat loss for temperature or conversion for concentration), the response is one sided, which means the process variable can only go in one direction. These processes will overshoot if reset action is used. Also, model predictive control (MPC) doesn’t work because it assumes it can drive the controlled variable in both directions (up and down).

For more details implications for PID control of batch processes see Advanced Application Note 4, the April 2005 Control Talk column in Control magazine, and the article “Life is a Batch" in the June 2005 issue of Control magazine. For information on the need and method of translation of variables for model predictive control of batch processes, see chapter 4 in the book New Directions in Bioprocess Modeling and Control published by ISA in 2006.

Time is precious so here is your chance to learn in five minutes what took me five weeks of investigation. While most of these thoughts were banging around in my mind for last couple of decades, they might never have congealed if not for some triggering thoughts from my colleagues Terry Blevins and Willy Wojsznis and some knowledge discovery in my favorite laboratory, the virtual plant. All of this stuff has been discussed to some degree in last year’s blogs with more detail available in my Control magazine articles and Control Talk columns, the book New Developments in Bioprocess Modeling and Control, and Advanced Application Notes 1-4. The notes and presentations based on my ISA books as they become available are free for the downloading from the website: http://www.EasyDeltaV.com/ControlInsights/

(1) All the most popular tuning rules reduce to the same equation for the controller gain for maximum load rejection.

(2) While the ultimate performance of a loop is proportional to the dead time squared, the actual performance is set by the tuning (reset time and controller gain).

(3) Nearly all studies on the beneficial effect of improving loop dynamics retune the controller for better performance. If the controller was not retuned, there would be no immediate recognizable benefit in most cases.

(4) You can estimate the amount of dead time you can add before the loop performance deteriorates for unmeasured disturbances by comparing the present controller gain to the maximum controller gain for maximum load rejection.

(5) I would be out of a job if there was no dead time or disturbances, because barring any extenuating circumstances the controller gain could be set higher than you have ever seen or the control valve just sequenced to predetermined positions.

(6) Continuous temperature, concentration, and pH control loops on large well mixed volumes are best treated as “near integrators” for tuning.

(7) The use of dynamic reset limiting and a delayed external reset can provide dead time compensation that is easier to implement and more robust than a Smith Predictor. If the valve position PV for single loops and the secondary loop PV for primary loops is used for external reset, it prevents the controller from outrunning the valve or secondary loop and the dead time compensation is more accurate.

(8) If the model dead time used for the Smith Predictor is 100% larger than actual, the Smith predictor can break out into rapidly growing oscillations. A model dead time that is too large besides too small can cause instability in this predictor.

(9) The controller gain setting must be significantly increased beyond the normal maximum controller gain to realize the benefit from dead time compensation.

(10) A zero discharge flow causes the mass to increase as a batch progresses, which causes concentration and pH control to have an integrating response. The integrating process gain here is inversely proportional to level. For vessel pressure control where the vent valve pressure drop is large or critical, the pressure response’s integrating process gain is proportional level because the vapor space volume is decreasing. However, for temperature control where there is significant heat release and cooling capability, vessel level has little effect on the controller gain except when it is above or below the heat transfer surfaces (e.g. coils) because the effect of more mass is cancelled out by more heat transfer area covered by liquid.

Last year we had disturbing thoughts on how fast upsets were particularly disruptive and anything you can do to slow them down makes the job of a loop much easier. In real processes, step disturbances are quite rare. However, there are some noticeable cases (e.g. on-off level control, interlocked and sequenced valves, and compressor surge control) where stuff comes at the loop too fast.

If level switches are replaced with a Hart or Fieldbus properly applied level transmitter, correctly tuned level controller, and throttling control valve with a digital positioner, you will make everything smoother downstream. The cost of the better automation system will pay for itself in terms of better reliability and visibility and reduced variability.

On-off valves must in many cases be sequenced and interlocked. The effect of these valves may be underestimated. Even for large valves with slow stroking times, most of the time is spent on the flat portion of the installed characteristic. For example a reactor air feed isolation valve had been deliberately slowed down by a restrictor on the actuator to take 145 seconds to stroke to allow the air compressor surge control system time to open its vent valve. An analysis of the installed characteristic revealed that there was actually only 1.7 seconds between when the flow dropped below the anti-surge controller set point and the flow hit the compressor surge line. The total time on the steep portion of interest in the installed characteristic was less than 3 seconds. The speed of the upset could have been regulated much better by a programmed partial reduction in air feed flow set point followed by a fast closing of the on-off valve to prevent reverse flow. I am must make it very clear at this point that a control valve should not be considered as a replacement for an on-off valve, or vice-versa. They serve distinctly different purposes. A control valve needs to have minimal seating friction for throttling and an isolation valve tight shutoff for isolation, which may be conflicting objectives.

Once compressor surge starts, not much can be done by a flow controller because it is like going over a cliff. The precipitous drop in flow occurs in less than 0.03 seconds. This was mistakenly interpreted as requiring a special microprocessor with a scan time of less than 0.05 seconds when really the control valve on big compressors often didn’t do much of anything for a second or more because you physically couldn’t move enough air out of the big actuator for the fail open vent valve. Also, the feedback controller needed to do something before it hit the surge curve. Once a compressor is in surge, an open loop back up is used to get out of surge because a flow reversal occurs every second or so totally confusing the controller.

For more details on compressor surge control see the books Centrifugal and Axial Compressor Control and A Funny Thing Happened on the Way to the Control Room (reprints available through UMI). Next week we will talk about the use of a simple open loop backup configured in a DCS to assist a PI loop for those applications were you need fast recovery for property and environmental protection.

Compressor anti-surge control is an extreme case but there are many applications particularly for parallel trains of equipment and batch to continuous transitions where it is advantageous to slow down disturbances by a coordinated startup and shutdown of flow set points.

Pressure waves travel at the speed of sound in the fluid (e.g. 1100 fps) whereas composition changes travel at the pipeline fluid velocity (e.g. 5 fps). The pressure waves can also reflect back and forth (e.g. water hammer), which like surge can be totally disruptive. Whether you are talking about pressure or composition changes it is wise here as well to slow them down by ramping the flow controller set point in the DCS rather than restricting the air flow to the actuators of on-off valves in the field. It is also beneficial to use pressure transmitters instead of pressure gages. Operator typically cannot outrun a pressure wave to get to the right one. In some cases we don’t know even whether a pressure upset is originating upstream or downstream let alone where it specifically starts. If you think about it, some field pressure regulators are also best replaced by pressure loops in a DCS to provide more visibility and control over the propagation of pressure waves and the allocation of pressure drops to prevent interaction and cavitation.

Momentum balances, normally not a part of dynamic process models, are required to simulate pressure waves and surge.

Top Ten Stuff That Comes at You Too Fast

(10) On-off level controlled flows
(9) Sequenced and interlocked flows
(8) Strong acids and bases
(7) Pressure waves
(6) Compressor surge
(5) Dunk shots
(4) Ice pucks
(3) Late night car commercials
(2) Corporate Restructuring
(1) Retirement

If you ever wondered if you are agitated enough, then this blog may help stir up some thoughts. Specifically, how does the relative type and degree of mixing in the plant design affect your job as an automation professional? If the process engineer tells you the project is installing a radial instead of an axial agitator, do you shudder with profound disappointment or just utter a sigh? What if the agitated vessel is replaced with a static mixer? Do you turn the project over to the intern and take early retirement?

I will first continue my role in life as a pH stalker but then move on to other processes and more general considerations.

I became sensitized to mixing because of the extreme sensitivity of pH loops to plant design. I have talked before about how pH processes are the best known indicators of valve stick-slip, particularly near the neutral point, A control valve resolution of 0.1% (exceptional by any standards) can cause a pH swing that is more than noticeable.

Similarly, pH processes are the best known indicators of the uniformity of mixing. Concentration fluctuations in hydrogen ion concentration as small as 0.0000001 normality can cause noise with a 1 pH amplitude at the neutral point. The only study I have seen on the mixing required for pH was a cop out because it was done at 4 pH where the sensitivity (slope of the titration curve) was 1000 times less.

Additionally, the consequences of mixing delays are most severely felt in pH processes. An increase in loop dead time increases the excursion in pH for a given load upset, which increases the nonlinearity seen by the control loop. The operating point nonlinearity for pH can be extreme. The process gain is proportional to the slope of the titration curve and inversely proportional to the total flow and can change by a factor of ten for each pH unit deviation from the neutral point in a strong acid and base system.

The game in a pH loop more than any other loop is to minimize noise and dead time.

For other processes, the required degree of mixing is a lot less, but whether you are talking about temperature or concentration control, poor mixing still shows up as more noise and more dead time. The percent nonuniformity from mixing multiplied by the conversion factors to get to percent of the measurement scale gives you the noise amplitude seen by the controller algorithm. The dead time from mixing in a well designed agitated vessel is roughly the turnover time, which in turn can be approximated as the liquid volume divided by the sum of the feed flow, recirculation flow, and agitator pumping rate. The average dead time is ideally more like ½ of the turnover time whereas the maximum dead time is the turnover time. This helps explain why you see ½ to 1 times the turnover time in the literature as the mixing delay. Since we are generally short on our dead time estimates because there are so many sources of process dead time, I don’t like to skimp on the mixing delay. See my Nov 20 2006 blog “Without Dead Time and Disturbances I Would Be Out of a Job” in the Plant Design category for a list of sources.

Unfortunately, the above assumes the liquid height is about the same as the vessel diameter (unless there a multiple levels of impellers), baffles every 90 degrees to prevent swirling, and an axial agitator to pull down liquid (not air) from the surface. If a camera shows the surface not being broken or swirling, or there is foaming, you can say “Houston we have a problem”, particularly if the vessel vendor or design firm is in Houston. In processes that cannot withstand high agitation because crystals or cells may be broken by the blades, there may be an opportunity to increase the recirculation flow and use a jet or eductor to amplify the effect of the flow (e.g. jet fermentors).

Bob Heider, adjunct Washington University professor, wisely pointed out that baffles cannot be used for biomass, crystals and particles when the baffles cause the solids to dam up or break up. Bob also provided the following memory dump on agitation.

Agitation Info

A bigger potential source of dead time is the injection delay from dip tubes for small manipulated flow (e.g. nutrient, reagent, reactant, or additive). The normal design practice is to have a robust sized dip tube go about halfway down the liquid to the impeller. Unfortunately, this creates a dead time when the manipulated flow is shutoff for a prolonged period of time that is the submerged dip tube volume divided by the flow. For example, just a gallon volume will cause a dead time of 1 hour for a 1 gph flow when the control valve reopens. There can be an even larger dead time because to see the final effect of stopping the flow, you have to wait till the concentration inside the dip tube drains and migrates into the mixture in the vessel. Various method of reducing injection and mixing delays are discussed in the ISA book Advanced pH Measurement and Control, 3rd edition, 2005.

This brings us to one grand generalization. For concentration changes, the residence time (volume divided by flow) becomes a process dead time for a pipe but becomes a process time constant for a well mixed vessel. Check out next week’s blog for the effect on tuning and loop performance. In the mean time, stay agitated.

Most of the literature makes valve nonlinearity seem bad or just plain ugly. However, if you take into account process dynamics and valve stick-slip, you might actually consider valve nonlinearity as good in some applications. So before we continue on to a discussion of the implications of mixing on tuning next week, let's consider the role played by the valve.

If we consult the last page of Advanced Application Note 4, we see the controller gain is proportional to the process time constant and inversely proportional to the open loop gain and dead time for maximum disturbance rejection. If we further consider that this same note shows that the open loop gain is the product of the valve gain, process gain, and measurement gain, we have the principles to be more intelligent in our valve trim choices and gain scheduling. Most people call the “open loop gain” a "process gain" even though it depends upon the valve flow characteristic and measurement scale besides the process.

The equal percentage trim has an inherent flow characteristic whose valve gain (curve slope) is proportional to flow. A lot has been published on how bad this is for controller tuning. For flow loops, this is true, but for temperature and composition (including pH) control in pipelines, static mixers, and heat exchangers (plug flow), this valve gain helps cancel out the process gain that is inversely proportional to flow. If you further consider the dead time is inversely proportional to flow, this valve nonlinearity is good.

How about well mixed vessels? Well the valve nonlinearity does the same thing so far as canceling the effect of flow on the process gain. However, since the residence time becomes a time constant rather than a dead time for a back mixed volume as discussed last week, the numerator for the controller gain is inversely proportional to flow so the benefit of canceling it out of the denominator is not such a good deal. Now the dead time from mixing is set by the turnover time and is only a weak function of feed flow. If there is a pipeline or dip tube with a significant transportation delay or poor mixing, then we are back to the case of the dead time being inversely proportional to flow. So for vessels, an equal percentage characteristic may be good or bad from a controller gain view point. There are other considerations that make this trim choice the right choice.

While a linear trim supposedly has the greatest rangeability based on best conformity of the flow coefficient to a designated curve, if you consider the effect of stick-slip which is greatest near the seat, the equal percentage trim has the best turndown (smallest controllable flow), which is more important to me than conformity of the trim characteristic. If you also consider that this trim can deal with the diminishing valve pressure drop caused by higher piping system pressure drops at higher flows and can thus prevent the installed characteristic from flattening out at large lifts, you have clues as to why an equal percentage trim is so popular.

In my book, quick opening trim is mostly ugly because the valve gain is so high near the seat it accentuates stick-slip and the process gain nonlinearity for temperature and composition control. This trim also hastens the premature flattening of the installed characteristic from a diminishing valve drop. However, for liquid pressure control, the process gain is proportional to flow so you can make a case a quick open trim is good for this loop. In fact pressure regulators tend to have this characteristic. Also, for anti-surge control and pressure relief, quick opening trim is used to establish a vent flow as quickly as possible.

Finally, I have to admit quick opening trim can help flush solids out of the seat. However, I would rather preprogram some kicker action in the DCS configuration to provide this burst so I can have a good flow characteristic to work with. I can also switch to pulse duration control for small controller outputs to prevent small flow areas that would plug.

In my December 18 blog “Linear in a Nonlinear World” we discussed the use of signal characterization to compensate for the installed characteristic of the control valve where the valve gain depends upon the operating point on the control valve characteristic. In part II we are looking at the use of signal characterization to compensate for a nonlinear process gain by translation of the original nonlinear process variable to a new linear one to enable adaptive controllers to better focus on other nonlinearities such as feed flow. Here in part II the process gain depends upon the operating point of the process variable. Examples of this translation to a new controlled variable are:

(1) Conductivity to % acid, base, or salt concentration
(2) Column top temperature to % reflux demand
(3) pH to % reagent demand

For conductivity, there is a peak in a plot of conductivity versus the acid, base, or salt concentration. The new process variable scale must be on one side or other of the peak. There is uncertainty in the exact location of the peak. If the operating point were to cross the peak, the process gain would go to zero and then change sign, which is disastrous to a control loop. The operating point must steer well clear of the peak.

For all of these examples, concentrations of other components in the feed can shift or change the shape of the curve but often the translation is better than no compensation at all for the process nonlinearity. For conductivity and pH, the effect of process temperature based on lab samples should be part of the calculation. For temperature, the effect of column pressure should be included (e.g. shift in boiling point with pressure).

The implementation involves first plotting the original versus the new process variable. For the examples noted this would be conductivity versus ion concentration for various temperatures, column temperature versus % reflux to feed ratio for various pressures, and pH versus % reagent to feed ratio for various temperatures. Since you are getting the X axis from the Y axis (the opposite of what is being done by the process), the data points for signal characterization are entered as Y,X pairs with a nonlinear bias to Y from a fit to the shift in the family of curves. Since the Fieldbus signal characterizer allows variable space of data points, closer points are used in the area of greatest curvature near the set point. This translation must be done for both the set point and the process variable. The original and new set points and process variables must be displayed and historized.

The benefits are most noticeable in pH loops because of their extreme sensitivity nonlinearity, and rangeability where changes in process gain of 100:1 and of reagent demand of 1000:1 are routine. Signal characterization has been shown to make dramatic reductions in startup time by the loop’s recognition that the acid or base reagent flow is really decades away from set point. It also prevents pH from zipping right through the neutral point (e.g. 7 pH) and banging between the flat portions of the titration curve, offering a settling time where there was none. The characterization restores the process time constant by slowing down the excursion rate and helps a continuous pH loop look more self-regulating by removing the acceleration from movement to steeper slopes on the titration curve. Thus, you see and realize the benefit from an investment in a well mixed vessel where the residence time is a process time constant that slows down concentration disturbances as discussed in blogs from the past few weeks.

There are some issues besides inaccurate curves and confusion in the operator interface. If your set point is always on a flat portion of the curve and the control system can keep the operating point close to the set point for the largest disturbance, the benefit from linearization is minimal. Additionally, if an excursion to the steep portion of the curve represents an extremely undesirable situation for equipment or environmental protection, then the elimination of the overreaction of the loop by removal of the acceleration through linearization may be the wrong thing to do even though it reduces overshoot and wasted reagent when returning the pH to its set point.

While you increase the dead time from valve dead band and resolution limits when the set point is on the steep part of the curve because you are slowing down the rate of change of the process variable and the overreaction of the controller output, this normally is much less important then the suppression of oscillations. The increase in dead band for operating points on the steep portions of the valve characteristic can be a concern for control valves because there is usually no stability issue from the much less severe nonlinearity of a valve.

These and other considerations and an application for pH control are shown in the attached file on “Linear Reagent Demand Control” which is an excerpt from my ISA pH Web Seminar at 2:00 EDT on May 16.

Linear Reagent Demand Control

I conclude with a top ten list that will appear in a future “Control Talk” column.

Top Ten Reasons Not to Use Linear Reagent Demand Control

(10) How do you know it is a pH loop if it is not oscillating
(9) You can better see if the pH sensor is still alive
(8) You can better tell if the operator is still alive
(7) You like bang-bang control
(6) Gives you chance to try out the manual mode
(5) The titration curve from the lab shows a straight line through the set point
(4) You like seeing the full effect of valve stick-slip
(3) Retuning loops is job security
(2) You can eat more doughnuts while waiting for a loop to startup and settle out
(1) Linear loops are for wimps

I have seen two control loops that did not go digital during the migration to a DCS. These electronic analog controllers stick out like a sore thumb in a modern day control room. The user would like to get rid of them along with the parts, maintenance, and operator interface issues. What keeps these relics from the 1970s hanging around?

The two analog control holdouts I have seen had 4 things in common: a variable speed drive, zero process dead time, a critical process impact, and an inability to run in manual.

If you don’t have time to read on to get details on the particular loops, the most important “insights” are:

(1) Controller tuning tools and methods that rely on an open loop test cannot be used
(2) Digital adaptive controllers that identify tuning from set point changes are needed
(3) Must faster measurement update, communication, and controller scan and execution time intervals must be developed for valid holdouts to go digital
(4) If a loop has a control valve, it is rarely a valid holdout

The first application was polymer melt pressure controlled by the manipulation of melt pump speed. The melt pressure was important for throughput and relative viscosity control. An analog trend chart recorder showed what would appear to be a lot of noise. However, if the loop was taken out of auto, the amplitude of the fluctuations got so much worse you could not afford to stay in manual for more than a few seconds. The loop was reacting and compensating for incredibly fast disturbances. The process time constant can be estimated from the fluid inertia and viscosity and typically varies between 50 and 500 milliseconds. The process dead time can be estimated as the time it takes a pressure wave traveling at the speed of sound in the fluid to propagate from the final element to the first major resistance to change the pressure difference that is the driving force for the acceleration of a basically incompressible column of fluid. In other words, the process dead time was essentially zero. The dead time in the loop was all due to the automation system. The dead time of a variable speed drive (VSD) is nearly zero if the following conditions are met in the VSD application.

(1) The change in speed is larger than the resolution limits of the VSD A/D card
(2) The change in speed is larger than any dead band introduced by the user into the VSD configuration to suppress reaction to noise
(3) The rate of change of speed is smaller than any rate limiting introduced into the VSD configuration to reduce motor load and upsets to down stream equipment
(4) The rate of change of speed is smaller than any rate limiting from rotor inertia

These conditions are met for a well designed VSD for liquid pressure control, which leaves the measurement and controller as the sources of dead time. The process is self-regulating but it takes a high speed recorder to see any sort of time constant unless a signal filter is added. Note that I am not advocating replacing control valves with a VSD. There are practical problems when a control valve is omitted, such as the reversal of flow and the creation of incredibly fast flow upsets to other loops and unit operations.

There is an important exception to zero process dead time for liquid flow control. For highly viscous flows, a “ketchup bottle effect” has been observed where there is a huge dead time to initially start a flow through a small injection orifice as described in the first chapter of my book titled A Funny Thing happened on the Way the Control Room available soon on http://www.EasyDeltaV.com/ControlInsights/

We all know about aliasing from digital communication, even more important here is introduction of a delay into a control loop that has essentially no process dead time.

Why am I obsessed with dead time? The ultimate performance (IAE) achievable for unmeasured disturbances with the fastest tuning is proportional to the dead time squared and the ultimate period for this dead time dominant loop is twice the dead time.

The second application was incinerator pressure controlled by the manipulation of an induced daft (ID) fan speed. The loop behaved like an integrator. If the controller was put in manual, the pressure could ramp and hit the interlock trip point in less than a second. Since open loop testing for exact quantification was not reasonable, a dynamic simulation was used to show it could occur in 0.25 seconds. While the residence time was on the order of 0.1 minute, the process gain was incredibly large because the measurement scale span was just a few inches of water column. The simulation also showed the decoupler between the forced draft (FD) fan and ID fan speed (air flow feedforward) was doing more harm than good because of the inverse response associated with the cold air flow. After elimination of the decoupler and retuning, the frequency of furnace trips was reduced but trips still occurred every couple of days. This process was controllable only because the process dead time associated with the furnace volume was essentially zero. For gas pressure systems the process dead time originates from gas volumes in series separated by flow resistances. The pressure sensor was seeing and the ID fan was acting on the same gas volume. The dead time in the loop was all due to the automation system.

Summarizing, a digital controller with a 0.1 second execution time was tried on startup but the furnace trips were excessive despite the best tuning and strategy. In 0.1 second, the pressure was almost half way to the trip point. When the digital controller was replaced with an analog electronic controller the pressure trips were eliminated.

This application and a phosphorous furnace application are discussed in chapter 3 titled “Pressure Control – Without Dead Time I would be out of a Job” in the aforementioned book A Funny Thing happened on the Way the Control Room.

Next week I will share my experiences with making control valves respond faster. With this under our belts, I will offer how fast digital devices and communication and data historians need to be once you get these valves to “turn on a dime or at least a quarter”.

Note that the equations for computing process dead time and time constants for these systems is in Tuning and Control Loop Performance – 3rd edition, but is out of print.

I have seen some of the best and the worst of control valves. I have also found out that not all positioners are created equal and responsible responsiveness depends upon the whole package.

Here is a checklist for making sure your valve is nimble enough to respond to every demand of your loop if you don’t have time to read my positioner horror stories.

Checklist for Responsive Responsible “Right On” Control Valves

(1) Digital positioner with high gain and rate action that is properly tuned
(2) Low friction packing that is properly tightened
(3) Low seating and sealing friction at closure
(4) Diaphragm or floating piston actuator that is properly sized
(5) Spline shaft connections for rotary valves
(6) Ball, disk, or plug stem (not actuator shaft) position feedback for rotary valves

If you are an optimist and don’t really want to know whether a control valve is really moving as the controller output changes, then make sure there is no positioner and check the valve stroke manually in the field for 0%, 50%, and 100% signals. This is a time honored tradition that started in the 1980s when the old timer instrument engineers who knew the importance of a positioner and how to adjust them had retired. I have seen a large plant responsible for the most profitable herbicide in modern times demand that projects not buy control valves with positioners because they were too troublesome to calibrate and maintain. If there is no position feedback in the control room and the smallest change made by the instrument technician is 50%, it may seem like the right conclusion. The project manager is not going to argue ($500 or more per valve savings adds up). It is amazing to me no one ever reasoned that hopefully your loop is not making 50% changes but more like 0.5% changes from execution to execution even during the most upsetting times (except for surge control).

Today we have position feedback from Hart or Fieldbus positioners and we are much more enlightened or are we? On a 1999 project, the plant insisted on using a valve that was great as an isolation valve that was also relatively inexpensive, which for a 18 inch valve was certainly appreciated by the project. Other loops with these valves oscillated continually but this was attributed to some strange behavior in the process. As a concession to the fact this on-off valve really needed to be a throttling valve, a digital positioner with all types of internal analysis and plotting capability and a customer witnessed acceptance test at the manufacturer was specified.

When I arrived for the test, days of fancy plots and data had been gathered to show the positioner feedback for changes as small as 0.5%. Wow, it looked great until I noticed the disk wasn’t moving. A travel gauge installed on the disk itself revealed that the disk would not budge until the change in signal was 8% or more. The positioner feedback was actuator shaft position. Even though the shaft had two pin connections to the disk stem position there was enough play that the high sealing (rubbing) friction for the tight shutoff valve meant the when the shaft moved the disk did its own thing, which was usually nothing. Also, there was a tendency for the stem to eventually jump form shaft windup. The application here was air so there were plenty of choices for real control valves.

In another application, a similar thing occurred but it was even worse in that the ball seal was so tight that the ball stem connection didn’t even follow the actual ball position. The digital positioners in the control room said everything was OK but the loop PV said otherwise. Unfortunately, since this was liquid phosphorous, it was difficult to find a throttling control valve that would not seize up. However, one plant did get a special v-notch ball valve to work after I configured a periodic pulse to flush it out. In other applications I have avoided solids accumulation, fouling, and plugging from low flow throttling by a transition to pulse duration control for throttle positions less than 25%. In other words, there often ways to make a control valve do its job without having to go to valves that piping has specified as isolation or block valves. Next week we will offer the top signs your control valve is really an on-off valve in disguise.

In my younger days I was presented with the critical need to make air compressor surge and electrical phosphorous furnace pressure control loops able to handle very abrupt and extreme disturbances. I vastly prefer the present to that present. These applications offered more excitement than engineers should be allowed to have.

The compressors provided the air feed to multiple exothermic reactors whose flow could drop enough on a reactor trip to trigger a surge in about 2 seconds. A surge every month would cause the other reactors to trip and cause accumulating damage and loss of efficiency in the compressor besides reactor downtime and a subsequent challenging startup of reactors and the associated waste oxidation boilers.

The phosphorus furnaces had to deal with what was called “controlled explosions” from sudden shifts in the ore around the electrodes (slag slides) that caused bursts of water vapor and CO2 besides phosphorus vapors and particles. The slag slides caused a pressure spike large enough each shift to blow the water out of the electrode seals. There were tubs of water around the furnaces to jump in if the hot phosphorous landed on you. Little fires would break out when you walked by the furnaces from your shoes scraping the phosphorous residue on the floor.

These were big problems in terms of both size (18 to 24 inch pipelines) and the safety implications besides the process efficiency and capacity considerations.

High speed recorder measurements of the of the compressor flow and furnace pressure response confirmed that the process dead time in both cases was essentially zero and the observed dead time was due entirely to the components in the automation system. I installed transmitters with a sensor response time constant of less than 0.1 seconds and controllers with a special scan rate of 0.05 seconds. I had to take some special precautions in the configuration to insure the controller loading would never have negative free time (a lesson as well for our personal lives).

The control valves were the largest source of dead time. The pre-stroke dead time and stroking time for the big actuators were estimated as the fill or exhaust factor for the actuator supplied by the valve manufacturer divided by effective fill or exhaust flow coefficient of the existing positioner. This yielded pre-stroke dead times ranging from 1.0 to 2.0 seconds and stroking times of about 10 to 20 seconds. A booster had a fill and exhaust flow coefficient that was 10 times larger than the positioner and therefore offered dead times and stroking times that were 10 times faster. However, I knew the actuator connection and air tubing would then become the restricting limitation, so I had these sizes judiciously enlarged in the field, I also added a position transmitter (before the days of Hart and Fieldbus positioners with position read back).

Armed with the rule “boosters instead of positioners should be used on fast loops” and a copy of the theoretical frequency response studies to back it up I arrived onsite for the compressor application and boldly insisted against the advice of the well seasoned instrument maintenance technician to replace the positioner with a booster on the compressor vent valves.

My confidence was shattered the morning the first surge valve was put in service. The flow transmitter showed the impending surge and the controller asked the valve to open. The valve responded by doing the worst possible thing. It slammed shut before the forward flow to any of the reactors had been established.

The technician who wanted the positioner took me to the surge valve and showed that he could move it to any desired position by tugging on the actuator shaft. Obviously, the buffeting action of the turbulent flow could cause the disc to wander and eventually close. The actuator size was checked and found to be adequate; the spring rate was increased but the results remained the same. Subsequent tests showed that the stem resisted movement considerably better if the actuator was fed directly from an I/P transducer and that it could not be budged at all if a positioner was installed.

We still needed speed, so I installed the booster on the outlet of the positioner. Unfortunately the po