This piece could have been titled "Exceptional Failures in Process Control." Despite my 25 years of explaining the importance of using middle signal selection, I don't see much evidence of what I have said has taken root outside of Monsanto and its spin-off Solutia, where it became a part of the culture and best practices to use middle signal of three pH electrodes for all important pH control loops. We ended up taking a view that all pH loops are important because if they are unimportant why should we go through the maintenance headaches and the risk of control system failure by installing a loop dependent upon the integrity of a single electrode. I think the main hurtle besides hardware and installation cost is the feeling that if one electrode requires so much effort, why should I add more? If the electrode life expectancy is too short, the feeling is right. We should not add more of a bad application or installation. Instead, we need to find a better design, technology, implementation, and location verified in testing via wireless pH or an alternate measurement (e.g. conductivity for concentrated acids or bases).

As an important side note, the use of three transmitters and middle signal selection on all of the important measurements (e.g. flow pressure, temperature, and level besides pH) used in the control system and safety system for a large intermediates plant has consistently eliminated false trips saving several million dollars per year.

What boggles my mind is that the risk of poor product quality and an environmental violation do not provide a wakeup call that the lifecycle cost of the measurement itself is insignificant in comparison to the risk. A simple quick ball park benefit versus cost analysis would show the absurdity of when the dollars of events likely to occur each year is more than 1000 times the dollars of the additional automation to prevent them. Unfortunately, we tend to get too focused on short term costs. Consider bioreactor batches worth millions of dollars each of a sold out pharmaceutical that are dependent pH upon control to within 0.02 pH. A second electrode is added but I am not sure it helps or just adds to the confusion. I find it almost bizarre how favorite electrodes are picked for the loop's PV and the stories that ensue about which one is best. It is a "fact of life" that electrodes will not agree in the short term due to non-ideal effects too numerous to get into here (check out my ISA book Advanced pH Measurement and Control for more info and previous blogs by searching for "pH"). The continual disagreement between two electrodes often leads to calibration adjustments chasing calibration adjustments. If left alone, the electrode that reads high today may in a couple of hours or at least by tomorrow read low. Electrodes can fail anywhere on the scale (including the most insidious failure of all type where the bad electrode signal is stuck at the pH set point). I maintain that the correct use of middle signal selection will actually reduce the long term maintenance cost by simple observation and the use of more intelligent practices eliminating unnecessary calibration and removal of electrodes.

A middle signal selection inherently ignores a single failure of any type and avoids the slowest electrode (e.g. coated electrode). This selection reduces noise and eliminates spikes without any addition of a signal lag like what you get from signal filtering. Middle signal selection also ignores an electrode with lower efficiency (shorter span) or that is drifting. Theoretically, electrodes of different "in service" time should be used to reduce the occurrence of concurrent failures. The middle electrode is the best signal on the average, but please don't use the average. I have seen some very smart attempts of computing average signals with built in intelligence on signal rejection that were out foxed by a single electrode failure scenario. You would think you could devise something smarter than the simple middle signal selection when in fact inherently it is impossible for a single failure. There can be additional intelligence for more than three electrodes or for protection against multiple concurrent failures.

To summarize, middle signal selection can improve process quality and on-stream time, reduce maintenance, and prevent environment violations by adding understanding and ignoring spurious signals, inaccurate measurements, and failures.

A prolonged deviation from the middle should be alarmed because if you don't fix the first failure or sustained error, middle signal section has a fifty-fifty chance of preventing the next failure or electrode inaccuracy. I could go on and on but I suspect you are pressed to move on. Before I go let's be frank with closing remarks in recognition of an engineer named Frank who was particularly astute at telling it the way it is.

There is an opportunity to use a statistical or first principle model based on titration curves to generate a third signal. Even if the model is wrong, it will be ignored by middle signal selection. There appears to be here mostly an upside where you will at least learn more about your process by developing a model.

I have no illusions as to whether this blog will change one person's mind enough to install middle signal selection even though it is a feature of a standard function block. I also have no expectations that enough users will see the need to take advantage of wireless measurements to eliminate the wiring installation and maintenance costs of going to three measurements. Even more unlikely is that users will end up using middle signal selection enough that it will be offered in a smart transmitter that inputs three electrodes even though electrodes are the weak link in regards to accuracy and reliability.

I will be doing the presentation McMillanISABostonExceptionalOpportunities.pdf next week at the Boston ISA section meeting. I will be giving out 10 free copies of my book The Funnier Side of Retirement for Engineers and People of the Technical Persuasion to balance out the serious stuff.

When?
Tuesday, October 20, 2009
6:00 - 7:00 Reception and registration
7:00 - 8:00 Dinner
8:00 - 9:00 Presentation

Where?
Best Western, Waltham, MA
380 Winter Street, Waltham, Massachusetts, 02451-8700, US
Phone: 781/890-7800 Fax: 781/890-4937

The opportunity afforded by online metrics is worth summarizing in this series even though it has been discussed in several entries on this website.

The need to cut costs has translated to an increased emphasis on process efficiency and the ability to justify software, hardware, and personnel. Increasingly these need to be hard benefits (e.g. reduction in raw material, downtime, and energy costs).

When I worked in process control improvement (PCI) in the technology department of a large chemical company, we had to show new benefits each year that were at least twice our salary to justify our job. By the end of the five year process control improvement effort we had 75 million dollars per year in savings documented. The PCI core group had 5 modeling and control specialists working with 20 or more process control engineers at key plants. The benefits reported depended upon the skills of particularly one person Glenn Mertz) who was extremely proficient in cost sheet analysis and working with operations and process technology.

Some companies are fortunate enough to have PCI as part of their culture as seen in the Control Talk Columns "Going, Going, Gone - Part 2" (September) and Part 3 (October) for examples. For many companies, benefits need to be reported in order for PCI and our profession to move forward or even exist. See the December 1 and 5, 2008 entries on this website "Past, Present, and Future of Automation - Part 5 (Benchmarking and Opportunity Assessment)" and Part 6 (Operator Interface) and the December 28, 2007 entry "Biggest Opportunities in Process Control Improvement - The Operator (Online Metrics) for more discussion of the aspects and importance of identifying and showing PCI benefits.

There are a lot of initiatives in the plant to improve plant operation by better operating procedures, equipment, and maintenance. All of these people take great pride in their work and are naturally eager to attribute better process operation to their efforts. Process technology often has the last say. The best way for PCI to get credit for improvement in plant operation is for the improvement and change to be visible in the data historian. A visible change in capacity, efficiency, or quality after a change in the process control system provides the documentation needed. If the PCI could be turned on and off, the correlation would be irrefutable but this is usually not practical. If no other events occurred when the PCI went online, a beginning of improved plant operation coinciding with the completion of the PCI, and a good explanation of cause and effect, will normally suffice for PCI to get credit. To help guide management and operations, comments should be entered in the historian and event makers for PCI provided.

PCI metrics for continuous process capacity are generally available from product flow measurements, downtime due to trips, and the time to startup or make a product grade transitions. PCI metrics for batch process capability can be generated from batch size, end point concentration, batch cycle time, and time in between batches. Quick and dramatic improvements in batch capacity have been achieved be the elimination of operator attention requests, manual actions, trips, and wait times for resource allocation (e.g. utility or charge systems), lab results, and reaction completion. Model predictive control and override control applications have been very successful for fed-batch processes. Reductions of 25% or more in batch cycle time are common for PCI. For a summary of some of the many possible batch control opportunities see BatchCycleTimeReduction.pdf from my PCI days.

PCI metrics for process efficiency are best expressed as a ratio of kilogram (pound) of input used (e.g. feed, fuel, reagent, and utility) per kilogram (pound) of product produced. For fuels, the numerator in the ratio may be expressed in thermal units, such as kilojoules (BTUs). For batch processes, the totalized input flow is divided by the batch size multiplied by the fractional product end point concentration. For continuous processes the instantaneous input flows are divided by intermediate or final product flow multiplied by the fractional product concentration. Synchronization of input flows to output flows can be done by the addition of a time constant equal to residence time and a time delay equal to the transportation delay. The flows can be totalized to compare shifts and periods of operation. Online process efficiency measurements require online or at-line analyzers or inferential measurements from first principle, neural network, polynomial, or statistical (e.g. PLS) models. These models in turn require flow measurements because nonlinear valve characteristics, backlash, and stick-slip make the use of controller outputs directly as model inputs ineffective and misleading. While reactant and fuel flows are typically measured, utility and reagent flows are often not. This short sightedness by plant projects (figuratively and literally), severely limits the ability to make improvements in the efficiency of use of these process inputs. I would wager a 10% reduction in the use of these inputs would more than pay for the flow meters. The old saying, you cannot control what you don't measure holds true for process efficiency. If I was a project manager, I would have a flowmeter on any input flow whose usage cost per year exceeded twice the installed cost of the flowmeter. I would at least provide the process connections for inserting a mobile wireless flowmeter. Where energy heat transfer rate calculations (e.g. heat removal rate as an inference of reaction rate) would be useful, I would install wireless RTD temperature transmitters on the streams entering and exiting the coils, exchangers, and jackets. Wireless transmitters allow the user to find during actual process operation the applications with the maximum benefit.

The first document you have on a project is typically a process flow diagram (PFD). The PFD defines the process. It is the ultimate source of information and sets the plant performance and design. It is interesting to me that we really don't know how well existing operations match the PFD. When I have posed the question of what is really the mass flow, pressure, temperature, and density of a plant stream to university students and even seasoned engineers in industry, the usual reply is that the stream conditions are what is shown in the PFD. As humans we are naturally optimistic and want to think everything is as believed and stated. As engineers we are accustomed to numbers being accurate to several significant figures. Alas, if you had the knowledge of what is really going in the process there would be a rude awakening. While uncomfortable, the awareness leads to better process control improvements.

In the PFD and in chemical engineering courses, the plant is assumed to be at steady state. Of course this does not work for batch processes. Less obvious is that it doesn't work well for continuous processes with merging and diverging trains of equipment and recycle streams. Even if a plant was at steady state, I doubt it would be within 10% of the PFD design conditions on all of the PFD process variables due to non ideal and unknown effects in the process calculations or simulations that generated the PFD. Maybe things have changed a lot, but in my days working at a large chemical company, the process engineers manually updated personalized spreadsheets that attempted to close the material and energy balances (unless we are talking about nuclear reactions, energy and mass are conserved - neither created or destroyed).

What if a plant had a live online PFD? What if we had live online material and energy balances? What if we had temperature, pressure, mass flow, and inferential measurements of the composition in every important process stream?

Coriolis flowmeters offer a true mass flow measurement that does not depend upon composition, density, velocity profile, Reynolds number, or viscosity. The physics of the measurement afford a rangeability and accuracy that is unexcelled (for an excellent perspective see the article by Peter Ginn "It's the Physics!", InTech, Feb 1996). Coriolis also provides a direct density measurement, a tube temperature measurement, and when coupled with an accurate differential pressure transmitter (DP) for viscous fluids, an inferential viscosity measurement. In the last couple of years, major improvements have been made in Coriolis technology. For example, Coriolis meters can measure two phase flow and can infer void fraction. Meter sizes can be as small as 2 millimeters making them ideal for labs and pilot plants. For slurries and clingy sticky fluids, straight tubes and higher velocities can be used to prevent coatings and accumulation of material. Coriolis meters can potentially provide more accurate batch charges than weigh tanks because Coriolis meters retain a better long term installed accuracy than load cells since Coriolis does not suffer from drift or installation effects. For more information on Coriolis see the EssentialBookCoriolisExcerpt.pdf.pdf from the new ISA book Essentials of Modern Measurements and Final Elements

When a Coriolis meter is put on a stream, the only process variable missing for a live online PFD is pressure, which could be easily added via a wireless pressure transmitter. For streams with acids and bases, wireless conductivity and pH transmitters could provide additional information on stream composition. For example, in absorbers for CO2 capture, wireless pH and conductivity measurements in concert with a Coriolis density and temperature measurement can provide inferential measurements of solvent concentration and CO2 loading important for optimizing absorber flow distribution.

There is a lot of talk about online process metrics but as far as I can see, what is done is loop metrics principally on process variability. A live PFD would enable online process efficiency metrics (e.g. yield) for each unit operation besides tighter mass balances. The stream variables would also lead to better data analytics and prediction of product quality.

I spent the first 7 years of my career in instrument design and construction. After being responsible for the calibration, installation, and commissioning of instruments for a half dozen plants in the 1970s, I became painfully aware that the actual performance of the measurements and control valves was largely unknown. These were the days before the advent of smart instrumentation. We didn't know the effect of stiction and backlash on valve position or the effect of impulse line, process and ambient conditions on sensors. We didn't know what was the installed accuracy of measurement or if a valve or measurement had a timely and sensitive response. We shifted set points and just shook our heads when the material and energy balances did not close. Since we were mostly interested in capacity we just pushed on to make more product. Operating efficiency and turndown were not as much an issue, which was fortunate because we didn't have the spectrum and accuracy of instruments for knowing process performance. The time I spent in the 1980s working on pH, furnace pressure, and compressor surge loops were the ultimate test of sensor and valve sensitivity and speed. My perspective on the importance of the field devices was solidified in the 1990s, when I was part of a corporate wide process control improvement program, most of the opportunities involved tuning loops and adding feedforward control and loops for fed-batch operation. A lot of great ideas went by the "way side" because of missing or imprecise measurements and unresponsive valves.

An important point is that if you don't have the capability of determining actual capability and benefits of the automation system, projects will seek the lowest cost alternatives. A classic example of capital cost superseding performance was the proliferation of rotary piping valves that were posed as throttling valves by the addition of spool type positioners to modulate a piston actuator, linkages, and stem connections fundamentally designed for on-off service. The leakage specs and price were attractive. Deadband and resolution limit were not considered. Since the valve specification didn't require the valve actually move in response to the small changes in signal commonly incurred in a control loop and there was no position feedback measurement either locally or remotely, the user did not know the real price paid. Aggravated by noisy flow measurements with poor turndown, increased process variability was attributed to mysterious sources. Without online loop metrics, there was little recognition of the deterioration in loop performance. Since the normal practice of testing whether a valve worked was to make 25% or larger changes in signal, instrument technicians and engineers where unaware that the valve did not respond to the small changes in controller output each scan. Putting a smart positioner on a piping valve with the feedback measurement of actuator shaft rather than ball or disk stem position in a rotary piping valve only added to the confusion. The actuator shaft would move in response to the positioner but the ball or disk did not due to extensive seal friction, ball or disk shaft windup, and backlash in the connection and linkages. It was only after actual tests in the flow labs of control valve manufactures was the true cost of these valve recognized. The publication of the lab test results and the subsequent ISA standards developed on valve step testing, the availability of position feedback as a secondary process variable on digital signals, and the analysis of resolution (e.g. stick-slip) and deadband (e.g. backlash) lead to an increased awareness and hence dramatic improvement in valve dynamics.

Today we have smart transmitters and control valves with a rangeability, resolution, and sensitivity that is an order of magnitude better than the typical fare of the last century. A combination of embedded intelligence and new sensor, transmitter, valve, and positioner technology have resulted in dramatic improvements. Combined with the ability to have additional process variables, diagnostics, and alerts reported to the control room by digital signals and the mobility afforded by wireless communication, we can increase the spectrum and flexibility of the field automation system including finding the optimum locations for process analysis and control. Doors will open for online data analytics, process performance metrics (e.g. energy, quality, and yield) and increased opportunities for basic and advanced control improvements to address the increasing needs of process efficiency, flexibility, and rangeability. My recent Control Talk column "Downturn Turndown" digs into the increased importance of sensor and valve performance with of course a top ten list to cap it off.

Recently it was realized that research and development could greatly benefit from the advanced performance, intelligence, and historization of smart industrial automations systems. The future is best exemplified by the lab optimized industrial distributed control system with industrial pH, dissolved oxygen, pressure, temperature, and mass flow measurements for bench top and pilot plant bioreactors that was pioneered by Broadley-James Corporation. The portability and reduced installation cost of wireless instrumentation increase the already significant advantages of moving advanced industrial automation system capability upstream in the commercialization process.

The foundation of a process automation system is the measurements and final elements. If you don't get these right not much else matters. Measurements provide the only window into the process and final elements provide the only means of affecting the process. The height of the pyramid consisting of increasingly more advanced layers of process analysis and control depends upon the integrity and breadth of the foundation. The goal of the book I just finished is to create a foundation where the sky is the limit for automation. The book royalties go to the Center for Energy and Environmental Resources at the University of Texas where tests are being conducted on the use of wireless conductivity, flow, pH, pressure, and temperature measurements for carbon dioxide capture research.

The new book titled Essentials of Modern Measurements and Final Elements makes no assumptions other than the reader has some technical background. In Chapter 1 Modern Measurement Fundamentals, special care has been taken to explain technical terms and concepts on the use and performance of measurements in the process industry. There is a special emphasis on the advances in wireless instrumentation and communication. Chapters 2 through 6 focuses on the details needed for the best implementation of specific types of measurements that would be used on automation upgrade and new plant projects today in the process industry. Chapter 7 on Final Element Fundamentals follows an approach similar to Chapter 1 in assuming no industrial experience so the material on control valves, dampers, guide vanes, and variable speed drives is beneficial to students and new employees. Chapter 8 gets into the details on the types of control valves that are used in 95% of the applications in the chemical and petrochemical industry. The book concludes with the latest details on WirelessHART automation systems in Chapter 9. The questions at the end of each chapter are designed to stimulate the thought process involved for a successful application.

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 is set to be less than 1/10th of the sum of the minimum loop deadtime and process time constant with some consideration as to maximum practical controller gains to reduce valve cycling and noise amplification. For any loop with a control valve, the minimum loop deadtime is about 1 second for an unmeasured disturbance so the ultimate limit on sample time is about 0.1 second. The practical limit reflects current tuning practices. 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.

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.

Unlike self-regulating processes that will line at a steady state after disturbances have died out, integrating processes will ramp until a physical limit is hit. The ramping response is caused by the lack of negative feedback (e.g. self-regulation) in the process as defined in Advanced Application Note 4. In other words an increase in the process variable does not increase a counteracting effect to make the response bend over and reach a equilibrium.

The most common integrating process is level. Since the discharge flow is not appreciably affected by level (except for the rare case of gravity flow), any difference between the feed and discharge flows causes the level to ramp. The low limit is the vessel running dry and the high limit is the vessel spilling over or flooding a vent system.

Other common examples are

(1) Gas pressure control of columns, furnaces, and vessels when changes in operating pressure does not appreciably affect the vent flow rate

(2) Batch temperature control when changes in vessel temperature does not appreciably change the heat transfer rate

(3) Batch pH control when there is no reagent reaction or consumption or reagent concentration does not appreciably change reagent reaction or consumption rate

(4) Batch dissolved oxygen control when the change in oxygen absorbed does not appreciably change the oxygen transfer rate

(5) Batch product composition control when a change in product concentration does not appreciably affect side reaction or degradation rate

(6) Vessel solids concentration control when changes in solids concentration does not affect the evaporation or precipitation rate

(7) Bioreactor biomass or cell density control before the stationary and death phases

Many processes due to a long process time constant or large process gain, will appear to ramp because the steady state is beyond the time range or control region, respectively. What the user sees on the trend charts and what the controller sees as a response from the process variable is a ramp. These processes called "near-integrating" or "pseudo-integrating" processes are better analyzed and tuned as if they were integrating rather than self-regulating processes. Temperature control of any continuous process with a large residence time (volume/flow) can be treated as a "near-integrating" process.

Most of the more important loops have an integrating or "near-integrating" response. Furthermore the ramp rate (%/sec) for a % change in controller output (integrating process gain) is often incredibly slow. These slow ramp rates require exceptionally high controller gains and large integral times.

The test results for a single use bioreactor (SUB) with what would appear to be a small volume (100 liters), revealed an integrating gain of 0.000008 %/sec/%, that was 30 time slower than a bench top bioreactor. The SUB volume was about 30 times larger than the bench top bioreactor volume. The relative size of the volumes is a strong factor but the relative size of other parts such as heat transfer area play a role. This was the first time temperature control was tried on a SUB in this lab. Fortunately an adaptive controller was in service that identified the unexpectedly slower integrating process gain. The best response was achieved with a controller gain of 80 and an integral time of about 10,000 seconds. A Lambda factor of 0.05 was needed. The test results are shown in "BioreactorTemperatureTuningTestResults.pdf."

The principle opportunity for integrating processes is realizing and using higher controller gains and larger integral times. We tend to use too much integral action (too small of an integral or reset time) because we are impatient and integral action provides a continual driving action to eliminate error. We don't normally think of using higher gains because the problem of instability from high gains is drilled into us in all our courses and books on process control, our older measurement systems often gave flaky signals, and before we had structure and set point filter options, high controller gains caused the controller output to peg on a set point change. Properly installed smart transmitters with integral sensors and primary elements have a noise level that is low enough and a sensor sensitivity and repeatability high enough so that the amplification of small changes provides corrective actions rather than amplification of noise or extraneous actions. The proper use of the many PID parameter, control options, and structure today allows the user to minimize the disruption to the operator and other loops.

Most people don't realize there is a window of allowable controller gains. As I mentioned we all know too high of a gain causes instability. For many integrating processes, this controller gain is way above our comfort level (e.g. gain > 100). More often we run into the low limit for controller gain (e.g. gain < 10). Too low of a controller gain causes overshoot and slow rolling oscillations. The correction is non intuitive. You need to increase the controller gain. Even with a high gain and integral time and rate action, it is difficult to prevent overshoot with an integrating process unless you take a very slow approach by using a PID structure that provides no step change in the controller output on a set point (e.g. proportional and derivative action on PV and integral action on error). The overshoot and speed of approach problem was the primary motivation for the simple control strategy for making a temperature go as fast as possible and then stop right at set point as discussed in the article "Full Throttle Batch and Startup Response"

The Lambda tuning equations for integrating processes automatically makes the controller gain large enough to stay above the low limit in the window of allowable controller gains. This is accomplished by keeping the product of controller gain and integral time to larger than 4 divided by the integrating process gain as seen the last slide of "LambdaTuningEquations.pdf." However to get an acceptably fast enough response, Lambda factors much lower than the user is accustomed to must be used. Not shown is the fact that derivative action is helpful. The rate time should be set to the next largest time constant for a self-regulating process and the largest time constant in an integrating process. These rules are consistent for a "near-integrating" since the integrating process gain is the process gain divided by the largest process time constant leaving the next largest time constant as the one used to set the rate time.

Temperature control of exothermic reactors where the reaction rate increases with temperature and particle or crystal size control where the formation rate increases with particle or crystal size can have an integrating followed by a runaway (positive feedback) response where is it is critical to maximize the controller gain and integral time.