*Posts on this page are from the Control Talk blog, which is one of the ControlGlobal.com blogs for process automation and instrumentation professionals and Greg McMillan’s contributions to the ISA Interchange blog.
The post What are New Technologies and Approaches for Batch and Continuous Process Control? first appeared on the ISA Interchange blog site.
The following technical discussion is part of an occasional series showcasing the ISA Mentor Program, authored by Greg McMillan, industry consultant, author of numerous process control books, 2010 ISA Life Achievement Award recipient and retired Senior Fellow from Solutia Inc (now Eastman Chemical). Greg will be posting questions and responses from the ISA Mentor Program, with contributions from program participants.
What is the technical basis and ability of technologies other than PID and model predictive control (MPC)? These technologies seem fascinating and I would like to know more, particularly as I study for the ISA Certified Automation Professional (CAP) exam.
Michel Ruel has achieved considerable success in the use of fuzzy logic control (FLC) in mineral processing as documented in “Ruel’s Rules for Use of PID, MPC and FLC.” The process interrelationships and dynamics in the processing of ores are not defined due to the predominance of missing measurements and unknown effects. Mineral processing PID loops are often in manual, not only for the usual reasons of valve and measurement problems, but also because process dynamics between a controlled and manipulated variable radically change, including even the sign of the process action (reverse or direct) based on complex multivariable effects that can’t be quantified.
If the FLC configuration and interface is set up properly for visibility, understandability and adjustability of the rules, the plant can change the rules as needed, enabling sustainable benefits. In the application cited by Michel Ruel, every week metallurgists validate rules, make slight adjustments, and work with control engineers to make slight adjustments. A production record was achieved in the first week. The average use of energy per ton had decreased by 8 percent, and the tonnage per day had increased by 14 percent.
There have been successful applications of PID and MPC in the mining industry as detailed in the Control Talk columns “Process control challenges and solutions in mineral processing” and “Smart measurement and control in mineral processing.”
I have successfully used FLC on a waste treatment pH system to prevent RCRA violations at a Pensacola, Fla. plant because of my initial excitement about the technology. It did very well for decades but the plant was afraid to touch it. The 2007 Control magazine article “Virtual Control of Real pH” with Mark Sowell showed how you could replace the FLC with an MPC and PID strategy that could be better maintained, tuned and optimized.
We used FLC integrated into the software for a major supplier of expert systems in the 1980s and 1990s but there were no real success stories for FLC. There was one successful application of an expert system for a smart level alarm but it did not use FLC. However, a simple material balance could have done as well. There were several applications for smart alarms that were turned off. After nearly 100 man-years, we have not much at all to show for these expert systems. You could add a lot rules for FLC and logic based on the expertise of the developer of the application, but how these rules played together and how you could tell which rule needed to be changed was a major problem. When the developer left the production unit, operators and process engineers were not able to make changes inevitably needed.
The standalone field FLC advertised for better temperature setpoint response cannot do better than a well-tuned PID if you use all of the PID options summarized in the Control magazine article “The greatest source of process control knowledge,” including PID structure such as 2 Degrees of Freedom (2DOF) or a setpoint lead-lag. You can also use gain scheduling in the PID if necessary. The problem with FLC is how you tune it and update it for changing process conditions. I wrote the original section on FLC in A Guide to the Automation Body of Knowledge but the next edition is going to have it omitted due to common agreement between me and ISA that making more room to help get the most out of your PID was more generally useful.
FLC has been used in pulp and paper. I remember instances of FLC for kiln control but since then we have developed much better PID and MPC strategies that eliminate interaction and tuning problems.
So far as artificial neural networks (ANN), I have seen some successful applications in batch end point detection and prediction and for inferential dryer moisture control. The insertion of time delays on inputs to make them coincide with measured output is required for continuous operations. For plug flow operations like dryers, this can be readily done since the deadtime is simply the volume divided by the flow rate. For continuous vessels and columns, the insertion of very large lag times and possibly a small lead time are needed besides dead time. No dynamic compensation is needed for batch operation end point prediction.
You have to be very careful not to be outside of the test data range because of bizarre nonlinear predictions. You can also get local reversals of process gain sign causing buzzing if the predicted variable is used for closed loop control. Finally, you need to eliminate correlations between inputs. I prefer multivariate statistical process control (MSPC) that eliminates cross correlation of inputs by virtue of principle component analysis and does not exhibit process gain sign reversals or bizarre nonlinearity upon extrapolation outside of test data range. Also, MSPC can provide a piecewise linear fit to nonlinear batch profiles, which is a technique we commonly do with signal characterizers for any nonlinearity. I think there is an opportunity for MSPC to provide more intelligent and linear variables for an MPC like we do with signal characterizers.
For any type of analysis or prediction, whether using ANN or MSPC, you need to have inputs that show the variability in the process. If a process variable is tightly controlled, the PID or MPC has transferred the variability to the manipulated variable. Ideally, flow measurements should be used, but if only position or a speed is available and the installed flow characteristic is nonlinear, signal characterization should be used to convert position or speed to a flow.
I implemented a neural network some years ago on a distillation column level control. The column was notoriously difficult to control. The level would swing all over and anything would set it off, such as weather or feed changes. The operators had to run it in manual because automatic was a hopeless waste of time.
At the time (and this information might be dated) the neural network was created by bringing a stack of parameters into the calculation and “training” it on the data. Theoretically the calculation would strengthen the parameters that mattered, weaken the parameters that didn’t, and eventually configure itself to learn the system.
The process taught me much. Here are my main learning points:
1) Choose the training data wisely. If you give it straight line data then it learns straight lines. You need to teach it using upset data so it learns what to do when things go wrong. (Then use new upset data to test it.)
2) Choose the input parameters wisely. I started by giving it everything. Over time I came to realize that the data it needed wasn’t the obvious. In this case it needed:
3) Ultimately the system worked very well – but honestly by the time I had gone through four iterations of training and building the system I KNEW the physics behind it. The calculation for controlling the level was fairly simple when all was said and done. I probably could have just fed it into a feedforward PID and accomplished the same thing.
The experience was interesting and fun, and I actually got an award from ISA for the work. However when all was said and done, I realized it wasn’t nearly as impressive a tool as all the marketing brochures suggested. (At the time it was all the rage – companies were selling neutral network controller packages and magazine articles were predicting it would replace PID in a matter of years.)
Thank you, this is a lot more practical insight than I have been able to glean from the books.
I imagine the batch data analytics program offered by a major supplier of control systems is an example of the MSPC you mentioned. I think I have some papers on it stashed somewhere, since we have considered using it for some of our batch systems. What is batch data analytics and what can it do?
Yes, batch data analytics uses MSPC technology with some additional features, such as dynamic time warping. The supplier of the control system software worked with Lubrizol’s technology manager Robert Wojewodka to develop and improve the product for batch processes as highlighted in the InTech magazine article “Data Analytics in Batch Operations.” Data analytics eliminates relationships between process inputs (cross correlations) and reduces the number of process inputs by the use of principal components constructed that are orthogonal and thus independent of each other in a plot of a process output versus principle components. For two principal components, this is readily seen as an X, Y and Z plot with each axis at a 90-degree angle to the each other axis. The X and Y axis covers the range of values principal components and the Z axis is the process output. The user can drill down into each principal component to see the contribution of each process input. The use of graphics to show this can greatly increase operator understanding. Data analytics excels at identifying unsuspected relationships. For process conditions outside of the data range used in developing the empirical models, linear extrapolation helps prevent bizarre extraneous predictions. Also, the use of a piecewise linear fit means there are no humps or bumps that cause a local reversal of process gain and buzzing.
Batch data analytics (MSPC) does not need to identify the process dynamics because all of the process inputs are focused on a process output at a particular part of the batch cycle (e.g., endpoint). This is incredibly liberating. The piecewise linear fit to the batch profile enables batch data analytics to deal with the nonlinearity of the batch response. The results can be used to make mid-batch corrections.
There is an opportunity for ANN to be used with MSPC to deal with some of the nonlinearities of inputs but the proponents of MSPC and ANN often think their technologies is the total solution and don’t work together. Some even think their favorite technology can replace all types of controllers.
Getting laboratory information on a consistent basis is a challenge. I think for training the model, you could enter the batch results manually. When choosing batches, you want to include a variety of batches but all with normal operation (no outliers from failures of devices or equipment or improper operations). The applications as noted in the Wojewodka article emphasize what you want to have as a model is the average batch and not the best batch (not the “golden batch”). I think this is right to start detecting abnormal batches but process control seeks to find the best and reduce the variability from the best so eventually you want a model that is representative of the best batches.
I like MSPC “worm plots” because they tell me from tail to the head the past and future of batches with tightness of coil adding insight. The worm plot is a series of batch end points expressed as a key process variable (PV1n) that is plotted as scores of principal component 1 (PC1) and principal component 2 (PC2)
If you want to do some automated correction of the prediction by taking a fraction of the difference between the predicted result and lab result, you would need to get the lab result into your DCS probably via OPC or some lab entry system interfaced to your DCS. Again the timing of the correction is not important for batch operations. Whenever the bias correction comes in, the prediction is improved for the next batch. The bias correction is similar to what is done in MPC and the trend of the bias is useful as a history of how the accuracy is changing and whether there is possibly noise in the lab result or model prediction.
The really big name in MSPC is John F. MacGregor at McMaster University in Ontario, Canada. McMaster University has expanded beyond MSPC to offer a process control degree. Another big name there is Tom Marlin, who I think came originally from the Monsanto Solutia Pensacola Nylon Intermediates plant. Tom gives his view in the InTech magazine article “Educating the engineer,” Part 2 of a two-part series. Part 1 of the series, “Student to engineer,” focused on engineering curriculum in universities.
For more on my view of why some technologies have been much more successful than others, see my Control Talk blog “Keys to Successful Control Technologies.”
The post, Common Mistakes not Commonly Understood - Finale, first appeared on the Controlglobal.com Control Talk blog.
Here is our finale just in time to serve as a momentous process control gift for the Holidays. Just don’t try to re-gift this to anyone unless they are into the automation profession or you big time.
(21) Misuse and Missing Use of Setpoint Filter. The use of a setpoint filter on a secondary loop setpoint will slow down the ability of the primary loop to make a correction for changes in load to or the setpoint of the primary controller in cascade control. For this reason, there has been a general rule of thumb that a setpoint filter should not be used on secondary controllers. However, as with most rules of thumb there are important exceptions derived from a deeper understanding. The setpoint filter on the secondary loop does not interfere with the ability of the secondary loop to reject disturbances and to deal with nonlinearities within the secondary loop that is often the most frequent and important role. If the setpoint filter is judicially applied so that it is less than 10% of the primary loop dead time, the effect on the ability of the primary loop to reject disturbances originating in the primary loop is negligible. The use of a judicious setpoint filter can ensure there are no temporary unbalances from changes of multiple flows under ratio control by enabling all the flows to move in concert. This is critical for reactant flows and the inline mixing of any flows. Often this unbalance was prevented by tuning the secondary flow loops to have the same closed loop time constant. Unfortunately, this forces the tuning of loops to be as slow as the slowest or most nonlinear flow loop. This detuning reduces the ability of the other loops to deal with their pressure disturbances and nonlinearities of their installed flow characteristics. Also, the use of setpoint rate limits that are different up and down gives directional move suppression to provide a fast approach to a better operating condition and a fast getaway to prevent undesirable operating point in the process variable (PV) or manipulated valve position. This is important to provide a fast opening and slow closing surge valve for compressor control, to optimize user valve position to improve process efficiency or maximize production by a valve position control, and to prevent oscillations across a split range point. For primary loops, a setpoint filter time equal to the reset time is the same as a PID structure of Proportional and Derivative on PV and Integral on Error (PD on PV, I on Error) so that the setpoint and load response are the same. The addition of lead time to the setpoint that is about 25% of the setpoint filter where the filter is lag time of a lead-lag block enables a faster setpoint response.
(22) Choosing and Achieving Level Control Objective. We are increasingly becoming aware that level loops are tuned too aggressively causing rapid changes in the manipulated flow upsetting downstream unit operations. The solution for level loops that need loose control is not to simply reduce the level controller gain because this can cause slow rolling oscillations. The tuning objective is to so minimize the transfer of variability of level to the manipulated flow and is more often stated as the maximization of the absorption of variability. The solution is to first increase the reset time dramatically, like one or two orders of magnitude, and then decrease the PID gain so that the product of the PID gain and reset time is greater than twice the inverse of the integrating process gain whose units are %/sec/% (1/sec) as discussed in Mistake 2 in Part 1. For surge tank level control, the objective is obviously maximization of absorption of variability. This objective has gained such popularity that the cases where the level controller must be tuned tightly are not recognized and addressed. In fact some may say there are no such cases and that feedforward control can take care of providing tighter level control when needed. There are exceptions. The biggest one that comes to mind is the distillate receiver level controller that manipulates reflux flow. Tight level feedback control achieves internal reflux control where changes in column top temperature, particularly from blue northerners, causes a change in overhead distillate flow and hence distillate receiver level that results in a correction of manipulated reflux flow in the direction to minimize the disturbance. Another case is where tight residence time control in continuous reactors provides enough time to complete a reaction but not so much time as to cause side reactions or polymer buildup. A change in production rate must result in a change in level setpoint that must be reached quickly by tight level control so that residence time (level/flow) is as constant as possible. A similar concern may exist for continuous crystallizers. For multiple effect evaporators, changes in discharge flow from the last stage to control product solids concentration must be translated to changes in feed coming into each stage by its level controller to affect product concentration. There are similar requirements whenever an upstream flow is manipulated to control a level, such as a raw material makeup flow to deal with changes in a recovered recycle flow. While feedforward flow and ratio control can help, good level control deals with the inevitable errors that cause unbalances in stoichiometry.
(23) Misunderstanding of Load Disturbances. There is a huge disconnect in the literature and what really happens in a plant in terms of the supposed location of a disturbance. The literature and consequently many tuning methods and new algorithms supposedly better than PID are based on the disturbance being on the process output downstream of time constants and dead times in the process and even in most cases ignoring any time constants or dead times in the measurement. This view is convenient for thinking model predictive control and internal model control are best for disturbance rejection and that tuning for setpoint changes is sufficient since a disturbance on the process output is as quick as a change in setpoint. The reality is that nearly all disturbances occur as a process input and are delayed by process dead times or slowed by process time constants. For lag dominant processes, this recognition is particularly important and is the basis of switching from self-regulating tuning rules where lambda is a closed loop time constant for a setpoint change to integrating process tuning rules where lambda is an arrest time for rejection of a load disturbance on the process input. Of course there are a few exceptions where the disturbance is on a process output that would benefit from a larger reset time but this can be identified by tuning the controller for a setpoint response. Also, when in doubt a larger reset time is always a good thing to try since integral action is destabilizing. More proportional action can be stabilizing as discussed in Mistakes 2 and 3 in Part 1.
(24) Missing Automated Startup. Often loops cannot be simply put in automatic for startup.The controller approach to setpoint is often not as smooth and consistent with other controllers approach to setpoint. Often the operator manually position valves to get the process to a reasonable operating state before going to automatic control. The best practices of the best operators can be automated and implemented with much better timing and repeatability enabling continuous improvement by better recognition of what is left to be addressed. If operators say the situation is too complex or conditional on their expertise to be automated, it is even a greater opportunity and motivation for automation. For much more on how procedural automation can be used for startup and dealing with abnormal situations, see the Sept 2016 Control Talk column “Continuous improvement for continuous processes”
(25) Missing Ratio Control. Nearly all process inputs are flows that have a specific ratio to each other for a unit operation as seen on a Process Flow Diagram. The simple use of ratio control is inherently powerful where a “leader” flow is chosen that is often a major feed flow and the other flow controller setpoints designated as “followers” are ratioed to the “leader” flow controller setpoint. If the flows need to work in concert with each other, a filter is applied to each flow setpoint including the “leader” flow for reasons noted in Mistake 21. The actual ratio must be displayed for the operator based on measured flows and the operator must be given the ability to change the ratio for startup and abnormal operating conditions via a ratio controller for each “follower” flow. The ratio often has a feedback correction by a primary temperature or composition controller output. For plug flow volumes, conveyors and sheet lines, the feedback correction changes the ratio setpoint. For back mixed volumes, the feedback correction biases the ratio controller output. For more on ratio control, see the 1/31/2017 Control Talk Blog “Uncommon Knowledge for Achieving Best Feedforward and Ratio Control”
(26) Misleading response time statements. The term response time has no value unless a percentage of the final response is noted. For linear systems, the 63% response time is the dead time plus one time constant. The 86% response time often used for valve response is the dead time plus two time constants. The 95% and 98% response times are the dead time plus three and four time constants, respectively. Waiting for the 98% response takes a lot of time making the test vulnerable to changing conditions and disturbances. For large distillation columns, it could take days to see a 98% response.
(27) Not knowing the dead time to time constant ratio. The tuning and performance of a control loop for self-regulating processes depends heavily upon the dead time to time constant ratio. Most studies in the literature are for loops that are smart enough to include dead time have a dead time to time constant ratio between 2 and 0.5 that are termed “balanced” self-regulating processes. When the dead time to time constant ratio is much larger than 1, the process is termed “dead time dominant” and the reset time can be significantly decreased and PID gain should be decreased to reduce the reaction to noise and abrupt response due to the lack of a significant time constant. For more on these processes see the 12/1/2016 Control Talk Blog “Deadtime Dominance - Sources, Consequences and Solutions”
When the dead time to time constant ratio is less than ¼, the process is termed “lag dominant” and “near-integrating”. Integrating process tuning rules should be used that increase the reset time and PID gain to account for the reduced degree of negative feedback in the process. In the time frame of a major PID reaction (4 dead times), the process ramps and appears to be similar to an integrating process. The integrated absolute error (IAE) for all processes is proportional to the ratio of the reset time to controller gain. The peak error for “dead time dominant” processes approaches the open loop error (error if controller was in manual). The peak error for “lag dominant” processes is inversely proportional to the controller gain since the PID gain can be quite high dominating the initial response. For more on how the dead time to time constant ratio affects performance see the 7/17/2017 Control Talk Blog “Insights to Process and Loop Performance”
(28) Unnecessary crossings of split range point. The valve stiction and nonlinearity and process discontinuity is greatest when switching from the manipulation of one valve and stream to another. Once a controller output crosses a split range point, the tendency is to oscillate back and forth unless there is a predominant need for one stream versus the other. Putting a dead band into the split range point will cause oscillations if there are two integrators either due to integrating action in a process or from the integral mode cascade loop or a positioner. The best way to prevent an unnecessary crossing of the split range point is to put up and down rate limits on a valve or flow controller setpoint where the rate limit is slow in the direction of going back to the split range point if there is no safety issue. For cooling and heating, the movement toward heating across split range point may be slowed down for temperature control. For venting and gas inlet flows, the movement toward more inlet flow across split range point may be slowed down for pressure control. For mammalian bioreactor pH control, the movement toward adding a base across the split range point, such as sodium bicarbonate, is slowed down to reduce sodium ion accumulation that increases cell osmotic pressure and cell lysis. External reset feedback to the primary PID of valve position or flow should be used so that primary PID (temperature, pressure or pH) does not try to change a valve or flow faster than it can respond. This is a great feature in general for cascade control and to provide directional move suppression for valve position control and surge control. For more on external reset feedback see the 4/26/2012 Control Talk Blog “What is the Key PID Feature for Basic and Advanced Control” A corrected and improved block diagram in time domain for a PID with the positive feedback implementation of integral action for the ISA Standard Form is
(29) Minimizing instrumentation cost. We get hung up on saving a few thousand dollars putting millions of dollars often at stake in terms of poor process performance and inadvertent shutdowns. A big mistake is allowing a packaged equipment supplier to choose the instrumentation since the supplier is seek to win the low bid contest. Similarly allowing purchasing to decide which instruments to buy is fundamentally bad. We need to take our knowledge about instrumentation performance and insist on the best even when the justification is not clear and pressure is put on to lower system cost. My cohort Stan Weiner would purposely increase the initial estimates for projects to give him the freedom to choose the best instrumentation. He favored inline meters such as Magnetic Flow Meters and Coriolis meters over differential head meters because of the better rangeability, accuracy, and maintainability and insensitivity to piping system. Similarly, for temperatures less than 400 degrees C, I use RTDs with “sensor matching” instead of thermocouples to reduce drift and improve accuracy and sensitivity by orders of magnitude despite a few hundred dollars more in cost.
(30) Lack of middle signal selection. The best way to avoid unnecessary shutdowns, eliminate the reaction to any possible type of single failure including a measurement stuck at setpoint, reduce the reaction to noise, spikes, and drift and provide intelligence as to what is wrong with a measurement is middle signal selection of three independent measurements. For pH this is almost essential. To me it is bizarre how multimillion bioreactor batches are put at risk from using two instead of three pH electrodes resulting in anybody’s guess as to which electrode is right. Some batches can be ruined by a pH that is off by just a few tenths yet engineers are reluctant to spend a couple of thousand dollars upfront not realizing that even if you disregard the cost of a potentially spoiled batch, the reduction in unnecessary maintenance more than pays for the extra electrode. For one large intermediate continuous process, the use of middle signal selection on all of the measurements used by the Safety Instrumented System (SIS) reduced the number of shutdowns from 2 per year to less than one every 2 years saving tens of millions of dollars each year. The risk of a disastrous operator mistake was also realty reduced because startup is the most difficult and hazardous mode for operations.
I could keep on talking but I think this is enough to start your “New Year Resolutions”. Hopefully, you will be better at keeping them than me.
The post What Are Best Practices and Standards for Control Narratives? first appeared on the ISA Interchange blog site.
The following technical discussion is part of an occasional series showcasing the ISA Mentor Program, authored by Greg McMillan, industry consultant, author of numerous process control books, 2010 ISA Life Achievement Award recipient and retired Senior Fellow from Solutia Inc (now Eastman Chemical). Greg will be posting questions and responses from the ISA Mentor Program, with contributions from program participants.
At the place I work we are typically good at documenting how we configure our controls in the form of DDS documents but not always as good at documenting why they have been configured that way in the form of rigorous control narratives.
We now have an initiative to start retrospectively producing detailed control narratives for all our existing controls and I am looking for best practice, standards and examples of what good looks like for control narratives.
I wondered if you had any good resources in this regard or you could point me in any direction. (I did look at ANSI/ISA-5.06.01-2007 but this seems more concerned with URS/DDS/FDS documents rather than narratives).
We are mainly DeltaV now.
We do a lot of DeltaV systems and we use 3 different ways to “document” the control system. As a system integrator “document” for me may mean something than different than for you so let me explain that these documents are my way to tell my programmers exactly how I want the system to be configured. These documents fully define the system’s logic so they can program it and I can test against it.
As I said there are three parts:
Obviously batch flowsheets do not apply if your system isn’t batch but the same flow sheets can be used to define an involved sequence.
The tag list is simply a large excel spreadsheet that includes all of the key parameters – module name, IO Name, tuning constants, alarm constants, etc . It also includes a “comment” cell that can include relatively simple logic like “Man only on/off FC valve with open/close limits and 30 sec stroke” or “analog input”, or “Rev acting PID with man/auto modes and FO valve” etc. Most of the modules can be defined on this spreadsheet.
The logic notes are usually a couple of paragraphs each and explain logic that is more complicated. Maybe we have an involved set of interlocks or ratio or cascade logic. If I have a logic note I’ll reference it in the tag list so the programmer knows to look for it.
The flow sheets are the last part. I usually have a flow sheet for every phase which defines the phase parameters, logic paths, failures, etc. (See Figure 1 for an example of an agitate phase.) Then I create a flow chart for every recipe which defines what phases I am using and what parameters are being passed. (See Figure 2 for an example of a partial recipe.)
Figure 1: Control Narrative Best Practices Agitator Phase
Figure 2: Control Narrative Best Practices Recipe Sample
Hiten Dalal’s Pipeline Feed System Example
I find the American Petroleum Institute Standard API RP 554 Part 1 (R2016) “Process Control Systems: Part 1-Process Control Systems Functions and Functional Specification Development” and the ISA Standard ANSI / ISA 5.06.01-2007 Functional Requirements Documentation for Control Software Applications to be very useful. ANSI/ISA95 also offers guidance on “Enterprise-Control System Integration.” These types of documents in my opinion help include the opinion of all stakeholders in the logic without the stakeholder having to be familiar with flow charting or logic diagrams or specific control system engineering terminology. The functional specification in my opinion is a progressive elaboration of a simple process description done by the process engineer. Once finalized, the functional specification can be developed into a SCADA/DCS operations manual by listing normal sequence of operation along with analysis of applicable responsibility such as operator action/responsibility, logic solver responsibility, and HMI display. You may download my example of a pipeline control system functional specification: Condensate Feed Pump & Alignment Motor Operated Valves (MOVs).
The post When and How to Use Derivative Action in a PID Controller first appeared on the ISA Interchange blog site.
Derivative action is the least frequently used mode in the PID controller. Some plants do not like to use derivative action at all because they see abrupt changes in PID output and lack an understanding of benefits and guidance on how to set the tuning parameter (rate time). Here we have a question from one of the original protégés of the ISA Mentor Program and answers by a key resource on control Michel Ruel concluding with my view.
Is there a guideline in terms of when to enable the derivate term in a PID?
Derivative is more useful when dead time is not pure dead time but instead a series of small time constants; using derivative “eliminate” one of those small time constants.
You should use the derivative time equal to the largest of those small time constants. Since we usually do not know the details, a good rule of thumb is adjusting Derivative time to half the dead time.
Adding derivative (D) will increase robustness (higher gain and phase margin) since D will reduce apparent dead time of the closed loop.
A good example is the thermowell in a temperature loop: if the thermowell represents a time constant of 10 s, using a D of 10 seconds will eliminate the lag of the thermowell.
Hence, the apparent dead time of the closed loop is reduced and you can use more propositional, shorter integral time; the settling time will be shorter and stability better.
When you look at formulas to reject a disturbance, you observe that in presence of D, proportional and integral can be stronger.
We recommend using derivative only if the derivative function contains a built-in filter to remove high frequency noise. Most DCSs and PLCs have this function but some do not or there is a switch to activate the derivative filter.
What does having a higher phase margin increase the robustness?
Robustness means that the control loop will remain stable even if the model changes. Phase and gain margin represents the amplitude of the change before it becomes unstable, i.e. before reaching -180 degrees or a loop gain above one.
Ta analyze, we use open loop frequency response, the product of controller model and process model. On a Bode plot, gain are multiplied (or added if plot in dB) and total phase is the sum of process phase and controller phase.
Phase margin is the number of degrees required to reach -180 degrees when the open loop gain is 1 (0 dB). If this number is large (high phase margin), the system is robust meaning that the apparent dead time can increase without reaching instability. If the phase margin is small, a slight change in apparent dead time will bring the control loop to instability.
Adding derivative adds a positive phase, hence increases phase margin (compare to adding a dead time or a time constant that reduces the phase margin).
The use of derivative is more important in lag dominant (near-integrating), true integrating, and runaway processes (highly exothermic reactions). The derivative action benefit declines as the primary time constant (largest lag) approaches the dead time because the process changes become too abrupt due to lack of a significant filtering action by a process time constant.
Temperature loops have a large secondary time constant courtesy of heat transfer lags in the thermowell or the process heat transfer areas. Setting the derivative time equal to the largest of the secondary lags can cancel out almost 90 percent of the lag assuming the derivative filter is about 1/8 to 1/10 the rate time setting. Highly exothermic reactors can have positive feedback that causes acceleration of the temperature. Some of these temperature loops have only proportional and derivative action because integral action is viewed as unsafe.
If a PID Series Form is used, increasing the rate time reduces the integral mode action (increases the effective reset time), reduces the proportional mode action (decreases effective PID gain or increases effective PID proportional band) and moderates the increase in derivative action. The interaction factors moderates all of the modes preventing the resulting effective rate time from being greater than one-quarter the effective reset time. This helps prevent instability if the rate time setting approaches the reset time setting. There is no such inherent protection in the ISA Standard Form. It is critical that the user prevent the rate time from being larger than one-quarter the reset time in the ISA Standard Form. While in general it is best to identify multiple time constants, a general rule of thumb I use is the rate time should be the largest of a secondary time constant identified or one-half the dead time and never larger than one-quarter the reset time.
It is critical to convert tuning based on setting units and PID form used as you go from one vintage or supplier to another. It is best to verify the conversion with the supplier of the new system. The general rules for converting from different PID forms are given in the ISA Mentor Program Q&A blog post How Do You Convert Tuning Settings of an Independent PID with the last series of equations K1 thru K3 showing how to convert from a series PID form to the ISA Standard Form.
In general, PID structures should have derivative action on the process variable and not error unless the resulting kick in the PID output upon a setpoint change is useful to get to setpoint faster particularly if there is a significant control valve or VFD deadband or resolution limit.
A small setpoint filter in the analog output or secondary loop setpoint along with external reset feedback of the manipulated variable can make the kick a bump. A setpoint lead-lag on the primary loop where the lag time is the reset time and the lead is one-quarter of the lag or a two degrees of freedom structure with the beta set equal to 0.5 and the gamma set equal to about 0.25 can provide a compromise where the kick is moderated while getting to the primary setpoint faster.
Image Credit: Wikipedia
The post, Common Mistakes not Commonly Understood - Part 2, appeared first on the ControlGlobal.com Control Talk blog.
Here we continue on in our exploration of what we should know but don’t know and how it hurts us.
(11) Ignoring actuator and positioner sensitivity. Piston actuators are attractive due to smaller size and lower cost but have a sensitivity that can be 10 times worse than diaphragm actuators. Many positioners look fine for conventional tests but increase response time to almost a 100 times larger for step changes in signal less than 0.2%. The result is extremely confusing erratic spikey oscillations that only get worse as you decrease the PID gain. My ISA 2017 Process Control and Safety Symposium presentation ISA-PCS-2017-Presentation-Solutions-to-Stop-Most-Oscillations.pdf show the bizarre oscillations for poor positioner sensitivity. Slides 21 through 23 show the situation, the confusion and a simple tuning fix. While the fix helps stop the oscillations, the best solution is a better valve positioner to provide precise control (critical for pH systems with a strong acid or strong base due to amplification of oscillations from sensitivity limit).
(12) Ignoring drift in thermocouples. The drift in thermocouples (TCs) can be several degrees per year. Thus, even if there is tight control, the temperature loop setpoint is wrong resulting in the wrong operating point. Since temperature loops often determine product composition and quality, the effect on process performance is considerable with the culprit largely unrecognized leading to some creative opinions. Some operators may home in on a setpoint to get them closer to the best operating point, but the next shift operator may put the setpoint back at what is defined in the operating procedures. Replacement of the thermocouple sensor means the setpoint becomes wrong. The solution is a Resistance Temperature Detector (RTD) that inherently has 2 orders of magnitude less drift and better sensitivity for temperatures less than 400 degrees C. The slightly slower response of an RTD sensor is negligible compared to the thermowell thermal lags. The only reason not to use an RTD is a huge amount of vibration or a high temperature. Please don’t say you use TCs because they are cheaper. You would be surprised at the installed cost and lifecycle cost of a TC versus an RTD. See the ISA Mentor Program Webinar “Temperature Measurement and Control” for a startling table on slide 4 comparing TCs and RTDs and a disclosure of real versus perceived reasons to use TCs on slide 7.
(13) Not realizing the effect of flow ratio on process gain. The process gain of essentially all composition, pH and temperature loops is the slope of the process variable plotted versus the ratio of the manipulated flow to the main feed flow. This means the process gain is inversely proportional to feed flow besides being proportional to slope of plot. In order to convert the slope that is the change in process variable (PV) divided by the change in ratio to the required units of the change in PV per change in manipulated flow, you have to divide by feed flow. The plot of temperature or composition versus ratio is not commonly seen or even realized as necessary. The same sort of relationship holds true where the manipulated variable is an additive or reactant flow for composition or a cooling or heating stream flow for temperature. For temperature control, the slope of the curve is often also steeper at low flow creating a double whammy as to the increase in process gain at low flow. Also, for jackets, coils and heat exchangers, the coolant flow may be lower creating more dead time for a sensor on the outlet. Fortunately for pH, we have pH titration curves where pH is plotted versus a ratio of reagent volume added to sample volume although often the sample volume added is just used on the X axis. In this case, you need to find out the sample volume so you can put the proper abscissa on the laboratory curve. In the process application the titration curve abscissa that is the ratio of reagent volume to sample volume is simply the ratio of volumetric reagent flow to volumetric feed flow if the reagent concentrations are the same. You can then use this plot with application abscissa in terms of flow ratios to determine process gain and valve capacity, rangeability, backlash (deadband) and stiction (resolution) requirements. An intelligent analysis of the amplification of oscillations by the slope of limit cycle amplitude from deadband and resolution limitations determines the number of stages and size of reagent valves needed. Often for strong acid or bases, two or three stages of neutralization with largest reagent valve on first stage and smallest reagent valve on last stage are needed due to valve rangeability or precision limitations. For more details, check out the 12/12/2015 Control Talk Blog “Hidden Factor in Our Most Important Control Loops”.
(14) Ignoring effect of temperature on actual solution pH. We are accustomed to using the built-in temperature compensator that has been in pH transmitters for 60 or more years to account for the effect of temperature seen in glass electrode Nernst Equation. What we don’t tend to do is quantify and take advantage of the solution pH compensation in smart transmitters. The dissociation constant for water, acids and bases are a function of temperature. If you express the water dissociation constant as a pKw and the acid as well as the base dissociation constant as a pka, whenever the pH is within 4 pH of the pKw or pKa, there is a significant effect of temperature on actual pH. Physical property tables can detail the pKw and pKa as a function of temperature but the best bet is to vary the temperature of a lab sample and note the change in pH after correction for the Nernst equation (some lab meters don’t do even Nernst temperature compensation).
(15) Replacing a positioner with a booster. Extensive guidelines dating back to Nyquist plot studies in the 1960s concluded that fast loops should use a booster instead of a positioner. I still hear this rule cited today. This is downright dangerous due to positive feedback from the high outlet port sensitivity of the booster and flexure of the diaphragm actuator causing valve to slam shut. The volume booster should be placed on the output of the positioner with its bypass valve slightly open to stop any high frequency oscillations as seen in the ISA Mentor Program Webinar “How to Get the Most out of Control Valves”. You can fast forward to slides 18 and 19 to see the setup.
(16) Putting VFD speed control in the DCS. We like putting controls and logic as much as possible in the control room for adjustment and maintenance. While this is normally a good idea, a speed loop in the VFD instead of DCS is orders of magnitude faster enabling much tighter control. In fact if the speed loop is put in the DCS for flow and pressure control, you will violate the cascade rule that the secondary loop (speed) must be 5 times faster than primary loop.
(17) Putting a deadband into split range block for integrating processes or cascade control or integral action in positioner. The deadband creates a limit cycle just like deadband from backlash in a control valve when there are two or more integrators in loop whether in process, PID or positioner.
(18) Not taking into account temperature and pH cross sectional profile in pipelines. The temperature and pH varies extensively across a pipeline especially for high viscosity feeds or reagents. The tip should be near the centerline. For small pipelines, this may require installing the sensor in an elbow preferably facing into flow unless too abrasive. The pH sensor tip must, of course, be pointed down preferably at about a 45 degree angle so that bubble in internal fill of electrode does not reside in tip. The angle prevents the bubble from residing at the internal electrode (relatively low probability but possible).
(19) Not preventing measurement noise from phases and mixing. Thinking we need to have the sensor see a process change as fast as possible, we fail to realize a few seconds of transportation delay is better than a poor signal to noise ratio. To prevent a sensor seeing bubbles or undissolved solids in a liquid and droplets or condensate in a gas or steam, you need to locate a sensor sufficiently downstream of static mixer or exchanger or desuperheater outlet or where any streams come together. You need to keep sensor away from a sparge and avoid top or bottom of vessel or horizontal line. For temperature control of a jacketed vessel with split ranged manipulation of cooling water and steam, you should use jacket outlet instead of inlet temperature measurement to allow time for water to vaporize and for steam to condense. An even better solution is to use a steam injector to heat up the cooling water eliminating the transition of phases from going back and forth between steam and cooling water in the jacket. The injector provides rapid and smooth transitions from cooling to heating over quite a temperature range going from cold to hot water.
(20) Tuning to make smooth approach of PID output to final resting value in near and true integrating chemical processes. The main task of composition, temperature and pH loops in chemical processes is to be able to effectively reject load disturbances at the process input. This requires a maximization of controller gain and significant overshoot by the controller output of the final resting value to balance the load. Many experts in tuning who worked mostly on self-regulating processes don’t realize this requirement and may even say you should never tune the controller output to overshoot the final resting value failing to realize near-integrating processes will take an incredible long time to recover and true integrating processes will never recover from load disturbance. To understand the necessity of overshoot in PID output, think of a level loop where the level has increased because the flow into the vessel has increased. To bring the level back down to setpoint, the outlet flow manipulated by the level controller must be greater than the inlet flow to lower the level to setpoint before outlet flow settles out to match the inlet flow (final resting value of PID output).
Always remember to …………………………………………………………..………….. Oh shoot, I forget ... senior moment.
The post How to Manage Pipeline Valve Positioner and PID Tuning first appeared on the ISA Interchange blog site.
I have been trying to get a handle on small ripples in one of the pipelines by using a rule of thumb to successively reduce proportional action by 20 percent and integral action by 50 percent. Using the same rule, I could stabilize the ripples on Friday. On Sunday, the product changed in the pipeline and with that back came those 4 percent ripples. There is one control valve that impacts line pressure. I could stretch ripples a bit but could not eliminate them. Output going to zero is natural scheduled shutdown of pipeline. I know it is a lot of information that I am providing but perhaps you can glance through and pinpoint something that stands out. I am learning since I started tuning the control valve that it is product sensitive as well.
Since I don’t know if there is a trend of valve signal and valve flow, I am not sure what is happening. If the considerable decrease in gain does not help or makes it worse, I am wondering if there is some valve stiction or backlash, respectively. Is the valve the same for both products? Could a product be causing more stiction due to buildup or coating on valve seating or sealing surfaces or stem? Could the Sunday valve be closer to the shutoff where friction is greatest?
It sure looks like you have too much proportional (P) action for the new product. The integral action is already greatly reduced and most of the overcorrection is occurring very quickly due to proportional action. I would try decreasing the proportional mode action (proportional mode gain) by 50 percent (cut gain in half). If this helps, reduce the proportional gain again. Based on the very small integral (I) action, you may be able to increase integral action once you decrease proportional action. However, I reiterate that if decreasing the gain simply increases the period of the oscillation, you have backlash or stiction. If amplitude stays the same, you have stiction.
Please make sure there is no integral action in the digital valve controller.
When you say no integral action, do you mean in valve positioner or in controller? I don’t think our positioner has any PID setup. Only PID action is in controller. Since it is liquid pressure and flow, we use P&I. Are you suggesting we use only P action in my controller?
I meant no integral in the valve positioner that for Fisher is called a digital valve controller (DVC). You should use integral action in most process controllers (e.g., flow and pressure). Integral action in the process controllers is essential for the PID control of many processes. So far as tuning the process controller for pipeline control, the integral time also known as reset time (seconds per repeat) should generally be greater than four times the deadtime for an ISA Standard Form. You must be careful about what PID form, structure and tuning setting units are being used. If the integral setting is an integral gain, such as what is used in the “parallel” PID form depicted in textbooks and used in some PLCs, the integral setting may not just be a simple factor of the deadtime (e.g., four times deadtime) but will also depend upon other dynamics. Also, some integral settings are in repeats per minute instead of seconds.
Please make sure you extensively test any tuning settings by making small changes in the setpoint with the controller in automatic or in the controller output by momentarily putting the controller in manual. There should be little to no oscillation. The tests should be done at different valve positions particularly if the valve installed flow characteristic is nonlinear. Oscillations may be most prone near the shutoff positioner where stiction is greatest from seat/seal friction.
If there is interaction between loops, the least important loop must be made slower or decoupling used by means of a feedforward signal. If you are going to do some optimization via a controller that seeks to minimize or maximize a valve position, the proportional gain divided by the reset time for this controller doing optimization must be an order of magnitude smaller than process controller to prevent interaction. These PID controllers used for optimizing a valve position are called “valve position controllers” (VPC). I hesitated to mention this to avoid confusion because these are not valve positioners and are only used for optimization. Also, nonlinear or notch gains and directional move suppression via external reset feedback are used to keep the VPC from responding too much or too little so the process controller does not oscillate or run out of valve.
Many newer smart positioners have added integral action to positioners in the last two decades. In some cases, integral action is enabled as the default. This prompted me to write the Control Talk blog post “Getting the Most Out of Positioners.” This blog does not address setting integral action in process controllers (e.g., flow and pressure controllers).
Do you teach a control valve tuning class? Is there a specific method you recommend for a pipeline control valve?
I do not offer a class on tuning positioners. Supplier courses on tuning positioners are good but you will need to insist on turning off integral action. You can have them talk to me if they disagree. In general you should make sure you do not use integral action and that you use the highest valve positioner gain that does not cause oscillation since for pipeline flow and pressure control, oscillations are not filtered. If you have an Emerson Digital Valve Controller (DVC), I recommend “travel control” with no integral action and with the highest gain that still gives an overdamped response. The valve must be a true throttling valve and not an on-off valve posing as a throttling valve as discussed in the Control Talk blog “Getting the Most out of Valve Positioners”. Note that in this blog we are going for a more aggressive response than what you need. Because of the lack of a significant process time constant in a pipeline, you need a smooth valve response. In the blog, the valve positioner gain is described to be set high enough to cause a slight overshoot and oscillation that quickly settles out. Oscillations in the valve response are useful to get a faster response for vessels and columns since there is a a large process time constant to filter out oscillations. You want to still use a high gain and no integral action in the positioner but seek an overdamped (non-oscillatory) response of valve position.
I have bought Tuning and Control Loop Performance Fourth Edition. I reference tables from there for suggested PID values. I have removed derivative from several pressure and flow loops and observed them to be equally efficient. In the process of tuning I have learned that operations installations have impact on loop tuning. I have made the following types of corrections,
(1) As installed, the logic had the PID getting initiated as soon as block valve #1 was fully opened but block valve #2 was getting commanded to open after #1 causing PID output to ramp off to high output limit since the control valve was not seeing full flow. We solved this by setting temporary upper clamp in PID output at safe limit to avoid overshoot until block valve #2 was fully opened.
(2) Transmitter range was high and margin of error was not acceptable by operations. Re-ranged transmitter to suitable range and brought error within acceptable margin.
(3) EIM Controls Electric and REXA electrohydraulic actuators have a limit on number of actuations. I added an acceptable dead band to reduce number of actuations.
The post, Common Mistakes not Commonly Understood - Part 1, first appeared on ControlGlobal.com's Control Talk blog.
There are many mistakes but some are repeated over and over again even though the automation engineer is attentive and experienced and has the best intentions. Part of the problem is overload in terms of tasks and the time crunch. It is highly unlikely engineers today read even a smattering of the thousands of pages in books, handbooks, white papers and articles. The knowledge to prevent the following mistakes may be buried in this literature but I am not so sure of even this. In any case, one probably could not find it. Here is my effort to get straight to the point of realizing and fixing mistakes.
(1) Reset time set too large for deadtime dominant processes. Most tuning algorithms don’t recognize what Shinskey found is that the reset time could be decreased by a factor of 8 or more from 3 to 4 times the dead time to 0.4 to 0.5 times the dead time for the same controller gain setting for severely dead time dominant processes. Lambda tuning can accomplish a dramatic reduction in reset time since the reset time is the time constant that for dead time dominant processes is by definition less than the dead time. The controller gain is also proportionally reduced providing stability despite a much smaller reset time, which is generally good since these processes are more likely to have noise and a jagged not so smooth response due to the lack of a significant time constant. The criticism that the controller reset time and gain becomes too small leading to integral-only type of control for severely dead time dominant processes is avoided by simply putting a limit of ¼ the dead time on the reset time that is then used in the equation for the controller gain as discussed in the June 2017 Control Talk Column “Opening minds about controllers, part 1”. This column is also a good resource for understanding the next common mistake where the reset time is set too small.
(2) Reset time set too small for lag dominant (near-integrating) processes, integrating and runaway processes. These processes lack self-regulation in the process and depend more upon the gain action in the PID to provide the negative feedback missing in the process. Since engineers are not comfortable with controller gains greater than 5 and operators object to sudden movements of PID output, the controller gain is often an order of magnitude or more too small. Since the product of the reset time and gain must be greater than twice the inverse of the integrating process gain to prevent the start of slow oscillations, the reset time is an order of magnitude or more too small. Since we are taught in control theory classes how too high a PID gain causes oscillations, the PID gain is typically decreased making the problem worse. For more on this pervasive problem and the fix, see the 9/14/2017 Control Talk Blog “Surprising Gains from PID Gain” which leads us to the next common mistake.
(3) PID gain set too small for valves with poor positioner sensitivity and excessive dead band from backlash and variable frequency drives with a large dead band setting. Not only is the PID gain too small per last mistake but is also too small to deal with valve problems as seen in my ISA 2017 Process Control and Safety Symposium slides ISA-PCS-Presentation-Solutions-to-Stop-Most-Oscillations.pdf that details a lot of cases where a counterintuitive increase in PID gain reduces or stops oscillations.
(4) Split ranged valves used to increase valve rangeability. The transition from the large to small valve is not smooth since the friction and consequently stiction is greatest near shutoff as plugs rub seats and balls or disks rub seals. Since the stiction in percent stroke translates to a larger abrupt change in flow and amplitude in the limit cycle, small smooth changes in flow are not possible especially near shutoff but also whenever the large valve is open. The better solution is a large and small valve stroked in parallel either where a Valve Position Controller manipulates the large valve with directional move suppression to keep the small valve near an optimum position or by simultaneous manipulation as detailed in the November 2005 Control feature article “Model Predictive Control can Solve Valve Problem.”
(5) Ignoring effect of meter velocity on flow measurement rangeability. The maximum velocity for a given meter size rarely corresponds to the velocity at the maximum flow in a process application. Often the maximum process velocity is less than half the maximum meter velocity for line size meters. Thus, the velocity at the minimum process flow is so far below the minimum flow for a good meter response that the actual rangeability is less than half what is stated in the literature.
(6) Ignoring the effect of noise on flow measurement rangeability. The signal to noise ratio often deteriorates before the meter reaches the low flow corresponding to its rangeability limit. The flow measurement can essentially become unusable for flow control making the actual rangeability much less than what is stated in the literature.
(7) Ignoring the effect of stiction, backlash and pressure drop on valve rangeability. There are many definitions of valve rangeability that are erroneous, such as those that define rangeability as the ratio of maximum to a minimum flow coefficient (Cv) where the closeness of the actual to theoretical inherent flow characteristic determines the minimum Cv leading to the conclusion that a rotary valve offers the greatest rangeability. The real rangeability should be the ratio of maximum to minimum controllable flow. Deadband from backlash and resolution from stiction near the shutoff position should determine the minimum position that gives a controllable flow. Since stiction is greatest as the plug moves into the seat or ball or disk moves into the seal particularly for tight shutoff valves, the minimum controllable position can be quite large (e.g., 2% to 20%). The flow at this position needs to be computed based on the installed flow characteristic. A ratio of valve to pressure drop less than 0.5 will cause a linear characteristic to distort to quick opening increasing the flow at the minimum controllable position causing a significant loss in rangeability. For equal percentage valves there is also a loss in the minimum controllable flow due to excessive flattening of the installed characteristic. There may also be significant flattening of the installed flow characteristic for rotary valves resulting in any rotation past 50 degrees being ineffectual due to the flatness of the installed flow characteristic, which shows up as the controller output through integral action wandering about above 50 degrees. My book Tuning and Control Loop Performance Fourth Edition published in 2014 by Momentum Press gives the equations to compute the real rangeability. It turns out sliding stem valves with diaphragm actuators and smart positioners have the best rangeability.
(8) Ignoring the effect of static head, motor and frame type, and inverter type and control algorithm on VFD rangeability. Since the inverter waveform is not purely sinusoidal, it is important to select motors that are designed for Pulse Width Modulation (PWM). These “inverter duty” motors have windings with a higher temperature rating (class F). Another option that facilitates operation at lower speeds to achieve the maximum rangeability offered by the PWM drive is a higher service factor (e.g. 1.15). To help prevent motor overheating at low speeds, larger frame sizes and line powered ventilation fans are used. In the process industry, totally enclosed fan cooled (TEFC) motors are used to provide protection from chemicals and ventilation by a fan that is run off the same power line as the motor. The fan speed decreases as the motor speed decreases. To reduce the problem from motor overheating at low speeds, an AC line power constant speed ventilation fan and a larger frame size to provide more ventilation space can be specified. Alternately, a separate booster fan can be supplied. For very large motors (e.g. 1000 HP), totally enclosed water cooled (TEWC) motors are used to deal with the extra heat generation. For low static head pump applications, the overheating at low speeds is not a problem because the torque load decreases with flow. Turndown also depends upon the control strategy in the variable frequency drive. All of the control strategies discussed here use pulse width modulation to manipulate the frequency and amplitude of voltage and current to each phase. Open loop voltage (volts/hertz) control has the simplest algorithm but is susceptible to varying degrees of slip. Most of the drives provided for pump control use this strategy in which the rate of change of flux and hence speed is taken as proportional to voltage. At low speeds the motor losses are larger making the difference between the computed and actual speed (slip) much larger. Some drives make a correction to the voltage to account for estimated motor losses. Ultimately these drives depend upon the DCS to correct for dynamic slip through proportional action and to correct for steady state slip through integral action in process controller(s). The rangeability is normally 40:1 with 0.5% speed regulation. Closed loop slip control has a speed loop cascaded to a torque loop. Speed (tachometer) and torque feedback are shown to be from sensors. The torque feedback may be calculated from a current sensor. A DCS process controller output is the speed set point for the speed controller whose output is the set point to a torque controller. PI rather than P-only controllers can be used since sticktion and resolution limits are negligible, eliminating any concern about limit cycles from integral action. The control system in the VSD is analogous to the cascade control system in a digital positioner. The speed controller plays a role similar to the valve position controller and the torque controller serves a similar purpose as the relay controller. However, in the digital positioner the relay response is inherently much faster than the valve position response. In the VSD, the torque controller can have a relatively sluggish response. To prevent a violation of the cascade rule that requires the secondary loop (torque) to be 5x faster than the primary loop (speed), the speed loop is slowed by decreasing the speed controller gain and integral time. Since the speed set point comes from process controller in the DCS, there is at least a triple cascade. In many cases there is a quadruple cascade control system, for vessel temperature to jacket temperature to coolant flow to speed cascade. The detuning of the speed controller causes detuning of the flow controller, which in turn may cause detuning the temperature controller. As a result, the ability to reject fast process disturbances may be compromised. The rangeability is normally 80:1 with 0.1% speed regulation. However, if the static head approaches the total pressure rise, the rangeability can deteriorate by an order of magnitude for all VFDs with a resulting installed flow characteristic that is quick opening.
(9) Ignoring changes in fluid composition in thermal mass flow meters. Changes in fluid composition cause a change in the assumed thermal conductivity and specific heat capacity of the fluid and the viscosity for liquids introducing a significant error. If you are trying to use a thermal flow meter on air/gas/vapors never install it in a service where it can ever see a gas/vapor approaching dew point. Fouling also causes an error due to thermal lags. Thermal mass flow meters are generally only successfully used on small dry pure gas flows such as oxygen or air for lab or pilot plant bioreactors (very controlled environment) where the fluid is clean single phase and composition is fixed and the remaining 1% or 2% error is corrected by a primary dissolved oxygen controller manipulating the secondary oxygen or air flow controller setpoint.
(10) Ignoring changes in emissivity in optical pyrometers. Two-color or ratio pyrometers measure the radiation at two wavelengths. If the change in emittance at each wavelength with temperature is identical (gray-bodies), the effect of emittance can be cancelled out by ratio calculations. In reality, the change in emittance with temperature varies with wavelength (non-gray-bodies). Additionally, the change in emittance with changes in surface, operating conditions, and the composition of the intervening space may vary with wavelength. In one comparison test on a blackbody, a single-color and two-color pyrometers exhibited errors of 2 and 30 degrees C, respectively. Equal changes in emittance due to surface and operating conditions and intervening gases, particles, and vapors may make a two-color ratio pyrometer more accurate than a single-color pyrometer but it puts into question any accuracy statements for two-color pyrometers that are much better than 30 degrees C.
Please take my advice. I am not using it anyway. I have mostly retired into the virtual world.
The post, Surprising Gains from PID Gain, first appeared on ControlGlobal.com's Control Talk blog.
We learned in control theory courses that too high a PID gain causes oscillations and can lead to instability. Operators do not like the large sudden changes in PID output from a high PID gain. Operators may see what they think is the wrong valve open in split range control as the setpoint is approached when PID gain dominates the response. Most tuning tests and studies use a setpoint response rather than a load response for judging tuning. A high PID gain that is set to give maximum disturbance rejection in load response will show overshoot and some oscillation for a setpoint response. Internal Model Control or any control algorithm or tuning method that considers disturbances on the process output will see a concern similar to what is observed for the setpoint response because the load and setpoint change are immediately appearing at the PID algorithm as inputs. There are many reasons why PID gain is unfavorably viewed. Here I try to show you that PID gain is undervalued and underutilized.
First, let’s realize that the immediate feedback correction based on the change in process variable being controlled may be beneficial in some important cases. The immediate action reduces the deadtime and oscillations from deadband, resolution and sensitivity limits in the measurement, control valve and variable frequency drive. The preliminary draft of ISA-PCS-2017-Presentation-Solutions-to-Stop-Most-Oscillations.pdf for the ISA 2017 Process Control and Safety Symposium show how important PID gain is for stopping oscillations from non-ideal measurements and valves and also from integrating processes.
PID gain does not play as important a role in balanced self-regulating process often shown in control theory courses and publications. The primary process time constant is not very large compared to the dead time. Consequently, there is more negative feedback action seen in these self-regulating processes. When the time constant becomes more than 4 times the dead time, we consider these processes to be near-integrating in that in the time frame of the PID they appear to ramp losing self-regulation. For these processes and true integrating processes, PID gain provides the negative feedback action missing in the process to halt or arrest the ramp. Integrating process tuning rules are used where lambda is an arrest time. Not readily understood is that there is a window of allowable gains, where too low of a PID gain cause larger and slower oscillations than too high of a PID gain. The problem is even more serious and potentially dangerous for runaway processes (highly exothermic reactors). Most loops on integrating and runaway processes have a reset time that is orders of magnitude too small and a PID gain that is an order of magnitude too low. A PID gain greater than 50 may be needed for a highly back mixed polymerization reactor. Many users are uncomfortable with such high gain settings.
For integrating and runaway processes, the PID output must exceed the load disturbance to return the process variable to setpoint. This is more immediately and effectively done by PID gain action. We can often see an immediate improvement in control by greatly increasing the reset time and then the gain. The gain must be greater than twice the inverse of the product of the open loop integrating process gain (%/sec/%) and reset time (sec) to prevent the start of slow oscillations from violation of the low gain limit.
As the process variable approaches setpoint, there is an immediate reduction in the PID contribution to the output from the PID gain. Reset has no sense of direction and will continue to change the output in the same direction not reversing direction till the process variable crosses setpoint reversing the sign of the error. Operators looking at digital displays waiting for a temperature to rise to setpoint will think a heating valve should still be open if the temperature is just below setpoint when in fact the cooling valve should be open to prevent overshoot. If the loop waits till the PV crosses setpoint, the correction is too late due to deadtime. PID gain provides the anticipatory action missing in reset action.
A setpoint filter equal to the reset time or a PID structure of proportional action on process variable instead of error will eliminate overshoot of the setpoint for PID tuning that maximizes disturbance rejection where the peak error and the integrated error are both inversely proportional to PID gain.
Finally, let’s realize that the use of external-reset feedback allows us to use setpoint rate of change limits on the valve signal or secondary loop signal that will prevent the large abrupt jump in PID output form a high PID gain that upsets operators and some related loops. External-reset feedback will also prevent the PID output from changing faster than a valve or secondary loop can respond. No retuning needed.
These days we personally don’t have time to wait for taking corrective action in our lives and need to seek more anticipation based on where we are going. We also benefit from external-reset feedback. We need to realize the same for PID loops. There is a lot to be gained from PID gain.
The post, How to Motivate Management and Millennials (M&Ms), originally appeared on ControlGlobal.com's Control Talk blog.
The key to a much brighter future of our profession depends upon management providing the funding and support and millennials seeking to improve process performance by the use of the best automation and process control. There are some common approaches to these seemingly very different groups.
Management is focused on the bottom line that may be as short term as quarterly results. If management has only business degrees, this may be the primary and perhaps the only motivation. If management also has a technical degree, there can be an additional motivation to advance the knowledge and technology used to make processes safer and more productive. Technical people are intrigued and attracted to new more powerful developments in technology. An upcoming Control Talk column with Walt Boyes will give considerable insight into how management thinks.
Millennials choose engineering as a major because they were interested in technology and having a positive impact on people and systems by using and advancing the latest technologies. Unfortunately, students have a negative image that industry is seemingly low tech, routine, and “down and dirty”. Peter Martin aptly discussed the negative view by engineering graduates of working in industry and the missing understanding of opportunities that meet their altruistic motivation in his July 2017 ISA Interchange post “The Challenges of Attracting Millennial to Industrial Careers”. Another goal of engineers may be to make money and seem important by moving on and becoming a manager.
The use of the best and highest tech hardware and software in automation and process control can yield impressive improvements in process safety and performance that are much faster and less expensive than changing process equipment. Unfortunately, management is often not aware of this as discussed in the May 2017 Control Talk column “The Invisibility of process control”.
What can possibly work to impress both management and millennials are demos that show the benefits and use of new technologies. The “before” and “after” cases should show benefits of increases in process performance in terms of dollars with a running quarterly total moving forward showing continuous improvement from knowledge gained. Since time is precious and concentration spans are short, the summary of dynamic runs can be presented in terms of trend charts of benefits for changes in demand and supplier feeds. “Seeing is believing.” The reality of being able to adjust to a dynamic world is inspiring. For engineers, a chance to see live demos may increase interest because of the dynamics. The emphasis should be on how fast and beneficial are the results from using the best technology. See the August feature article “Virtual Plant Virtuosity” for how to develop high tech solutions that impress the M&Ms satisfying altruistic and monetary motivations.
I love M&Ms and there are so many flavors now. See if you can enjoy the M&Ms that determine our profession as much as the M&Ms that satisfy your cravings. Maybe M&Ms can be a totally sweet deal.
The post Webinar Recording: How to Use Key PID Features first appeared on the ISA Interchange blog site.
Héctor Torres, a protégé of the ISA Mentor Program from its inception, provides a detailed view of how to use key PID features that can greatly expand what you can achieve. The setting of anti-reset windup ARW limits, dynamic reset limit, eight different structures, integral dead band, and set-point filter. Feedforward and rate limiting are covered with some innovative application examples.
The post, Insights to Process and Loop Performance, originally appeared on ControlGlobal.com's Control Talk blog.
Here we look at a myriad of metrics on process and control loop performance and show how to see through the complexity and diversity to recognize the commonality and underlying principles. We will see how dozens of metrics simplify to two classes each for the process and the loop. We also provide a concise view of how to compute and use these metrics and what affects them.
Let’s start with process metrics because while as automation engineers we are tuned into control metrics, our ultimate goal is improvement in the process and thus, process metrics. The improvement in profitability of a process comes down to improving process efficiency and/or capacity. Often these are interrelated in that an increase in process capacity is often associated with a decrease in process efficiency. Also an increase in the metrics for a particular part of a process may decrease the metrics for other parts of the process. The following example cited in the April 2017 Control Talk Column “An ‘entitlement’ approach to process control improvement” is indicative of the need to have metrics and an understanding for the entire process:
“In a recent application of MPC for thermal oxidizer temperature control that had a compound response complicating the PID control scheme, there was a $700K per year benefit clearly seen in reduced natural gas usage. However, the improvement also reduced steam make to a turbo-generator, reducing electricity generated by $300K per year. We reached a compromise of about $400K per year in net benefit because of lost electrical power generation from less steam to the turbo-generators. We spent many hours to align the benefit with measureable accounting for the natural gas reduction and the electrical purchases. Sometimes the loss of benefits is greater than expected. You need to be upfront and make sure you don’t just shift costs to a different cost area.”
Process efficiency can be increased by reducing energy use (e.g., electricity, steam, coolant and other utilities) and raw materials (e.g., reactants, reagents, additives and other feeds). The efficiency is first expressed as a ratio of the energy use per unit mass of product produced (e.g., kJ/kg) or energy produced (kJ/kJ) and then ideally in terms of ratio of cost to revenue by including the cost of energy used (e.g., $ per kJ) and the value of revenue for product produced (e.g., $ per kg) or energy produced (e.g., $ per kJ). The kJ of energy and kg and mass are running totals where the oldest value of mass flow or energy multiplied by a time interval between measurements is replaced in the total by the current value. A deadtime block can provide the oldest value. The time interval between measurements and the deadtime representative of the time period for the running total should both be chosen to provide a good signal to noise ratio. The deadtime block time period should also be chosen to help focus on the source of changes in process efficiency. For batch operations, the time period is usually the cycle time of a key phase in the batch and may simply be the totals at the end of the phase or batch. For continuous operations, I favor a time period that is an operator shift to recognize the key effect of operators on process performance. This time period is also suitable for evaluating other sources of variability, such as the effect of ambient conditions (day to night operation and weather) and feeds and recycle and heat integration (upstream, downstream and parallel unit operations). The periods of best operation can be used to as a goal to be possibly achieved by smarter instruments or better installations less sensitive to ambient conditions or smarter controls thru procedural automation or state based control as discussed in the in the Sept 2016 Control Talk Column “Continuous improvement of continuous processes”.
The metrics that affect process capacity are more diverse and complicated. Process capacity can be affected by feed rates, onstream time, startup time, shutdown time, maintenance time, transition time, spectrum of products and their value, recycle, and off spec product. An increase in off spec product that can be recycled can be taken as a loss in product capacity if the raw material feed rate is kept the same or taken as a loss in process efficiency if the raw material feed rate is increased. If the off spec product can be sold as a lower revenue product, the $ per kg must be correspondingly adjusted.
For batch operations, an increase in batch end point in terms of kg of product produced and a decrease in batch cycle time including time in-between batches can translate to an increase in process capacity. If a higher endpoint can be reached by holding or running the batch longer, there is a likely increase in process efficiency assuming a negligible increase in raw material but there may be an increase or decrease in process capacity. The optimum time to end a batch is best determined by looking at the rate of change of product formation (batch slope) and if necessary the rate of change of raw material and energy use to determine the optimum time to end the batch and move on. A deadtime block is again used to provide a fast update with a good signal to noise ratio to compute the slope of the batch profile and the prediction of batch end point. Of course whether downstream units for recovery and purification are able to handle an increase in batch capacity and their metrics must be included in the total picture. For example in ethanol production, a reduction in fermenter cycle time may not translate to an increase in process capacity because of limitations in distillation columns downstream or the dryer for recovery of dried solids byproduct sold as animal feed. For more on the optimization of batch end points see the Sept 2012 Control feature article “Getting the Most Out of your Batch”.
The metrics that indicate loop performance can be classified as load response and setpoint response metrics. The load response is often most important in that the desired setpoint response can be achieved for the best load response by the proper use of PID options. The load response should in nearly all cases be based on disturbances that enter as inputs to the process whereas many academic and model based studies are based on disturbances entering in the process output. For self-regulating processes where the process deadtime is comparable to or larger than the process time constant, the point of entry does not matter because the intervening process time constant does not appreciably slow down input disturbances in the time frame of the PID response (e.g., 2 to 4 deadtimes). However, most of the more interesting temperature and composition control loops in my career did not have a negligible process time constant and in fact had a near-integrating, true integrating or runaway open loop response.
The load metrics are peak error and integrated error. The peak error is the maximum excursion after a load upset. The integrated error is most often an absolute integrated error (IAE) but can be an integrated square error. If the response is non oscillatory, the integrated error and IAE are the same. There are also metrics indicative of oscillations such as settling time and undershoot. The ultimate and practical limits to peak error are proportional to the deadtime and inversely proportional to controller gain, respectively. The ultimate and practical limits to integrated error are proportional to the deadtime squared and the ratio of controller reset time to controller gain, respectively.
For setpoint metrics, there is the time to get close to setpoint, which I call rise time, important for process capacity. I am sure there is a better name because the metric must be indicative of the performance for an increase or decrease in setpoint. The other setpoint metrics are overshoot, undershoot and settling time that can affect process capacity and efficiency. The use of a setpoint lead-lag or PID structure that minimizes proportional and derivative action on setpoint changes can reduce overshoot, despite using good load disturbance rejection tuning. A setpoint lag equal to the reset time (no lead) corresponds to a PID structure of Proportional and Derivative on the Process Variable and Integral action on the Error (PD on PV and I on E).
See the Sept and Oct 2016 Control Talk Blogs “PID Options and Solutions - Part 1” and “PID Options and Solutions - Parts 2 and 3” for a discussion of loop metrics in great detail including when they are important and how to improve them. Also look at the presentation for the ISA Mentor Program WebExs “ ISA-Mentor-Program-WebEx-PID-Options-and-Solutions.pdf ”.
My last bit of advice is to ask your spouse for metrics on your marriage. Minimizing the deadtime while still having a good signal to noise ratio is particularly important. For men, the saying “Happy wife, happy life” I think would work the other way as well. I just need a rhyme.
The post How to Get the Most out of Control Valves first appeared on the ISA Interchange blog site.
The data that is really needed when selecting and sizing a control valve is rarely understood and specified, which leads to excessive variability originating from the valve. In this presentation, ISA mentor Greg McMillan discusses pervasive problems and rampant misconceptions. He then provides guidance—supported by test results—on how to select a good throttling control valve. He also explains PID tuning adjustments and a key PID feature that can be utilized to provide precise, smooth, and fast control.
The post Webinar Recording: How to Get the Most out of Control Valves first appeared on the ISA Interchange blog site.
The post, Fixes for Deadly Deadband, first appeared on ControlGlobal.com's Control Talk blog.
While there are some cases where deadband is helpful, in most applications the effect is extremely detrimental and confusing. Deadband can arise from any sources either intentionally or inadvertently. Deadband creates deadtime and for certain conditions excessive and persistent oscillations.
The increase in loop deadtime is the deadband divided by the rate of change of controller output. The increase in deadtime can increase the peak error and integrated error from a load disturbance. If there are two or more integrators in the system due to integral action in the valve positioner, variable speed drive, controller(s), or process, a limit cycle will develop.
The biggest and most troublesome source of deadband is backlash from an on-off or isolation valve (tight shutoff valve) posing as a throttling valve. The positioner seeing feedback from the actuator shaft of such rotary valves often does not realizer the internal closure member (e.g. ball or disk) is not responding due to backlash from the connections between the shaft, stem and ball or disk or the shaft windup from seal friction. The positioner diagnostics say everything is fine even meeting the requirements set by the ISA-75.25.01 Standard for Measuring Valve Response. Creative story telling develops to explain the oscillations in the process.
An on-off or isolation valve offers a great advantage when used in series with a throttle valve. Besides achieving tight shutoff, the placement of a quickly stroked completely open or closed on-off or isolation valve close-coupled to the connection into the process eliminates the deadtime and any unbalance between ratioed flows during the start and stop of reactant and reagents enabling more precise composition and pH control. The throttle valve is located at a position that is more accessible for better maintenance and with some straight runs upstream and downstream. The throttle valve straight run requirements are rather minimal but can give a more consistent flow relationship between valve position and flow.
For the throttle valve, the best solution is to get rid of the excessive deadband. Given that you are literally and figuratively, stuck with deadband principally when the source is a big valve, an increase in the PID gain will reduce the peak and integrated absolute error (IAE) by increasing the rate of change of the PID output and thus decreasing the additional deadtime from deadband. If there is a limit cycle, increasing the PID gain reduces the amplitude and period of the limit cycle, decreasing the persistent IAE and increasing the ability of downstream volumes to filter out the oscillations. Open loop step tests don’t reveal the additional deadtime but show a decrease in process gain upon a reversal of direction of step change. A filter time can be judiciously added that is less than 20% of the total loop deadtime seen in the test to prevent changes in the PID output from noise exceeding the deadband of the valve. For more on the effects of backlash see the May 2016 Control article “How to specify control valves that don’t compromise control” and the recording of the YouTube recording to be posted in June on the “ISA Mentor Program Webinar Playlist” of my ISA Mentor WebEx “ ISA-Mentor-Program-WebEx-Best-Control-Valve-Rev0.pdf ”. The article white paper and presentation also shows that an increase in PID gain eliminates an oscillation from poor positioner sensitivity by making changes in the valve signal larger than sensitivity limit.
A simple algorithm can be configured to increase the change in PID output by an amount slightly less than the deadband when the output changes direction and the change is greater than the noise band seen in the PID output. The kick of the output upon a change in direction eliminates the deadtime and lost motion from backlash. The practical issue is the deadband may vary with valve position, time, operating conditions, and positioner tuning. These algorithms are often used for Model Predictive Control besides PID control.
A lead-lag on the valve signal can reduce the effect of deadband, resolution and positioner sensitivity but the valve movement can quickly become erratic for a lead much larger than the lag time and noise.
Often deadband is a parameter in a Variable Speed Drive (VSD) setup to reduce changes in speed from noise. Often deadband is set too large because of a lack of understanding of the detrimental effect. The deadband should be just slightly larger than ½ the noise band seen in the VSD setpoint.
Dynamic simulation with a backlash-stiction block and a PID with external reset feedback can show this and much more. The virtual plant is my lab to rapidly explore, discover, prototype and test solutions.
I recently went to a Grateful Dead tribute band concert. The “dead heads” were grateful the music of the band was not dead. Keep your control system alive by not succumbing to the deadly deadband.
The post How to Overcome Challenges of PID Control and Analyzer Applications via Wireless Measurements first appeared on the ISA Interchange blog site.
This article was authored by Greg McMillan, industry consultant, author of numerous process control books, 2010 ISA Life Achievement Award recipient and retired Senior Fellow from Solutia Inc. (now Eastman Chemicals).
Wireless measurements offer significant life-cycle cost savings by eliminating the installation, troubleshooting, and modification of wiring systems for new and relocated measurements. Some of the less recognized benefits are the eradication of EMI spikes from pump and agitator variable speed drives, the optimization of sensor location, and the demonstration of process control improvements. However, loss of transmission can result in process conditions outside of the normal operating range. Large periodic and exception reporting settings to increase battery life can cause loop instability and limit cycles when using a traditional PID (proportional-integral-derivative) for control. Analyzers offer composition measurements key to a higher level of process control but often have a less-than-ideal reliability record, sample system, cycle time, and resolution or sensitivity limit. A modification of the integral and derivative mode calculations can inherently prevent PID response problems, simplify tuning requirements, and improve loop performance for wireless measurements and sampled analyzers.
The combination of periodic and exception reporting by wireless measurements can be quite effective. The use of a refresh time (maximum time between communications) enables the use of a larger exception setting (minimum change for communication). Correspondingly, the use of an exception setting enables a larger refresh time setting. The time delay between the communicated and actual change in process variable depends upon when the change occurs in the time interval between updates (sample time). Since the time interval between a measured and communicated value (latency) is normally negligible, on the average, the true change can be considered to have occurred in the middle of the sample time. This delay limits how quickly control action is taken to correct changes introduced by process disturbances.
Since ultimately what you often want to control is composition in a process stream, online analyzers can raise process performance to a new level. However, analyzers, such as chromatographs, have large sample transportation and processing time delays that contribute to the total loop deadtime and are generally not as reliable or as sensitive as the pressure, level, and temperature measurements.
The sample transportation delay from the process to the analyzer is the sample system volume divided by the sample flow rate. This delay can be five or more minutes when analyzers are grouped in an analyzer house. Once the sample arrives, the processing and analysis cycle time normally ranges from 10 to 30 minutes. The analysis result is available at the end of the cycle time. If you consider the change in the sample composition occurs in the middle of the cycle time and is not reported until the end of the next cycle time, the analysis delay is 1½ times the cycle time. This cycle time delay is added to the sample transportation delay, process deadtime, and final control element delay to get the total loop deadtime. The sum of the 1½ analyzer cycle time plus the sample transportation delay will be referred to as the sample time.
Most of the undesirable reaction to discontinuous measurement communication is the result of integral and derivative action in a traditional PID. Integral action will continue to drive the output to eliminate the last known offset from the setpoint even if the measurement information is old. Since the measurement is rarely exactly at the setpoint within the A/D and microprocessor resolution, the output is continually ramped by reset. The problem is particularly onerous if the current error is erroneous.
Derivative action will see any sudden change in a communicated measurement value as occurring all within the PID execution time. Thus, a change in the measurement causes a spike in the controller output. The spike is especially large for restoration of the signal after a loss in communication. The spike can hit the output limit opposite from the output limit driven to from integral action. The spike from large refresh time can also cause a significant spike, because the rate of change calculation uses the PID execution time.
A smart PID has been developed that makes an integral mode calculation only when there is a measurement update. The change in controller output from the proportional mode reaction to a measurement update is fed back through an exponential response calculation with a time constant equal to the reset time setting to provide an integral calculation via the external reset method. For applications where there is an output signal selection (e.g., override control) or where there is a slowly responding secondary loop or final control element, the change in an external reset signal can be used instead of the change in PID output for the input to exponential response calculation. The feedback of actual valve position as the external reset signal can prevent integral action from driving the PID output in response to a stuck valve. The use of a smart positioner provides the readback of actual position and drives the pneumatic output to the actuator to correct for the wrong position without the help of the process controller.
For a reset time set equal to the process time constant so the closed loop time constant is equal to the open loop time constant, the response of the integral mode of the smart PID matches the response of the process. This inherent compensation of process response simplifies controller tuning and stabilizes the loop. For single loops dominated by a large time in between updates (large sample time), whether due to wireless measurements or analyzers, the controller gain can be the inverse of the process gain.
In the smart PID, the time interval used for the derivative mode calculation is the elapsed time from the last measurement update. Upon the restoration of communication, derivative action considers the change to have occurred over the time duration of the communication failure. Similarly, the derivative response to a large sample time or exception setting spreads the measurement change over the entire elapsed time. The reaction to measurement noise is also attenuated. This smarter derivative calculation combined with the derivative mode filter eliminates spikes in the controller output.
The proportional mode is active during each execution of the PID module to provide an immediate response to setpoint changes. The module execution time is kept fast so the delay is negligible for a corrective change in the setpoint of a secondary loop or signal to a final control element. With a controller gain approximately equal to the inverse of the process gain, the step change in PID output puts the actual value of the process variable extremely close to the final value needed to match the setpoint. The delay in the correction is only the final control element delay and process deadtime. After the process variable changes, the change in the measured value is delayed by a factor of the measurement sample time. Consequently, the observed speed of response is not as fast as the true speed of process response, a common deception from measurements with large signal delay or lag times.
Communication failure is not just a concern for wireless measurements. Any measurement device can fail to sense or transmit a new value. For pH measurements, the broken glass electrode or broken wire will result in a 7 pH reading, the most common setpoint. The response of coated or aged electrodes and large air gaps in thermowells can be so slow to show no appreciable change. Plugged impulse lines and sample lines can result in no new information from pressure transmitters and analyzers. Digitally communicated measurements can fail to update due to bus or transmitter problems.
If a load upset occurs and is reported just before the last communication, integral action in the traditional controller drives the PID output to its low limit. The smart PID can make an output change that almost exactly corrects for the last reported load upset, since the controller gain is the inverse of the process gain.
The wireless measurement sample time and transport delay associated with sample analyzers must be taken into account when using these measurements in control. A minimum wireless refresh time of 16 seconds is significant compared to the process response for flow, liquid pressure, desuperheater temperature, and static mixer composition and pH control. The sample time of chromatographs makes nearly all composition loops deadtime dominant except for industrial distillation columns and extremely large vessels. To eliminate excessive oscillations and valve travel caused by sample time and transport delay, a traditional PID controller is tuned for nearly an integral-only type of response by reducing the controller gain by a factor of 5. Increasing the reset time instead of reducing could also provide stability, but the offset is often unacceptable especially for flow feedforward and ratio control.
The smart PID can be aggressively tuned by setting the gain equal to the inverse of the process gain for deadtime dominant loops. The result is a dramatic reduction in integrated absolute error and rise time (time to reach setpoint). The immediate response of the smart PID is particularly advantageous for ratio control of feeds to wild flows and for cascade and model predictive control by higher level loops. The advantage may not be visible in the wireless or analyzer reported value because of the large measurement delay. The improvement in performance is observed in the speed and degree of correction by the controller output and reduced variability in upper level measurements and process quality. A similar deception also occurs for measurements with a large lag time relative to the true process response due to large signal filters and transmitter damping settings, and slow sensor response times. An understanding of these relationships and the temporary use of fast measurements can help realize and justify process control improvement. The ability to temporarily set a fast wakeup time and tight exception reporting for a portable wireless transmitter could lead to automation system upgrades.
Level loops on large volumes can use the largest refresh time of 60 seconds without any adverse affect because the integrating process gain is so slow (ramp rate is less than 1% per minute). Temperature loops on large vessels and columns can use an intermediate refresh time (30 seconds) and the maximum refresh time (60 seconds), respectively, because the process time constant is so large. However, gas and steam pressure control of volumes and headers will be adversely affected by a refresh time of 16 seconds because the integrating response ramp is so fast that the pressure can move outside of the control band (allowable control error) within the refresh time. Furnace draft pressure can ramp off scale in seconds. Highly exothermic reactors (polymerization reactors) can possibly run away if the largest refresh time of 60 seconds is used. To mitigate the effect of a large refresh time, the exception reporting setting is lowered to provide more frequent updates.
Measurements have a limit to the smallest detectable or reportable change in the process variable. If the entire change beyond threshold for detection is communicated, the limit is termed sensitivity. If a quantized or stepped change beyond the threshold is reported, the limit is termed resolution. Ideally, the resolution limit is less than the sensitivity limit. Often, these terms are used indiscriminately.
Wireless measurements have a sensitivity setting called deadband that is the minimum change in the measurement from the last value communicated that will trigger a communication when the sensor is awake. In the near future, the wakeup time in most wireless transmitters of 8 seconds is expected to be reduced. pH transmitters already have a wakeup time of only 1 second enabling a more effective use on static mixers.
A traditional PID will develop a limit cycle whose amplitude is the sensitivity and resolution limit, whichever is larger, from integral action. The period of the limit cycle will increase as the gain setting is reduced and the reset time is increased. A smart PID will inherently prevent the limit cycle.
Wireless and composition measurements offer a significant opportunity for optimizing process operation. A smart PID can dramatically improve the stability, reliability, and speed of response for wireless measurements and analyzers. The result is tighter control of the true process variables and longer battery and valve packing life.
A version of this article originally was published at InTech magazine.
The post, Deadtime, the Simple Easy Key to Better Control, first appeared on ControlGlobal.com's Control Talk blog.
Deadtime is the easiest dynamic parameter to identify and the one that holds the key to better control. Deadtime found visually or by a simple method can tell you what is limiting the ability of the loop and what the remedy is. In most loops, you as the automation engineer can gain a much greater understanding and make a dramatic improvement. You can become famous by Friday (assuming you read this on a Monday).
You can make a small setpoint change in automatic or a small output change (e.g., 0.5% ) by momentarily putting the loop in manual. The time to a change in the process variable in the correct direction is the deadtime. To detect the deadtime and noise visually, compression must be turned off. The remaining principal limit to identifying the deadtime in fast loops is the update time of the historian. For loops with a relatively large deadtime (e.g., greater than 10 sec), the deadtime can be visually identified assuming the update time is 1 sec or less. For loops with smaller deadtimes, you can put a few function blocks together in a module executing as fast as possible (e.g., 0.1 sec for many DCS) to tell you the deadtime. Of course, good tuning software that is executing very fast can tell you the deadtime and a lot more. The point here is that just knowing the deadtime offers exceptional insight and power to do what is right. So without delay let’s explore how we all can become more responsive.
The sum of all the discrete update times such as PID module execution rate and wireless update time and all the signal filter times and transmitter damping time should be less than 20% of the total loop deadtime to limit deterioration in achievable loop performance to be appreciably less than 20%. The valve response time should also be less than 40% of the deadtime. This is going to be difficult to achieve and to measure if you don’t have a true throttling valve and is nearly impossible in pressure systems because the process deadtime is usually so small (e.g., less than 1 sec). Almost as difficult but perhaps less important is the size of the valve response time compared to process deadtime in level systems assuming liquid flows into or out of the volume are manipulated for level control and the sensitivity limit and noise in the level measurement is extremely small enabling a detection of a small level change. The time for the level to get through a sensitivity limit or noise band is additional deadtime.
Consider the case where a loop in manual has no oscillations and develops oscillations when the loop is put in automatic. If the period of oscillations is 3 to 4 times the deadtime, the PID gain is too high. If the period is 6 to 10 times the deadtime, the reset time is probably too small. If the period of a level, gas pressure, or temperature loop on a vessel or column is more than 20 times the deadtime and decaying, it is most likely due to small of a PID gain - actually the product of PID gain and reset time is too small but it is most often caused by a PID gain needed for a normal reset setting being much greater than what is used due to the comfort zone of operations (e.g., many of these loops should have a PID gain that is 50 to 100 unless the reset time is greatly increased). If the period is more than 20 times the deadtime and the amplitude is constant indicating a limit cycle, the source is deadband, backlash, stiction, or resolution limit. If the source is deadband or backlash, increasing the PID gain should be able to reduce the oscillation amplitude and period.
Now let’s look at the situation where a loop in manual has an existing oscillation. If the oscillation period is less than the deadtime, it is essentially noise and the PID should not react to it. If the period is between 2 and 10 times the deadtime, the PID gain must be considerably reduced to prevent amplification of the oscillation due to resonance. If the period is more than 10 times the deadtime, the PID gain should be made as aggressive as possible to reduce the amplitude of the oscillation. Of course, the best solution is to find and eliminate the source of the oscillation.
What the controller sees in the first four deadtimes is most important in terms in controller tuning because unless the PID is seriously detuned, the PID should have reacted to arrest the response from a load disturbance. This corresponds to a lambda setting of 3 or less deadtimes. For a near-integrating, integrating, and runaway process, the maximum ramp rate in % of PV scale (%/sec) in the first four deadtimes divided by the step % change in PID output is approximately the integrating process gain (1/sec) that can be used with the deadtime to tune the PID using integrating process tuning rules. If there is a compound response, having the PID appreciably do its job within 4 deadtime simplifies the tuning and reduces what the PID sees and has to deal with in terms of the consequences of a later response typically due to recycle effects.
The reset time should be greater than 3 deadtimes for a PID with the exception being a truly deadtime dominant process (a rather rare case). The more likely scenario as mentioned before, is that the reset time must be increased because the product of the PID gain and reset time is too small for near-integrating, true integrating and runaway processes. For many loops on vessels and columns, the reset time is several orders of magnitude too small.
Deadtime also determines the limit as to loop performance even if the loop is tuned aggressively. The minimum peak error is proportional to the deadtime and the minimum integrated absolute error is proportional to the deadtime squared. If a PID gain is detuned, the effect can be equated to an increase in an effective deadtime greater than the actual deadtime. In other words, if you spend money to decrease deadtime in the process by better equipment or piping design or in the automation system by faster valves, measurements and discrete actions, if the PID is not tuned to match the decrease in actual deadtime, you do not see an improvement due to an effective deadtime from sluggish PID control.
For more on how deadtime limits performance, see slides 12-14 of ISA-Mentor-Program-WebEx-PID-Options-and-Solutions.pdf and for much more see the associated ISA Mentor Program Webinars.
The effect of an operating point non linearity is reduced by decreasing the total deadtime. My goal in pH control on difficult systems was to make the total deadtime as small as possible by better mixing and reagent injection so that the excursion on the nonlinear titration curve was as small as possible from tighter pH control. In other words, an increase in deadtime causes an increase in the nonlinearity seen, which causes a further deterioration in control (a spiraling effect, literally and figuratively).
If the deadtime is zero, the controller gain could be theoretically infinite and control perfect. Without deadtime, I would be out of job. The good news is that negligible deadtime only exists in simulations. Even if the process deadtime is extremely small, just having an automation system creates a deadtime that must be dealt with. The bad news is that deadtime is extremely detrimental and is not given the proper consideration as to what it is telling you and the PID. Also, the misuse of the term of “process deadtime” rather than “total loop deadtime” leads people into missing the important opportunities to reduce deadtime in the valve, measurement and controller, which is usually more readily done, more in your realm of responsibility and typically much less expensive than reducing deadtime in the process.
The post PID Tuning Rules first appeared on the ISA Interchange blog site.
Nearly every automation system supplier, consultant, control theory professor, and user has a favorite set of PID tuning rules. Many of these experts are convinced their set is the best. A handbook devoted to tuning has over 500 pages of rules. The enthusiasm and sheer number of rules is a testament to the importance of tuning and the wide variety of application dynamics, requirements, and complications. The good news is these methods converge for a common objective. The addition of PID features, such as setpoint lead-lag, dynamic reset and output velocity limits, and intelligent suspension of integral action enable the use of disturbance rejection tuning to achieve other system requirements, such as maximizing setpoint response, coordinating loops, extending valve packing life, and minimizing upsets to operations and other control loops.
The purpose of a control loop is to reject undesired changes, ignore extraneous changes, and achieve desired changes, such as new setpoints. PID control provides the best possible rejection of unmeasured disturbances (regulatory control) when properly tuned. The addition of a simple deadtime block in the external reset path can enhance the PID regulatory control capability more than other controllers with intelligence built-in to process dynamics, such as model predictive control. In plants, unknown and extraneous changes are a reality, and the PID is the best tool if properly tuned. The test time has been significantly reduced for the most difficult loops. Simple equations have been developed to estimate tuning and resulting performance for a unified approach. (Equation derivations and a simple tuning method are in the online version.)
The foremost requirement of a PID is to prevent the activation of a safety instrumentation system or a relief device and the prevention of an environmental violation (RCRA pH), compressor surge, and shutdown from a process excursion. The peak error (maximum deviation from setpoint) is the most applicable metric. The most disruptive upset is an unmeasured step disturbance that would cause an open loop error (Eo) if the PID was in manual or did not exist. The fraction of open loop error seen in feedback control is more dependent upon the controller gain than the integral time since the proportional mode provides the initial reaction important for minimizing the peak error. Equation (1) shows if the product of the controller gain (Kc) and open loop gain (Ko) is much greater than one, the peak error (Ex) is significantly less than the open loop error. The open loop gain (Ko) is the product of the final element, process, and measurement gain and is the percent change in process variable divided by the percent change in controller output for a setpoint change. For most vessel and column temperature and pressure control loops, the process rate of change is much slower than the deadtime. Consequently, the controller gain can be set large enough where the denominator becomes simply the inverse of the product of the gains. Conversely, for loops dominated by deadtime, the denominator approaches one, and the peak error is essentially the open loop error.
The peak error is critical for product quality in the final processing of melts, solids, or paste, such as extruders, sheet lines, and spin lines. Peak errors show up as rejected product due to color, consistency, optical clarity, thickness, size, shape, and in the case of food, palatability. Unfortunately, these systems are dominated by transportation delays. The peak errors and disruptions from upstream processes must be minimized.
The most widely cited metric is an integrated absolute error (IAE), which is the area between process variable and the setpoint. For a non-oscillatory response, the IAE and the integrated error (IE) are the same. Since proportional and integral action are important for minimizing this error, Equation (2) shows the IE increases as the integral time (Ti) increases and the controller gain decreases.
Equation (2) also shows how the IE increases with controller execution time (Δtx) and signal filter time (τf). The equivalent deadtime from these terms also decreases the minimum allowable integral time and maximum allowable controller gain, further degrading the maximum possible performance. In many cases, the original controller tuning is slower than allowed and remains unchanged, so the only deterioration observed is from these terms in the numerator of Equation (2). Studies on the effect of automation system dynamics and innovations can lead to conflicting results because of the lack of recognition of the effect of tuning on the starting case and comparative case performance. In other words, you can readily prove anything you want by how you tune the controller.
IE is indicative of the quantity of product that is off-spec that can lead to a reduced yield and higher cost ratio of raw material or recycle processing to product. If the off-spec cannot be recycled or the feed rate cannot be increased, there is a loss in production rate. If the off-spec is not recoverable, there is a waste treatment cost.
A controller tuned for maximum performance will have a closed loop response to an unmeasured disturbance that resembles two right triangles placed back to back. The base of each triangle is the total loop deadtime and the altitude is the peak error. If the integral time (reset time) is too slow, there is slower return to setpoint. If the controller gain is too small, the peak error is increased, and the right triangle is larger for the return to setpoint.
The major types of process dynamics are differentiated by the final path of the open loop response to a change in manual controller output assuming no disturbances. (The online version shows the three major types of responses and the associated dynamic terms.) If the response lines out to a new steady state, the process is self-regulating with an open loop time constant (τo) that is the largest time constant in the loop. Flow and continuous operation temperature and concentration are self-regulating processes. If the response continues to ramp, the process is integrating. Level, column and vessel pressure, batch operation temperature, and concentration are integrating processes. If the response accelerates, reaching a point of no return, the process has positive feedback leading to a runaway. Batch or continuous temperature in highly exothermic reactors (e.g., polymerization) can become runaway processes. Prolonged open loop tests are not permitted, and setpoint changes are limited. Consequently, the acceleration is rarely intentionally observed.
The three major types of responses have an initial period of no response that is the total loop deadtime (θo) followed by the ramp before the deceleration (inflection point) of a self-regulating response and the acceleration of the runaway response. The percent ramp rate divided by the change in percent controller output is the integrating process gain (Ki) with units of %/sec/%, which reduces to 1/sec.
For at least 10 years, slow self-regulating processes with a long time to deceleration have shown to be effectively identified and tuned as “near integrating” or “pseudo integrating” processes, leading to a “short cut tuning method” where only the deadtime and initial ramp rate need to be recognized. The tuning test time for these “near integrating” processes can be reduced by over 90% by not waiting for a steady state. Recently, the method was extended to runaway processes and to deadtime dominant self-regulating processes by the use of a deadtime block to compute the ramp rate over a deadtime interval. Furthermore, other tuning rules were found to give the same equation for controller gain when the performance objective was maximum unmeasured disturbance rejection. For example, the use of a closed loop time constant (λ) equal to the total loop deadtime in Lambda tuning yields the same result as the Ziegler Nichols (ZN) ultimate oscillation and reaction curve methods if the ZN gain is cut in half for smoothness and robustness. Equation (3) shows the controller gain is half the inverse of the product of integrating process gain and deadtime.
The profession realizes that too large of a controller gain will cause relatively rapid oscillations and can instigate instability (growing oscillations). Unrealized for integrating process is that too small of a controller gain can cause extremely slow oscillations that take longer to decay as the gain is decreased. Also unrealized for a runaway process is that a controller gain set less than the inverse of the open loop gain causes an increase in temperature to accelerate to a point of no return. There is a window of allowable controller gains. Also realized is too small of an integral time will cause overshoot and can lead to a reset cycle. Almost completely unrealized is that too slow of an integral time will result in a sustained overshoot of a setpoint that gets larger and more persistent as the integral time is increased for integrating processes. Hence a window of allowable integral times exists. Equation 4a provides the right size of integral time for integrating processes. If we substitute Equation 3 into Equation 4a, we end up with Equation 4b, which is a common expression for the integral time for maximum disturbance rejection. Equation 4a is extremely important because most integrating processes have a controller gain five to 10 times smaller than allowed. The coefficient in Equation 4b can be decreased for self-regulating processes as the deadtime becomes larger than the open loop time constant (τo) estimated by Equation 5.
The tuning used for maximum load rejection can be used for an effective and smooth setpoint response if the setpoint change is passed through a lead-lag. The lag time is set equal to the integral time, and the lead time is set approximately equal to ¼ the lag time.
For startup, grade transitions, and optimization of continuous processes and batch operations, setpoint response is important. Minimizing the time to reach a new setpoint (rise time) can in many cases maximize process efficiency and capacity. The rise time (Tr) for no output saturation, no setpoint feedforward, and no special logic is the inverse of the product of the integrating process gain and the controller gain plus the total loop deadtime. Equation 6 is independent of the setpoint change.
Fast changes in controller output can cause oscillations from a slow secondary loop or a slow final control element. The problem is insidious in that oscillations may only develop for large disturbances or large setpoint changes. The enabling of the dynamic reset limit option and the timely external reset feedback of the secondary loop or final control element process variable will prevent the primary PID controller output from changing faster than the secondary or final control element can respond, preventing oscillations.
Aggressive controller tuning can also upset operations, disturb other loops, and cause continual crossing of the split range point. Velocity limits can be added to the analog output block, the dynamic reset limit option enabled, and the block process variable used as the external reset to provide directional move suppression to smooth out the response as necessary without retuning.
The different closed loop response of loops can reduce the coordination, especially important for blending and simplification of the identification of models for advanced process control systems that manipulate these loops. Process nonlinearities may cause the response in one direction to be faster. Directional output velocity limits and the dynamic reset limit option can be used to equalize closed loop time constants without retuning.
Final control element resolution limits (stick-slip) and deadband (backlash) can cause a limit cycle if one or two or more integrators, respectively, exist in the loop. The integrator can be in the process or in the secondary or primary PID controller via the integral mode. Increasing the integral time will make the cycle period slower but cannot eliminate the oscillation. However, a total suspension of integral action when there is no significant change in the process variable and when the process is close to the setpoint can stop the limit cycle. The output velocity limits can also be used to prevent oscillations in the controller output from measurement noise exceeding the deadband or resolution limit of a control valve preventing dither, which further reduces valve wear.
Controllers can be tuned for maximum disturbance rejection by a unified method for the major types of processes. PID options in today’s DCS, such as setpoint lead-lag, directional output velocity limits, dynamic reset limit, and intelligent suspension of integral action, can eliminate oscillations without retuning. Less oscillations reduces process variability, enables better recognition of trends, offers easier identification of dynamics, and provides an increase in valve packing life.
The post, When and How to use Pulse Width Modulation of Controller Outputs, first appeared on ControlGlobal.com's ControlTalk blog.
In some applications, throttling of the manipulated flows is difficult or impossible. In the biochemical industry, where precise (good resolution and sensitivity) throttling valves without any crevices (to meet sanitary requirements) are rather limited (there are exceptions, such as the Fisher Baumann 83000-89000 series). Often, pulse width modulation (PWM) is used to turn nutrient and reagent pumps on and off. In the chemical industry, PWM is used to open and close valves whose trim would plug or whose stem would stick if throttled. The sudden burst of flow from on-off action helps flush out the trim and wipe the stem clean. PWM is correspondingly used for small reagent flows, corrosive fluids, and slurries. It is also used to prevent flashing by a valve position that ensures a pressure drop above the critical pressure drop. PWM is also used in temperature loops to turn heaters on and off. Here, it is commonly called “time proportioning control” but the action is principally the same. Temperature loops for extruders, silicon crystal pullers and environmental chambers often use this technique.
All the applications of PWM have one thing in common; a capacity to filter or dilute the pulses so that they do not appear as measurement noise in the controlled variable. PWM provides a train of pulses that show up as a sawtooth in the measurement unless attenuated. The mass of fluid and metal in a reactor, extruder, or crystal puller and mass of air in an environment chamber must be large enough and the maximum pulse width small enough so that the amplitude of the sawtooth seen is negligible.
The rangeability achieved by PWM is basically equal to the maximum pulse width divided by the minimum pulse width. Since a valve must reach a set position and the pump must reach a set speed during the pulse, the minimum pulse width is fixed by the pre-stroke deadtime and stroking time of the valve or the rate limiting of the speed and acceleration time of pump. Usually, four seconds is adequate for small valves and pumps. The maximum pulse width is the pulse cycle time when the pulse is almost continuously on. Since the pulse cycle time also sets the time between successive pulses, it adds a maximum deadtime to the loop that is about equal to the cycle time when the pulse is almost continuously off. For an average controller output of 50%, the deadtime added is about ½ the cycle time. Thus, the cycle time chosen represents a compromise of the desire to maximize rangeability and minimize the sawtooth amplitude seen in measurement and minimize loop deadtime. An additional consideration is the wear and tear on the final control element. Pumps, agitator and motor driven valves have a maximum duty cycle that must not be exceeded. Also, heaters in the motor starter will trip for too short a cycle time because the temperature rise from lack of cool down is equated to an overload current. For valves, periodic opening and closing will eventually cause packing, seat, seal or trim failure.
The consequences and methods of mitigation of pulses are discussed in the 12/15/2014 Control Talk Blog “Controller Attenuation and Resonance Tips”. A simple equation to predict the amplitude of pulses after attenuation by the process or a filter that are seen by the controller is discussed in the 12/02/2104 Control Talk Blog “Measurement Attenuation and Deception Tips”.
The generation of a pulse train is done by special output cards or by the configuration of function blocks on the PID output. The heart of the configuration is a ramp that resets itself periodically. An integrator (INT) function block is employed to generate the ramp. The configuration depends upon what version of Distributed Control System (DCS) or Programmable Logic Controller (PLC) is used. For an integrator that will ramp up towards a setpoint, the input to the integrator is set equal to 100% divided by the desired cycle time. The integrator setpoint is set slightly larger than 100%. The ramp “on” time or pulse width is determined by comparing percent ramp value (integrator output) to the percent controller output via a high signal monitor (HSM) block. When the ramp value exceeds the controller output, a discrete is see equal to one (true), which opens a discrete output or transfers in an analog output value that corresponds to the closed valve position or minimum speed. If the controller output drops below the minimum pulse width, the pulse is turned off by transferring in a negative value before the ramp value is used as the operand of the HSM block on the output of the INT block. The PID low output limit should be set to be slightly less than this minimum pulse width. The functionality of blocks depends upon the DCS or PLC used and any configuration must be extensively tested before being used online.
For viscous fluids and slurries, a precise control valve may be continuously throttled until the valve position gets so small that laminar flow or plugging can occur. At this point (e.g., below 20% PID output), PWM starts. The throttling valve position then stays open (e.g., 20%) and an inexpensive on-off valve in series with the control valve is open and closed by PWM.
There are many other applications of PWM. Pulsed flows have been shown to increase the yield of reactors, the separation in distillation columns and the combustion efficiency of burners. Pulse reagent flow has been very successfully used to mimic a titration for batch pH control. While many well designed pulsed strategies can work for this application, PWM on a proportional-only pH PID controller retains a conventional operator interface via the PID operator faceplate and tuning via the PID gain setting. Also, the controller output can be transferred in for the analog output to reduce batch processing time by providing pulses that are not only longer but that are also larger (pulse width plus pulse amplitude modulation). The gain of the manipulated variable is now nonlinear and is proportional to the controller output. However, for proportional-only control of batch pH processes, this gain change may be advantageous and offset the low pH process gain from the operating point being on the flat portion of the titration curve at the beginning of the batch cycle moving to the steep part of the curve at the end of the batch. This is an example of how a continuous control technique is also useful for batch processing.
PWM also dramatically reduces the effects of deadband and resolution limit in the control valve or variable speed drive assuming the pulse amplitude is at least 5 times as large as the suspected deadband and resolution limit. This normally is the case if the amplitude is > 5%. However, for valves designed for tight shutoff, the backlash and stiction may be as large as 10% requiring 50% amplitude.
You may want to check your pulse now to see how excited you are about PWM opportunities.
The post, Getting the Most Out of Valve Positioners, first appeared on ControlGlobal.com's Control Talk blog.
Today’s smart digital valve positioners have incredible capability and flexibility as to tuning, performance and diagnostics. Here we look at how to get the most out of these positioners by tuning and by making sure the valve assembly does not hinder performance and gives the position feedback needed.
First of all, the positioner needs to know the actual position of the internal flow element (e.g., plug, ball, or disk). “High Performance” control valves often have the lowest cost and least leakage and often a straightforward compliance with the piping specification for isolation valves. The appearance of a win-win situation is the root cause of poor performance that often cannot be fixed by even the best valve positioner. The feedback measurement is often on the actuator shaft. Since “High Performance” valves tend to be rotary valves, there is backlash and consequently deadband in the linkages and connections that translate actuator movement into valve stem movement. Then due to the high sealing friction particularly near the closed position from the plug, ball or disk rubbing against the seal, there is considerable friction and a poor resolution. In many cases, the valve stem may be moving but the actual plug, ball or disk is not. This stem windup (twisting) may eventually cause the internal flow element to jump to a new position overshooting the desired position. The smart valve positioner that is measuring actuator shaft position doesn’t see what is really happening in terms of stem and internal flow element position.
So what you need is a true throttling valve and not an isolation or on-off valve posing as a control valve. A true throttling valve has a diaphragm actuator, splined shaft to stem connections, a stem that is integrally cast with the internal flow element (no stem to element connections), and no rubbing of the element against the seal once it opens. To achieve isolation you then install a cheap low leakage on-off valve in series downstream of the throttling valve. For pH control, the on-off valve should be close to the injection point to reduce reagent delivery delay upon opening.
Diaphragm actuators are now available with much higher actuation pressures enabling their use on larger valves and higher pressure drops. If you still need to go to a piston actuator, the one with the least sliding friction giving the best resolution (best sensitivity) is the best choice (given the reliability is good). Again, the position shaft connections need to be splined (keylock and pinned connections have a surprising amount of play causing backlash and shaft windup). The positioner feedback mechanism must be properly adjusted to give as accurate an indication of position as possible.
Not all smart positioners have a good sensitivity and sufficient air flow capacity. Spool type positioners and low air consumption positioners will cause long response times for small and large changes in signal, respectively. Most tests done for step changes in signal of 2% to 10% don’t reveal a problem. High performance valves will be lying to even the best positioners, making diagnostics and supposed step response capabilities invalid. For faster response on fast loops, volume boosters should not replace positioners but be used on the output of the positioner with its bypass cracked open for stability.
Given that you have a good actuator, throttling valve and positioner you are still not home free until you tune the positioner. We know from PID control, that a loop’s performance is only as good as the PID tuning. In fact the Integrated Absolute Error and Peak Error are functions of the tuning settings. A loop with great valve, process and sensor dynamics will perform as bad as a loop with poor dynamics if the controller is poorly tuned.
Positioners have traditionally been high gain proportional only controllers. If a high gain sensitive pneumatic relay is used in the positioner, position control can be quite tight since the offset from setpoint for a proportional only controller is inversely proportional to the controller gain. The offset is also of little consequence, since the effect is rather minor and short term with the process controller correcting the offset. What the process controller needs is an immediate fast total response. There are much larger nonlinearities and offsets that the process controller has to deal with. The original idea of cascade control is to make the inner loop (in this case the positioner) as fast as possible by maximizing inner controller gain, which means going to proportional or proportional plus derivative control. Integral action in the inner loop is hurtful unless we are talking about a secondary flow loop for ratio control or feedforward control. The advent of smarter positioners has led to much more complex control algorithms that include integral action. The use of integral action may make the valve step response tests look better by the final position more closely matching the signal. Not realized is that the positioner gain has to be reduced and that integral action in the positioner increases the instances of limit cycles. In fact, with the process controller in manual (positioner signal constant) a limit cycle will develop from stiction in the valve unless an integral deadband is set. Also, the increase in the number of integrators in the control system means that the process controller with integral action will develop a limit cycle from backlash since there are now two integrators. So here we have the common situation where an attempt to make appearances look better we have created a problem due to lack of a fundamental understanding. Many positioners now come with the integral action turned on as a default.
The solution is to omit integral action and use the most aggressive gain setting. For the Digital Valve Controller, this means going to travel control instead of pressure control. Overshoot of the setpoint is not a problem as long as the oscillation quickly settles out. Some overshoot helps in terms of working through deadband and resolution limits faster and increasing the size of the step seen by the positioner algorithm with its sensitivity limits. In fact, a lead/lag with lead time greater than lag time on the input signal is sometimes used to accomplish the same result. You should not get hung up on the exact change in position for a change in signal. For small signal changes, the linearity due to resolution limits is going to look bad because the resolution as a fraction of a small step is large. The really important thing is that the position changes quickly and the 86% response time is fast. Positioners with poor sensitivity and tuning may have a response time that is an order of magnitude larger than possible. What the process loop really wants is the manipulated variable to respond quickly to its demands and corrections. Also, for backlash, limit cycles can be eliminated or at least the amplitude reduced by a higher gain.
For much more on how to get good valve rangeability and a sensitive and fast response see the 5/01/2016 Control Talk Blog “Sizing up Valve Sizing Opportunities” and the Control May 2016 feature article “How to specify valves and positioners that don’t compromise control”.
Make sure this is not just talk. Put yourself into a position to get the most out of valve positioners.
The post, Uncommon Knowledge for Achieving Best Feedforward and Ratio Control, first appeared on ControlGlobal.com's Control Talk blog.
Here are key relatively straightforward principles and practices not commonly known that can make a world of difference in the performance of feedforward and ratio control systems. Also, guidance is offered for the setup and automatic identification of the parameters needed by the use of software designed to tune PID feedback controllers. Even though this capability may be well within the existing functionality of the software, often the instructions needed are insufficient or missing in action.
The desired and undesired changes in process outputs can be largely be traced back to changes in the flow rate of process or utility streams. These changes are relatively fast and can propagate downstream as changes in stream temperature and composition besides flow. In comparison, the changes in raw material composition and temperature in storage tanks are relatively slow and small.
Feedforward and ratio control seek to maintain the stream flow and consequently the stream composition and temperature at the values that would be seen in an accurate Process Flow Diagram (PFD). Unrealized is that most PFDs and the feedforward gains and ratios used could be updated online as the flow rates are corrected by the feedback controller by better visibility of the correction. Also, a PI controller similar to a valve position controller (VPC) can be simply set up to automatically adapt the feedforward gain or ratio by seeking to gradually and smoothly minimize the feedback correction.
We must first define some nomenclature. The “leader” flow leads the way and may be a disturbance flow or a manipulated flow that first sets the direction and production rate. Feedforward or ratio control sets the “follower” flow that will seek to stay in a given ratio to the “leader” flow. This simple strategy enables preemptive disturbance rejection to reduce load upsets, coordination for changes in production rate or state (e.g., startup or transition) to reduce unbalances, and decoupling to reduce interaction.
It is most important that the feedforward or ratio correction in a manipulated “follower” flow arrive in the process at the same time at the same point as the “leader” flow. The correction in the “follower” flow must not arrive too soon causing inverse response or too late creating a secondary disturbance. The criterion for what is too soon and too late is tight and loose, respectively. Arriving just 10% of the loop deadtime too soon is significantly disruptive whereas arriving 100% of the deadtime too late may just require a reduction in the ratio or feedforward gain.
Overcorrection can cause the opposite response and is much worse than under correction. Since we never know the feedforward gain or ratio exactly due to uncertainties in measurements or process stream conditions and our feedforward timing is never right, the full theoretical feedforward gain or ratio is not used in the setting. Often the actual value used is less than 80% of the theoretical value.
To get the timing of the “follower” flow better, dynamic compensation of the feedforward or ratio control is done by insertion of a Deadtime block and Lead/Lag block in the “follower” flow required for feedforward or ratio control. The feedforward or ratio control block deadtime is equal to deadtime for a change in “leader” or load flow minus the loop deadtime for a change in “follower” flow resulting in a change in the controlled process variable (PV). If the “leader” deadtime is significantly less than the “follower” deadtime, you are out of luck unless you can insert a deadtime in the setpoint changes to the “leader” flow. In this case, the feedforward or ratio control deadtime is the total deadtime from a “leader” setpoint change minus the deadtime from a “follower” flow change.
The Lead time in the Lead/Lag block is set equal to the lag from a change in the “follower” flow in terms of affecting the PV (Lead time setting = “follower” lag). The Lag time in the Lead/Lag block is set equal to the lag from a change in the “leader” flow resulting in a change in the PV (Lag time setting = “leader” lag). The Lag time in the Lead/Lag block should be at least 1/5 the Lead time to prevent an erratic and noisy “follower” flow. The feedforward gain or ratio is set to make the contribution of the “follower” flow cancel out the “leader” flow. The feedforward scale of a feedforward summer for direct manipulation of a final control element should be set to match to the final control element capacity with signal characterization added to reduce nonlinearity.
You can use software for tuning a PID controller to tell you what is the “follower” and “leader” deadtime, lag time and process gain. The process gains should be expressed in terms of the change in the PV divided by the change in “leader” flow and “follower” flow realizing that the feedforward scale or secondary controller takes care of the final control element gain. If there is no “leader” flow controller, a dummy controller can be set up as if there was such a flow controller and the tuning software used. Of course, changes in the "leader" flow must be made for identification of dynamics.
Due to the lack of guidance on how to identify the feedforward gain and dynamic compensation for PID control, Model Predictive Control (MPC) may be used because it automatically does this during the test and identification process that is largely automated. If there are multiple feedforwards, complex or compound dynamics due to recycle streams or heat integration or full decoupling is needed, MPC is a much better choice unless you have exceptional PID expertise. However, when ratio control is needed for visibility and accessibility, many of the same practices sited here for PID control are also needed for an MPC solution. Furthermore, the temptation to use a controlled variable that is a ratio of two flows only works over a narrow flow range because of the extreme nonlinearity created by division of one flow by another flow. As the devisor significantly decreases, the process gain greatly increases.
For fast processes where the process deadtime or secondary lag is small or are nearly the same for both the “leader” or “follower”, the dynamic compensation largely depends upon dynamic response of the secondary flow or speed loop and final control element (control valve or variable frequency drive). While lags and delays in the process response of the PV may be similar or negligible for both “follower” and “leader” flows, the “follower” flow correction must work through the secondary flow or speed loop and a final control element with uncertain and non-ideal dynamics. Thus, poor secondary controller tuning and/or control valve or VFD nonlinearity, slow response time, deadband and resolution limits deteriorates feedforward and ratio control besides feedback control.
If there is no secondary flow loop and the correction is done by direct manipulation of the final control element, a feedforward summer is best despite what might be said in the literature about whether the intercept or slope of feedforward changes. These plots do not take into account that bias errors exist in valve positions and leader flow measurements and that uncertainty and nonlinearity is more robustly handled by a bias correction as seen in model predictive control and neural networks. A feedforward multiplier introduces a nonlinearity that is problematic if there is no secondary flow or speed loop. Most gas and steam pressure systems do not have a secondary loop and therefore use feedforward summers.
For steam header pressure controller a feedforward summer can be used to provide decoupling and minimize upsets to upper headers from lower headers. The change in steam flow through a letdown valve to the respective header is a feedforward input to the letdown valve of a higher pressure header. If the valves are linear and the feedforward scale is set based on the valve flow capacity, the theoretical feedforward gain is one. If the letdown valves have similar dynamics, no dynamic compensation of this feedforward signal for decoupling is needed. This simple half decoupling can be carried up through multiple headers so that changes in the lowest pressure header steam use or generation does not upset any of the higher pressure headers. Changes in steam use and generation are also a feedforward signal to the letdown valve for the respective header. Again, linear valves with a feedforward scale set based on valve flow capacity makes the theoretical feedforward gain one if consistent steam flow units are used. If the steam suppliers entering the header and the steam users exiting the header are within a few thousand feet of each other, the process deadtime and secondary lag is the same for both changes in supplier or user flow and letdown flow. In this case, the dynamic compensation needed is simply based on the letdown valve response time. For a letdown valve 86% response time (T86) of 4 seconds, a Lead time setting of 4 seconds and a Lag time setting of 1 second is a good starting point to help get through the resolution limit for the integrating or dominant lag process response of steam header pressure.
If there is a secondary flow or speed loop, a ratio control block and bias/gain block is used to provide operator visibility and accessibility important for changes in process conditions and states. This is a significant consideration often lacking in the operator interface. For level, composition, pH and temperature control of back mixed volumes (e.g. liquid vessels with recirculation and/or agitation and columns), the bias is corrected by feedback control effectively being a feedforward summer. For composition, quality, temperature and pH control by manipulation of feeds for plug flow volumes (e.g., fluidized catalyst bed reactors, static mixers, clarifiers, inline polymer reactors, conveyors, settling tanks, extruders, and sheet lines), the ratio is corrected by feedback control effectively being a feedforward multiplier. Whether the ratio or bias is corrected is determined by whether a significant process time constant is present due to back mixing or missing due to plug flow. The manipulation of heating or cooling for control may necessitate the correction of the bias due to the introduction of a heat transfer lag.
Whatever is not corrected by feedback control can be adapted by the setup of a simple PI controller with mostly integral action and effectively directional move suppression that seeks to make the feedback correction negligible (bias and ratio correction factor approach zero and one, respectively).
For ratio control of reactants, dynamic compensation is not normally used. However, unbalances in the stoichiometric ratio of reactants can accumulate in back mixed volumes with prolonged delayed effects and promote more immediate and abrupt effects in plug flow volumes for changes in production rate, from correction by analyzers and changes in state. The more commonly cited solution is to tune all the reactant flow loops to have the same closed loop time constant of the reactant loop with the poorest and slowest response. However, this degrades the ability of the faster and better loops to deal with disturbances such as pressure, temperature or composition changes to mass flow loops particularly when coriolis meters are used. The better solution is to tune all the reactant flow controllers for the fastest disturbance rejection and use an equal setpoint filter on the reactant flow loops to make the “follower” loops respond in unison via ratio control with changes in the “leader” loop setpoint.
For ratio control of column process flows for temperature or composition control, the relative location that the process flow enters the column has a great effect on the Lead/Lag settings.
For level control, the dynamic compensation of the ratio control is usually not needed if a process flow is manipulated that directly enters or exits the volume. However, if level control is achieved by the manipulation of heat transfer, overhead condensation or by the addition of a stream to another stream or volume introducing delays and lags in the “follower” response, dynamic compensation is necessary.
Final control element and flow measurement rangeability can be a severe limiting factor. Oversized meters and control valves and improper installation make the actual rangeability much less than the stated rangeability by the supplier. Sliding stem low friction packing control valves with diaphragm actuators and digital valve controllers and VFDs with pulse width modulated inverters and speed to torque control and low static head, and magmeters and coriolis flow meters offer the best rangeability. If there is good final control element rangeability, insufficient flow meter rangeability can be dealt with by an inferential flow measurement computed from valve position or VFD speed at low flows. There must be a smooth transition from measured to inferred flow. The feedforward gain or ratio factor should be reduced since the inferential measurement is less accurate. Note the VFD rangeability and linearity dramatically deteriorates as the static head increases and approaches the system pressure drop from friction.
For a better understanding of this and much more see a preview of a future ISA Mentor Program WebEx presentation “ ISA-Mentor-Program-WebEx-Feedforward-and-Ratio-Control.pdf ” and of course, buy one of my books.
I have found a book to update that will showcase a great mind. Now I just need to find the great mind.
The post, Keys to Successful Process Control Technologies, first appeared on ControlGlobal.com's Control Talk blog.
Why do many process control technologies fail to make prime time being relegated to special applications that are few and far between? Here I give what I see as the keys to a technology being successful and widely used in plant applications. How and to what degree each of the major technologies achieves these keys is discussed along with what is left on the table.
Keys to Success
(1) Technology addresses the actual dynamics of industrial process applications (open loop steady state or integrating process gain, open loop time constant(s) and deadtime without which I could retire per my last Control Talk Blog)
(2) Technology successfully makes automatic adjustments to process without operator intervention
(3) Technology achieves benefits in terms of an increase in process efficiency and/or capacity
(4) Tools exist to identify the adjustments needed to achieve the best process performance
(5) Technology can be adjusted to perform well for different production rates and process conditions
(6) Operator can understand what the technology is doing to the process
(7) Application can be readily monitored for performance and problems easily identified and fixed
(8) Technology can be applied, maintained and improved by the average process control engineer
(9) Technology is widely applicable
(10) Technology does not distract operator and does not create false information
(11) Technology identifies unsuspected relationships
(12) Technology identifies unsuspected causes and effects
The PID is widely applicable and can be applied by the average process control engineer. Studies have shown that for control of a single process variable, the PID is near optimal in terms of rejecting load disturbances as documented in a study by Bohl and McAvoy in a landmark paper in 1976. This has not stopped thousands of papers and millions of hours being spent on trying to invent a controller to replace the PID for single loop control. If the process has a runaway response (e.g., highly exothermic reactor temperature) or requires very fast execution (e.g., < 1 sec), the PID is the only safe solution.
The PID can be readily applied but it is amazing how many of these controllers are not properly tuned. Part of this is due to the lack of really understanding the modes and options. To confuse users, the integral mode gives the type of response by humans in terms of never being satisfied accentuated by digital displays to excessive resolution and having no anticipation or understanding of the trajectory. The result is most loops have an order of magnitude or more too much integral action (too small of a reset time) except for deadtime dominant loops as discussed in my last Control Talk Blog. Confusing the situation are over a 100 tuning rules with the advocates having a closed mind to the relevance of other tuning rules as discussed in my Control Global Whitepaper “So Many Tuning Rules, So Little Time”. While the average user can apply almost any set of tuning rules to get good performance, a consultant is often useful to get the best performance. Tuning software can automatically identify the actual dynamics for feedback control and provide recommended settings that can be scheduled online for different production rates and process conditions. PID output for startup, batch operation and abnormal conditions can be scheduled automatically by the use of the Track or the Remote Output modes. Cascade control can be effectively used to compensate for fast secondary disturbances and nonlinearities. Advanced Regulatory Control techniques such as feedforward, override and valve position control can quickly implemented to increase process capacity or efficiency. When override control is used, the selected controller is evident to the operator. The external reset feedback feature must be used so that the unselected PID offset in output from the selected PID output is the unselected PID error multiplied by its gain providing predictability as to when an unselected PID will take over control. An enhanced PID can be used to enable the PID to use at-line and even off-line analyzer results for composition control with sensitivity of tuning settings to cycle time eliminated for self-regulating process responses and minimized for integrating process responses as described in the 7/06/2015 Control Talk Blog “Batch and Continuous Control with At-Line and Offline Analyzers Tips”.
While the PID can be configured and tuned to provide automatic control for all sorts of situations resulting in its use in over 99% of the feedback control systems, there is a lack of a straightforward general approach to increase its effective. While my previous blogs this past year supporting my YouTube Videos on PID Options and Solutions give a useful perspective and a lot details, what is needed is for someone to step back and give a clear step by step approach that allows for different objectives and tuning methods. By the way, these videos are now posted as part of a very practical and useful playlist of ISA Mentor Program Webinars that should be significant value to every automation engineer.
There are limitations to PID control that move users to consider Model Predictive Control (MPC). Most tuning software does not effectively and automatically identify the feedforward dynamic compensation needed by a PID. Also, more than half decoupling for a PID is confusing and too challenging for the average user. Compound or complex responses where the initial response is different than the later response due to recycle or competing effects, tuning of the PID is difficult. Additionally the simultaneous predictive honoring of the multiple constraints is not within the normal capabilities of the PID. These situations all lead us to MPC to achieve greater process performance, especially since the MPC software may be integrated into the DCS and implemented by a simple configuration making its use less expensive, faster and easier. Also, let’s face it, doing an article or presentation on a PID application is not going to get you as much recognition as doing one on an MPC application.
Model Predictive Control
MPC can be potentially used as a replacement for any PID control loop in continuous operation if the PID gain needed is less than ten, the execution rate does not need to be faster than 1 second and derivative action is not essential for safe tight control for unmeasured disturbances. The temperature control of many liquid reactors falls into this category. For highly exothermic reactors, the positive feedback response can create a dangerous runaway condition with a point of no return. While technically processes should be designed so this cannot happen, the acceleration in the response from a batch polymerization reaction can exceed coolant capabilities leading to relief valves popping and blowing over the reactants to a flare system. I have been in a control room when this occurred.
MPC can be extended to fed-batch if models are switched as the batch progresses. The control of a temperature profile or product composition profile can be done by the translation of the controlled variable from temperature or concentration to rate of change of temperature and concentration, which is the slope of the batch profile. This creates a pseudo steady state and the ability to readily make changes in the controlled variable in both directions.
The abilities to handle nonlinearities and provide operator understanding are more challenging for MPC than PID control. Consultants and proficient users can address these needs by switching in different models for different production rates and plant conditions and adding graphics that enable operator to see future trajectories and understand the relative contribution of each controlled variable, disturbance variable and constraint variable. For multivariable applications proficiency in knowing and improving matrix condition number and using the optimizer (e.g., linear program) is useful.
My experience personally and after doing 14 years of Control Talk Columns with industry experts, is that the most successful and extensively used technologies are PID and MPC because they address Keys 1-10 with a track record for Keys 1 and 2 far exceeding the remaining technologies. Where PID and MPC fall short is in the ability to find unsuspected relationships and cause and effects particularly when there are a large number of variables at play. The variables for PID and MPC are chosen and classified as to controlled, manipulated, disturbance, constraint and optimization variables based on knowledge of person applying the technology. Also, regimented testing is used to identify the relationships. On the plus side, PID and MPC identify all of the necessary dynamics, whereas the other technologies leave a lot to the imagination. Thus, the PID and MPC solutions are narrowed dramatically compared to the remaining technologies to be discussed when looking at orders of magnitude number of variables with very little preconceptions and regimented testing as proposed in the Industrial Internet of Things (IIoT).
Multivariate Statistical Process Control (Data Analytics)
Data analytics can find relationships between process inputs and process outputs in very large sets of data. Data analytics eliminates relationships between process inputs (cross correlations) and reduces the number of process inputs by the use of principal components constructed that are orthogonal and thus independent of each other to each other in a plot of a process output versus principle components. For two principal components, this is readily seen as a X, Y and Z plot with each axis at a 90 degree angle to the each other axis. The X and Y axis cover the range of values principal components and the Z axis is the process output. The user can drill down into each principal component to see the contribution of each process input. The use of graphics to show this can greatly increase operator understanding. Data analytics excels at identifying unsuspected relationships. For process conditions outside of the data range used in developing the empirical models, linear extrapolation helps prevent bizarre extraneous predictions. Also, there are no humps or bumps that cause a local reversal of process gain and buzzing.
Batch data analytics does not need to identify the process dynamics because all of the process inputs are focused on a process output at a particular part of the batch cycle (e.g., endpoint). This is incredibly liberating. The piecewise linear fit to the batch profile enables batch data analytics to deal with the nonlinearity of the batch response. The results can be used to make mid batch corrections. Potentially batch data analytics can address most of the keys to success listed.
The application of data analytics to continuous processes requires the synchronization of the process inputs with the process outputs predicted. This can be problematic because there are many delays and time constants of instrumentation and control valves and unit operations that are unknown. A deadtime is identified to be applied to each process input to deal with these dynamics. As you can imagine, this may work on a single unit operation with negligible automation system dynamics and no recycle but is a big challenge for most plant applications
A word of caution for data analytics and the next technology discussed Artificial Neural Networks (ANN). What are identified are correlations and not causes and effects. The review by people who understand the process (e.g., process and research engineers) and the possible use of first principle simulations are needed to validate causes and effects.
Both data analytics and ANN have been much more successful in predicting human responses that tend to be much less deterministic and much more situational. There is a possibly a greater future in terms of predicting operator responses than process responses when it comes to continuous unit operations.
ANN offer predictive behavior that addresses nonlinearities much more effectively than data analytics particularly for continuous processes. However, ANN does not necessarily use principle components to prevent cross correlations and to reduce the number of inputs. Consequently, an ANN may have order(s) of magnitude greater number of process inputs than data analytics without the ability to determine numerically the contribution of each input beyond a relative sensitivity analysis. Also, the response may have humps or bumps conditions within the test range and strange results outside of test data range.
Special techniques can be used to address these concerns. The future of ANN to me is more as a supplement rather than a replacement to other technologies. There has been considerable success in the use of a combination of an ANN and MPC. Nothing to me precludes the use Principal Component Analysis (PCA) extensively employed by data analytics to eliminate cross correlations and to dramatically reduce the number of process inputs to ANN. The main problem is that the ANN and data analytics camps see their technology as the best taking an adversarial view rather than trying to get the best out of both.
Fuzzy Logic Control (FLC)
FLC has been successfully used where a process model cannot be obtained (e.g., mining industry). FLC automates the best of the relative operator responses. Given a model, it has been shown that a PID gives as good or better performance for a single loop if all of the PID options are used as needed. I expect the same is true in terms of MPC being a better solution for multivariable control if models can be identified. FLC is undesirable from the standpoint of tuning, understanding and technical support.
I implemented a successful FLC for a neutralization system. The plant was afraid to touch the FLC because no one understood it. Decades later, I showed how an MPC could provide better optimization with a more continuously improvable and maintainable system as documented in the November 2007 Control article “Virtual Control of Real pH”.
Seems like a great idea and maybe there is a future for it but the company I spent my career with had 10 people working on expert systems for ten years with only some limited success. Often the expert systems complained too much and sometimes about the wrong things. Operators saw them as a distraction. False alarms totally subjugated confidence in them. The expert systems fell into disuse and were turned off within a few years after the developer left the area. This was back in the 1990s. My hope is that technology has advanced and addressed these issues. I strongly suggest one check how well they meet Keys 1-12.
The expert system we used also had fuzzy logic. I remember entering a lot of rules and then wondering how did they play together and what was the order of execution. It was very easy maybe too easy to add rules but extremely difficult to analyze the consequences for all the possible conditions.
Expert systems do not have dynamics and are thus not used to predict a process variable but rather an abnormal condition and a possible steady state correction. I think there is a place for them in helping provide better diagnostics based on problems identified by other technologies.
As a closing note, the one New Year’s Resolution that I can guarantee I am going to keep is “I will not keep all my New Year’s Resolutions.” Happy New Year!
The post What Are Best Cutover Strategies for Upgrading an Industrial Plant? first appeared on the ISA Interchange blog site.
ISA Mentor Program, authored by Greg McMillan, industry consultant, author of numerous process control books, 2010 ISA Life Achievement Award recipient and retired Senior Fellow from Solutia Inc (now Eastman Chemicals). Greg will be posting questions and responses from the ISA Mentor Program, with contributions from program participants.
Here we see how a general and open-ended question can lead to a very insightful, useful, and at times, humorous discussion on a critical phase of migration projects where automation is transferred from the old to a new system with many safety and performance implications. The question was posed by Marsha Wisely, a relatively new protégée. The answer is provided by Hunter Vegas, ISA Mentor Program co-founder, who has the most extensive practical experience of the program resources in automation project design and execution.
I am looking for general information on the different types of cutovers and what the impact of each type may be. If this topic is too big, I would appreciate even some quality documentation on the topic such that I can read up and maybe come back with more specific questions.
It IS a pretty open-ended question, and the topic is rather huge. Books by practitioners are more useful because they reflect actual plant experience. Greg adds that academics who have partnered with industry have expanded our understanding and principles through theory supporting practice. However, you get good at cutovers by doing them – lots of them – and learning (sometimes the hard way) what works and what does not. Older folks in automation have lots of gray hair, or none at all, and plant startups are most likely to blame. There are several universal constants that you must know about startups:You get good at cutovers by doing them – lots of them – and learning (sometimes the hard way) what works and what does not. Older folks in automation have lots of gray hair, or none at all, and plant startups are most likely to blame. There are several universal constants that you must know about startups:
a) Instrumentation and automation are almost always the last thing to install and check out because you can’t hook up instruments on pipes that don’t exist, and you can’t talk to instruments if there are no wires. Invariably the civil, mechanical, and electrical installations all get delayed on a cutover but the start date does not slide so instrumentation is ALWAYS critical path. You start off with three weeks to install your equipment and that usually gets whittled down to three hours. “You mean you aren’t DONE yet?!?!?!”
b) You can’t run the plant without instrumentation so if your manager keeps hounding you about when you’ll be done, tell him that you absolutely promise that you’ll be done before he starts up.
c) Regardless of who is on what shift, the night shift is almost always half as productive as the day shift.
d) Good, Fast, and Cheap (pick any two).
e) There is no such thing as too much coffee on a startup.
The different “cutover types” isn’t nearly cut and dried or simple. I’ll start by talking about a control system retrofit as opposed to a new plant or a major plant expansion. Usually a retrofit has a couple of different types:
1) Software upgrade – same instruments, same hardware (or maybe a few upgraded computers), and a new revision of software.
2) Hardware upgrade – same field instruments, new hardware and software. (Maybe same vendor, maybe not)
3) Total revamp – upgrade/replace a lot of field instruments and the entire control system.
All of those can be done in various ways that have different impacts on the production plant. Nearly all COULD be done without shutting down the plant. It would be expensive, and it would take a while, but it can certainly be done. I’ve converted three or four plants from pneumatic controllers on panelboards to full-blown distributed control systems (DCS) and never shut them down. Ultimately, the economics of a plant production drives the decision. If they are making a million dollars of product a day, they can afford to pay an extra $200,000 to cutover on the fly. Similarly, if they are barely running eight hours a day, a two- or three-day outage doesn’t matter. Often the cutover is a mix – you do as much work as possible prior to shutting down and then you come down and finish the rest.
That totally depends on the plant, the process, and the timing. When I am looking at a retrofit project, I start by talking to the plant and understanding a few things:
1) Their budget
2) Their schedule
3) The financial justification for the project
4) What their “pain points” are
Based on that, I then start going over the project – looking at panels, mulling my options, determining what I have to do, and how I might go about it. I also investigate what options I have and what is required to change over the system. Is the I/O compatible? Are there interconnect options and do they work? (Just because a company SAYS they offer interconnect options, does NOT mean they actually work as advertised!) Eventually, I pull it all together and hatch a plan, figure what it will cost, and present it to the client. If they can afford it and like it, we are off to the races. If they don’t, then we tweak it accordingly. Maybe they can afford to take a longer outage and save on installation costs. Maybe they are willing to pay more to bring the plant on line faster. Maybe there is a way to jury rig things to get the bulk of the benefit now and do the rest of the cutover later when they have more time.
Well, that’s a start… let’s see what questions this generates.
Thanks for the information! When you have only seen one of something, it’s hard to know what the norm is or what the other options are. Also, I really like your universal constants section! After reading your comments, I started thinking about how the information applied to my experience. Below I have a summary of my experience to give you some background on me, and it inspired some questions as well.
The cutover and startup both took a good deal longer than previously anticipated. Loop checkouts were taking longer than anticipated, the plant discussed how to improve, but management never prioritized timeliness over safety and quality of work, which I respect and appreciate (I know that isn’t always the case). There was some unexpected equipment issues, namely valves were backwards (fail open when they should be fail closed and vice versa).
Startup was also slower due to some equipment issues – air lines to pneumatic valves got leaks and equipment that was working before startup needed to be repaired before being brought back up. This leads to some frustrations and finger pointing at the software – “The valves aren’t working; they worked before. It must be the software.” Definitely a good lesson in patience. In those cases, we basically did an additional loop check, which quickly found the issue. Shutdowns are in the hands of operations and can be pretty hard on equipment.
This leads to the following additional questions.
1) If you do a hardware and software cutover, are those types of equipment issues more likely to get caught? And are these types of equipment configuration errors common?
2) Are there any common issues caused by shutting down equipment that are good to check before starting back up?
3) In your notes, you mention asking the customer about their “pain points.” Could you provide some examples?
4) My next couple projects are for new plants – one of which I will be leading. Do you have any advice specific to new plants? It looks like your “HOW to cutover section” covers a lot, but are there any additional challenges associated with new plants?
Thanks again for your help! I greatly appreciate your insight into the field since I haven’t had much field experience. My next couple projects, in addition to being new plants, also have a larger equipment and instrumentation component to my assignment, and I am looking forward to that exposure.
Let me try to answer your questions:
1) If you do a hardware and software cutover, are those types of equipment issues more likely to get caught? And are these types of equipment configuration errors common?
If you are doing a hardware/software cutover you have to be fanatical about the details. Picking up failure modes of valves, ranges, square root versus linear, interlocks, etc. is pretty much a given. Only our most experienced people generally do the decoding of the existing system because it is so crucial to success. If you toss it to some inexperienced engineers, you will pay a wicked price on startup and rework.
2) Are there any common issues caused by shutting down equipment that are good to check before starting back up?
One thing I learned very early on was, if it all possible, obtain historical data of the running plant before you shut down and have it available for access during the start up. It is a very common trick for plant personnel to get the automation company to fix instrumentation that hasn’t worked for years. If you can point to the fact that the transmitter has been flat-lined since 2006, it is a pretty easy argument to say that fixing that transmitter is out of scope.
Pain points are what keeps them up at night. Are they struggling with quality? Production? Throughput? Reliability? Does something break and it takes the techs 2 days to figure it out? Is there a particular area in the plant that is always in manual because the controls never work? Is the messaging awful so the batch stops and nobody knows why? Etc.
Often I can fix a lot of those things fairly easy and make them very happy. Or I can offer solutions that increase the work scope some but has very big payback.
New plants can be very painful and difficult for a whole new set of reasons. Specifically:
a) How good is the engineering contractor? The large Architectural & Engineering (A&E) firms can be awful or wonderful; it all depends on what team you get. If you get the “A” team, things will be in pretty good shape. Unfortunately, they might bait you with the “A” team but swap you the “C” team later in the project or you end up with the “C” team due to turnover. Either way, the engineering is just wrong. Wrong pipe, wrong materials, wrong instruments, wrong drawings, wrong…wrong…wrong. And it’s your job to make it work using BIFF principles (Big Improvements for Free) because the money has already been spent.
b) Was anything reviewed by someone other than the engineering contractor? The best option is for the plant personnel to review and approve everything; however, the plant often lacks either the expertise or the time to do that and the engineering contractors are infamous for dropping 1,000 pages on your desk and demanding you approve it overnight or “the project will be delayed and it was your fault.” The second-best option is for a third party to at least look things over and catch the worst stuff. If nobody looks it over, then you better have the “A” team or it will be bad.
c) Does anyone even know how this is supposed to work? Is the plant a copy of an existing plant and the process is well understood (and ideally you have people from that plant helping you) OR is this Serial #1 and the only people who have a vague clue are some lab chemists who had a tiny pilot reactor running somewhere? Obviously, Serial #1 is going to be tricky because nobody knows what they don’t know and nobody has the answer.
d) What is the schedule/budget like? Was it stupid aggressive to start? If so, it will be tough, as nothing ever goes to plan on a new plant and both the budget and schedule are bound to suffer. I have had no-bid projects that I knew were badly estimated because failure was virtually assured, and I was much happier having my competitor get the black eye than me!
Now don’t get me wrong – large projects are executed successfully all the time and with the right team things can go very smoothly. But if the client goes with an unproven, low-bid entity, they often regret the decision. I once was part of a large project that was more than half-way completed but things were going so badly that the client fired the A&E firm mid project and went with a new one. They lost $10 million in engineering but probably saved $20 to $30 million in extra startup costs because the first firm was doing so poorly.
Discussion and answers are provided by Greg McMillan, Hunter Vegas (co-founder of the ISA Mentor Program and project engineering manager at Wunderlich-Malec), Brian Hrankowsky (consultant engineer at a major pharmaceutical company), Michel Ruel (executive director, engineering practice at BBA Inc.), Leah Ruder (process systems automation group manager at the Midwest Engineering Center of Emerson Process Management), Nick Sands (ISA Fellow and Manufacturing Technology Fellow at DuPont) and Bart Propst (Process Control Leader for the Ascend Performance Materials Chocolate Bayou plant).
The post, Deadtime Dominance - Sources, Consequences and Solutions first appeared on ControlGlobal.com's Control Talk blog.
If the total loop deadtime becomes larger than the open loop time constant, you have a deadtime dominant loop. There is a lot of misunderstanding posed by this difficult challenge including a lot of prevalent myths. Here we look at how we get into this situation, what to expect and what to do.
Deadtime is the ultimate limit to loop performance. A controller cannot correct for a disturbance until it starts to see a change in the process variable and can start to change the manipulated variable (e.g., flow). If the deadtime was zero, the controller gain could be set as large as you want and the peak and integrated errors made as small as you want provided there is no noise. The PID gain could theoretically approach infinity and lambda approach zero. Without deadtime, I would be out of job.
Fortunately, deadtime dominant loops where the deadtime is much larger than the open loop time constant, is usually relegated to a few loops for processes involving liquids and gases. Deadtime dominance typically occurs due to transportation delays (deadtime = volume/flow), cycle times of at-line analyzers (deadtime = 1.5 x cycle time, or improper setting of PID execution rates or wireless update rates (dead time = 0.5 execution or update rate time interval). Deadtime dominance due to transportation delay is more common in the mining industry and other processes that are primarily dealing with solids.
Large transportation delays result from conveyors or feeders or extruders where the feed of solids are manipulated at the inlet and the quality or quantity is measured at the outlet (discharge) assuming the senor lag is much less than the transportation delay (i.e., total loop deadtime). The use of a gravimetric feeder can eliminate the delay for feed flow control making a much more precise and responsive loop. A kiln has a temperature sensor lag that creates an open loop time constant that is usually larger than the transportation delay. However, the use of an optical sensor (e.g., pyrometer) for temperature control eliminates this lag possibly making the loop deadtime dominant.
In vessels and pipes where there is no axial mixing, also known as back mixing due to an agitation or recirculation or sparging, there is no mixing up and down in the vessels and mixing forward and backwards in the pipe. We call this plug flow. There may be radial mixing, to make the concentration or temperature in a cross section of the vessel or pipe more uniform or to keep solids dispersed. Radial mixing does not reduce transportation delays whereas axial mixing eliminates plug flow decreasing transportation delays. Any remaining delay from axial mixing depends upon the degree of mixing. For vessels with good geometry and axial agitators, the delay is approximately the turnover time that is the volume divided by the summation of the agitator pumping rate, recirculation rate and feed rate.
Deadtime dominance usually doesn’t occur in temperature loops except for inline systems, such as plug flow liquid or polymer reactors where there is negligible heat transfer lag in the equipment or sensor.
For a deadtime dominant system, the peak error for a step load disturbance approaches the open loop error (error if loop is in manual). Better tuning and feedback control algorithms cannot do much to improve the peak error. There can be a significant improvement in the integrated error by better tuning and feedback algorithms for a load disturbance but the ultimate limit to the integrated error is the peak error multiplied by the deadtime.
The best solution is to eliminate the deadtime that is the source of deadtime dominance by better process, equipment or automation system design. The next best solution is feedforward control given a reliable timely measurement of the disturbance. The feedforward measurement does not need to be extremely accurate. An error of 20% will still result in an 80% improvement in integrated error if the feedforward timing is right. The correct timing cannot be achieved when the correction by feedforward has a deadtime greater than the deadtime in terms of the actual disturbance affecting the process variable that is the controlled variable. There is no compensation for too much deadtime in the feedforward path. A feedforward correction arriving too late must have the feedforward gain reduced. If the feedforward correction arrives after most of the feedback correction has been made, a second disturbance is created making the response worse than no feedforward. Even more disruptive is a feedforward correction that arrives too soon due to deadtime in the feedforward path being less than the deadtime in the disturbance path because this creates inverse response confusing the feedback controller. However, this situation is readily corrected by inserting a deadtime in the feedforward (FF) signal equal to the disturbance path deadtime minus the feedforward path deadtime. A lead lag is used as well in the FF dynamic compensation where the FF lead time is set equal to the lag in the feedforward path and the FF lag is set equal to the lag in the disturbance path.
The solution cited for deadtime dominant loops is often a Smith Predictor deadtime compensator (DTC) or Model Predictive Control (MPC). There are many counterintuitive aspects in these solutions. Not realized is that the improvement by the DTC or MPC is less than for deadtime dominant systems than for lag dominant systems. Much more problematic is that both DTC and MPC are extremely sensitive to a mismatch between the compensator and model deadtime versus the actual total loop deadtime for an overestimate besides an underestimate of the deadtime. The consequences for the DTC and MPC are much greater for an overestimate. For a conventional PID, an overestimate of the deadtime just results in more robustness and slower control. For a DTC and MPC, an overestimate of deadtime by as little as 25% can cause a big increase in integrated error and an erratic response.
A better general solution first advocated by Shinskey and now particularly me is to simply insert a deadtime block in the PID external reset feedback path (PID+TD) with the deadtime updated to be always be less slightly less than the actual total loop deadtime. Turning on external reset feedback (e.g., dynamic reset limit) on and off enables and disables the deadtime compensation. Note that for transportation delays, this means updating the deadtime as the total feed rate or volume changes. This PID+TD implementation does not require the identification of the open loop gain and open loop time constant for inclusion as is required for a DTC or MPC. Please note that the external reset feedback should be the result of a positive feedback implementation of integral action.
There will be no improvement from a deadtime compensator if the PID tuning settings are left the same as they were before the DTC or by a deadtime block in external reset feedback (PID+TD). In fact the performance can be slightly worse for even an accurate deadtime. You need to greatly decrease the PID integral time toward a limit of the execution time plus any error in deadtime. The PID gain should also be increased. The equation for predicting integrated error as a function of PID gain and reset time settings is no longer applicable because it predicts an error less than the ultimate limit that is not possible. The integrated error cannot be less than the peak error multiplied by the deadtime. The ultimate limit is still present because we are not making deadtime disappear.
If the deadtime is due to analyzer cycle time or wireless update rate, we can use an enhanced PID (e.g., PIDPlus) to effectively prevent the PID from responding between updates. If the open loop response is deadtime dominant mostly due to the analyzer or wireless device, the effect of a new error upon update results in a correction proportional to the PID gain multiplied by the open loop gain and open loop error. If the PID gain is set equal to the inverse of the open loop gain, the correction is perfect and takes care of the step disturbance in a single execution after an update in the PID process variable. The integral time should again be set to be very small and the positive feedback implementation of integral action must be used with external reset feedback enabled. The enhanced PID greatly simplifies tuning besides putting the integrated error close to its ultimate limit. Note that you do not see the true error that can have started at any time in between updates but only see the error measured after the update.
For more on the sensitivity to both under and over estimates of the total loop deadtime and open loop time constant, see the ISA books “Models Unleashed” pages 56-70 for MPC and “Good Tuning : A Pocket Guide 4th Edition” pages 118-122 for DTC. For more on the enhanced PID, see the July-Aug 2010 InTech article “Wireless: Overcoming challenges in PID control & analyzer applications” and the July 6, 2015 Control Talk Blog “Batch and Continuous Control with At-Line and Offline Analyzers Tips”.
If you can make the deadtime zero, let me know so I can retire.
Here we finish up the extensive presentation of how to get the most out of your PID. We start with a look at the contribution of each mode and show how to estimate performance metrics from tuning settings and how excessive integral action and insufficient proportional action create oscillations. The key role of dead time, external reset feedback and different PID Forms and Structures is given.
The following key points support and augment the PID Options and Solutions - Part 2 and PID Options and Solutions - Part 3 recordings for the ISA Mentor Program.
2-1: To see the effect of each mode, a setpoint change can be made to a PID with a Structure of “PID on Error” (no setpoint filter or lead-lag) and freezing the process variable PV so that there is no feedback and so what is seen is the PID response solely due to the setpoint change. The same results could be observed by stepping the PV via the Analog Input block Simulate input keeping the SP constant.
2-2: The proportional mode provides a step change in output proportional to the step change in error. The derivative mode with a built in filter with a time constant that is about 1/8 the rate time provides a kick that decays out to zero. The integral mode ramps the output and at the transpired time equal to the reset time the integral mode contribution to the output will equal the output step from the proportional mode. Hence integral time settings are in terms of seconds per repeat or minutes per repeat where “per repeat” means the integral mode has repeated the contribution from the proportional mode. Often the term “per repeat” is dropped. Tuning settings that are the inverse of the integral time, usually include the “repeat” term, such as “repeats per second” and “repeats per minute”.
2-3: The integral mode will continue to increase its contribution to the PID output in the direction to cross the setpoint (SP) as the process variable approaches the SP. It only reverses direction after the PV crosses the SP, too late considering all processes have deadtime. This causes overshoot and hunting. Integral action is never satisfied and is always ramping the output since the error is never exactly zero.
2-4: The proportional mode will decrease its contribution to the PID output that is in the direction of driving the PV to cross SP, effectively working to stop the PV as the PV gets closer to the SP. This anticipation helps prevent the PV from crossing the SP. Derivative action is providing a contribution to the output that is proportional to slope of the approach in a direction to help stop the PV.
2-5: Most loops have too much integral action and not enough proportional action resulting in a SP overshoot to achieve a new setpoint or return to an existing setpoint in the recovery from a load upset. If there is too much proportional action relative to integral action, the approach to setpoint will falter (PV trend momentarily flattens out and then resumes the approach). This is a rate case but can occur for deadtime dominant processes that inherently need more integral and less proportional action.
2-6: Humans looking at digital displays of PV and SP are prone to advocate and even manually do what the integral mode does. Looking at the reactor temperature loop example faceplate or numerical values of PV and SP, humans expect for a PV value below SP that the steam valve should be open whereas a trend chart with an intelligent time scale would show the cooling water valve should be open based on the PV slope. See the 6/28/2012 Control Talk Blog “Future PV Values are the Future” for how to easily provide a PV value one deadtime into the future to help operators and engineers gain some anticipation and recognition that processes cannot stop on a dime.
2-7: The ultimate limit to the error from a step load upset to the process input for a perfectly tuned PID would approximate two right triangles placed back to back where each base is the dead time. The amplitude of the triangles that is the peak error is proportional to the ratio of the dead time to the dead time plus time constant. The area of the triangles that is the integrated error is proportional to this peak error multiplied by the dead time and thus the dead time squared. Dead time can come from the valve, process, piping, equipment, sensor, transmitter, PID execution time and filter time. Decreasing the dead time enables decreasing lambda and hence more aggressive PID tuning. A lambda equal to 3 dead times is typical unless there is significant nonlinearity or unknowns in the control loop response dynamics requiring a larger lambda. This lambda can tolerate an increase in total loop dead time or open loop gain of up to 300% without starting a significant oscillation. This lambda provides a gain margin of about 6 and a phase margin of about 76 degrees for self-regulating processes.
2-8: The practical limit to the error for a step load upset depends upon the PID tuning. Decreasing the PID gain affects both the peak error and integrated error whereas increasing the reset time mostly affects the integrated error. The performance for a load upset can be easily tested by momentarily putting the PID in manual, making a step change in the PID output and immediately putting the PID back in automatic. The open loop error can be seen for self-regulating processes by leaving the PID in manual.
2-9: The ultimate limit to the rise time (time to reach setpoint) for a setpoint change is about 2 dead times and is achieved by a very large PID gain (lambda slightly less than the dead time) or setpoint feedforward if output limits do not restrict the change in output. Usually, the prevention of overshoot is more important than minimizing the rise time. The tracking of a PID output that is the output limit and then using a future PV value to determine when to put the output at the final resting value and then releasing the PID from the track mode after one dead time can minimize the rise time to less than 2 dead times with no overshoot as explained in the May 2006 Control article “Full Throttle Batch and Startup Response”.
2-10: The use of a setpoint filter equal to the reset time enables the use of load disturbance tuning without causing much overshoot for a setpoint change. The use of a lead time equal to about ¼ the setpoint filter time will reduce the rise time but possibly create some overshoot. Alternatively, a Two Degrees of Freedom (2DOF) PID structure can be used with beta and gamma equal to about 0.25.
2-11: Near-integrating, true integrating and runaway processes will develop oscillations from a PID gain too small that are greater than the amplitude and period of oscillations for a PID gain too high unless the reset time is increased. The proportional mode provides the negative feedback action missing in these processes that do not have sufficient self-regulation.
2-12: The amplitude and period of oscillations from backlash increases as the PID gain decreases similar to what occurs for violation of the low PID gain limit for processes without sufficient self-regulation.
3-1: Lambda tuning prevents violation of the low PID gain limit by keeping the product of the PID gain and reset time greater than the inverse of the equivalent integrating process gain that is critical for near-integrating, true integrating and runaway processes. Since the permissible PID gain is much higher than users are accustomed to (e.g., 80 for batch reactor temperature control), users tend to use a much lower PID gain not realizing they need to proportionally increase the reset time. Often the reset time needs to be increased by two orders of magnitude for these processes. Not realized is that external reset feedback can enable the use of high PID gains without upsetting other loops as noted in item 3-3.
3-2: The maximum desirable level error can be achieved by estimating the corresponding arrest time. For crystallizer, evaporator and reactor composition control and distillation overhead receiver reflux control, the error is minimized. For most other volumes, the error is maximized (the absorption of variability in volume is maximized) so that feed changes to downstream equipment are minimized.
3-3: The power of external reset feedback is incredible. I find more and more potential applications and benefits including directional move suppression (fast opening and slow closing surge valves, gradual smooth optimization and fast getaway for abnormal conditions, prevention of unnecessary crossings of the split range point), prevention of oscillations from deadband or a slow secondary loop or control valve, prevention of upsets to other loops from aggressive tuning settings, and proper positioning of the output of unselected PID controllers in override control (difference in outputs between selected and unselected controller is the unselected PID gain times its error so that it takes over when its PV hits its setpoint).
3-4: The enhanced PID can tolerate variable and large analyzer cycle times. For self-regulating processes where the cycle time is much greater than the 63% response time, the PID gain can be increased to be almost the inverse of the open loop gain. External reset feedback is turned on in the enhanced PID.
3-5: PID structure options help the PID do well in many diverse applications.
3-6: The positive feedback implementation of integral action is essential for external reset feedback. Most DCS systems think they have this capability but don’t. Unless the integral action is achieved by a filter whose input is the PID output or feedback of the manipulated variable and whose filter time is the reset time with the filter output added to the output from the proportional mode creating positive feedback, the PID does not really have external reset feedback. For more info see the May 2006 article by Greg Shinskey “The power of external-reset feedback”.
This is “no repeat” day. I will not repeat a Control Talk Blog today. I repeat that I will not repeat a Control Talk Blog today.
The post Why Do Most Vessel Control Loops Need the Reset Time Increased by Two or More Orders of Magnitude? first appeared on the ISA Interchange blog site.
The proportional, integral, derivative (PID) controller is the common key component of all control loops. Basic control systems depend upon the PID to translate the measurement signals into set points of secondary loop controllers, digital valve controllers, and speed controllers for variable frequency drives. The success of advanced control, such as model predictive control, depends upon the basic control system foundation and hence the PID.
Elmer Sperry developed the first example of the PID in 1911, and Nicolas Minorsky published the first theoretical analysis in 1922. Ziegler and Nichols published papers on the ultimate oscillation method and reaction curve method for controller tuning in 1942 and 1943. While the parameters chosen as factors in the tuning settings provided overly aggressive control, the basic premise of an ultimate gain and ultimate period is essential to the fundamental understanding of the limits of stability. The identification of the slope in the reaction curve method is a key to the use of the near-integrator concept that we will find here to be critical for most composition, temperature, and pH loops to improve tuning settings and dramatically shorten test times.
Bohl and McAvoy published a paper in 1976 that showed the PID can provide nearly optimal control for unmeasured load disturbances. Shinskey wrote many books detailing the knowledge of process dynamics and relationships essential to the best application of PID control. Shinskey developed the original equation for integrated error from disturbances as a function of tuning settings, as detailed in the January/February 2012 InTech article “PID tuning rules.” Shinskey also published a book dedicated to PID controllers that showed the simple addition of a dead time block in the external reset feedback path could further enhance the PID performance by dead time compensation.
Internal model control (IMC) and lambda tuning rules were developed based on pole and zero cancellation to provide a good response to set points and disturbances at the process outlet. However, most of the improvement in set point response could have been achieved by a set point lead-lag or PID structure. Also, these tuning rules do not perform well for the more common case of disturbances on the process input (load upsets), particularly for lag dominant processes. Skogestadt developed significant improvements to IMC tuning rules. Bialkowski showed that always using lambda rather than lambda factors, relating lambda to dead time, and treating lag dominant processes as near-integrators enable the PID to provide good nonoscillatory control for load upsets besides dealing with the many different difficulties and objectives for which lambda tuning was originally designed. Not realized is that most methods converge to the same basic expression for the PID gain and reset time when the objective is load disturbance rejection and that a tuning parameter that is the closed loop time constant or arrest time is set relative to dead time. Also not recognized is how PID features, such as structure, external reset feedback, enhanced PID for analyzer and wireless, simple calculation of a future value, valve position controller, and “full throttle” set point response, can increase process efficiency and capacity, as noted in the ISA 2013 book 101 Tips for a Successful Automation Career.
The user is confronted with considerable disagreement of tuning rules as seen in the 400 pages of tuning rules in the 2006 book by O’Dwyer, not realizing most of them can be adjusted by factors or a near-integrator concept to achieve good control.
The modern PID has many more options, parameters, and structures that greatly increase the power and flexibility of the PID, but most are underutilized due to insufficient guidance. Additionally, ISA standard form used in most modern control systems is not the parallel form shown in most textbooks or the PID series form pervasively used in the process industry until the 1990s.
All of this can be quite overwhelming to the user, particularly because tuning is often done by a generalist faced with rapid changes in technology and with many other responsibilities. My goal in my recent articles, books, and columns (including blogs), which are more extensive and less supplier-specific than white papers, is to provide a unified approach and more directed guidance based on the latest PID features that are missing in the literature. The recently completed Good Tuning: A Pocket Guide, Fourth Edition seeks to concisely present the knowledge needed and to simplify the tuning by switching between just two sets of tuning rules, largely depending upon whether the PID is a primary or secondary controller. A primary PID for vessel or column composition, gas pressure, level, pH, and temperature control uses integrating process tuning rules where the lambda arrest time is set. A secondary PID for liquid pressure, flow, inline pH, and heat exchanger temperature control, uses self-regulating process tuning rules where the closed loop time constant is set. In both situations, lambda rather than a lambda factor is used and chosen relative to the dead time to provide the degree of tightness of control and robustness needed.
The best thing a user can do is to use good tuning software, attend supplier schools, and get a consultant in the plant for on-site solutions and practice. It is also important to take responsibility for avoiding common tuning mistakes. Here we step back to make sure we are not susceptible to oversights and misunderstandings. The following compilation has the most common, disruptive, and potentially unsafe mistakes first, but all can come into play and be important.
1. Using the wrong control action: In analog controllers and in many early distributed control systems (DCSs) and programmable logic controllers (PLCs), the valve action affected only the display of the output on the station or faceplate. The specification of an “increase-to-close” valve action for a fail-open valve reversed the display but not the actual output. Consequently, the control action had to take into account the valve action besides the process action. If the valve was “increase-to-open” (fail close), the control action was simply the reverse of the process action (direct control action for reverse-acting process and vice versa). If the valve was “increase-to-close” the control action was the same as the process action (direct control action for direct-acting process and vice versa) if not reversed in the current-to-pneumatic (I/P) transducer or positioner. In today’s systems, the user can specify “increase-to-close” in the PID block or analog output block besides the digital valve controller, enabling the control action to be set as the opposite of the process action. The challenge is realizing this and making sure the increase-to-close valve action is only set in one place. If you do not get the control action right, nothing else matters (the PID will walk off to its output limit).
2. Using PID block default settings: The settings that come with a PID block as it is dragged and dropped into a configuration must not be used. When first applying PID to dynamic simulations of new plants, typical settings based on process type and scale span can be used as a starting point. However, tuning tests must be done and settings adjusted before operator training and loop commissioning.
3. Using parallel form and series tuning settings in the ISA standard form: A parallel form that uses integrator gain and derivative gain settings that are put into the ISA standard form as reset time and rate time settings can be off by orders of magnitude. A series form can provide good control with the rate time equal to or greater than the reset time. This is because interaction factors inherently reduce the PID gain and rate time and increase the PID reset time to prevent oscillations from the derivative mode contribution being greater than the contribution from the other modes. Using a rate time equal to or greater than the reset time in an ISA standard form can cause severe fast oscillations.
4. Using the wrong units for tuning settings: Here we consider just the series form and ISA standard form. Controllers can have a gain or proportional band setting for the proportional mode. The gain setting is dimensionless and is 100 percent divided by the proportional band. Some PID algorithms in control studies and actual industrial systems have the gain setting in engineering units, which leads to a very bizarre setting. The integral mode setting can be repeats per second, repeats per minute, minutes per repeat, or seconds per repeat. The units of these last two settings are commonly given as just minutes or seconds. The omission of the “per minute” can cause confusion in the conversion of settings. The conversion of the rate time is simpler, because the units are simply minutes or seconds.
5. Using the wrong units for output limits and anti-reset limits: In analog controllers and in many early DCS and PLC systems, the output and consequently the output limits and anti-reset windup limits were in percent. In modern control systems, the output is in engineering units, and the limits must be set in engineering units. For valves, the units are usually percent of valve stroke. For a primary (upper) PID that is sending a set point to a secondary (lower) PID, the primary PID output is in the engineering units of the secondary PID process variable.
6. Tuning level controllers: If you calculate the product of the valve, gain, process gain, and measurement gain, where the process gain is simply the inverse of the product of the fluid density and vessel cross-sectional area, you realize the open loop integrating process gain is very small (e.g., 0.000001 1/sec), leading to a maximum PID gain for stability that is more than 100. For surge tank level control, a PID gain closer to unity is desired to absorb fluctuations in inlet flows without passing them on as changes to a manipulated outlet flow that will upset downstream users. Users do not like a high PID gain even when tight level control is needed. Decreasing the level controller gain without a proportional increase in the reset time will cause nearly sustained slow rolling oscillations. Further decreases in the PID gain only make the oscillations worse. Most oscillations in production plants and poor performance of distillation columns can be traced back to poorly tuned level controllers. The solution is to choose an arrest time (lambda for integrating processes) to either maximize the absorption of variability (e.g., surge tanks level control or distillate receiver level control where distillate flow is manipulated) or maximize the transfer of variability (e.g., reactor level for residence time control or distillate receiver level control where reflux flow is manipulated for internal reflux control). The integrating process tuning rules prevent the violation of the window of allowable PID gains by first setting the arrest time and using this time to compute the reset time and finally the PID gain.
7. Violating the window of allowable controller gains: We can all relate to the fact that too high of a PID gain causes oscillations. In practice, what we see more often is oscillations from too low of a PID gain in primary loops. Most concentration and temperature control systems on well-mixed vessels are vulnerable to a PID gain that violates the low PID limit, causing slow rolling, nearly undamped oscillations. These systems have a highly lag dominant (near-integrating), integrating, or runaway process response. All of these processes benefit from the use of integrating process tuning rules to prevent the PID gain from being less than twice the inverse of the product of the open loop integrating process gain and reset time, preventing the oscillations shown in the figures. The oscillations in the figures could have been stopped by increasing the reset time. In industrial applications, the reset time in vessel control loops often needs to be increased by two or more orders of magnitude. Note that the oscillations get worse as the process loses internal self-regulation, going from a near-integrating (low internal negative feedback) to an integrating (no internal feedback) and to a runaway (positive feedback) open loop response. For runaway processes, there is also a minimum gain setting independent of reset time that is the inverse of the open loop runaway process gain. The identification of the open loop integrating process gain can generally be done in about four dead times, greatly reducing the test time and reducing the vulnerability to load upsets.
Click here to continue reading Greg McMillan’s article at InTech magazine.
This is the official online community site of the Emerson Global Users Exchange, a forum for the free exchange of non-proprietary information among the global user community of all Emerson Automation Solution's products and services. Our goal is to improve the efficiency and use of automation systems and solutions employed at members’ facilities by sharing our knowledge, experiences, and application information.
User Groups |
World Areas |
Community Guidelines |
Legal Information |
Contact Community Manager
Website translation provided by
© 2015 Emerson Global Users Exchange. All rights reserved.