Modeling and Control

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

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

Practices Offered to Improve the Performance of pH Electrodes

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

The following file provides some background information on these practices.

Info on pH Measurement Practices

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

Top Ten Signs of a Rough pH Startup

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

In the early 1990’s, I helped establish Emerson’s advanced control program. As part of this initiative, a technical advisory committee was formed to periodically review the program and to help provide technical direction. This committee was made up of leaders in process control such as Professors Karl Astrom, Lund University, Tom Edgar, University of Texas, and Tom McAvoy, University of Maryland. Some of the initial products that came out of this research and development program were initially introduced in the Provox and RS3 product lines. Later, with the introduction of DeltaV, these products for control tuning, fuzzy logic control, model predictive control, and property estimation were embedded in the DeltaV system.

The DeltaV architecture supports abstraction of system software components from the physical hardware. This capability allowed the advanced control team to introduce a family of control system simulation products, DeltaV Simulate. These products allow all system features to be combined and executed on a single workstation or distributed between multiple workstations without the requirement for a physical controller.

While Greg McMillan was at Solutia, he and I often interacted on beta tests of advanced control products and the control system simulation environment supported by DeltaV.

The technical basis for the advanced control and simulation products available in DeltaV are described in a series of over thirty technical papers. These papers were written and presented by the advanced control team at various control conferences. As customers applied these products we received detailed technical questions that were often best addressed by referring the customer to one or more of these technical papers. Thus, Dr. Willy Wojsznis, Emerson Process Management, and I began to discuss writing a book that incorporated information from these papers that could serve as a reference for control engineers working in this area. In addition, we felt it was important to address the benefits and application of this technology on various processes. Thus, we invited Greg McMillan and Mike Brown to co-author this book. The book, Advanced Control Unleashed, was published in the fall of 2003 and that year was ISA’s best selling book

If you are interested in exploring the benefits of advanced control, then Advanced Control Unleashed is a good starting point in learning about this technology. With ISA’s permission, I have created a brief overview of the book.

Overview of Advanced Control Unleashed

You can purchase this book directly from ISA. As stated in the book introduction, all royalties from the books are donated by the authors to “universities, consortia, and educational programs to promote and enhance the development and use of advanced process control.”

When the Fieldbus Foundation was established, I lead the team that wrote the Function Block Application Process Specifications. This was a unique experience since it offered a rare opportunity to work with and to get to know some of the best control engineers from Siemens, Yokogawa, Honeywell, Leeds and Northrup, Foxboro, ABB, Smar and others leading companies in our industry.

In the committee meetings, we were able to put aside the commercial barriers that often prevent open discussions. In many cases, we used the ground breaking work by the SP50 User layer committee as a starting point in the development of the function block specifications. However, significant changes were made to incorporate the best ideas and features from existing products. Also, the architecture and formal model defined by Part 1 of the specification were written and structured to support object oriented design and implementation of field devices – something important if you are a manufacturer of field devices.

Basic architectural components such as status and mode definitions were discussed by the function block team at length and in the end reflected the contribution of more than one company. For example, the definition of status was strongly influence by Tom Kinney, Foxboro, and Bill Hodson, Leeds and Northrup. Similarly, the basic function block set defined in Part 2 of the specification were reviewed in detail to insure the blocks addressed the requirements of each company. The advanced function block set defined in Part 3 of the specification and the specialized blocks in Part 4 and 5 were defined by members of the team that had direct experience with the functionality encapsulated by the block.

When work on parts 1-3 of the Foundation Specification for the Function Block Application Process was completed in 1994, there were very few fieldbus products on the market and even fewer control systems that supported these devices. In the function block team we discussed the need to communicate the advantages of fieldbus to inform the industry of this technology. Thus, Tom Kinney, Foxboro, and Marcos Peluso, Smar, and I worked together to put on a fieldbus tutorial at ISA1995 conference. In this tutorial, we gave an overview of the technology to a standing room only audience. To show how fieldbus would be applied, we took a waste water process example and analyzed the benefits that fieldbus had over a traditional installation. For the next seven year, Marcos – who is now with Emerson - and I continued to host fieldbus a tutorial each year at ISA. In each tutorial, we focused on a different aspect of fieldbus installation and application. Much of the material that we developed for these tutorials can be found at Fieldbus Tutorial. If you are unfamiliar with fieldbus and want to quickly get up to speed on the basics then you may find the information provided at this site to be helpful.

Professor Tom Edgar of the University of Texas and Joseph Alford from Eli Lilly and Co. have been collecting data and opinions regarding the current syllabus of the typical undergraduate Chemical Engineering Process Control Course and its relevance to the skills and knowledge needed in today's industrial process control environment. An upcoming issue of Intech magazine will have an article by Edgar on how process control education in the universities should be changed. There will be replies from a key group of professors and practitioners. My contribution is the first 250 words of the following:

Terry Tolliver and I have taught a course on dynamic modeling and control at Washington University (WU) in Saint Louis since 2002 that is a requirement for a degree in chemical engineering. The course uses an industrial virtual plant and the ISA book Advanced Control Unleashed. The students are very computer literate and pick up on the use of industrial software from just a few screen prints put into the laboratory exercises. The knowledge gained is generally applicable since the function blocks are based on Foundation Fieldbus used in millions of devices and by over a hundred manufacturers. The configuration environment is also consistent with the international standard IEC 61804. The students learn how to intelligently discuss and use an industrial process simulation, DCS, and data historian that form a virtual plant on their desk. There is companion course taught by Bob Heider where an actual hardware version of the same DCS is used to control the temperature, pressure, and level of vessels in a hardware lab. The three professors have a total of more than 100 years experience in industry.

Most of the chapters in Advanced Control Unleashed start with an introductory section on “Practice,” continues with sections on “Opportunity Assessment” and “Application” and concludes with “Theory”. The strategy is to provide the relevance and practical considerations before getting into the theory that offers a deeper understanding. For example, in Chapter 2 - “Setting the Foundation”, the student gets an overview and perspective, list of opportunities, examples, application detail, and rules of thumb before getting into the theory where the focus turns to the set up of the differential equations for the material and energy balances to enable the student to learn the source of process time constants and gains in terms of process parameters. The students are not asked to solve or integrate these equations. Instead, the students graphically create a dynamic simulation of processes for unit operations commonly encountered on the job. Blocks for filters, dead times, noise, periodic disturbances, and backlash and sticktion are added to make the challenge of process control more realistic. Additionally the students configure an actual control system that can be downloaded into a real DCS. The students apply industrial embedded tools for auto tuning, statistical analysis, and model predictive control (MPC). The course centers on time response because this is what they see on the trend charts in the control room but there is a session to show how to go from the time domain to the frequency domain.

When I recently went back to WU and gave a guest lecture on the use of PID and MPC for fed-batch control of a fermenter, a student asked “what is a batch?” I knew that students were taught to think in terms of a steady state and the material and energy balances on a Process Flow Diagram (PFD) for continuous operations but I didn’t fully realize the implications until the question.

I have had chemical engineers in industry ask, why do you need a PID or MPC when you can just set the flow shown on the PFD? In fact, the batch sequences defined by process engineers today often try to set a predetermined step sequence of flows instead of using feedback control to sort it out. I have also have had experienced instrument engineers ask why do you need a Coriolis density measurement when the composition is constant as shown on the PFD? I also see ads for pressure and temperature compensated differential pressure orifice meters that claim to offer an unqualified mass flow measurement. If only the composition in all the pipelines were constant. This would sure make life easy. Product quality would be a non issue. Obviously the importance of dynamics and disturbances for process measurement and control is often missing in action.

In a batch process, the product concentration follows a profile. In some case there is also a temperature profile and in almost every case where a PID or MPC is used, the transfer of variability for a constant set point means there is a profile in the controller output. This understanding is lacking when chemical engineers are taught to think steady state. The lessons from batch would also be useful for the automated startups and grade transitions in continuous operations.

To add a bit of levity, I offer the following Top Ten List:

Top Ten Reasons Why an ISA Book on Control is not a University Text

10. Costs less than $100
9. The authors spent too much time in industry
8. Contains top ten lists and cartoons
7. Shows flow sensors upstream instead of downstream of the control valve
6. Discusses stick-slip and backlash
5. Shows unmeasured load upsets as inputs to the process
4. Includes field implementation considerations
3. Estimates tuning settings to just two significant digits
2. Doesn’t use tensor analysis for flow loops
1. Depicts signal lines as electronic instead of the pneumatic

Well, you may not become rich and famous in the conventional sense. You may not make it into the best seller list or your favorite book club but you can impress your friends, relatives, and associates and increase the marketability of your skills. More important is the sense of accomplishment in adding your expertise to the common body of knowledge. Here is the chance to share what you have learned and insure it is not lost.

I didn’t start out as a particularly a good writer and I am still learning. You are not trying to win any literary awards. You would be taking way the job of the ISA copy editor if you were to write perfect sentences. Like anything, writing becomes much easier with practice. A side benefit is it seems to have made me more articulate in public speaking.

I listen to music while writing but I have to admit I prefer peace and quiet when I first get started on a new subject. Once I get a flow going then I can rock. The inspiration and feeling I get from my favorite artists has helped make the writing process more fun and creative. The following is a summary of what has worked for me.

• Think that you are trying to say what is most important to a good friend
• Just get started and get a stream of ideas out
• Do figures last so it doesn’t interfere with the flow
• Start each paragraph with introductory sentence
• Keep sentences short
• Explain each point in detail
• Define nomenclature and engineering units for equations
• Give assumptions and offer examples
• Don’t obsess about wording – this is the ISA copy editors job
• Get associates in your field and related field to proof read everything
• Double check equations and figures

I am a terrible proof reader of my own stuff, because I read what I want to say and not what I actually have in print. Before you send a rough draft to ISA, it is wise to get several of your associates to proof read the text, equations, and figures. All technical comments are useful because even the ones that are off-base reflect a misunderstanding that should be addressed. Note that ISA for the last 3 years requires that you submit drafts of chapters as they are completed for technical review. This process should start about a year before the next ISA Expo. A schedule showing the planned submission date of each chapter is submitted with the outline and brief description of the book’s uniqueness and audience in the proposal. Below is an example of a proposal for my most recent book. The schedule shown was too tight.

Book Proposal Example

I have found it useful to put key concepts (insights) in italics for emphasis and easy reference. So often we get lost in the details and lose track of the underlying principles and governing generalities that are so important for dealing with new situations.

The following presentation at ISA Expo 2006 in Houston in the “Automation Connection Series” on Oct 18 offers a Smorgasbord of books, advantages of being an author, and the steps. Contact me if you want to get started on the road to fame and fortune.

ISA Expo 2006 Presentation

When designing a control strategy you may be faced with the challenge of there being an extra degrees of freedom. One of the most common examples is where one control parameter may be maintained at setpoint through the adjustment of two manipulated parameters. Often the solution is to address the control design using split range or valve position control. Through the use of these techniques, the two actuators appear as one actuator to the PID control. However, there are some significant differences in the resolution and dynamic response that may be achieved using either technique. An alternate approach is to implement a strategy that combines the best of split range and valve position control.

I once was responsible for the design of the 400 # header pressure controls for a new power house in a pulp and paper mill. Under normal operating conditions, the header pressure was to be maintained by the turbo-generator extraction to the 400# header. However, if the turbine was to trip or be taken off-line for maintenance, then two pressure reducing valves (normally closed) were to be used to let down steam from the 1475# header to the 400# header. Under a trip condition, it was important that the full dynamic range of the pressure reducing valves be used to make up for the steam that had been supplied by the turbine extraction. This objective could be achieved through the use of split range control. However, if the turbo-generator was to be off-line for an extended period of time for maintenance, then it would be advantageous to provide the precise pressure control that may be achieved by taking advantage of the operating characteristics of valve position control. After some work, I came up with a network that combined the best of split range and valve position control. I commissioned and tested the header controls as the power house was brought on-line. The 400# header control proved to be quite effective and after over 20 years is still in use at the plant- migrated to new controllers.

The work I did on the revised header pressure control strategy was documented in a paper that I wrote and presented shortly after the power house startup, "Improving PRV Pressure Control", ISA 31st Annual Southeaster Conference, April, 1985. The technique was later used within Emerson’s pulp and paper group to address a variety of applications e.g. furnace draft control variable speed ID fan in combination with damper, variable speed pump in combination with a regulating valve for recovery boiler liquor flow control, forced-draft fan control pressure control using a variable speed fan with inlet vanes. Many of these applications were documented by Bill Love, Forney International, in an article “Innovative control technique that improves control rangeability and resolution in paper mill applications”, Tappi Journal, February, 1994.

The tools that are available in most modern control systems are sufficient to implement the network that I originally designed for the header pressure control. The basic network design is shown in the following:

Combining Split Range and Valve Position Control

Also, this material includes an example of how the network may be implemented as a re-usable composite block in DeltaV. Some process examples are iprovided that allow you to compare the dynamic response of this network to that achieved using spilt range control and valve position control. If your control objectives can not befully met by valve position or split range control, then you may want to consider this network for your application.

Terry described an innovative technique of using the PID block for combining split ranged control and valve position control (see Terry’s Oct 16 entry). This technique eliminates the limit cycle at the split range point caused by the increase in nonlinearities and the decrease in resolution imposed by backlash, backfilled pipes and dip tubes, rangeability limits, and friction particularly associated with starting a flow from zero. This technique also eliminates the conceptual and tuning problems with valve position control. People tend to confuse valve position control with valve positioners or digital valve controllers. The tuning of the integral-only controller for valve position control is much more critical than most people realize to prevent interaction but provide a fast enough response to reject large load upsets. The best quantitative analysis I have seen on the severity of the tuning issues with “valve position control” is the article by Cheng-Ching Yu and William L. Luyben titled “Analysis of Valve-Position Control for Dual-Input Processes” (Ind. Eng. Chem. Fundam. Vol. 25, No. 3, 1986 pp 344-349).

Instead of a special network for PID control, a standard Model Predictive Control (MPC) block can be configured to eliminate the need for split ranged control and valve position control. The MPC is simply set up for two manipulated variables (MV), one controlled variable, and one optimization variable. The optimization variable is the manipulated variable that provides the finest control (e.g. set point of the fastest and most precise control valve or variable speed drive). The optimization objective is to gradually return the “fine” MV to a mid range (e.g. 50%) after helping the “coarse” MV reject a load upset or minimize overshoot of a new set point. To insure the optimization takes a back set to tight regulation and set point response of the controlled variable, the “penalty on error” (PE) of the optimization variable is decreased (e.g. optimization variable PE=0.1).

When the MV have different process dynamics, the advantage of MPC is greater. By the automatic identification and incorporation of the MV dynamics in MPC, better feedback, feedforward, and constraint control is possible. The longer term view of the MPC also makes it less sensitive to resolution limits. Additionally, the “maximum MV rate” parameter can be written to zero when the controlled variable is close enough to set point to eliminate the limit cycle from the “coarse” MV. The following white paper discusses in more detail this use of a MPC to eliminate split ranged control and valve position control. The article titled “A Fine Time to Break Away from Old Valve Problems”, in the November 2005 issue of Control magazine provides more background and a perspective.

White Paper on Dual MV MPC

When planning or designing a fieldbus installation, it is important to consideration where the control will reside. The function block set supported by fieldbus devices may be used to address many common control applications. If you are dealing with a slow process then where the control is done may have little impact on the observed control performance. However, when addressing faster processes such as liquid pressure and flow, then this decision will directly impact control performance. A common misconception is that best control performance will be achieved by using the control system controller for all control and thus only use fieldbus to access IO in field devices. It turns out that in many cases that the best performance is achieved by using the control functionality of the fieldbus devices.

One of the primary reasons for this difference in performance rest with the fact that IO access and control execution can not be synchronize when a control loop is split between the controller and fieldbus devices. In a traditional control system it is possible to highly over-sample the IO and thus minimize delay introduced between an input being scanned and the value being used in control. For example, in some cases, the information from a traditional analog or discrete card is refreshed in controller memory every 50 msec. Thus, the measurement is always fresh when control is executed without the need for synchronization. However, the relatively low speed of the H1 physical layer (31.25Kbaud) does not permit the control system to over-sample the fieldbus analog input blocks. This is especially true if multiple transmitters are installed on the same segment since only one fieldbus device is allowed to communicate at any given time. Thus, fieldbus measurements used in the controller may be as old as one macro-cycle. From a control standpoint, this added delay directly impacts PID tuning and control performance.

When the Fieldbus Foundation specifications were originally developed, a considerable amount of time was spent addressing the issue of control performance. The system management capability and publisher/subscriber services of the fieldbus communication stack allow the execution of blocks and associated communications to be precisely synchronized. Time is periodically broadcast to insure that all devices on a fieldbus segment have the same sense of time. When control is done in the field using devices on the same segment, then each device that participates in the loop execution is given its portion of the schedule. Communications used in control always occur as scheduled. Thus, when control is done in the field it is possible to minimize any delay associated with communications and scheduling.

The loop execution speed that may be achieved using control in the field depends on a number of factors. For instance, the execution period will get longer as more loops are added to the same segment. Also, there is a large different between manufacturers (3X or more) in the time that is required for a function block to execute in a fieldbus device. Some of the latest generation fieldbus devices have significantly shorter block execution times. Also, if the control system supports the assignment of function blocks to the H1 interface card, then the execution of these blocks can be synchronized with other blocks on the segment since the H1interface card is just another fieldbus device. The power available to the H1 interface card allows much faster processors and thus blocks assigned to the H1 interface card execute in less time than the same blocks in field devices.

Marcos Peluso, Emerson Process Management, and I put together a presentation last year in which we addressed many of the factors that influence control performance when using fieldbus devices.

Achieving Target Control Performance Using Fieldbus Devices

If you are planning or designing a fieldbus installation then you may find the information contained in this presentation of interest.

Since we tend to learn by examples and we can all relate to the challenges of parenting, I came up with the following discussion of the relative advantages of PID and model predictive control (MPC) of a teenager for my Control Talk column in the September issue of Control Magazine.

Would Proportional-Integral-Derivative (PID) or Model Predict Control (MPC) be better as a parental advisory control algorithm? The PID excels at handling the unknown upsets, non-stationary behavior, and unpredictable nature of a teenager’s response and derivative action provides some preemptive action. However, the abrupt action by PID could amplify noise in a teenager’s behavior. But then there is the dead band from backlash and the resolution limit from friction. Maybe a lot of dither from the PID could keep the teenager off balance and minimize inaction if it doesn’t wear you out. There is inverse response to contend with where the teenager starts out doing the opposite of what was requested. Theoretically, MPC could build this behavior into its knowledge of the teenager’s future trajectory. Also, the patience of a MPC is inline with modern parenting ideas. But I not sure your can do enough pseudo random binary sequences (PRBS) or know the time to steady state of a teenager so my initial thought is proportional plus derivative (PD) with an adjustable bias to deal with the potentially unstable response. But then what do I really know anymore? I am an empty nester heading out for one of many vacations without children. All I need to know now is how to enjoy the antics of grandchildren and quickly return them to their rightful owner for process control.

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

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