Are you ready to take a trip down memory lane to learn how to achieve the best process efficiency and capacity for the most important batch and continuous control applications? My journey to find and implement the best process control by realizing the best instrumentation, control strategies and PID control has been a bumpy ride at times, but we learn the most from mistakes.
Electrical and instrument design and construction
I started out in electrical and instrument (E&I) design and construction in 1969. After a 12-week training course and some preliminary plant work helping my mentor, I moved to an onsite position at a plant in West Virginia. It was responsible for installation, calibration, checkout and startup of the E&I systems for a new production unit and control system upgrade of three existing production units. I oversaw E&I construction using union electricians and pipefitters who, like me, had no previous experience in instrumentation. We learned the hard way.
Plant E&I technicians didn’t show up until after the plant was up and running at design specifications. The biggest problem I experienced was rotary control valves with rusted piston actuators from outside storage, pulley systems and rudimentary positioners. These valve assemblies provided by an on/off valve supplier were cheap. I replaced the valve assemblies with actuators, direct links and high-gain positioners designed for throttling valves. The startups were tortuous, but successful.
I became the company’s representative at a contractor in Massachusetts responsible for designing the world’s largest specialty chemicals plant in Texas. I subsequently moved to the plant construction site, in charge of two newbies with no experience. Fortunately, the Texas plant’s E&I technicians took the lead in instrumentation calibration.
In our water batching application, temperature loops were too slow. and valves didn’t open until PID output was 25% or more. Thermowells had sand in them from them being installed with no covers during piping and equipment sandblasting, and valves with a terrible response didn’t have positioners. The contractor followed the prevailing rule that valves in fast loops shouldn’t have positioners, which was based on a theoretical study that didn’t consider bench settings, poor resolution from stiction, lost motion from backlash and shaft windup.
After we cleaned out the thermowells and installed high-gain pneumatic positioners, the startup quickly and greatly improved. I developed a first-principle dynamic simulation to help understand the challenging objectives of an axial compressor, and the simulation attracted the engineering technology department. I was invited to be a modeling and control engineer in 1977.
Engineering technology (ET) modeling and control
I was incredibly lucky to be working in ET because it featured about 100 of the world’s experts in modeling and control. The department head was University of Texas Professor Emeritus James Fair. We were all given the freedom to advance technologies to improve operations by process modeling and control, and publishing results, an incredible opportunity that I doubt exists anywhere today.
I worked on control problems for large axial compressors with steep surge curves, which were prone to more damage than centrifugal compressors. I found one that had a pitot tube in a box for compressor suction flow measurement, which is a key process variable for surge control. A venturi tube replacement made a big improvement in operating point relative to surge point.
I thought I could enable the surge controller to recover from surge by making the surge valves, measurement and controller work faster. However, I learned once in surge, the jumps in flow were so large and fast, due to positive feedback in compressor curve to left of surge point, that feedback control is unable to recover. So, I designed an open-loop backup to equip the controller with a surge valve opening large and long enough to recover from surge. I provided some simple additional logic to predict a potential crossing of the surge curve that would later become a future value prediction block.
I tried following the rule of using boosters instead of positioners on fast loops. On startup, the surge valves assigned to be open suddenly slammed shut. An E&I technician showed me that with a booster and no positioner, he could manually move the diaphragm actuator shaft of a 24-inch surge valve. So, I immediately reinstalled the positioner with the booster on its output with a booster bypass valve slightly open to prevent the oscillations from the positioner seeing the very small booster volume instead of the large actuator volume. The original theoretical studies that advised replacing positioner with booster didn’t account for the booster’s low inlet port sensitivity and high outlet port sensitivity to slight changes in diaphragm actuator volume and pressure from diaphragm flexure leading to positive feedback.
A few years later, when I showed up at supplier’s lab to test the pressure control valve response time for sensitive furnace pressure control in inches of water column, I was surprised to find the 16-inch valves with diaphragm actuators had boosters but no positioners. I showed them how I could manually position the valves by grabbing the shaft. The supplier put positioners on the valves with boosters on the positioner outputs with slightly open bypass valves, and found an article written a decade earlier by their company expert detailing the need for boosters on positioner outputs with a slightly open bypass valve that’s now an integral part of some boosters. The pressure control valves then all worked well with a response time of less than one second.
A key to my many more intelligent compressor control applications, including dealing with compressors in series, was the momentum balance showing the compressor installed characteristic to the left of the surge point. This was rarely seen in the literature and seldom described by compressor suppliers that positive feedback could cause jumps in suction flow. My book Centrifugal and Axial Compressor Control is free to download and provides the calculations that reveal the critical hidden curve, and the control strategies and instrumentation for best compressor control including response time requirements.
One plant asked me to help with some new, large valves in the piping spec they bought to improve pressure control. They were inexpensive, large-capacity, tight-shutoff valves. When I went to the plant lab to check out the valves’ response, the person in the lab room looking at the computer results from a smart positioner said the valves were responding to a 0.5% change in signal. I put a travel gauge on the butterfly disc and found that the valve didn’t respond to signal changes smaller than 8%. The positioners were being lied to due to shaft windup from high-seal friction from tight-shutoff design.
I subsequently found the same problem in another plant installing new piping spec valves. These rotary valves are attractive because they’re cheap with high capacity and tight shutoff. They’re sometimes referred to as high-performance valves. Often, technicians are clueless because smart positioner readback and diagnostics are lies due to shaft windup from seal friction, stem friction from graph oil packing, lost motion from key lock connections, rack and pinion actuator resolution limitation, and backlash from link-arm and Scotch-yoke connections.
For much more on how to ensure the best valve dynamic response and avoid falling for on-off valves posing as throttling valves, read Annex A that I wrote for ISA-TR 75.25.02-2024 and my 2016 Control article and whitepaper “How to specify valves and positioners that do not compromise control.”
Automated startups
A difficult plant compressor startup triggered several trips on startup attempts. The plant was subject to six or more startups each year. When I told operators I’d automate the startup, they said it wasn’t possible because they always experienced trips. To me, this was motivation for startup procedure automation.
I used dynamic simulations to develop and thoroughly test the logic needed. The first and subsequent automated startups were successful. Next, we needed to reduce plant shutdowns because it had been pushed beyond original design capacity due to so many recycle streams. By installing three measurements with middle signal selection, improving downstream operations, and flow control on key recycle streams to prevent positive feedback, plant trips went from more than six per year to one per decade. Plant production capacity and safety greatly increased.
My colleague didn’t think it was a good idea to automate a hazardous raw material transfer that was extremely difficult for operators. So, he sat in the control room with gas masks on, while I commissioned the procedure automation that was thoroughly tested by a dynamic simulation. It went extremely well. I’ve successfully used procedure automation in difficult plant operations and used state-based control to proactively deal with abnormal plant operating conditions.
Pressure control
I found some analog control holdouts. One was a furnace pressure controller that could go off-scale in 0.1 seconds, and the other was a polymer pressure controller. The manipulated variable in each case was an incredibly fast variable speed drive (VSD) set by a fast analog signal. I remember looking at the polymer pressure controller and thinking the fast oscillations were noise but later finding out they were load disturbances.
I improved pressure control in less demanding, but fast, unit operations by using DCS controllers with 0.05 second execution rates, valves with boosters on positioner outputs with a slightly open bypass valves, and eliminating impulse lines with new, direct-mounted transmitters with no damping. The plots showed greater amplitude oscillations because they weren’t filtered by the new transmitters. Fortunately, I kept the older, slower transmitters, with trend plots that showed significantly tighter control from faster valve, controller and measurement.
Reactor control
I helped improve the temperature control of some highly exothermic reactors. In one fed-batch reactor, we used temperature rate of change as the controlled variable to prevent overshoot of batch end-point temperature. In another reactor, high-gain and high-rate action without any reset action was used to prevent a runaway condition. When I was in the control room to improve overhead receiver control, I was asked to wear a gas mask because, if the reaction started a runaway due to an increase in temperature, there was no point of return, and the plant depended on relief valves and flare stacks when a reactor’s contents burst. I developed first-principle, ordinary, differential equations to model the positive feedback time constant and process gain for such highly exothermic reactors. This revealed the point-of-no-return and the need for closed-loop testing for tuning and high-temperature controller gains, even to the point of causing great changes in the chilled water secondary temperature controller setpoint. Sometimes nearly full-scale, secondary temperature control oscillations are needed to keep reactor temperatures close to setpoint.
I detailed how there’s a controller gain window where a controller gain that’s too low can cause rapid acceleration and runaway that’s worse than the increasing oscillation amplitude from a controller gain that’s too high. I was able to compute a window of allowable gains from the positive feedback time constant and process gain. A study by William Luyben showed how the minimum controller gain increased with loop deadtime and heat transfer lag, concluding that the ratio of maximum to minimum controller gain should be more than six for robustness.
I also showed that a constant maximum jacket recirculation system is essential to minimize secondary temperature control delay. For a high-recirculation flow, changes in heat transfer are seen quickly in the jacket inlet. For cooling, a chilled water makeup valve on the inlet is used in conjunction with a jacket return flow by jacket discharge pressure control. For startup heating, an inline steam injection unit on jacket inlet flow provides fast transition to and from heating.
A fed-batch reactor pushed way beyond its design limits was experiencing delays due to exceeding permissible operating ranges of a large variety of equipment pressures and temperature. I installed override controllers on each of the process variables suspected of causing halts in batch feeds. On operator graphics, the override controller taking control of reactant feed rate was flagged, and total time an override controller had been limiting feed for each batch was also displayed. The batch cycle time was reduced by more than 25%. Override controllers that never came into play were eliminated to simplify the override system. These controllers were secondary effects. The root causes of production limits were addressed by key override controllers.
Get your subscription to Control's tri-weekly newsletter.
pH control
A bigger part of my career was addressing pH control problems. Fortunately, I developed a charge balance with a simple and fast integral-halving solution to provide dynamic simulation to study and improve pH systems. I also had the luxury of a pilot plant with eight types of electrodes installed to study response time and short- and long-term accuracy and reliability. The plant revealed a lot of hype from suppliers, and showed the best electrode was a glass bulb with a liquid reference junction.
Subsequent improvements in glass formulations and offering a replaceable double junction made the decision easier. I also detailed how middle signal selection could improve not only reliability but also response time, and reduce noise by inherently ignoring a slow and erroneous electrode response. The many lessons I learned about electrodes, piping, equipment and valves were the result of dynamic pH simulations and plant applications and consultations with experts of electrode suppliers. They’re captured in my ISA book, Advanced pH Measurement and Control – Digital Twin Synergy and Advances in Technology 4th Edition. The fact that the electrode supplier experts and I are all retired makes this book more important.
I was able to dramatically improve pH system performance by reducing delays stemming from reagent injection, transportation to electrodes, mixing and electrode lags. I was able to show via dynamic simulations confirmed by plant results how reagent injection of typically low flow rate reagent (e.g., 0.1 gph) in recirculation lines, rather than via dip tubes, could reduce delays from hours to seconds.
Also, I also showed the use of pH control in a recirculation line of an existing tank with just eductor mixing could be used to eliminate the considerable cost of installing a new vessel with stringent axial mixing and dimensional requirements. I also detailed reagent demand control by translating the controlled variable from the Y axis (pH) to the X axis (0 to 100% reagent demand) of the titration curve by a conventional piecewise signal characterizer. This linearization by signal characterization is particularly important for strong acid and strong base systems that can deteriorate a beneficial 20-minute vessel process time constant to less than 10 seconds due to acceleration on the steep portion of the titration curve.
Linearization also reduces the appearance of pH measurement noise that’s distracting to operations.
I used dynamic simulations in nearly all my process control improvement studies, which were even more critical for pH control due to challenges posed by extreme nonlinearity and sensitivity. To match lab titration curves, I found I needed to include a small amount of carbonic acid from absorption of carbon dioxide in the air that was never discussed in the literature. The effect of carbonic acid was extraordinary for strong acid and strong base systems, reducing the titration curve slope and hence the pH process gain by orders of magnitude.
Safety glass extruder and sheet line control
I learned through dynamic simulation how to improve extruder product quality by specific energy consumption (SEC) focusing on pre-mix speed and heating by reducing zone heaters to be just a warm blanket. The simulations were also extended to sheet-line quality control to improve sectional sheet thickness PID controllers with deadtime compensation with setpoints improved to optimize cross sectional thickness related to optical clarity.
Opportunity sizing and assessment
ET that was developed when the CEO was an engineer was dissolved when the CEO was a lawyer. A handful of modeling and control specialists formed a new group for opportunity sizing and assessment. Plant accounting records were studied to detail current operation and find examples of best operations. The financial potential advantage was detailed and reviewed in a subsequent meeting with plant engineers, operators, technicians and researchers. Control solutions gained by the openness and synergy of the wide spectrum of meeting participants were proposed and their potential value agreed upon.
I, and a few other modeling and control specialists, often did studies using model results to propose process control improvements triggered by the opportunity assessment often by better PID control strategies. We were often able to implement them with the help of key plant people within a week or two on site. We ended up with production efficiency and capacity benefits of more than $20 million per year. After about eight years, we started to run out of readily recognizable applications. Also, since the chemical company had incurred lots of debt in its spin-off, while a very profitable business was sold, it was headed for bankruptcy. I retired in 2001 to protect my retirement account and went to work as a contractor and eventually a part time employee in PID and dynamic simulation research and development (R&D).
PID and dynamic R&D
I was given the opportunity during 2001-23 to work part time in a leading automation supplier’s R&D division. I developed first-principal dynamic simulations to greatly improve batch profile slope and endpoint control, plant waste treatment pH control, bioreactor product quality and cycle time, kiln production schemes developed in the 1980s by for better oxygen control, dryer production by inferential moisture control, and dramatically improve the performance and simplify the tuning of control loops using analyzers and wireless measurements with an enhanced PID using external-reset-feedback. Here are some key articles documenting methods and results:
- “Full Throttle Batch and Startup Response,” Control, May 2006
- “Virtual control of Real pH,” Control, Nov 2007
- “Unlocking the Secret Profiles of Batch Reactors,” Control, July 2008
- “Virtual Plant Provides Real Insights,” Chemical Processing, Jan 2009
The opportunity of valve position control (VPC) to improve process capacity and efficiency has been greatly increased by advances in PID functionality, such as external-reset feedback (e.g., dynamic reset limit) and an enhanced PID (e.g., PIDPlus). In “Don’t Over Look PID in APC,” Control, Nov 2011, I was able to show how to more effectively realize VPC applications to maximize production rate and minimize energy, reactant, and reagent use and plantwide flow feedforward (ratio control) to maximize plant flexibility.
In “Get the Most Out of Your Batch,” Control, Sept 2012, I developed some innovative, easy-to-implement general solutions to increase front-end and parallel batch efficiency and capacity that take advantage of analyzers and inferential measurements. The test case is the front end of an ethanol plant with batch fermenters, but much of the methodology is applicable to batch reactors for food, beverages, drugs and chemicals.
I was able to dramatically improve the dynamic modeling accuracy and ease of biological reactors by using Convenient Cardinal Equations to model the dramatic effect of temperature and pH and Michelis-Mentor Equations to model the effect of glucose and glutamine on cell growth rate and product formation rate documented in my ISA book, New Directions in Bioprocess Modeling and Control 2nd Edition.
These are just some of the highlights of my career. Much more of what I’ve learned about how to improve PID control is in my book, Tuning and Control Loop Performance 4th Edition, that’s free to download. The book reveals how external-reset-feedback can reduce oscillations from slow secondary loops and slow valves. It includes tests showing process, controller, measurement, valve, disturbance and interaction effects on loop performance.
To increase process engineers’ awareness of potential opportunities, I recently wrote a Chemical Processing article: “Find Missed Opportunities in Process Control,” Chemical Processing, Oct 2020.
I received the ISA Mentor Program Achievement Award. Best Control Valves, Feedforward and Ratio Control, Strange but True Process Control Strategies, PID Options and Solutions, plus other presentations by ISA Mentor Program leaders. They’re viewable here.
There was a very successful use of dynamic simulation to greatly improve a steam header system plagued by interactions between headers and large changes in steam generation and use. The addition of 30-plus feedforward signals without dynamic compensation greatly improved the ability the headers to handle large disturbances, reducing peak errors from feedback control by 40% or more. The addition of dynamic compensation provided a further reduction of 30% of the remaining peak error. Benefits included eliminating operating equipment trips due to steam pressure upsets, and a significant reduction in energy use by eliminating unnecessary vent valve openings and utility boiler use. Slide 18 of the ISA Mentor Program Feedforward and Ratio Control Webex gives a simplified view (less headers and details on steam generators and users).
I’m fortunate to have Peter Morgan continue this journey with my collaboration, using dynamic simulations to test and improve override control and dead time compensation by PID controllers with external-reset-feedback. Next up are studies to improve variable frequency drive (VFD) resolution and rangeability. We received ISA Standards Achievement Awards for ISA Technical Report ISA-TR5.9-2023, “Proportional-integral-derivative (PID) algorithms and performance,” that documents many of the advances in PID technology and uses to improve process control.
