By now, virtually every industrial process in the world is under automatic control, so the logical next step is advanced process control (APC). APC has already been implemented in many facilities, albeit with varying degrees of success. But when APC works, the results can be nothing short of spectacular.
Dr. James Ford, senior consultant at Maverick Technologies, says that Maverick engineers recently implemented a set of model-based controls for BASF in Geismar, La. "The application was on a series of fixed-bed catalytic reactors in which the reaction was highly exothermic while the catalyst was very heat-sensitive," Ford explains (See Figure 1).
"The temperature had to be kept high enough to promote the reaction, but just a short high-temperature excursion could rapidly deactivate this very expensive catalyst. Several APC algorithms were used to implement the solution in a Foxboro DCS. The project was very successful, especially at controlling the reactor temperature profile during bed catalyst changes," concludes Ford.
"The controls implemented by Maverick outperform the best of my operators, even with me at the console instructing them exactly how to handle a change-out," says David A. Weatherford, technologist at BASF. "They saved us a lot of money in terms of catalyst cost and paid back the investment in a couple of months."
Enterprise Products in Houston, Texas, is an energy company that operates pipelines, 25 natural gas processing plants, 20 NGL and propylene fractionators, and import/export terminals. "When I took on my new role, I didn't think I'd be able to justify replacing advanced regulatory controls, which I'd developed and supported over a four-year period," says Tom Lyndrup, a staff engineer in the Advanced Process Control group. "But two years later, that's exactly what I am doing."
With some success, he adds, especially in projects where APC replaces existing controls. "In some cases, 24 hours after a new APC routine was started up, we achieved significantly higher total production than the day before, despite a change in feed composition that normally results in lower rates." For more on what Enterprise Products is accomplishing with APC, see the story, "Justifying APC."
Terrance Chmelyk, manager of process solutions at system integrator Spartan Controls in Burnaby, British Columbia, Canada, installed an Emerson Process Management energy management system (EMS) at a cogeneration facility in a pulp and paper mill.
Such an application poses many problems that can't be handled well by conventional controls. "The sale price of electricity may change based on the time of day, the month of year, the market price of power and/or the current rate of power production," he explains. "Other complicating factors include variable fuel costs and availability, dynamic unit efficiencies, fluctuating mill heat requirements and many production-rate-limiting process constraints."
The energy management system employs facility-specific business models based on model-predictive control (MPC), linear programs and advanced energy metering to continually manage the energy contract and determine the optimum operating parameters to maximize profitability.
"The integrated EMS is able to achieve almost 6.5% improvement in overall cogeneration profitability, or about $3000 per day," he says. "Based upon a typical 350-day operating year for the plant, we'd expect that profits could be over $1million per year due to a real-time optimized plant."
Chmelyk continues, "The open-loop operational strategies that existed before saw the human operators make very infrequent changes to unit processes and didn't consider the changing efficiencies or economic models. This is very typical of most cogeneration operations, and the complexity and number of combinations are very difficult for a human operator to process on a real-time basis. In contrast, the integrated EMS with closed-loop control of the boilers and pressure reducing valves is able to make real-time adjustments to find the optimal combinations minute by minute."
Tom Kinney, product manager at Invensys Operations Management, says, "MPC [model-predictive control] moves the operating process closer to multiple constraints simultaneously to realize the maximum profit from an operating unit. Units typically pay for themselves within six months and equip companies for ongoing financial, productivity, quality or other strategic advantages." Table 1 lists the benefits of APC.
Given these success stories, why isn't APC implementation more widespread?
APC Is Incredibly Hard
If you saw the movie "Moneyball" or read the book, you probably remember Billy Beane telling Scott Hatteberg how easy it would be for him to switch from catcher to first base. For confirmation, Beane looks to infield coach and colleague Ron Washington, who doesn't miss a beat when he tells Hatteberg, "It's incredibly hard."
According to end users, some vendors sound like Beane, particularly when they are trying to sell end users on overly complex APC implementations that turn out to be incredibly hard to implement and nearly impossible to maintain.
Dan Cox, director of engineering at AOC Resins in Collierville,Tenn., has reservations about off-the-shelf APC. "Most APC products don't work well in our batch environment," he says. "This is in contrast to the vendor's claims, but I still feel this is the case. The amount of engineering required for MPC or neural nets tend to outweigh the benefit."
So AOC Resins implements APC with custom coding and reaps the fruits of its labors, as detailed in the story, "Roll Your Own APC."
Lou Bertha, engineer at RDI Controls in Lower Burrell, Pa. is also skeptical. "From my experiences, APC is a marketing term that big DCS companies use to sell a bunch of preconfigured subroutines that either provide generic routines or process specific routines such as soot blowing, compensated gas flow calculations, etc.," he says. "Advanced is questionable. Sometimes these canned routines are just an easy way to develop a system, since the specific routine has been developed, but typically end users cannot adjust these routines for their specific application. The ironic thing is what the big DCS houses call APC is what most smaller engineering companies call the next project."
Ford of Maverick agrees. "The glory days of implementing MPC on every process in a refinery or chemical plant are over," he says. "Too many failures. The future will see a renewed interest in APC solutions tailored to fit each specific control problem and will rely on a combination of traditional advanced regulatory control, inferred properties and MPC techniques when appropriate and justified."
APC implementation is difficult and so is maintaining its full effectiveness. APC will only work as designed for a limited period of time, as the underlying process conditions on which the original implementation was based are continually changing to one degree or another in every process plant.
In a refinery, for example, a switch might be made in the product mix in response to customer demand. Raw material input properties can vary significantly, particularly if suppliers are changed. Adding new equipment can significantly change the process, as can normal deterioration of existing equipment.
As changes occur within the process, corresponding adjustments need to be made to the APC algorithms and models. Without this on-going maintenance, the original APC application will lose effectiveness, often to the point where it must be decommissioned.
Without Maintenance, Benefits Fade Away
Whether model-based or some other underlying technology, any APC implementation will lose its effectiveness over time if it's not properly maintained. Maintenance doesn't require as much effort as initial implementation—step-testing isn't generally required, for example—but adjustments must nevertheless be made to maintain effectiveness.
Richard McCormick, automation engineer and industry consultant with Mick Automation in
Quebec City, Canada says MPC maintenance is and will continue to be the biggest issue. "Degradation of performance over time after start-up is seen everywhere, thus benefits decrease over the years," he points out. "Automatic monitoring tools for MPC applications are oriented for detecting things like models mismatch, the percent of the time constraints are hit, and other measurements to flag and identify causes for MPC performance degradation over time. Lack of experienced personnel for good support remains the biggest challenge for users. This is one reason why using specialized firms seems to be popular."
Mick Automation implemented MPC on FCC-polymerization units in the late 1990s, and it took a couple of years to complete it. That was the easy part. "The major thing after implementation is to maintain this huge application with knowledgeable process engineers and to have the operators understand and believe in the moves made by the optimizer," he says. "When the process engineer who participated in its design and implementation changed position, the application slowly and continuously degraded to a point where it was decommissioned. This is why I would now stay away from this type of optimization unless you have an impressive and knowledgeable staff to maintain it."
For more from Mick Automation, see "Under the Hood with MPC."
An automation engineer at a major refiner, who wishes to remain anonymous, says getting a multi-variable control (MVC) application running isn't too much of a problem. "Just spend big money to hire an expert," he advises. "The problem has always been keeping it running once the expert goes home. If the vendors were listening to their customers, you would see more support tools for running applications, better visibility into why a controller did something unexpected, the ability to identify poor performing models in real time and user-friendly tools to update individual models inside a running application."
APC implementation and maintenance are difficult, as end users well know. But at least both end users and vendors recognize the problem and are actively addressing the issues.
Making APC Easier
End users are making APC easier by going back to basics, often implementing APC functionality by creating custom configurations using standard programming tools provided by automation system vendors.
In the story, "Roll Your Own APC," Dan Cox explains how AOC Resins uses math blocks in DeltaV and procedures in SQL to crunch numbers, rather than relying solely on APC-specific software routines provided by their supplier.
Todd Gionet, process control engineer at Agrium Carseland Nitrogen Operations in Denver, Colo., uses an Invensys Connoisseur advanced process control application (http://tinyurl.com/9mgq3xp) with an MPC configuration and a linear programming optimization layer that provides target information to the constraint controller.
In the midst of all the advanced APC software, Gionet uses some straightforward controls too. "As an example, we run an advanced servo control cascade incorporating feed-forward control. Dead time compensation is faked with an adaptive first-order filter block. We currently have one important PID loop running a self-tuning algorithm."
"I might be too liberal in terms of what most control engineers consider advanced control, but keep in mind that cascade control was once considered advanced control," Gionet says.
Bertha of RDI Controls, agrees. As he explains in the story, "Aero-Derivative Advanced Control," he used user-defined subroutines to control combustion turbines. While the integration of multiple PID loops was a challenge, the individual control loops are fairly simple. "The blocks we created wouldn't be considered advanced by most engineers, but they allow us to easily add functionality and simplify the overall logic configuration," he says.
"Is this an APC type of system?" he asks. "Depends on your definition of APC. If you are looking for advanced from a control perspective then yes; if you're looking for canned routines that say they are advanced, then no."
In many cases, the more esoteric varieties of APC aren't needed. "We often make advanced control more difficult than it has to be," our anonymous engineer opines. "For example, I do pass-balance control with PID blocks—very simple. I am continually amazed at the number of people who poo-poo the idea, don't believe it can be done, and/or insist on writing some fancy code or use an MVC to do what is essentially integral-only control. And very slow integral control at that."
"The user community has to take greater ownership of technology selection," he says. "If you rely on a vendor, you are going to get what he is selling. If management has been sold that MVC is APC, then it makes it impossible to apply simpler, possibly better—or at least more cost-effective—tools to do the job."
End users are doing their part to make APC easier, and vendors are moving in the same direction. Table 2 lists some of the improvements APC vendors could make to keep end users happier.
Vendors on Board
Vendors are very aware that APC needs to be easier, and they are responding by adding features that speed and simplify implementation and by providing maintenance tools that alert end users when adjustments to existing applications are required.
Kinney says Invensys is working to make its software easier. "In general we foresee tools which make it easier to configure and maintain controllers getting more intuitive and easier to use. A part of this trend is an improvement in performance monitoring and diagnostic tools to help both the engineer as well as the operator to understand what the controller is doing and why. Some processes require nonlinear techniques, and we see this as improving in the future as well."
Robert Golightly, APC manager at AspenTech, says, "Traditional maintenance issues will be eliminated with built-in technology that automates much of the maintenance cycle and keeps controller models matched with plant performance, all without the need for the level of expertise required today."
Perry Nordh, product manager at Honeywell Process Solutions, says Honeywell has a system in place to help end users implement APC. "Benefits Guardianship Maximum (BG Max) is a flexible and comprehensive service program that provides advanced process control (APC) performance management services to help plant managers maximize the return on investment from Honeywell's advanced control and optimization solutions," he explains. "BG Max Services use an innovative and collaborative workflow approach to extend current customer engineering capabilities. This is done by actively monitoring APC solutions, detecting and prioritizing issues and proactively initiating resolution actions."
Other APC vendors offer similar support services, but users also want vendors to make APC easier for them to use without having to rely on supplier support.
"Opto 22 doesn't focus on many application-specific algorithms or HMI setup tools, although many users and vendors sell or share industry-specific tools for our PAC Control programming software," says Tom Edwards, Opto 22 senior technical advisor. "The tools that users and vendors provide for the PAC Control programming software are primarily in the form of subroutines, which are imported into PAC Control and incorporated into a control program."
"Opto 22 provides subroutines as free 'integration kits' on the website," he adds. "Other subroutines are shared by customers in the online OptoForums."
"APC will become available anytime everywhere, because the basic operation data and software are easily available at the DCS and PC level," says Saravanan Prabakaran, solutions consultant at Yokogawa. "The implementation cost and the maintenance cost of the APC will become cheaper than what it is currently. The tools that will be available for APC implementation will help in easy implementation of advanced process control."
The optimal solution is elusive, but perhaps the most promising path is to start with the simplest APC possible, one that delivers an acceptable percentage of theoretical benefits, but without stultifying complexity. Users and vendors alike must keep in mind that implementing APC is just a start, as without proper maintenance any solution will soon lose its effectiveness. Even with complex APC technologies, simpler is often better.