Advanced Process Control Finds Future Through Past

Sept. 1, 2004
Models, control and optimization algorithms, and real-time data combine to predict healthier processes

It’s not, “You say ‘tomato’ and I say ‘tomahto.’ ” With advanced process controls (APCs), you either know what they are or you don’t. If you do know, you know they give for-real, money-in-your-pocket rewards to manufacturers—rewards so impressive that very few will speak publicly about using them. If you don’t know, it’s time you did, don’t you think?

As used today, the technology traces its beginnings to the late 1950s. It predicts process outcomes of varied operating conditions. APC relates manipulated variables (e.g., condition of a process heater) and control variables (e.g., temperature), providing multivariate control and also adaptive tuning and predictive/process diagnostics. Generically, APCs consist of three components: a computer-simulation model that integrates process knowledge and historical data; control and optimization algorithms; and current, real-time process information.

Justification for implementing advanced process controls comes from improved performance, because they stabilize operations. They remember different operating scenarios, so your operators won’t have to. With adaptive control, you could update an existing control model. For example, if the tool you’re using is a proportional-integral-derivative (PID) tuner, then you could develop new tuning parameters. With predictive control, you handle time delays more effectively in the existing controller, to optimize throughput and reduce waste.

Be advised, though, the real APC performance-killer is inadequate monitoring and maintenance. So before installing and then ignoring them, consider this: If you use model predictive control (MPC), it “will degrade 10 percent per year. Typically, at 30 percent degradation, users re-step,” says Warren Mitchell, director of advanced applications for Edmonton, Alberta, Canada-headquartered Matrikon Inc. “The controllers will either die a smoldering death in three to five years, or very quickly, within six to 12 months, if there’s no proper support.”

And support is what manufacturers have less of now, due to downsizing and squeezed budgets. In general, with fewer control engineers, there’s a pinch on plant performance monitoring and maintenance. A condition-based approach can overcome this, Mitchell believes. His company’s ProcessDoctor and ProcessACT “fit into preventive and predictive maintenance strategies.” Combine those technologies with Matrikon’s Web-based ProcessNET and you get a view into operations via a dashboard, says Gail Powley, Matrikon’s senior industry adviser. She says that the product trio has reduced Matrikon users’ maintenance costs by 30 percent and increased profitability by 5 percent.

As to causes of maintenance headaches, Mitchell says, “75 percent of regulatory controls (PID loops) in a facility are underperforming. Many control professionals spend most of their time in fire-fighting, or reactive, mode. What we’re trying to do is turn the whole of maintenance on its head,” he says.” That would mean turning from a reactive/run-to-failure model and possibly the preventive model, in which assets are often replaced too early, to the forward-looking predictive-maintenance model.

Better process forecasts come through a breakthrough that Pittsburgh-based Adaptive Resources says it has made with its model predictive control software. In the identification algorithm of the modeling engine component of its core technology, “we can identify open-loop process models from closed-loop, dynamic process data,” explains Director of Engineering Chetan Chothani. Most users have to break the PID loop and then artificially excite the process to generate response data to identify the open-loop process models, he says. But his modeling engine takes process data, without artificial excitation, and extracts the open-loop process model, Chothani says.

Forward to the future also comes through the other half of his company’s core technology. It’s a control engine that he says uses the process model selected by the modeling engines’ identification algorithm “to predict the future direction of the process and generate a series of control outputs, based on the history of all the variables and predictions.”

Users find value

And with as many variables encountered in refineries, petrochemicals, food processing, water treatment and elsewhere, there are many specific ways APCs prove their worth. At Motiva Enterprises, LLC’s Norco Refining facility near New Orleans, a company-vendor team has worked for nearly 20 months on installing Shell-licensed-to-Yokogawa multi-variable control and on-line modeling packages at four process operations, including two ethylene units, says Merle Likins, a principal advanced control engineer in Yokogawa’s Houston office. The retrofit involves replacing a pneumatic-instruments system and an older distributed control system (DCS), he notes. Commissioning will be in late 2004 or early 2005, he says, and the system should be fully functional in 2005.

Another process example is at American Water Services Canada Corp., in Grand Bend, Ontario, Canada. The company uses Mantra advanced controls, from ControlSoft Inc., of Highland Heights, Ohio. “We have it in our pre- and post-chlorination,” explains Gary Maxwell, operations manager of the 120 million-gallons-per-day (MGD) water-treatment plant in Grand Bend and the 42 MGD Port Stanley water treatement plant. The PID loop picks up the pre-chlorination water concentrations. “We have a set point that Mantra automatically controls. It’s flow-paced, so if flows vary, Mantra changes the set point,” he says about the Grand Bend operations, the location where the technology is now being used.

The controller determines if the chlorine concentration in the end-of-treatment clear wells is too high or low, Maxwell explains. “If it’s either high or low, then the technology will adjust the level.” The technology, which connects pre- and post-chlorination as well as discharge values, provides an estimated 25 percent savings “because we’re not manually adjusting the chlorine.”

But regardless of the process in which such controls are used, you’d better monitor, say Matrikon’s Mitchell and Powley. The monitoring must reach into every level of control systems—from the first-level instrumentation layer, to regulatory controls, then analyzers and online controls and, finally, APCs themselves. “The issue is not with one or the other layer; it’s with all layers. If any of these break, then the others can’t deliver,” says Mitchell. Be cognizant of inherent process time issues, too, suggests Bill Gough, president of Richmond, British Columbia, Canada-based Universal Dynamics. “In every case in a plant where there is a product, there is a time-delay problem. All the subsystems need to be working perfectly well.”

Be aware of what causes poor operational performance—controls degrading to the lowest common denominator—Gough also advises. The result is “an insidious ineffectiveness,” he believes, “because there’s nothing patently obvious that something is wrong when you simply detune the PID by turning the knob. The symptom goes away. But you’ve made no progress toward finding the solution. From that day forward, you’re going to pay a price in poor performance.” Part of the problem, he remarks, is that working with PID-type tools means having to do something different each time you work with them. Also, you need to understand for which plant level you’re installing APCs. Historically, that meant something residing at the supervisory level to optimize targets—and setup has been fairly expensive and complex to implement, says Gough. “Bigger chemicals companies are finding that traditional APC technologies are improving premium applications, such energy use, products produced and economic value.”

To get into the guts of the enterprise, though, his company’s BrainWave technology fits across a broader and deeper base, he notes. “It allows it to be brought down, to let each process loop run better throughout the plant.” That includes, he says, “everything, starting from troublesome single-loop applications to small-dimension, multivariate problems.” One example is use in distillation columns. But what counts, no matter where the technology is used, is the users’ operational point of view, he adds.

Operational relief also comes because model-predictive technology (MPT) forces the right answer, he declares. “The controller needs the right model. That forces you down a better road. It forces you to understand what’s going on in the process.” So what’s the advanced control approach? Running different products that cause signal changes in the loop, says Gough. The controller retrains under the new conditions, which allows end-users to associate the change with the correlated process change. “And then you can save it (the control change) and recall it. You can then build up these sets of models so that, over time, your control systems run well over all conditions. It evolves,” he says.

APCs are overcomers and enhancers, with exceptional potential. They leverage existing knowledge and skill of operators and engineers, real-time process data, dynamic process variations, advanced mathematics and connectivity. They provide steadiness. They offset or eliminate poor-performance penalties.

APCs can help manufacturers trade the guess-work and headache involved in PID tuning for boosted performance. And, trade aggravation for success and higher profitability. In today’s produce-to-order, zero-defect, just-in-time global market, APCs can keep your company’s clients and shareholders happy.

See sidebar to this article: Pavilion produces polypropylene properly

Sponsored Recommendations

MSD-SLC16G

CLICK industrial memory card, 16GB microSD. For use with all products with microSD memory card slot.

C0-12DRE-D

CLICK Ethernet Analog PLC, 24 VDC required, Ethernet and serial ports, Discrete Input: 4-point, DC, Analog Input: 2-channel, current/voltage, Discrete Output: 4-point, relay, ...

C2-FILL

CLICK PLUS option slot cover.

USB-CBL-AMICB6

AutomationDirect programming cable, USB A to microB-USB, 6ft cable length. For use with Productivity1000 and Productivity2000 CPUs and most USB devices.