Double shot and score: Link your advanced control systems to predictive models to improve plant yield, profitability

Built in 2005, East Kansas Agri Energy's ethanol plant in Garnett, Kan., was designed to produce ethanol from locally grown corn. But soon after the plant was first commissioned—with demand for fuel ethanol soaring—management started looking for ways to increase production capacity and reduce energy costs.
By Malcolm Wheatley, Senior Contributing Editor March 1, 2009

Built in 2005, East Kansas Agri Energy ‘s ethanol plant in Garnett, Kan., was designed to produce ethanol from locally grown corn. But soon after the plant was first commissioned—with demand for fuel ethanol soaring—management started looking for ways to increase production capacity and reduce energy costs.

A further goal was to improve ethanol yield at the facility, which each year turned 13 million bushels of grain into 35 million gallons of ethanol.

The plant turned to Pavilion Technologies , a Rockwell Automation company, for its specialized predictive modeling capabilities. The plan: optimize dryer performance at the plant’s dry mill, where grain was dried and readied for fermentation, by predicting and controlling grain moisture levels.

Linked to an advanced process control system, East Kansas Agri Energy would then be able to dynamically optimize dryer production 24/7—maximizing throughput even as natural gas consumption was minimized.

Across industries, companies are turning to advanced control systems to deliver similar productivity improvements. Once the province of highly specialized engineers, today’s control systems are not only easier to use, but—as with East Kansas Agri Energy—capable of integration with predictive models that improve profitability.

The result? Plant-floor operators become better decision makers, and the links that control systems now provide to enterprise-level systems deliver more actionable information to the executive suite.

It’s a double boost to productivity and profitability.

At East Kansas Agri Energy, for example, the investment in Pavilion’s predictive control capability was credited with driving more than $2.5 million per year to the bottom line—an outcome described by plant manager Doug Sommer as “far exceeding our expectations.”

The increased revenue was achieved via a 12-percent increase in production coupled with a 9.9-percent increase in energy efficiency. In addition, ethanol yields climbed as a result of a 3.3-percent reduction in the standard deviation of moisture in the grain undergoing fermentation.

Truth will set you free

According to Pavilion’s senior product manager Ric Snyder, ethanol plants like East Kansas Agri Energy’s represent a good example of how modern control systems can deliver significant productivity improvements. The systems offer a “clean slate” perspective of how modern process plants should be manned and managed.

Snyder believes other industries are taking note. “The ethanol model might be becoming the norm,” he says.

Typically, for instance, ethanol plants are lightly staffed. “They are very lean operations, with maybe 50 people in the entire plant,” says Snyder. “Compared with other industries, [operators] are much more empowered to make changes. Give them the information that’s necessary, empower them to make those changes, and set them free to adjust temperatures, pressures, and set points as appropriate.”

That kind of empowerment hasn’t come about without significant strides in terms of ease of use—a message that comes from plant automation vendors. Emerson Process Management , for example, has worked hard to eliminate the need for skilled specialist engineers to set up and maintain advanced control systems.

“The idea is to change the paradigm,” says Bob Lenich, Emerson’s director of data management services and solutions. “By making advanced process control easier to use, you’re making it accessible to a greater cross-section of people in the plant.”

Aspen Technology has taken the ease-of-use principle one step further, launching a single software platform capable of handling all three of the company’s advanced process control ranges.

“It’s a huge step forward,” says Robert Golightly, advanced process control product marketing manager, Aspen Technology. “The traditional approach is separate software platforms for each product range. Now customers can buy one software package, and deploy and maintain multiple controllers—significantly lowering total cost of ownership.”

Is this for real?

Ethanol plants, it turns out, are in the vanguard of manufacturers embracing real-time cost and profit information—again using that information to make decisions about production parameters, aided in many cases by “what-if?” simulations that allow them to explore the impact of production changes without actually making them for real.

“They don’t have to wait until the end of the month to see what the cost per gallon was: They know it more immediately, and can make changes to production processes,” says Pavilion’s Snyder. “Changes to production parameters aren’t arbitrary ones made in isolation—they have a direct impact, in real time, and plants need to see the impact of those changes in real time.”

The result is a transformation of the role of the plant-floor operator, says Don Morrison, senior product marketing manager for Honeywell Process Solutions ‘ Profit Suite line of business, a collection of advanced process control and optimization products and services.

“Today the operator is more of a manager and less of a button-pusher,” Morrison asserts. “The task has evolved from exercising minute-by-minute control to one that is much more objective-driven. It’s about maximizing this parameter, and minimizing that parameter, and not just about keeping pressures and temperatures and flow rates within preferred limits.”

A recent example is the implementation of Honeywell’s Profit Controller advanced process control solution at the plants of Irish dairy products maker Glanbia Ingredients , where operations staff faced the challenge of maximizing efficiency and throughput without installing new and costly equipment.

The plant’s energy-intensive evaporation and drying process generates a powdered dairy product in accordance with specific characteristics like moisture content. Profit Controller was able to optimize dryer control to run closer to limits than was the case with purely manual control.

Product quality improved due to optimized moisture content. Additionally, a 5-percent increase in production throughput was attained as a result of the much more stable production process. Energy savings in turn decreased overall production costs, delivering a payback on the Honeywell investment in just six months.

The application provided yet another benefit: real-time reporting.

“We now have fast access to real-time production data,” states Donal Reilly, production manager at the plant. “We know how well we are doing on an hourly and daily basis in terms of production.”

Measure, track, improve

A similar story comes out of Irving, Texas-based ExxonMobil , where a long history of plant monitoring and optimization is being augmented by what the company calls “model-based performance monitoring.”

As at Glanbia, instead of comparing present performance against past performance—using data gathered from a plant historian, for example—the oil giant is more interested in comparing present performance with a theoretical “ideal” derived from what a software model of the process indicates is possible.

“If you can’t measure and track it, you can’t improve it,” says Apostolos Georgiou, online optimization section head at ExxonMobil. By defining key unit measures of a process, he explains, it is possible to identify lost opportunities and take corrective action.

The logic is clear, explains Harpreet Gulati, director of product marketing for Invensys Process Systems , from which ExxonMobil’s model-based control systems were sourced. Merely seeing a deviation from past performance doesn’t tell you what to do to get back on track, whereas a model provides much more insight to changes that will bring performance back where it should be.

“[Determining] the impact of cleaning, maintenance or process alterations, a model-based approach says: ‘This is your present configuration—and if you made these changes, then this would be the outcome,’?” says Gulati. “It shows people where the lost opportunities are, and where you are in terms of performance versus where you could be.”

That said, the “people” Gulati refers to have, to date, mostly been at plant level, directly involved in hour-by-hour operations. But it’s yet another characteristic of modern control systems that are reaching out to the executive suite as well.

“Senior-level executives in the enterprise don’t need to know information at the level of whether or not a given tank is filling up,” says Pavilion’s Snyder. “What they do need is aggregated performance information on a shift, grade, or product basis. They need to see this in real time instead of waiting for a daily or weekly report.”

The ease with which that communication can take place also is a factor. Standards like OPC and OPC-HDA are being overtaken by Web services, at last eliminating what Snyder describes as a control system’s “gotcha.”

Executive bookmarks

“There used to be something of a Catch-22 with control systems,” Snyder explains. “Any tool that could give an adequate amount of information was pretty much too complicated for the average executive to use. That’s no longer the case. There’s not much more to it than opening up a Web browser and pointing it at a bookmark.”

Or one can point to a manufacturing intelligence dashboard. According to manufacturing intelligence vendor Iconics , a specialist tool—BizViz Productivity Analytics—uses data mining technology to track real-time key performance indicators such as throughput, OEE, yield, availability, and uptime.

“It’s all about letting people make better decisions with confidence,” says Tim Donaldson, director of marketing for Iconics. “It allows them to say: This is what we need to do, and here are the metrics that say why.”

Developed with the intention of “delivering actionable information to people who can act on it,” Donaldson adds, Productivity Analytics is, by design, deliberately data-agnostic to provide as rich an information picture as possible—extracting data from plant historians, manufacturing execution systems, and OPC data from plant-floor devices and control systems.

This combination of ease-of-use and modeling capabilities is being credited with delivering competitive advantage through product differentiation. Aspen Technology’s Golightly, for instance, describes instances where customers are escaping traditional commodity-based businesses by customizing their products for particular applications.

“It can be difficult for makers of commodity polymers to compete with the lower costs of the Middle East,” says Golightly. “But they can tweak things at the catalyst and process-control layer to deliver particular polymer properties, thus differentiating their product from the commodity product—and do it again and again, supported by the simulation model, reprogramming their controllers on the fly.

The result is a low-margin “me too” business transformed into a polymer boutique, producing higher-margin specialist products to order.

All told, model-based performance monitoring, “what-if?” simulation, and improved ease of use add up to reveal one very handy edge over the predecessors of today’s advanced control systems: Indispensability.