Data Acquisition / Optimization

Model-Predictive Control Without the Lab Coats

Odds Are MPC Can Optimize Productivity, Materials and Energy in an Industry Near You.

By Paul Studebaker

AF14 Banner

As the advanced control and optimization arm
of Rockwell Automation, Pavilion is a combination of people, products and a process. "We have talented chemical engineers, process guys and a powerful
model-predictive controller (MPC) with an established presence and reputation in the traditional process industry applications like chemical plants and refineries," said Jim Miller, Pavilion business director, Process Business, Rockwell Automation. The company's MPC capabilities are on display this week at Automation Fair in Anaheim, California.

"Now we're extending MPC into other industries that Rockwell Automation serves, including natural gas liquids, mining and consumer packaged goods (CPG), where MPC hasn't been commonly used."

"The objective is to simplify MPC and offer it in different form factors," said Stephen Pulsifer, director of process marketing, Rockwell Automation, "without the lab coats."

Rockwell Software Pavilion8 Model-Predictive Control software makes MPC easier to implement. "When you look at advanced control, you typically see a lot of math," Miller said. "We put the engineering into the software and make it easy to tune, to select data sets and model sets. The customer is more comfortable, and they're able to use it."

PlantPAx MPC allows users to put MPC in the control layer—to use the control system instead of a server to process the data. "It's easy to understand and maintain, and you can afford to go after smaller projects where you might not think it's worthwhile to use MPC," Miller said. For example, applying a 3 x 3 matrix on a chocolate tempering machine has resulted in a 30% improvement in productivity.

PlantPAx MPC can handle as many as five 10 x 10 matrices. For some applications, MPC can be used to perform multi-variant regulatory control. "Running at sub-second speeds, it can replace PID to directly control a difficult process," Miller said.

MPC applications range from more traditional optimization saving a percentage of an expensive chemical used in a mining operation that added up to $100 million per year, to optimizing a multiple unit process line in a consumer packaged goods plant. "Where's the bottleneck? Is it the freezer? The fryer? It's not always the same. It moves around, depending on how the plant, equipment and product are running," Miller said.

In that CPG plant, using MPC for real-time constraint management increased production 6-8% while reducing energy consumption about the same amount, Miller said. "It functions as a ‘quality controller' that identifies the bottleneck and automatically does what's needed to relieve it."

Pavilion real-time optimization (RTO) has been used in many facilities to manage utilities. It can reduce energy costs, typically 10%, by optimizing the selection of turbines, fuels and loads to run at any given moment based on production requirements and the current prices of electricity and fuels.

In the works for future release are real-time models for chiller, boiler and turbine energy efficiency that users will be able to simply "attach to the data and retune on the fly," Miller said. Optimization runs automatically and operators are encouraged to keep them engaged by faceplates that show in real time the costs of operations and how much the advanced controls are saving.

Traditional issues with keeping models tuned and operating correctly are addressed by proactive "value-based support," Miller said. Pavilion experts provide quarterly reports to clients on how much their operations are saving, and can remotely dial in, monitor and tune systems as needed. "As we embed more of this capability into the controller, plants will be able to do this themselves."

Rockwell Automation is offering turnkey MPC as a "ValueFirst" methodology, making proposals that speak directly to return on investment (ROI). "We send in the engineers, they review the process and quantify the performance improvements available by reducing variability. They predict the economic ROI, and clients can see that before they buy," Miller said. Many cases show paybacks of less than a year, sometimes one or two months.

"The studies easily convince the process and controls people. The challenge is to articulate it to the C level. There, we use success stories—examples where we've done it before. We show them power of the technology, and they usually sign up," Miller said. Plants often start with a smaller project and then build on the results.