Deciding to use advanced process control (APC) should be a no-brainer. However, because it needs a little brain power to get it up and running, many users remain reluctant to implement it—despite all of the many benefits it could give them.
To combat this prejudice and maybe gain a few converts, Mark Leroux and Dr. Eduardo Gallestey presented "The Business Value of APC" on the third day of ABB Automation and Power World 2010 this week at the George R. Brown Convention Center in Houston. Leroux is collaborative production management (CPM) marketing manager in ABB's Process Automation division, and Gallestey is ABB's product manager for APC and CPM.
"In many typical applications, APC can improve production by 3%, reduce energy consumption by 3%, shrink your carbon footprint and improve product quality," said Leroux. "Now. ABB can help bridge the gap between automation and financial objectives, but it's important to understand the differences between APC and classic process control."
Leroux explained that, in classic process control, a controller brings a single value close to a setpoint using a single manipulated variable, or it can involve a reactive process based on currently measured conditions, such as manipulating a heater to set the temperature in a tank.
"There are several main advantages to APC," said Leroux. "Because several actuators act simultaneously to achieve the best trade-offs, the application can achieve its goals in the fastest or most efficient fashion. In addition, APC's predictive character allows early recognition of potential violations and timely implementation of remedies, which means improved stability and reduced defects. It also lets applications work nearer to process constraints, which further reduces costs. The disadvantages of APC are that it requires users to make a modeling effort and take on a computational load, but both of these can be solved."
Leroux added that APC's three main strategies are fuzzy logic, neural networks and model-predictive control (MPC). He explained that fuzzy logic is best used for cases where rules as to how to react to process conditions are known, while neural networks are best used for cases where online measurements are hard to obtain reliably, such as with soft sensors, and MPC is best used for processes with strong coupling among variables, competing optimization goals and limiting process constraints.
Basically, MPC consists of a method for handling disturbances and forecasting changes, and its main ingredients are a plant model and definitions of objective functions. The model is used to predict system behavior some steps into the future, but it requires solution of the optimization problem at every sampling time.
"However, an APC platform shouldn't dictate which strategy to use," cautioned Leroux. "Your process should! For example, an industrial steam plant might be able to use alternative fuels and then can use APC to balance the performance intermittence and variability cause by those fuels. This can be very difficult to handle with classic process control, but it's easier with APC.
"Beneficial results with APC can include best-case optimization of energy purchased and sold; optimal trade off among steam consumers to maximize profit; consistently using the lowest fuel mix that meets the constraints; and optimal reaction to disturbances such as alternative fuel failures and varying steam demands."
Gallestey added that users can deploy ABB's cpmPlus Expert Optimizer software to help create model drawings of applications and plants, including valves and other devices, to add in their dynamic performance values via an OPC server, to write in setpoints, and then to secure outputs to reflect what's actually happening on their processes. Consequently, when the resulting APC program is initiated, values about demand and costs can be plugged in, and the APC system will show how the application would respond—all without any additional programming required.
"Classic process control can do things automatically," Gallestey said, "but it can't respond to other kinds of variables, such as cost constraints."
