Improved Loop Control for Complex Systems

April 1, 2006
So-called “advanced loop tuning” produces the same results as traditional proportional-integral-derivative (PID) tuning, says Dr. George Cheng, co-founder and president of Cybosoft, General Cybernation Group Inc. (www.cybosoft), in Rancho Cordova, Calif. This advanced tuning focuses on making it easier for end-users to tune difficult plant loops, adds Don Morrison, ProfitSuite marketing manager for Honeywell Process Solutions (, Phoenix. “By its nature, advanced loop tuning makes it easier to tune single loops,” he states.

Advanced tuning also is a subset of broader loop management, Morrison suggests.“You could have a schedule. You’re going to tune every loop, one-by-one. But you’re potentially doing work that doesn’t need to be done,” he says. With loop management, control systems engineers and technicians can see what percentage of their loops are “bad actors that need to be at the top of my list,” he explains.

Noting that advanced loop tuning and management lessens disruptions to processes, Randy Miller, a Thousand Oaks, Calif.-based senior principal research engineer and manager of loop-management services for Honeywell Process Solutions, stresses the need for root-cause analysis. “It is so important because issues can get complicated, and can be much more than just tuning one loop,” Miller explains. “Root-cause analysis points in the direction of what corrective action—invasive or non-invasive—needs to be taken.” So what’s his bottom line for advanced loop tuning or management? Productivity. It’s not about advanced statistics or advanced this or that, Miller says, “but about solving important problems as quickly as possible.”

Complex systems

Even so,Cheng notes that the most widely used industrial controller is still the PID. Besides being simple, easily understood and easily implemented in hardware or software,it doesn’t require a precise process model to start up or maintain, he adds. However, Cheng says PID controllers can be frustrating and hard to use with complex systems that are typically nonlinear, time-variant, coupled, and possess parametric or structural uncertainties. Thus, on the factory floor, many loops are left in manual control because operators have trouble keeping them running smoothly in automatic, he observes.

But even newer PID auto-tuning and self-tuning methods that have been developed have problems, Cheng adds, because it’s difficult to keep a good process model online to retune the PID. Too, if the self-tuner is rule-based, “it is often difficult to distinguish between the effects of load disturbances and genuine changes in the process dynamics,” he says. That may cause controller overreaction to a disturbance, and creation of an unnecessary adaptation transition, he explains.

Cheng has developed a model-free adaptive (MFA) controller, based on the intelligent nature of neural networks, which has replaced and is replacing PID controllers and tuners. “It can adapt to new operating conditions and control complex systems without requiring process models,” he says. Some of the technology’s implementations include nonlinear, multivariable and other processes with large time delays, varying dynamics and changing operating conditions.

One company that is replacing its PID controllers with Cybosoft’s MFA controller is Siemens Building Technologies Inc.( ), Buffalo Grove, Ill.

With advanced loop tuning or management, the goal is increasing the productivity of the person doing the tuning, Miller emphasizes. While the aim is an excellent result, he notes, “it’s not good enough to achieve that result if it takes several hours.”

Kenna Amos, [email protected], is an Automation World Contributing Editor.