MARCH 31, 2008--Adaptive controllers are gradually breaking out of the confines of research environments and finding their way into various industries. The advent of microprocessors and advanced computing platforms has catalyzed the shift from proportional-integral-derivative (PID) controllers to adaptive controllers in process industries such as petrochemicals, mining, steel, paper, food and many others.
New analysis from Frost & Sullivan (www.technicalinsights.frost.com), Continuous Adaptive Control--Technology Developments, finds that adaptive control has found application in varied and numerous segments because of its ability to improve performance in mechanical and nonmechanical systems alike. Systems involving material, money, supply and demand, and many others can successfully incorporate adaptive controls.
Adaptive controllers evolved from a solution for low-bandwidth applications to serving higher-bandwidth applications such as robots, spacecraft, complex machining processes, and many others. Missile control and guidance, fluid drive, industrial process control, power drag, firepower-control system, ship navigation, and other nonlinear mechatronic systems also increasingly depend on adaptive control technology.
"While designing such complex and highly cognitive systems, developers need to be conscious of the time sensitivity of input and output data," says Frost & Sullivan research analyst S. Menaka. "Scientists will also have to consider other factors such as machine-human interface, ability to create cognitive solutions in a stipulated time, real-time performance control of the system, architecture independence, data normalization, and other such related factors."
Due to significant time delays in operation or installation in systems with unknown dynamics, controllers were not used in mechanical systems; however, they have come a long way since then. After the advent of PID self-tuners, adaptive controllers became popular among commercial goods manufacturers and in other industries.
The operation of traditional controllers made offline training imperative and compelled scientists to take stock of the complexity of the systems, since greater system complexity jeopardized the reliability of the controller. Moreover, there had to be a trade-off among the error occurrences, correction identification, and steady performance of the system.
Next-generation adaptive controllers use the model changes and the process output to compute the integrated square error (ISE) for each of the three process parameters. Thus, after analyzing the low, middle, and higher combinations of all three parameters, it can devise 27 models.
"Through continued iterations, each model is normalized to a total ISE and the best value computed for each of the parameters is used in the next iteration as the middle value," notes Menaka. "Thus, the model will undergo interpolation with re-centering of the parametric values to ultimately reach an optimum corrective model."
Intense research in the future will focus on other techniques, in addition to data analysis, inferential estimation, integration of neural network-based technologies into existing systems, and predictive control. There is also a keen focus on the development of newer modeling methodologies. The existing systems are under observation to integrate them with the newly developed methodologies.
Meanwhile, adaptive control is increasingly emerging as an embedded technology. It is finding application in a number of domains and expects to have the ability to control higher-level functions. The control algorithm enhances the safety, economy, and reliability of the entire system and is responsible for the success or failure of the system.
"As a high-end application, the control algorithm may also be embedded into an enterprise resource planning system," observes Menaka. "Eventually, scientists could develop a hybrid system comprising interacting subsystems."