| Many Control Systems today include advanced and sophisticated built-in control capability as they are deployed to use. They include finely tuned adaptive controllers. These control systems maintain a broad control region under which the system is controlled. However, as they continue to operate and function under different environments, scenarios and users, performance tends to adopt a biased profile with time. This persistence in biased operation increases the susceptibility of the control system to premature wear and tear, degraded performance and premature failure. This research proposes a framework to mitigate that risk by adapting the controller in real-time to the specific conditions surrounding the system, accounting for the actual user, environment, and scenario.;The proposed algorithm uses a triggered system identification approach to re-identify the current state of the system structure and parameters. It replaces the old identified system model with a better fit one. This mechanism allows the control system to continuously learn changing environment and operator behavior with minimal expert system knowledge. The algorithm also selects an optimal tuned controller based on a criterion that includes the newly identified system model as well as controller performance specification.;This research verifies and validates the hypothesis proposed through an experimental implementation of a real-time adaptive wind tunnel control system. The experiment represents a systems-engineered approach that comprises an integration of multiple modern hardware and software tools. |
