Automatic Tuning of Linear System Controller Parameters via Extremum Seeking

This thesis presents a study on the automatic tuning of linear system controller parameters via Extremum Seeking (ES), with application to a thermal process. The research integrates classical control principles with modern model-free optimization techniques to develop an adaptive framework for real-time PID parameter tuning. The theoretical foundation is established through an overview of control theory, highlighting the transition from classical and modern control paradigms to robust and adaptive approaches. Emphasis is placed on the Proportional–Integral–Derivative (PID) controller, the most widely used feedback mechanism in industrial systems due to its simplicity and reliability. Conventional tuning methods—such as Ziegler–Nichols, relay feedback, and iterative feedback tuning—are reviewed, underscoring their limitations in complex or uncertain systems. To address these challenges, discrete-time Extremum Seeking is adopted as a flexible, model-free optimization method capable of continuously adjusting controller gains to minimize a performance-based cost function. A detailed modeling and control design procedure is developed for a thermal plant driven by an electric heater. Simulation results demonstrate that the ES-based PID tuning achieves improved performance over manual tuning, offering faster settling time, reduced overshoot, and smoother control action. Finally, multi–set-point experiments validate the algorithm’s effectiveness across varying operating conditions.