An automatic tuner with short experiment and probabilistic plant parameterization
Research output: Contribution to journal › Article
A novel automatic tuning strategy is proposed. It is based on an experiment of very short duration, followed by simultaneous identification of LTI model parameters and an estimate of their error covariance. The parametric uncertainty model is subsequently exploited to design linear controllers with magnitude bounds on some closed-loop transfer function of interest, such as the sensitivity function. The method is demonstrated through industrially relevant examples. Robustness is enforced through probabilistic constraints on the H∞ norms of the sensitivity function, while minimizing load disturbance integral error (IE) to ensure performance. To demonstrate the strength of the proposed method, identification for the mentioned examples is carried out under a high level of measurement noise.
|Research areas and keywords||
Subject classification (UKÄ) – MANDATORY
|Journal||Int. Journal of Robust and Nonlinear Control|
|State||Published - 2017 Jul|