Minimum entropy-based performance assessment of feedback control loops subjected to non-Gaussian disturbances

Abstract

Control performance assessment (CPA) is a useful tool to establish the quality of industrial feedback control loops. While many current CPA techniques are developed solely for the controlled systems under the assumption of the Gaussian disturbance, the conventional minimum variance control (MVC) approach would not be applied when the disturbance distribution is non-Gaussian. In this paper, based on the information theory and the minimum entropy criterion, a more general CPA index for the feedback control loop subjected to unknown disturbance distribution is investigated. The fundamentals of MVC are first reexamined, and then an innovative performance index is given by incorporating the entropy. The feedback control algorithm based on minimum entropy, called MEC (minimum entropy control), is derived. MEC based CPA for the controlled system requires effective and systematic identification of the associated system models based on the closed-loop data. In this work, a new methodology based on the entropy criterion instead of the mean square error criterion is presented to estimate disturbances for the purpose of evaluating the performance of the control systems. To demonstrate the effective MEC based CPA method, both numerical and industrial examples are applied and compared with the MVC based CPA method.