Data-driven control performance degradation recovery as an extension of loop transfer recovery

Abstract

This article investigates the problem of data-driven control performance degradation recovery from performance degradation modeling and a concrete control structure associated with data-driven implementation methods. The control performance degradation recovery problem is inspired by and serves as an extension of the classical loop transfer recovery (LTR). In the first part, a loop performance degradation model is proposed where performance degradation is quantized, between the “ideal” full state-feedback control and the “real” dynamic output feedback control scenarios, by the error at the plant input and output. It is demonstrated that control performance degradation recovery is consistent with optimizing the robust stability, promoting the ability of disturbance rejection and improving the fault detectability. This consistency promises various applications of the model, for instance, to the design of fault-tolerant control systems. In the second part, a concrete realization of the control performance degradation recovery problem is presented as optimizing an $H_{\infty }$ norm that is dependent of the Youla parameter system. Data-driven formulation of the problem is introduced together with data-driven $H_{\infty }$ norm estimation involved in it. In the third part, data-driven optimization method for the $H_{\infty }$ norm is developed for real-time performance degradation recovery. The data-driven $H_{\infty }$ norm estimation and optimization method also links the classical model-based $H_{\infty }$ norm optimization to the data-driven field.