(305c) Data-Based Proportional-Integral Controller Assessment and Retuning

Proportional-integral (PI) control is the dominant control structure implemented within the process industries, often to less than acceptable effect.  In this work, a new algorithm is presented for performance assessment of PI control of self-regulating processes within single-input single-output (SISO) control loops wherein disturbance rejection is the main concern. Two methods are used for diagnosis of sluggish or aggressive tuning: (i) pattern classification of the estimated closed-loop load response to disturbances, and (ii) determination of the process output’s Hurst exponent through the method of detrended fluctuation analysis. The Hurst exponent gives a measure of long-standing positive or negative correlations in the system’s output, which are indicative of poor controller performance. In an earlier work, the authors applied the method of detrended fluctuation analysis for assessment of a more general class of control loops; however by focusing on the simple PI structure, the present work allows specific recommendations on retuning of controller parameters to be made. The methods used herein rely solely on the measured process output, and therefore, a priori model information on the plant to be controlled is not required. The diagnosis and retuning techniques are demonstrated using results from several simulation examples.