Monitoring and diagnosis of PI controllers

A PI controller is one of the most widespread controller types used in industry. If the structure and order of a controller is restricted (e.g. a PI controller as is the case in this paper), this may cause worse output variance as compared to the minimum variance (MV) benchmark. Finding the best achievable performance boundary for this controller is a challenging task as it leads to a non-convex optimization problem. This paper presents a fast, simple and comparably accurate method of estimating the PI performance boundary. The proposed method finds the minimum output variance with a PI controller (PIMOV), and this is then used to obtain an accurate PI trade-off curve for a SISO system which can be used in monitoring and diagnosis. A Full theoretical analysis is presented, and a simulation study is performed for the assessments using the Matlab optimization toolbox. Several simulation examples are incorporated to show the usefulness of the proposed algorithm.