Performance Assessment: User-Specified Benchmark

Abstract

Control loop performance assessment has been extended to many situations, and many approaches have been developed as discussed in the earlier chapters, e.g., performance assessment of: 1) SISO feedback control systems (Desborough and Harris 1992; Stanfelj, Marlin, and MacGregor 1993; Kozub and Garcia 1993; Lynch and Dumont 1993; Tyler and Morari 1996); 2) feedback control of nonminimum phase SISO systems (Tyler and Morari 1995); 3) MIMO feedback control systems (Huang, Shah, and Kwok 1995; Huang, Shah, and Kwok 1996; Harris, Boudreau, and MacGregor 1995; Harris, Boudreau, and MacGregor 1996). The portion of a process output that is feedback controller invariant determines the minimum variance achievable theoretically and characterizes the most fundamental performance limitation of a system owing to the existence of time-delays or infinite zeros. Practically, however, there are many other limitations on the achievable control loop performance. The existence of nonminimum phase or poorly damped zeros, sampling rate, amplitude and/or rate constraints on control action, robustness constraints etc. are examples of such limitations; therefore, a feedback controller that indicates performance reasonably close to minimum variance control does not require further tuning (if the variance is the most important issue) while a feedback controller that indicates poor performance relative to minimum variance control is not necessarily a poor controller. Further analysis of performance limitations and comparison with more realistic benchmarks are usually required. Performance assessment with minimum variance control as a benchmark requires minimum effort (routine operating data plus a priori knowledge of time-delays), and serves as the most convenient first-level performance assessment test, therefore, (if the variance is the main point of interest). Only those loops that indicate poor first-level performance need to be re-evaluated by higher-level performance assessment tests. A higher-level performance test usually requires more a priori knowledge than just a knowledge of time-delays. This chapter addresses practical issues that are of interest for such a higher-level performance assessment test.KeywordsTransfer Function MatrixNonminimum PhasePartial Fraction ExpansionMinimum Phase SystemDither SignalThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.