Control loop performance assessment using ordinal time-series analysis

E. C. MARTÍNEZ; C. DE PRADA

Computer-Aided Chemical Engineering

Elsevier

Lugar: Amsterdam; Año: 2007; p. 261 - 266

Both safety and economy of process plants are heavily dependent on the properfunctioning of several thousand control loops regulating or optimizing unit andplant-wide performance over a wide range of operating conditions. In this workordinal methods for analyzing controller´s error are proposed by focussing onmerely comparing error values beyond a proper control horizon. The local orderstructure of a well-performing controller is characterized as a symbolicsequence of equally probable permutations. A new performance index based onthe complexity of the error time series measured by the permutation entropy ofordinal patterns is proposed.