Comparative analysis of robust estimators on nonlinear dynamic data reconciliation

• Published in: Computer Aided Chemical Engineering Vol. 25; pp. 501 - 506
• Main Authors: Prata, Diego Martinez, Pinto, José Carlos, Lima, Enrique Luis
• Format: Book Chapter
• Language: English
• Published: 2008
• Subjects: Nonlinear dynamic data reconciliation; robust estimation and gross error; Nonlinear dynamic data reconciliation; robust estimation and gross error
• ISBN: 9780444532275; 0444532277
• ISSN: 1570-7946
• DOI: 10.1016/S1570-7946(08)80088-0

This paper presents a comparative performance analysis of various robust estimators used for nonlinear dynamic data reconciliation process subject to gross errors. Robust estimators based on cost functions derived from robust probability theory reduce the effect of gross errors on the reconciled data, avoiding the traditional iterative requirement procedures. The following robust probability functions were compared in this paper: Cauchy, Fair, Hampel, Logistic, Lorentzian, Normal Contaminated and Welsch. As a benchmark for this study it was adopted a nonlinear CSTR frequently reported in the process data reconciliation literature. The comparative analysis was based on the ability of the reconciliation approaches for reducing gross errors effect. Although the presence of constant biases has represented a problem for all the analyzed estimators, Welsch and Lorentzian cost functions, in this order, have shown better global performance.