Premium
Generalized likelihood ratios for gross error identification in dynamic processes
Author(s) -
Narasimhan Shankar,
Mah Richard S. H.
Publication year - 1988
Publication title -
aiche journal
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.958
H-Index - 167
eISSN - 1547-5905
pISSN - 0001-1541
DOI - 10.1002/aic.690340810
Subject(s) - identification (biology) , process (computing) , error detection and correction , feature (linguistics) , likelihood ratio test , computer science , mathematics , process control , control theory (sociology) , control (management) , mathematical optimization , algorithm , statistics , artificial intelligence , linguistics , philosophy , botany , biology , operating system
Gross error identification in dynamic processes is important in ensuring proper process control. This paper describes the application of a generalized likelihood ratio (GLR) method for identifying gross errors caused by biases in measuring instruments and controllers, process leaks, and failure of controllers. As shown in the application to steady state processes (Narasimhan and Mah, 1987), this method provides a general framework for identifying different types of gross errors whose effect on the process can be modeled. An important feature of the work is the treatment of closed‐loop dynamic processes. The formulation of the hypotheses of the GLR method proposed by Willsky and Jones (1974) is extended for this purpose. For estimating the time of occurrence of the gross error, a simple chi‐square test on the innovations (measurement residuals) is used, which is computationally more efficient than the method used by Willsky and Jones. Through simulation studies of a level control process the appropriate selection of parameters of the GLR method is investigated.