Premium
Input design for model order determination in subspace identification
Author(s) -
Misra Pratik,
Nikolaou Michael
Publication year - 2003
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.690490820
Subject(s) - subspace topology , identification (biology) , multivariable calculus , fractionating column , process (computing) , computer science , uncorrelated , distillation , control theory (sociology) , system identification , heat exchanger , mathematical optimization , mathematics , biological system , artificial intelligence , engineering , control engineering , chemistry , data mining , chromatography , statistics , botany , control (management) , biology , measure (data warehouse) , operating system , mechanical engineering
Subspace identification methods require that the inputs to the process being identified be persistently exciting. This may be inadequate for subspace identification of ill‐conditioned multivariable processes, because the process model order may be underestimated, leading to subsequent identification of poor models. To remedy the problem, it is proposed that inputs must be used that excite a process to be identified in a way that produces as uncorrelated process outputs as possible. This can be accomplished either in open or in closed‐loop fashion. Simulations on a high‐purity distillation column and on a heat exchanger illustrate the merit of the proposed approach.