文章基本信息

标题：Constrained model-based design of experiments for the identification of approximated models
本地全文：下载
作者：Marco Quaglio ; Eric S. Fraga ; Federico Galvanin 等
期刊名称：IFAC PapersOnLine
印刷版ISSN：2405-8963
出版年度：2018
卷号：51
期号：15
页码：515-520
DOI：10.1016/j.ifacol.2018.09.197
语种：English
出版社：Elsevier
摘要：AbstractThe identification of an approximated model, once an opportune mathematical structure is selected, requires both a precise estimation of its parameters and the determination of the range of conditions in which the model provides accurate predictions, i.e., the domain of model reliability. A variety of model-based design of experiments (MBDoE) techniques are available in the literature for designing highly informative trials for the precise estimation of model parameters. Available MBDoE methods assume that the model structure is exact in the formulation of experimental design metrics. Hence, in the presence of an approximated model, the employment of conventional MBDoE approaches may lead to the collection and fitting of data at conditions where the model performance is very poor, thus leading to the degradation of the fitting performance and a loss of model predictive power. In this work, an iterative framework for the identification of approximated models is proposed in which the MBDoE step is constrained to the domain of model reliability. The method is tested on a simulated case study on the identification of an approximated kinetic model of catalytic ethanol dehydrogenation.
关键词：KeywordsModel approximationParameter estimationInformation analysisOptimal experiment designAdaptive algorithms