摘要：This paper presents the results of the model assessment performed on an industrial predictive controller applied to a propylene/propane separation system at Braskem, in Brazil, using the methodology proposed by Botelho et al. (2015a, b, c). Besides identifying the controlled variables with modelling uncertainties that were degrading the controller performance, the methodology diagnosed the main cause of the problem, which was related to model-plant mismatch for seven controlled variables, and due to the effect of unmeasured disturbances for other three controlled variables. This diagnosis enabled a more focused approach to the identification work, increasing the productivity of the controller revamp.
关键词：Control applicationsModel Based ControlModellingSystem Identification
