Online multivariable identification of a nonlinear distillation column using an adaptive Takagi-Sugeno fuzzy model

In this paper, an online multivariable identification approach has been developed based on an adaptive Takagi-Sugeno fuzzy model. The approach utilizes an evolving rule-base structure and model parameters to adapt the identified fuzzy model to process dynamic changes. Two new schemes have been proposed to improve the rule-base structure evolution. The first scheme smoothes the rule generation in the initial uncertain commissioning period of the identification. The second scheme diagnoses the inactive generated rules by examining their past activation record to delete them, leading to a more compact and efficient rule-base. A weighted recursive least squares (WRLS) algorithm is employed to estimate the rules consequent parameters. The proposed identification approach has been evaluated by a nonlinear distillation column benchmark to demonstrate its effectiveness to identify compact and accurate multivariable fuzzy models.