Measurement structure design for multicomponent distillation column with specific estimation objective

In this work, the problem of designing the estimation structure to infer some of the effluent compositions in a multicomponent distillation column is addressed, with the aim to determine the structure that yields the best estimator functioning in the sense of a suitable compromise between performance versus complexity, regardless of the particular estimation algorithm employed. The structure design involves the selection of (i) the number of modeled components, (ii) the innovated components, and (iii) the number and locations of sensors. The consideration of the problem within the adjustable-structure geometric estimation framework in the light of the column characteristics leads to a methodology with an analysis stage, where structures are screened and candidate ones are selected, followed by a synthesis stage which yields the structure with the best estimator functioning. The proposed approach is applied through simulations to an industrial-type hexacomponent distillation column, finding that the estimation task can be adequately performed with an algorithm whose number of ODEs (77) is considerably smaller than the one (2849) that results from the direct application of the standard EKF.