Distillation is the most widely used separation process in the chemical and petrochemical industries. The principal function of a distillation column is to deliver products meeting certain composition specifications. The need to conserve energy and to produce products in an optimum manner provides incentives for composition control of fractionator products. Although on-line analyzers such as Gas Chromatography(GC) have the advantage of directly measuring the product quality, the composition control has not been preferred yet because the online analyzers still suffer from large measurement delays, high investment and maintenance costs and low reliability.
In that respect, the estimation of the product compositions is an important problem related to the inferential control of distillation columns. Inferential measurement of the desirable attribute from other, more easily monitored parameters has a particularly exciting potential. Effective use of the secondary measurements such as tray temperatures will help us solve the measurement problems related to the composition control of distillation columns.
In this thesis, the estimation problems which arise from the lack of the composition measurements for distillation control are studied: For designing the linear static estimator, two approaches (i.e. projection and regression type estimators) have been discussed. The relationships between the estimators have been analyzed by considering an inversion problem. Design guidelines on the design of composition estimator via Partial-Least- Squares (PLS) have been presented: (1) the determination of the number of factors; (2) the selection of the secondary measurements; (3) the effect of the transformation and scaling. These guidelines may help us design a robust composition estimator.
In order to overcome the limitations of the linear composition estimators, which came from the nonlinearities of distillation system, a nonlinear static estimator based on the open equation...