Combining first principles with black-box techniques for reaction systems

A control-oriented hybrid model structure combining first principles models with standard black-box techniques for modelling the nonlinear dynamics of reaction systems is presented in this paper. The approach is formulated in a general framework for continuous stirred tank reactors and analysed in details through a case study of a reactive distillation column. The approach is based on easily established mass balance equations, the stoichiometry of the system, as well as model reduction techniques. Using this approach, the nonlinear dynamic model for a reaction system can be represented as a (usually dominant) linear part in terms of a combined input and a nonlinear part. The choice of combined inputs and the model structure is motivated by some general control objectives for this class of systems. Progressive identification of this model structure can be performed when the linear part is dominant. Application of this modelling approach to real process data is presented. This model structure has been successfully used in an IMC scheme for an industrial reactive distillation column.