GREY BOX MODELING FOR CONTROL - QUALITATIVE MODELS AS A UNIFYING FRAMEWORK

Grey box modelling traditionally reflects that both a priori and exper imental knowledge are being incorporated into the model-building proce ss, where both of them may exhibit uncertain character. A brief invest igation into various grey box modelling approaches reveals that they d iffer mainly with respect to the required model accuracy. Moreover, th e goal of the model application has to be considered in the model buil ding, since this goal defines the desired accuracy of the model, which is represented as model uncertainty. This paper advocates the view th at grey box modelling is model building which incorporates uncertainty description. Qualitative differential and algebraic equations are pro posed in this paper as a unifying framework for development of dynamic models with uncertainty. The steps in the model development cycle are defined for this unifying framework, wherein the computational comple xity issues are addressed at each step. It is also shown how qualitati ve differential and algebraic equations can be specialized to importan t well-known grey box model forms such as robust models with parametri c uncertainty, constraint qualitative differential equations and digra ph models. The presented concepts and grey box model forms are illustr ated on a simple example: a heat exchanger with bypass.