MODELING AND CONTROL OF INDUSTRIAL SYSTEMS: A HYBRID APPROACH

ABSTRACT: Some problems of hybridization in the area of industrial systems are considered. The term “hybrid approach ” is used for variety of classes of systems and approaches. Firstly, it describes the aggregation of different intelligent techniques (Neural networks, Fuzzy logic, Rule based, Statistical, Analytical). Secondly, it allows several types of dynamics: continuous, discrete, semi-continuous and their combinations to be treated. Thirdly, an integration of several control approaches is possible. In the industrial systems the measurement data contain inaccurate information due to imperfect instruments and/or gross errors. Approaches and methods for steady state and dynamic fault measurement diagnosis and data reconciliation are presented. The industrial plants are characterized quite often with immeasurable or hardly measurable key variables. The problems of implementation of hybrid intelligent technologies for inference measurements are under consideration. An attempt of using a generalized approach in treating continuous, discrete and semi-continuous processes, which are typical for the industrial systems, is presented. The possibilities of aggregation of different intelligent technologies are considered. Using hybrid approach a regime depending multiobjective function is proposed in order to emphasize the relevant current situation and operational requirements (accuracy, safety, efficiency, ecological norms etc). It is shown that in the cases of variable constraints, depending on operational conditions and rectangle input/output scheme, the combination of intelligent control technologies gives very promising results