Identification of multivariable fuzzy systems

The paper deals with the basic principles of modeling of large production fuzzy systems with the properties of uncertainty. The problem of identification of fuzzy system as a problem of preselection of fuzzy model expressed by production rules and providing the at least some non-negative functional deviations of the current and the calculated output has been formulated. The problem of identification consists of two interacting problems: structural and parametric identification. To ensure the required accuracy of fuzzy model algorithms of parametric and structural identification are interacting in accordance with the terms of adequacy, the convergence and the transition from one the algorithm to another. The efficiency of hybrid identification largely depends on the quality of information received by the production, which depends on the availability of excess accurate data, noise and measurement errors, as well as data characterizing the non-stationarity of characteristics of the equipment. The proposed method allows for a limited array of data to receive an adequate fuzzy model of a fuzzy production system.