R.V. Litovkin

With the increase of incoming information in modern control systems is becoming increasingly difficult to control the process of transformation at different stages. Under the control means the possibility of obtaining information about the «correct» the system and management in order to guarantee the performance of the system. Typically, complex systems of information processing are based on the tree (hierarchical) structure. Leaves the structure are the primary converters, directly interacting with the object. This is followed by various functional transducers, performing in the general case, the reduction «complexity» of information. As a result, the output of a tree transformation (the root) has only the information with sufficient quality to carry out certain actions, such as diagnosis and management of the facility. Spatial branching tree structure of different levels can lead to either significantly change the qualitative properties of these levels, or their fault. Therefore required to control the characteristics of transformation and adaptation to carry out its parameters and structure. Technically, the adaptation can be carried out on several schemes. Opposite versions of the scheme are based on the «black» and «white box». The scheme of «black» box implies the absence of models of converters, but a sufficient system of test signals for direct exposure to the object or the input of the system changes. The process of adaptation occurs when the test signals at the entrance of or impact on the object directly, then by changing the properties of the system, making the necessary performance index system. The main disadvantage of this scheme is as follows: interruption of the system, the need for an input of physical quantities, the need for site management. The preferred scheme is the «white» box, which means the availability of adequate model systems, which clearly indicate «that as» to adjust to achieve the required quality in general. The advantage is the continuous adjustment during operation and the lack of impact on the inputs and the object. The difficulty in this case is only the construction of an adequate model of the converter. In a formal mechanism for adaptation in this case is the parametric optimization. At the same time a large number of nonlinear functional dependencies of functionality as a model transducer and the constraints leads to the existence of ravine surface and a large number of local function extremumies. The use of «standard» approach to the optimization seems ineffective. Proposed to use special methods of optimization based on the values of the parameters over optimization, but with a significant reduction in computational procedures. By one of these methods include the so-called genetic algorithms