A.V. Eliseev, R.M. Kalashnikov, D.A. Tyurin

The problem of synthesis of the algorithm of measurement filtration and software adaptation of information and measuring system (IMS) in the conditions of rigidity of the differential equation system (DES) describing the model of information process has been considered. It is shown that the complexity of the problems to be solved in real-time in terms of uncertainty requires the substantial strengthening of the intellectual support of the decision-making processes of the choice of a particular algorithm of processing depending on the complexity of measuring situation. One of the factors causing the uncertainty conditions of the solution IMS targets - namely, the stiffness of stochastic differential equations used to describe a model of the information process has been considered. It is shown that to improve the efficiency of IMS it is necessary to synthesize the filter algorithm that takes into account the stiffness of the DES as well as to solve the problem of increasing the level of automation of the selection process of the efficient processing algorithm measurement in real time directly in the performance of the IMS target intent by including the promptly advising expert system (PAES). The mathematical formulation of the problem has been made. We have considered some of the criteria of rigidity. We have also offered the algorithm of filtration synthesized on the basis of mathematical method allowing to reduce the rigid nonlinear filtering problem, by the transition to the new variable of integration, to a new non-rigid problem which permits to expand the boundaries of sustainability of the applied numerical methods and, therefore, to provide a solution to the problem of filtration of the state vector in the real time. We have designed PAES based on fuzzy production rules intended to issue recommendations on the selection of the classical or «light-handed» filters for processing of the measurement in the particular situation. PAES inclusion in the IMS will solve the problem of software adaptation to the conditions of the measurements. We have held computing experiment confirming the increase computational efficiency (for a specific example in 10 times) of the dynamic filtering in the stiffness of the DES.