On a computer research of K-models

The paper deals with a research of output prediction algorithm for a system of nonlinear stochastic static objects that can not be decomposed, in partial parametric and nonparametric uncertainty. The authors set the goal to perform modeling in conditions brought nearer to real technological industrial process conditions, due to the prior information obtained. In the approach the theoretical aspects of parametric and nonparametric identification of stochastic objects are used; term of compound vectors that allows to take account of intrastation complexity is introduced; problems of distinct discrecity of variables measurement and account of different levels of the prior information are considered. The authors present a new type of models which is based on the following triad: basic laws, parameterized and nonparameterized relations. The work contains new algorithms, models and computational investigations. The results can be applied when modeling of technological processes ofproduction and processing, as well as in description of objects characterized with multi-linked relations and complexity resulting from the lack of information.