DIAGNOSIS OF SYSTEMS UNDER CONDITIONS OF SMALL INITIAL DATA SAMPLING

Abstract

Object of the study is to assess systems state in conditions of a small sample of initial data. Relevance of the problem is as follows. The functioning of a significant number of real objects takes place under conditions of poorly predicted changes in the values of environmental factors affecting system efficiency. The resulting heterogeneity of the results of objects experimental study and the environment of their functioning leads to reduction in sample size. At the same time, the standard requirements regarding the correspondence of the number of experiments and the number of coefficients of regression equation determining system state are not met. Purpose of the study is to develop methods for assessing systems state operating in a changing environment, in conditions of small sample of initial data. Tasks to be solved to achieve the goal: the first is the equivalent transformation of the set of observed initial data forming a passive experiment in aggregate into an active experiment, which corresponds to an orthogonal plan; the second is the construction of a truncated orthogonal representative sub-plan of the general orthogonal plan obtained as a result of solving the first problem. Research methods: statistical methods of experimental data processing, regression analysis, method for solving a triaxial boolean assignment problem. The results obtained: orthogonal representative subplan of the complete factorial experiment being formed makes it possible to calculate a truncated regression equation containing all the influencing factors and their interactions. Analysis of the coefficients of this equation by known methods makes it possible to cut off its insignificant elements.