Modern development of technology is characterized by sharp complication of the tasks solved at production of products, high requirements to their reliability, short deadlines of creation and introduction in operation, aspiration to reduce costs of development of a product at satisfaction of the set conditions. The existing methods of an assessment and quality control of a product by results of tests are often inefficient in the conditions of the determined experiment or the diverse volume-limited statistical information on results of physical modeling, prototyping and tests of a small number of samples available. One of the solutions of the problem of ensuring quality and reliability at a design stage and production of products is an application of the theory of similarity and modeling. In the article the direction of development of the stochastic similarity theory for the effective solution of problems of an assessment and quality control of complicated systems in the course of their creation and operation is offered. The main theorems of similarity are formulated, the criterion of stochastic similarity in the form of the relation of functions of distribution of parameters (characteristics) of systems is theoretically reasonable and experimentally confirmed. Examples of the solution of applied tasks are given. Further development of the theory of stochastic similarity will allow us to advance in the research of non-limiting distributions which are a little described in classical probability theory and mathematical statistics, especially in that their part which will give the chance to manipulate selections of a small volume and censored selections obtaining maximum information from them.
In the article the effective non-parametric method to identify patterns of failures of electronic equipment automation of dynamic systems. The essence of the offered method is that on the small selections presented in the form of a variation row, practically always it is possible to find such transformation as a result of which the statistics which isn't depending on parameters of distribution of population will turn out. It is advisable to define function of distribution of such statistics as a result of statistical modeling if analytical construction it is complicated.
