:   .., .., .., ..
:  
:  118
:  
:  2025
:   .., .., .. . // . - 2025. - . 118. - .7-22.
:   , , -, , ,
(.):  one-shot device, reliability function, beta distribution, lower confidence limit, Bayesian approach, zero-failure
:   () (). : 50-476-84; -; -; . - . , - -. . , . . , .
(.):  The article is devoted to the problem of confirming the lower confidence limit of the reliability function of one-shot devices (systems) in the zero-failures at the element (sub-system) level. Systems with a series connection of different types of elements are considered and four approaches to determining confidence intervals are compared: the method from RD 50-476-84; Monte Carlo simulation; approximation of the distribution of the estimate of the reliability function the system by the beta distribution; the Bayesian approach. The necessity of using computational-experimental methods in assessing highly reliable one-shot devices is explained. Based on the experimental data of two test variants, with different and equal volumes of element test cycles, it is shown that the RD method forms higher boundaries of confidence intervals compared to the approach based on the beta distribution and the Monte Carlo method. At the same time, the latter are sensitive to a priori assumptions and can underestimate the estimate in the zero-failures. It is shown that the method of determining the lower confidence limit currently used in RD excludes part of the operating time of individual elements. Developing a mechanism for accounting for such operating time by using a priori information on the failure-free operation of elements will increase the lower confidence limit. The results are addressed to developers of highly reliable one-shot devices, managers and specialists of the quality management system.

PDF

: 87, : 38, : 14.


© 2007.