Open AccessEstimation of the gamma-percentile life for the binomial test plan
Author(s) -
В. С. Михайлов
Publication year - 2019
Publication title -
nadëžnostʹ
Language(s) - English
Resource type - Journals
eISSN - 2500-3909
pISSN - 1729-2646
DOI - 10.21683/1729-2646-2019-19-2-18-21
Subject(s) - percentile , binomial distribution , binomial (polynomial) , product (mathematics) , gamma distribution , simple (philosophy) , set (abstract data type) , test plan , statistics , mathematics , function (biology) , schedule , negative binomial distribution , plan (archaeology) , computer science , reliability engineering , weibull distribution , engineering , geography , philosophy , geometry , archaeology , epistemology , evolutionary biology , poisson distribution , biology , programming language , operating system
In today’s industry of highly dependable, unique, complex products, it is a commonplace situation when it is required to obtain an estimate of gamma-percentile storing life (GPSL) or gamma-percentile time to failure (GPTF) (hereinafter referred to as GPL, meaning GPSL or GPTF respectively) based on tests that did not produce any failures. Normally, a test plan is classified as binomial tests or tests with limited time and recovery. GPTF is understood as the operation time during which no failure occurs with the probability γ expressed as percentage. Similarly, GPSL is understood as the total schedule time of product storage during and after which the product is able to perform the required function with the probability γ expressed as percentage.The Aim of this paper is to find such estimate of GPL that would be simple and more efficient as compared to the conventional one and not significantly inferior to the efficient estimate, if such exists, in terms of proximity to the GPL estimate in case a binomial test plan is used.Methods of research. The research of efficient estimates is based on the integrative approach that uses the construction of the choice rule (criterion) of efficient estimate specified on the vertical sum of biases of estimates selected out of a certain set based on the function of the distribution law parameter.Conclusions. The obtained estimate of GPL is simple and more efficient as compared to the conventional one and not significantly inferior to the efficient estimate, if such exists, in terms of proximity to the GPL estimate in case a binomial test plan is used. The obtained estimate of GPL has significant advantages, i.e. the estimate is efficient within a sufficiently wide range of estimates, it allows obtaining a value based on the results of tests that did not produce any failures. The obtained estimate of GPL is recommended for failure-free tests conducted using a binomial plan.