РЕГРЕССИОННО-ВРЕМЕННЫЕ МОДЕЛИ ПРОГНОЗИРОВАНИЯ КАЧЕСТВА ИЗДЕЛИЙ ЭЛЕКТРОННОЙ ТЕХНИКИ

Abstract

Translate Article

Рассмотрены способы ускоренной оценки технического состояния изделий электронной техники на этапе эксплуатации методами теории планирования активного эксперимента. Предложен конкретный способ, сокращающий общую продолжительность испытаний для получения регрессионновременной модели прогнозирования качества изделий электронной техники. Introduction. The increase of requirements for reliability of electronic products results in development of methods, which allow estimating their technical state during the operation stage quickly and with a high confidence level. The development of such methods is possible, when the dependence of the technical state of electronic products on influencing factors and operating time is plotted. The mathematical expression, approximating such dependence, is called a regressiontime model. Therefore, the most efficient method of its construction is the theory of active experiment design. For obtaining a regressiontime model, specially planned tests are carried out. Meanwhile, there are two methods of including time in the model: to conduct each test in the experiment design during the time t to consider the time t as an additional factor along with factors electronic products operating conditions. The purpose of the work is to study the methods of the construction of a regressiontime model in accordance with the total duration of electronic products tests for reducing financial and time costs. Results. When constructing regressiontime models, using the method of the theory of active experiment design, experiment designs of the second order found the most common application in electronic instrument making: an orthogonal central composite design a rotatable central composite design a Box design a Hartley design a BoxBehnken design. These designs were studied in accordance with the duration of products tests. When comparing methods of the construction of the regressiontime model, the number of factors к, determining the numerical value of electronic products quality indicator was taken as equal to: 2 3 4 5 6. It was obtained that including time in the experiment design as an additional factor reduces the general duration of tests with any number of factors к. The experiment design of the orthogonal central composite order when к2 and к4, the Hartley design when к3 and к5, the BoxBehnken design when к6 have the maximum gain in the duration of tests. Conclusion. When constructing the regressiontime model of forecasting the dependence of electronic products quality indicator on conditions and operating time using the method of the theory of active experiment design, it is necessary to consider time as an additional factor in the experiment design. The inclusion of time into the experiment design as an additional factor reduces the general duration of electronic products reliability tests from 10 to 48 depending on the experiment design type and the number of factors, determining the numerical value of the electronic product quality indicator.