Assessing the reliability of switchgear based on censored statistical information

Abstract

Switching devices (starters, relays, etc.) are bulk products on which high reliability requirements are imposed. One reliable way to determine the quality of these products is different types of random testing, including tests of electrical endurance. When testing devices, there are situations when a certain part of the test objects do not fail over the period of observation. In these cases, it is necessary to conduct a statistical analysis of samples, which are characterized by a lack of information about the points of failure of controlled objects (or one object). This phenomenon is called censored data and the samples obtained as a result of these tests are called censored samples [1]. Note that specific methods for assessing reliability and testing statistical hypotheses should be applied to these sam ples. These methods are generally more complicated than traditional methods of mathematical statistics. The analysis of censored samples takes on special sig nificance in assessing the reliability of complex and expensive technical devices, as well as products with costly testing methods. The latter include switching devices because of the length and high energy expen ditures of the wearing tests.Three pole starters of the PM12 016 series of ver sion B with the nominal contact current of the main circuit of 16 A have been used as an object of research.The tests have been conducted in accordance with the requirements of [2]. The sample starters have been tested in the application category AC 4 at the nominal operating current of 4 A and the nominal operating voltage of 660 V, 50 Hz. The switching frequency is 600 cycles per hour, PV = 2.78%, and the nominal life is 200000 switching cycles.The starters cyclically commute the test circuit with the following parameters: U on = U off = 660 V, I on = P off = 24 A, cosϕ = 0.65. During the starter test ing, the contact wear has been controlled by changing the dip ω, which is a determining parameter of the performance reliability of contacts. This is due to the design features of starters and the availability of the dip measurements by the techniques set out in [3]. The failure of contacts has been measured periodically in a number of switching cycles that make up 15-20% of the nominal life of the starters. In deciding on the test plan, it was assumed that the starters are nonrepairable products. The tests have shown that, in this shipment, failures occurred outside the number of switching cycles guaranteed by the technological normative doc uments due to the gradual aging of contacts (according to the scheme of cumulative changes [4]). The wear on contacts eventually resulted in two cases of contact welding and one case of complete wear out. Test results are shown in Fig. 1.We believe that periodical stops of tests for the dip measurements have the nature of an a priori censor ship. Consequently, the results of observations can be regarded as censored sampling obtained under multi ple left censoring [1]. The purpose of this paper is to identify the most valid reliabili...