Jinhua MI An E-BAyEsiAn mEthod for rEliABility AnAlysis of ExponEntiAlly distriButEd products with zEro-fAilurE dAtA mEtodA

Yi-Chao YIN,Hong-Zhong HUANG,Weiwen PeNG,LI Yan-Feng

2016-01-01

Abstract:Reliability is the probability of a component or system working regularly. Reliability technology was firstly applied in the fields of nuclear industry, aviation and aerospace, and then was spread to the industries of electronic, metallurgy, mechanical and so on. The researches and analyses on the reliability theories and applications have lasted for decades. Now reliability has become one of the most important areas in the industry. With the improvement of reliability applications, products with high reliability require a long time to obtain the failure data in normal life tests. Actually, zero-failure data is often detected in Type I Censoring life tests. The early research about zero-failure data started by Bartholomew [1], who firstly used the test time as the estimate of the average life, which was obviously too small. Martz and Waller [12] developed a method for an exponential failure-time model and a gamma prior distribution on the failure rate. Chow and Shao [2, 3] used a regression model and the weighted least squares method for accessing the shelf-lives of drug products. Miller, Morell and Noonan [13] provided the formulae for the probability of failure with zero-failure data based on some assumptions of Bayesian prior distribution, and gave an example of failure rate estimation with zero-failure data. Han [4, 5] developed an E-Bayesian estimate method based on hierarchical Bayesian estimation, which was proved to be feasible for zerofailure data. Liu [9] investigated the applications of E-Bayesian estimation for exponential life distribution. Based on the researches of E-Bayesian estimation, we introduce a reliability analysis method for the seekers, which is more convenient and effective for zero-failure data of exponential distribution. The remainder of the paper is organized as follows. Section 2 presents a brief description of E-Bayesian method for the failure rate estimation of exponential distribution. In Section 3 we introduce the assumptions for reliability testing with zero-failure data. In Section 4 we provide an E-Bayesian method for exponential distribution with zero-failure data. In Section 5 we illustrate how the proposed E-Bayesian method can deal with zero-failure data in life testing, and compare it with different zero-failure analysis methods. In Section 6 We give some comments and conclusions.