ABSTRACT
By analyzing the whole process, if we can separate each steps, it can solve many practical problems employing stepwise analysis, such as dealing with precise time data and interval data. Using conditional probability, it can solve these problems
1 INTRODUCTION
For a long time, the quantitative assessment of the high-reliability and long-life product has always been a technical problem in the field of reliability engineering. With the rapid development of science and technology and constantly improvement of the level of reliability, this problem is increasingly outstanding. For products of high reliability and long life, it mainly through accelerated life testing to evaluate lifetime. The Accelerated Life Testing (ALT) was employed to find out the life characteristic of long-life and high-reliability product. Among all the accelerated life testing methods, SSALT (step-up or step-down) is considered to be an important research direction because of its relative mature theory, the smaller sample size it required, quickly failed specimen saving test cost and time in a larger extent [1]. In the SSALT all specimen under testing have experienced the effects of all the previous stress steps, which means that all of the failures besides the first step occurs in the condition of they are survived after all of the previous steps. If the product’s life obeys certain distribution under each different stress levels, the effect of the stresses may resulting in different test time in type-II censored ALT. For the product whose life obeys two-parameter Weibull distribution, many scholars believe that the acceleration efficiency of step-down test is higher than step-up
excellently. The presented method gives a form of conditional probability, which can transfer the cumulative damage problem, existing in SSALT to a mathematic problem. It provides an easier method to compare the acceleration efficiency of step-up ALT and step-down ALT. By analyzing the failures and survivals of each steps respectively, it provides a method to establish the maximum likelihood function of acceleration model parameters.
