Step Test Increases With The Conversion Time Of The Random Stress Statistical Analysis And Optimal Design

With the development of science technology and economy, step-stress accelerated life test becomes an important method. Moreover, step-stress life-testing with random stress-change times is studied because it is more useful in practice. In this dissertation, step-stress life-testing with random stress-change times is discussed and problems are solved as follows:1. The step-stress life-testing with random stress-change times under exponential distribution is studied, when the samples are censored. We set up the model of test and give the maximum likelihood estimates (MLE), moment estimates, s-confidence interval estimates and Bayesian estimates for model parameters. Those estimates are proved optimum by using numerical method. On the basis, an optimum test plan is explored by minimizing the asymptotic variance of the maximum likelihood estimates of the mean life at a design stress. The plan is proved effective through an example.2. We consider the more common situations: the tests with competing causes of failure and the tests under exponential distribution with two parameters. The models are set up when samples are completed and censored, MLE of models parameters are given. Simulative data show our method is accurate and effective. Then the optimum test designs to minimize the sum of asymptotic variance the MLE of the log mean life in each cause at a design stress. Also we obtain the optimum test plan which is minimized the asymptotic variance the MLE of the mean life at a design stress. At last, we illustrate the optimum test'plans on some examples.3. A new model (KH-M) is used to instead the CE-M. Some results are obtained under Weibull distribution, which are similar to those of exponential data.