This thesis is concerned with the problem of optimally designing accelerated life test plans for products with unequal size in which two levels, high and low, of stress and constantly applied. It is assumed that the failure rate of product is directly proportional to its size. Two life distributions, exponential and Weibull, are considered.
For exponential life distribution, it is assumed that the mean life is a log-linear function of(possibly transformed) stress. Minimizing the asymtotic variance of maximum likelihood estimator of the mean life at the design stress is used as an optimality criterion. Two cases where the test procedure is observed continously in time until a prespecified censoring time are considered:
1) Two stress levels and number of test products allocated to each are given, a heuristic algorithm which determines the sample allocation policy is given.
2) High stress level and the total number of test products are given, modified coordinate descent method which determines the low stress level and the number and size of test products allocated to it is given.
For Weibull life distribution, it is assumed that the scale parameter is a log-linear function of stress. Minimizing the asymtotic variance of maximum likelihood estimator of the 100p-th percetile at the design stress is used as an optimality criterion. Two cases where the test procedure is observed continously in time are considered:
1) Two stress levels and number of test products allocated to each are given, a simple algorithm which determines the sample allocation policy is given for complete observation.
2) High stress level and the total number of test products are given, using the algorithm for exponential life distribution, procedure to determine the low stress level and the number and size of test products allocated to it is given for type I censoring.
Optimal plans for test products with unequal size are compared with those for test products with equal size. Sensitivity analysis is given for Weibull life distribution. Optimal ALT plan for exponential life distribution is compared with that for Weibull life distribution.
