This dissertation proposed a method for designing optimal reliability testing plans for simple life test, accelerated life test, and simple step accelerated life test using the Bayesian approach and a multi-objective genetic algorithm. Two criteria, cost and prediction precision of the test, have been optimized simultaneously. The Bayesian approach was considered an attractive alternative to the traditional maximum likelihood method for minimizing uncertainty in the “planning values” of the model parameters. The multi-objective genetic algorithm was adopted to solve the planning problem and to produce a set of alternative plans from which a planner may choose the most desirable plan.

In this dissertation, the effect of the priors and the cost structures on the optimal plans was studied to investigate how the optimal plans will response when different priors and cost structures are used. In addition, MLE and Bayesian approaches were compared to investigate the similarities and the differences between those methods. The statistical optimal plans were investigated to see the effect of the model constraints on the statistical optimal plans for accelerated life tests and for simple step accelerated life tests.

This dissertation was based on Bayesian method and genetic algorithms, which are flexible methods, therefore the models proposed in this dissertation can be easily extended for future work to cover a larger spectrum of test structures, utility functions, life distributions, prior distributions, and cost structures.