The linear accelerated model is often used to the statistical
analysis of constant stress accelerated life test, whereas it does not relate well
with the facts. By adopting the power functional accelerated model, the relationship
of sample quantiles among different constant stress levels is obtained, which can
lead to the estimations of the parameters in accelerated model and the characteristic
coefficient vectors by virtue of the least square method, then the life-time data
transformation between different stress levels can be operated. For complete data
and censoring data, a Dirichlet process prior is introduced to gain the posterior
distribution and the nonparametric Bayesian estimation of the reliability function,
meanwhile, the consistency of the posterior estimators is proved. Finally, a real
life example of Metal-Oxide-Semiconductor capacitors is analyzed to illustrate the
effect of our model.