Integrating field failure data and accelerated life testing by Bayesian methods for product reliability inference

A Bayesian statistical approach is proposed to improve the prediction of product reliability by using field performance data and accelerated life testing results. This approach develops the calculation of a calibration factor that compensates the very broad variation in field conditions to the controlled conditions in a laboratory setting. An example, based on temperature stress and the Arrhenius function, is developed to instruct on how to estimate the calibration factor and other important life distribution parameters in different scenarios. The Winbugs program was used to do the simulations to find the parameter estimates when closed-form posterior distributions are not feasible