Reliability assessment and prediction during product development

In today´s environment, project managers are often called upon to assess the reliability of highly reliable systems both during and at the end of their development stages without the benefit of extensive test results. These assessments are used to determine whether the project is on schedule and/or if additional testing or development is required. The small amount of actual test data and the lack of the ability to assess the risk associated with the reliability prediction render the use of classical techniques doubtful. The authors address this problem for the case of attribute test data using a fully Bayesian approach. The use of the ordered Dirichlet distribution for modeling the reliability growth process is discussed. How the unique features of the distribution may be used to facilitate the incorporation of prior opinion into the analysis and further illustrate how this subjective information can be properly combined with test data so that reliability assessment and prediction of a product during its development stage can be accomplished