This paper describes a method for the development of a Bayesian prior time-to-failure distribution used as an early estimate of the reliability of a new product which is an evolution of a previous ancestor design. In new product development, it is desirable to estimate the reliability of that new product as early as possible in the development program, in order to quantify future financial risks as a result of service costs, and to facilitate risk-based decision-making. It is difficult to make these reliability estimates sufficiently early in the development cycle to be of use in decision-making because of a lack of information. However, this “lack of information” is often a perception only, particularly when a new design is an evolution of a previous design. Making use of the reliability-related knowledge from an ancestor design can improve the accuracy of reliability predictions of the evolutionary design before testing of that design even commences. The method proposed is as follows. First, develop the list of failure mechanisms for the new product. Second, develop the time-to-failure distribution parameters for each of those mechanisms from ancestor product data, including the uncertainty inherent in each of those parameters. Third, develop a list of all changes being made in the new product design which will affect the reliability of the new design for a particular failure mechanism. Fourth, quantify the impact of each design change as an improvement factor distribution. Fifth, combine the ancestor product time-to-failure distribution parameters and the improvement rate factors using propagation of uncertainty rules to produce an estimate of the time-to-failure distribution parameters for the evolutionary product for each mechanism. Lastly, use Bayes’ rule to incorporate new information as the design process progresses. This method will allow estimates of new product reliability to be made earlier in the product development cycle. In making these estimates, this method will formally use pertinent reliability-related information from the previous generations of a product, and information on the impact of proposed design improvements. The method is architected in a manner that facilitates a structured approach to capturing and managing changes in the state of knowledge regarding the reliability of the new product which occur as the design process progresses. As new information arrives during the design process, in the form of either new field reliability information from the ancestor design, or on the demonstrated effectiveness of design improvements, that information can be formally entered into the model, and the reliability predictions for the product of interest can be updated to reflect that new state of knowledge.

This content is only available via PDF.
You do not currently have access to this content.