Bayesian Approach for Design for Reliability

This paper provides an assessment method to compare design alternatives, the method starts as early as the concept development phase of a system. The method uses Bayesian techniques to quantify system reliability and reliability growth using system and subsystem test results, both historical and planned. The methods support collaboration with test planners to identify reliability drivers based on test history and plan testing to enable sufficient reliability growth to identify subsystems whose assessed reliability is reducing the system reliability below or near reliability requirements. The method results provide a reliability value with uncertainty bounds that reduces as the number of planned test successes increase. The results also support decisions to add additional tests required to reach desired reliability levels. End users, can use the results to compare design alternatives and down select on an alternative that reaches desired reliability and uncertainty with less costly tests. Current design reliability prediction methods use military standards to choose design alternatives. The resulting predictions are component based predictions using component reliability test history for similar technologies produced by various manufactures. The current analysis method does not consider the test history and/or planned testing of the design alternate manufacture. Additionally, current predictions tell you what the design can achieve but does not tell you if it will achieve it or inform you when the design reliability is in trouble. The assessment method described in this paper uses Integrated Reliability Growth Testing (IRGT) approach with environmental adjustments to reduce reliability uncertainty. The growth model uses Bayesian techniques with sub allocated priors to reduce accumulated conservatisms to assess reliability of a system at any given time with uncertainty.