Non-Parametric Bayesian Inference for Reliability Function Elicitation

Abstract

This project intends to make an application of Bayesian nonparametric method of elicitation of priors proposed by Oakley & O'Hagan (2007), Moala (2009) and Moala & O'Hagan (2010) to estimate the reliability function, considered completely unknown in form, based only on information supplied by the specialist. The objective is the application of Bayesian analysis and its ability to incorporate a priori knowledge on the theory of reliability, which would be extremely useful when testing time is limited to a formal procedure for obtaining results about the reliability of a product. This occurs mainly in high-reliability systems, which can not afford enough samples to meet the requirements for reliability testing. In this case the information from the expert may be more viable source of information. Thus, the development of a consistent methodology that adequately represents the information provided by specialists to get the reliability is of great importance for the application of Bayesian methods and may represent an important contribution to the theory elicitation and analysis of reliability.