Numerical methods for bayesian inference in case of aging

Carlo Alberto Clarotti, Benjamin Villain, Henri Procaccia

Research output: Contribution to journalArticle

1 Citation (Scopus)


In recent years, much attention has been paid to Bayesian methods for Risk Assessment. Until now, these methods have been studied from a theoretical point of view. Researchers have been mainly interested in : - studying the effectiveness of Bayesian methods in handling rare events ; - debating about the problem of priors and other philosophical issues. An aspect central to the Bayesian approach is numerical computation because any safety / reliability problem, in a Bayesian frame, ends with a problem of numerical integration. This aspect has been neglected until now because most Risk studies assumed the Exponential model as the basic probabilistic model. The existence of conjugate priors makes numerical integration unnecessary in this case. If aging is to be taken into account, no conjugate family is available and the use of numerical integration becomes compulsory. EDF (National Board for Electricity, of France) and ENEA (National Committee for Energy, New Technologies and Environment, of Italy) jointly carried out a research program aimed at developing quadrature methods suitable for Bayesian Inference with underlying Weibull or gamma distributions. The paper will illustrate the main results achieved during the above research program and will discuss, via some sample cases, the performances of the numerical algorithms which were implemented. The sample cases that will be considered are part of the study on the appearance of stress corrosion cracking in the tubes of Steam Generators of PWR French power plants.
Original languageEnglish
Pages (from-to)35 - 39
Number of pages5
JournalAmerican Society of Mechanical Engineers, Pressure Vessels and Piping Division (Publication) PVP
Publication statusPublished - 1996
Externally publishedYes


All Science Journal Classification (ASJC) codes

  • Mechanical Engineering

Cite this