The Development of Posterior Probability Models in Risk‐Based Integrity Modeling

There is a need for accurate modeling of mechanisms causing material degradation of equipment in process installation, to ensure safety and reliability of the equipment. Degradation mechanisms are stochastic processes. They can be best described using risk‐based approaches. Risk‐based integrity assessment quantifies the level of risk to which the individual components are subjected and provides means to mitigate them in a safe and cost‐effective manner. The uncertainty and variability in structural degradations can be best modeled by probability distributions. Prior probability models provide initial description of the degradation mechanisms. As more inspection data become available, these prior probability models can be revised to obtain posterior probability models, which represent the current system and can be used to predict future failures. In this article, a rejection sampling‐based Metropolis‐Hastings (M‐H) algorithm is used to develop posterior distributions. The M‐H algorithm is a Markov chain Monte Carlo algorithm used to generate a sequence of posterior samples without actually knowing the normalizing constant. Ignoring the transient samples in the generated Markov chain, the steady state samples are rejected or accepted based on an acceptance criterion. To validate the estimated parameters of posterior models, analytical Laplace approximation method is used to compute the integrals involved in the posterior function. Results of the M‐H algorithm and Laplace approximations are compared with conjugate pair estimations of known prior and likelihood combinations. The M‐H algorithm provides better results and hence it is used for posterior development of the selected priors for corrosion and cracking.