The selection of corrosion prior distributions for risk based integrity modeling

Abstract

The deterioration of the condition of process plants assets has a major negative impact on the safety of its operation. Risk based integrity modeling provides a methodology to quantify the risks posed by an aging asset. This provides a means for the protection of human life, financial investment and the environmental damage from the consequences of its failures. This methodology is based on modeling the uncertainty in material degradations using probability distributions, known as priors. Using Bayes theorem, one may improve the prior distribution to obtain a posterior distribution using actual inspection data. Although the choice of priors is often subjective, a rational consensus can be achieved by judgmental studies and analyzing the generic data from the same or similar installations. The first part of this paper presents a framework for a risk based integrity modeling. This includes a methodology to select the prior distributions for the various types of corrosion degradation mechanisms, namely, the uniform, localized and erosion corrosion. Several statistical tests were conducted based on the data extracted from the literature to check which of the prior distributions follows data the best. Once the underlying distribution has been confirmed, one can estimate the parameters of the distributions. In the second part, the selected priors are tested and validated using actual plant inspection data obtained from existing assets in operation. It is found that uniform corrosion can be best described using 3P-Weibull and 3P-Lognormal distributions. Localized corrosion can be best described using Type1 extreme value and 3P-Weibull, while erosion corrosion can best be described using the 3P-Weibull, Type1 extreme value, or 3P-Lognormal distributions.