Updating of Probabilistic Corrosion Model Based on Bayesian Procedure

Abstract

Corrosion condition is predicted based on the corrosion model. The corrosion model is necessary to be identified according to the corrosion data collected from the various vessels because corrosion phenomenon is stochastic. However, in order to predict corrosion condition of one specific vessel, such corrosion model is necessary to be modified to reflect the effect of specific corrosion environment of the subject vessel. In the study, procedure of updating corrosion model was investigated based on Bayesian inference on the parameters in the probabilistic corrosion model which utilizes the thickness measurements data. The developed procedure was demonstrated by the application of actual thickness measurements data of the vessel. Even though the amount of corrosion data was limited, the corrosion prediction model was well updated which could be verified by the concentration of posterior distribution which shows the degree of belief on the parameters in the probabilistic corrosion model. The estimated distributions of coating life and corrosion wastage were compared with the frequency distributions obtained by the corrosion data. The estimated distributions of coating life and corrosion wastage showed good agreement with the frequency distributions obtained by the corrosion data.