Internal corrosion modeling of oil and gas pipelines requires the consideration of interactions between various parameters (e.g. brine chemistry, flow conditions or scale deposition). Moreover, the number of interactions increases when we consider that there are multiple types of internal corrosion mechanisms (i.e. uniform corrosion, localized corrosion, erosion-corrosion and microbiologically influenced corrosion). To better describe the pipeline internal corrosion threats, a Bayesian network model was created by identifying and quantifying causal relationships between parameters influencing internal corrosion. One of the strengths of the Bayesian network methodology is its capability to handle uncertain and missing data. The model had previously proven its accuracy in predicting the internal condition of existing pipelines. However, the model has never been tested on a pipeline in design stage, where future operating conditions are uncertain and data uncertainty is high. In this study, an offshore pipeline was selected for an internal corrosion threat assessment. All available information related to the pipeline were collected and uncertainties in some parameters were estimated based on subject matter expertise. The results showed that the Bayesian network model can be used to quantify the value of each information (i.e. which parameters have the most effect now and in the future), predict the range of possible corrosion rates and pipeline failure probability within a given confidence level.
A Structure-variable Bayesian Network Model for Vehicle Threat Assessment
