Decision making through a Bayesian network for a pipeline in design

Decision making for investing in a new pipeline project can be a long and costly process. This is usually due to the uncertainty and missing information regarding the interactions of parameters (e.g. brine chemistry, flow conditions or scale deposition) during internal corrosion assessment. In addition, these interactions result in multiple forms of internal corrosion threats (i.e. uniform corrosion, localised corrosion, erosion-corrosion and microbiologically influenced corrosion). Currently, there are no corrosion models in the market that consider all the different corrosion threats, and the predicted corrosion rates are normally conservative, leading to high overall project cost from the usage of higher grade construction material or strict maintenance regime. To predict a much more accurate internal corrosion rate with consideration of all possible corrosion mechanisms in a pipeline, a Bayesian network (BN) model was created that identifies and quantifies the causal relationships between parameters influencing internal corrosion. The model had previously proven its accuracy in predicting the internal condition of operational pipelines where explicit knowledge is available. However, the model has never been applied for a pipeline in design stage, where the design is based on tacit knowledge. In this study, to evaluate the applicability of this BN model on the pipeline at design stage, an offshore pipeline was assessed for internal corrosion.