LAUSR

ABSTRACT Failure threats in subsea pipelines are hard to inspect, but parameters influencing them are easier to observe. Hence, nowadays, Bayesian network models became more relevant, as the model can be updated with the sparse observations while considering the underlying uncertainty. This holds for failure threat assessment of subsea pipelines, specifically for a highly random corrosion mechanism, which has not been captured in the current traditional assessments appropriately. However, a number of researchers stated that it is difficult to build the Conditional Probability Table (CPT) of the Bayesian networks. In such cases, it has been suggested to employ expert knowledge to determine the conditional probability distributions, which involves some uncertainties and high data deviation. This paper focusses on developing a dynamic Bayesian network-based framework to minimise the inputs from the expert domain in the CPT development, while providing an efficient option to analyse the pipeline residual life due to corrosion threat.