Abstract
Failure rates, which quantify the normalized likelihood of pipeline failure, are an integral part of assessing the reliability and risk of pipelines. The industry-wide trend of utilizing probabilistic methods for estimating failure rates raises the question whether the frequentist or Bayesian definition of probability is more suitable. The paper illustrates some limitations of the frequentist probability definition for pipeline risk assessment and supports the Bayesian approach for analyzing pipeline failure rates. The Bayesian quantification of probabilities leads to coherent uncertainty assessment and propagation even if evidence is combined from different sources either through a repetition of the prior-likelihood model or a multi-level / hierarchical approach that integrates all available data and information in one model. Selecting or disregarding data for estimating failure rates is no longer necessary as they all contribute to the result based on their relative uncertainties. Examples are provided in the paper to illustrate the benefits of the Bayesian probability approach.
