This paper describes a novel approach for modelling offshore pipeline failures, which has been developed by the authors, using the Discriminant Analysis Technique. From data for pipelines in the Gulf of Mexico and the North Sea, a methodology has been developed for predicting the probability of any pipeline failing. This method can accommodate the manifold variables which affect such failures and, in this respect, it is vastly superior to the conventional method which is based on overall failure rates.
The greatest limitation with any pipeline reliability criterion which is based on average failure rates (such as the number of failures per 1000 kilometre - years) is that it is unable to discriminate between the various causes and modes of failure and enable the failure characteristics of a unique pipeline to be predicted with accuracy. This results from the fact that pipeline failure rates are highly region specific, product specific and cause specific. In an attempt to resolve this problem, the authors have developed a methodology, for predicting the probability of a pipeline failing, which is based on the statistical technique known as Discriminant Analysis. The essence of this technique is that it accepts all the data which characterize a pipeline (such as its length, diameter, steel grade, geotechnics, shipping intensity etc.), without prejudice, for those pipelines which have failures and those which have survived and allows the "data to speak" thereby enabling all those complex inter-relationships and interdependencies which exist in reality and affect a pipeline's reliability performance, to be extracted from the data. By these means it is possible to rank those design variables which have the greatest influence on a pipeline's integrity. The research which is reported here is an extension of work previously reported by de la Mare and Bakouros (1989, 1993) which considered the case for pipelines in the North Sea only.