Quantifying the Probability of Failure During the Pre-Commissioning Hydrotest

Failures during the pre-commissioning hydrostatic test of a newly constructed pipeline are rare, but occasionally they do occur. Structural reliability techniques can be used to estimate the probability of failure during the precommissioning hydrotest, and to investigate the sensitivity of the probability of failure to the test pressure. This paper describes a study of the probability of failure during the hydrotest, based on data for the BBL Pipeline. The BBL Pipeline is a 36 in. outside diameter, approximately 235 km long pipeline designed to export natural gas from the Netherlands to the UK. The definition of failure is limited to failure of the line pipe due to internal pressure loading. Failure of fittings (e.g. flanges, valves, etc.) is not considered. With this definition of failure, three different scenarios are considered: 1. Failure of defect-free pipe. 2. Failure of pipe containing a ‘workmanship’ defect (i.e. a defect in the pipe body or a weld that is acceptable to the relevant specifications or standards). 3. Failure of pipe containing a defect not acceptable to workmanship levels (e.g. a crack, or a dent on a weld). Defects larger than workmanship defects encompass defects that would not fail at the design pressure, but would fail at higher pressures, and ‘gross’ defects that would fail at very low pressures. It is difficult to estimate the probability of failure of such defects because it is highly dependent on the probability of these defects being present in the pipeline. Conversely, it is relatively straightforward to obtain a reasonable upper bound estimate of the probability of failure due to defect-free pipe or a workmanship defect. Consequently, in this study, only the probability of failure due to (1) or (2) has been calculated using structural reliability techniques. Inferences about (3) are drawn from the results of (1) and (2), and from historical data on hydrotest failures. It is shown that there is a hydrotest level below which the probability of failure is predicted to be zero. For defect-free pipe this is at least equal to the level of the mill test. In general, this hydrotest level depends upon factors such as: the ratio of the axial stress to the hoop stress in the mill test and the hydrotest, and the size of defects in the pipeline.