Structural Reliability Analysis (SRA) methods have been applied to marine and offshore structures for decades. SRA has proven useful in life extension exercises and inspection planning of existing offshore structures. It is also a useful tool in code development, where the reliability level provided by the code is calibrated to a target failure probability obtained by SRA. This applies both to extreme load situations and also to a structural system under the influence of a time dependent degradation process such as fatigue.
The current analysis methods suggested for service life estimation of subsea wells are deterministic, and these analyses are associated with high sensitivity to variations in input parameters. Thus sensitivity screening is often recommended for certain input parameters, and the worst case is then typically used as a basis for the analysis. The associated level of conservatism embedded in results from a deterministic analysis is not quantified, and it is therefore difficult to know and to justify if unnecessary conservatism can be removed from the calculations.
By applying SRA to a wellhead fatigue analysis, the input parameters are accounted for with their associated uncertainty given by probability distributions. Analysis results can be generated by use of Monte-Carlo simulations or FORM/SORM (first/second order reliability methods), accounting for the full scatter of system relations and input variations. The level of conservatism can then be quantified and evaluated versus an acceptable probability of failure.
This article presents results from a SRA of a fictitious but still realistic well model, including the main assumptions that were made, and discusses how SRA can be applied to a wellhead fatigue analysis. Global load analyses and local stress calculations were carried out prior to the SRA, and a response surface technique was used to interpolate on these results. This analysis has been limited to two hotspots located in each of the two main load bearing members of the wellhead system.
The SRA provides a probability of failure estimate that may be used to give better decision support in the event of life extension of existing subsea wells. In addition, a relative uncertainty ranking of input variables provides insight into the problem and knowledge about where risk reducing efforts should be made to reduce the uncertainty.
It should be noted that most attention has been given to the method development, and that more comprehensive analysis work and assessment of specific input is needed in a real case.
Neural Networks Applied to the Wave-Induced Fatigue Analysis of Steel Risers
