Marine drilling risers are integral parts of the deep water offshore oil and gas industry. They are required to be designed for safe operations during their service lives with appropriate degree of reliability. With limited experience present in ultra-deep water, drilling risers are subjected to a range of uncertainties arising from untested environmental conditions. However, the current industry practice is limited to deterministic design of drilling risers which cannot account for uncertainties present in real life scenario. Under uncertain environmental conditions, deterministic methods may lead to undesired consequences, i.e. over conservative or unsafe design and misguided estimates of operability and down time of ultra-deep water drilling risers affecting the total life cycle cost. Thus, structural reliability analysis is particularly useful for prediction of the probabilities of downtime and disconnection of drilling risers incorporating the environmental uncertainties. In addition, structural reliability analysis can be used to reduce the total life cycle cost of ultra-deep water drilling risers.
In reliability analysis, many studies use uncorrelated random variables to represent uncertainties for simplification. Nevertheless, uncertainties in environmental conditions may be strongly correlated (for example wind and wave loads). If the correlation is not accounted for, it may lead to erroneous probability estimates. Thus, a joint environmental model is proposed in this paper using the conditional modeling approach where a joint density function is defined in terms of a marginal distribution and a series of conditional density functions. The joint density functions of environmental conditions are constructed in the current study using the recorded metocean data for Gulf of Mexico available from National Oceanic and Atmospheric Administration (NOAA) website. Then a computational model of connected ultra-deep water drilling riser system is constructed in ORCAFLEX to conduct time domain dynamic analysis. Thereafter, the correlated random variables in combination with the drilling riser computational model are utilized for conducting Monte Carlo Simulation (MCS) to evaluate the probabilities of downtime and disconnection. MCS is a widely accepted and robust approach and generally used as a benchmark to verify the accuracy of other reliability methods. But, in presence of large number of random variables representing environmental uncertainties, MCS is computationally demanding especially for the large number of simulations required to estimate small failure probabilities associated with extreme values. To this end, probability density functions of drilling riser responses are evaluated using Shifted Generalized Lognormal Distribution (SGLD) and Generalized Extreme-Value (GEV) Distribution both of which show similar accuracy (compared to MCS results) at a fraction of computing time (around 1/500 times).