A key aspect in the design of a mooring system for a floating production unit is the estimation of the extreme mooring line loads for a specified short-term sea state of typical duration equal to 3 hours. Commonly used design approaches today are based on time-domain simulations whereby each 3 hour sea state is run a number of times (typically 10–30 times) to represent the randomness of the sea. A maximum response is recorded from each simulation. Particular statistic of the maxima data (e.g. mean, most probable maximum or a percentile) is used to represent the extreme mooring load for which the lines are designed.
This paper studies and assesses the accuracy of obtaining design value from a population of maxima with reference to the mooring line load of a large ship-shaped floating production vessel. A coupled model, including all mooring lines and risers, has been developed, validated and used to generate responses for 100yr extreme condition and 10,000yr survival condition. To establish an accurate benchmark against which the results are compared, the time-domain analyses (duration 3 hours) are repeated 170 times, for each sea state, to represent different random realisations of each environment. It is examined how the accuracy of predicting the design mooring line load, from a sample of response maxima, improves as the number of simulations is increased progressively from 10 through to 170. The assessment is performed across different statistics of maxima that are usually chosen to represent the design response. Besides the mooring line load, other response parameters such as heave and turret excursion, are examined in this paper. The paper examines whether the severity of the response (100yr vs 10,000yr storm) or the response variable affect the number of maxima required to achieve statistical stability. The results indicate fitting a Gumbel distribution to the maxima from about 30–40 simulations can yield results that are statistically stable and accurate and are recommended as preferred methods of estimating the design response.