How to Account for Short-Term and Long-Term Variability in the Prediction of the 100 Years Response?
Current practices in offshore unit design are based on the prediction of the 100 years response (tension, offset, stress...). The methodologies described in various standards (ISO, API...) are all very similar: several design environments are described with a combination of sea state, wind and current. Usually envelope contours are used, describing a set of environmental conditions corresponding to a 100 years return period. These design conditions are supposed to produce the highest responses. A time domain (or sometimes frequency domain) simulation is done on each of these short-term conditions, and the 3h most probable maximum (MPM) is computed for each. The highest MPM over all the design conditions is taken as the 100 years response. This approach completely neglects the short-term variability of the response. This paper compares several design methods with the exact 100 years response. The exact 100 years response is computed by integrating the conditional short-term distributions with respect to the probability density function of the environmental conditions. The various design methods are all based on a simplification of an Inverse First Order Reliability Method (IFORM) approach, which requires computing one or several design conditions corresponding to one or several return periods, each of these conditions being associated with a given short-term quantile. Computations are done using two datasets. At first realistic line tensions of 7 offshore units are used, based on a large number of simulations with a mooring software. On a second stage a more general parametric model using a Weibull distribution to describe the long-term variability and a Gumbel distribution to describe the short-term distribution of the 3h maximum is used. It is shown that the current methods are unconservative with respect to the exact 100 years response. A more accurate method is proposed, based on a 40 years return period associated with the quantile 90% and a correction factor of 1.04.