Long term statistics of extreme environmental loads on fixed offshore structures

Abstract

The design and re-assessment of fixed offshore structures for operation in the harsh conditions of the North Sea is largely determined by extreme environmental loading. The ability to predict accurately the extreme storm load remains an important factor in the continued safe and economic exploitation of the hydrocarbon reserves in this area. The problem of specifying met-ocean design conditions (MOe) is one of estimating the environmental variables corresponding to some return period, typically 100 years, on the basis of measured or hind cast time series extending over a relatively short period, say 5 or 25 years. A new method for deriving met-ocean design conditions has been developed at KSEPL. Statistics are calculated for the whole storm rather than for three-hour intervals. By combining a model, for translating a set of environmental conditions into a structural load, and the environmental parameters for a sea state, statistics can be generated for Ioads rather than for single environmental parameters. These load statistics are then used in an extrapolation process to extrapolate to rare events. Once an extreme Ioad is derived, a characteristic set of environmental parameters (joint met-ocean conditions) which could have generated this load can be derived via "back calculation". This set of environmental parameters can be used as an input for a more detailed calculation of the distribution of the extreme load over the structure. The new method avoids the shortcomings of existing approaches and is distinguished from them by being equally applicable to the prediction of extreme waves, long term Ioad statistics and joint met-ocean conditions. Environmental conditions and long term Ioad statistics are directionally sensitive. To account for this phenomenon, the met-ocean data for a Iocation must be divided into sectors, containing the directions from which severe storms are coming. So far, for the North Sea, these sectors have been chosen on the basis of wave height. The resulting sectors are reasonably wide. From the NESS data base, sectors containing severe storms include around 100 storms (sufficient for the subsequent statistical analysis). Storms in a wide sector are assumed to have a uniform distribution over this sector. However, (tidal) current can be much more sensitive to direction than significant wave height, and can influence loads. To resolve the (tidal) current correctly, we would have to reduce sector sizes so much that the number of storms would be too small for a reliable statistical analysis. It is possible to increase artificially the sample size, by treating the angle between mean wave direction and (tidal) current direction as a parameter that can be varied. A" the storms in the wide sectors (based on wave height) can then be used to generate Ioad statistics in narrower sectors (based on current speed). This idea has been implemented within KSEPL methodology in this work. The effects of strongly directional current and artificially increased sample size on Ioad statistics and on a structural reliability are investigated. The data base used for this purpose contains a" the relevant environmental parameters for a location in the southern North Sea for a hind casting period of 25 years. The southern North Sea is selected, because of its relatively high (tidal) current speeds, compared to other areas in the North Sea. IV For the example considered the derived extreme met-ocean design conditions and calculated structural reliability are not very sensitive to the effects of a strongly directional current. The extra effort necessary to account tor the effect of those strongly directional currents seems therefore, based on the calculations as carried out in this report, not justified. Moreover, since the southern North Sea is the sector of the North Sea with the strongest and most directional currents, it is likely that this result pertains to the whole of the North Sea. However, further work (e.g. calculations tor other grid points) may be required to generalise this conclusion.