Offshore structures are typically required to withstand extreme and abnormal load effects with annual probabilities of occurrence of 10−2 and 10−4 respectively. For linear or weakly nonlinear problems, the load effects with the prescribed annual probabilities of occurrence are typically estimated as a relatively rare occurrence in the short term distribution of 100 year and 10 000 year seastates. For strongly nonlinear load effects, it is not given that an extreme seastate can be used reliably to estimate the characteristic load effect. The governing load may occur as an extremely rare event in a much lower seastate.
In attempting to model the load effect in an extreme seastate, the short term probability level is not known nor is it known whether the physics of the wave loading is captured correctly in an extreme seastate. Examples of such strongly nonlinear load effects are slamming loads on large volume offshore structures or wave in deck loads on jacket structures subject to seabed subsidence. Similarly, for structures which are unmanned in extreme weather, the governing load effects for the manned structure will occur as extremely rare events in a relatively frequent seastate.
The present paper is concerned with the long term distribution of strongly nonlinear load effects. Using a simple point estimate of the wave elevation correct to second order and a crest kinematics model which takes into account the possibility of wave breaking, the long term distribution of drag load on a column above the still water level is studied and compared with a similar loading model based on second order kinematics which does not include the effect of wave breaking.
The findings illustrate the challenges listed above. Model tests are useful in quantifying strongly nonlinear load effects which cannot be calculated accurately. But only a relatively small number of seastates can be run in a model test campaign and it is not feasible to estimate short term responses far beyond the three hour 90% fractile level. Similarly, Computational Fluid Dynamics (CFD) is increasingly useful in investigating complex wave induced load effects. But only a relatively small number of wave events can be run using CFD, a long term analysis of load effects cannot in general be carried out.
It appears that there is a class of nonlinear problems which require a long term analysis of the load effect in order for the annual probability of occurrence to be estimated accurately. For problems which cannot be estimated by simple analytical means, the governing wave events can be identified by long term analysis of a simple model which capture the essential physics of the problem and then analysed in detail by use of CFD or model tests.
Machine Learning-Based Modeling for the Duration of Load Effect in Wood Structural Members Under Long-Term Sustained Load