Page 254 - Dynamic Loading and Design of Structures

Page 225

Significant impulses from the non-linear wave loading will, by their nature, only be
induced by large waves. The response from the impulse will usually be damped out well in
advance before arrival of the next wave that is large enough to induce a new impulse. The
damping and scattered appearance of large waves indicate that the non-linear amplification is
larger for the extremes than for the standard deviation of the response.
Farnes et al. (1994) also compared calculations with test results, based on a tuned dynamic
model and wave surface elevation according to test samples. The accuracy was quite good.
However, later work on other structures (gravity platforms, TLPs with more complex
geometry) were not as encouraging. As mentioned in Section 5.3.5 there is no satisfactory
theoretical method for calculating ringing loads. Design loads would therefore have to be
obtained primarily by model tests, however, supported as much as possible by analyses.
Loads or load effects for final design should be established by recognizing that combined
−
2
steady state and (transient) ringing effects with an annual exceedance probability of 10 and
10 −4 for ultimate and accidental collapse limit states checks, respectively, are aimed at. Some
guidance on the determination of ringing loads by theoretical and experimental methods are
presented by DNV (1995). Ringing is known to be caused by random long crested waves in
sea states corresponding to a steepness of approximately 0.03–0.05 and Hm0
≥10m. It is important to have a sufficient number of ringing events within the time series of
observations in order to establish estimates of extreme values. This fact implies long time
series.
It is important to separate the steady state and ringing response by filtering and to compare
the observations with predictions. Close agreement between experimental and theoretical
values for steady state loads is expected. The analysis method for the ringing contribution
may be applied to tune the analysis model, which can be used for other predictions and, hence,
provide an additional reference for judging the uncertainties involved.
Load factors are applied to cover uncertainties in the environmental condition used, and
load estimation procedure, and would, hence, depend on whether loads are obtained by tests
or analyses or a combination. An uncertainty associated with selecting sea states from the
long term data basis is that critical conditions could be omitted. To limit this uncertainty,
analyses—despite their uncertainty—can be used without too much effort to screen important
conditions. Uncertainties in model tests may, for instance, be concerned with scaling effects,
model simplifications, non-uniformity of wave elevation across the basin, finite dimensions
causing wave reflections the and data acquisition system. The main uncertainty in the
theoretical model is concerned with the kinematics and hydrodynamic model, which in
general have been calibrated against experimental results to some extent. Statistical
uncertainties are present in both experimental and theoretical analyses.
Under these circumstances it is clearly not possible to set a general level of safety factors
for the ringing component. It must rather be set on the basis of the