Page 248 - Dynamic Loading and Design of Structures

Page 219

above, uncertainties may add to those in the wave loads themselves. The main source of
uncertainty is associated with damping. API (RP2A-LRFD API, 1993) specifies an additional
=
load factor γ D on inertia forces, before the load factor γ1.35 for wave loads is applied on the
excitation forces and factored inertial forces. The total load factor on dynamic load
contributions is therefore about 1.7. This factor was determined (Moses, 1985) based on an
estimate of the additional uncertainty associated with dynamic loads. This approach is limited
to jackets. No other codes for jackets, jack-ups and other fixed platforms include this kind of
additional load factor γ D. It is important to consider the load factor γin view of the possible
D
conservatism built into the procedure used to estimate load effects, and especially the
damping model assumed. Extreme dynamic load effects in fixed platforms are sensitive to
equivalent damping values below 1.0 per cent of the critical value (Karunakaran, 1993). By
conservative estimate of the damping in that range, no additional load factor would be
required. If the equivalent damping is more than 1 per cent, the sensitivity to damping is so
small that no γ is required despite the large uncertainty in estimating the damping ratio. This
D
is often the case in practice.

5.5.3 Calculation of stress ranges for FLS check

Fatigue design is commonly based on resistance data specified by SN-curves. In special cases,
fracture mechanics approaches may be applied. Stress ranges are based on expected long-term
distributions of stress ranges, without any load factor. Moreover, the design criterion is based
on linear cumulative damage, such as the Miner—Palmgren law, typically allowing damage
in the range of 0.1 to 1.0. The significant uncertainties in fatigue loads and resistance imply a
high failure probability. Acceptable safety is hence ensured by a proper inspection,
maintenance and repair strategy. For this reason simplified design analyses may also be
justified.
Fatigue estimates may be based on alternative approaches—in a hierarchy of procedures
with increasing accuracy and complexity. Here, three main alternatives are considered:
●Assume that stress ranges follow a two-parameter Weibull distribution, obtained by
estimating s corresponding to an exceedance probability of 1/n ; and assume γaccording to
0
0
guidance—including the effect of dynamics—mentioned in Section 5.4.6. Calculation of s 0
and selection of λobviously need to be conservative.
●FDA for each sea state (i) to determine response variance and assume narrowband response,
implying Rayleigh distribution of stress ranges. Moderate non-linearities may be accounted
for by determining a quasi-transfer function based on time domain analysis, or another
linearization approach. Factors may be introduced to correct for wideband or non-Gaussian
response.
●TDA combined with rainflow counting of cycles for a representative set of sea states that
are found (e.g. by frequency domain analysis) to contribute most to the fatigue damage.