Page 359 - Marine Structural Design

Chapter 17 Fatigue Capacity 335

While the influences of the environment and plate-thickness are explicitly account for in most
of the design codes, other items listed in the above may or may not be required to be
considered by some design codes.
17.1.6 Fatigue Strength Improvement
When the theoretically calculated fatigue life is less than the required, methods of justifying
the fatigue design include:
By improving the design of structural details (e.g. to reduce stress concentration, residual
stress and misalignment, and locally increase the wall-thickness)
By using improved analysis methods: Spectral fatigue analysis is usually more accurate
that simplified fatigue assessment. Time-domain analysis may be better than the spectral
fatigue analysis. The selection of sea states and loading conditions and quality of the
environmental data all will influence the fatigue analysis results.
From capacity point of view, the three most important factors that affect fatigue are: stress
concentration due to weld geometry, defect shape and distribution and residual stress. Hence,
methods for improving fatigue capacity through fabrication and repair include (BV, 1998):
Modification of the weld geometry by grinding or weld toe remelting
Improvement of the welding procedures and workmanship
Introduction of compressive stresses, for example by hammer or shot peening
Post-weld heat treatment
However, the most efficient methods re possible improvement of the design such as reducing
the geometric stress concentration factors (BV, 1998):
Improvement of the shape of cut-outs
Softening of brackets toes
Local increase in thickness
More detailed discussions on improvement of weld details and fatigue design are given in Part
III Chapter 22.

17.1.7 Experimental S-N Curves
Most of the S-N curves are determined in laboratories where test specimens are subjected to
constant amplitude until failure. The S-N curves are derived by their mean fatigue life and
standard deviation of log N. The mean S-N curve is defined as 50 percent of the specimens
will fail. The basic design S-N curve is given as:
log N = log K,, - m log S (1 7.8)

where K,, is obtained from the mean value of logK, . To derive the S-N curves, a large
number of tests are required. However, when the coefficient m is known, 10 tests may be
sufficient to accurately derive the S-N curve (BV, 1998):

. 5 at stress level corresponding to N = lo4
. 5 at stress level corresponding to N = 5 IO5