Statistical evolution of small fatigue crack behavior is analyzed. Three co
nsiderations an given, They consist of the total fit, the consistency of st
atistical parameters with test data, and the practice of commonly used dist
ributions, namely Weibull (two- and three-parameters), normal, lognormal, e
xtreme minimum, extreme maximum, and exponential. The crack growth rate dat
a of 1Cr18Ni9Ti stainless steel-weld metal reveal that fatigue damage is a
gradual process involving an initial non-ordered (chaotic) or load history-
independent random state to an ordered load history-dependent random state.
The seven distributions are examined. Different from existent conclusion,
appropriate assumed distribution for characterizing the evolution data is t
he normal or extreme minimum value model rather than the Weibull two- or th
ree-parameter. The other distributions do not fit the data as they violated
two or all three of the mentioned considerations. (C) 1999 Published by El
sevier Science Ltd. All rights reserved.