Evolutionary Statistical Characters of Fatigue Damage of Smooth Surface Samples by an Effective Short Fatigue Crack Criterion

Abstract

Evolutionary statistical characters of fatigue damage are investigated on the smooth surface samples of 1Cr18Ni9Ti welded metal. Previous effective short fatigue crack (ESFC) criterion by Zhao et al. is employed with a local viewpoint of fatigue damage. Main attentions are paid to three kinds of ESFC data, i.e., density values of ESFCs, dominant ESFC (DESFC) lengths, and growth rates of DESFC. Six possible statistical models, i.e., normal, lognormal, extreme maximum value, extreme minimum value, two-parameter Weibull, and three-parameter Weibull, are compared to examine their validity for describing these three kinds of data. In the comparisons, three aspects of the statistical models are considered synthetically, i.e., total fit effect, consistency with fatigue physics, and safety of prediction in right tail region. Results reveal that there is a significant damage character of the microstructural short crack (MSC) stage and the physical short crack (PSC) stage. The variation coefficient of the data increases in MSC stage, and then decreases in PSC and long crack (LC) stages. Extreme minimum value distribution is the appropriate model for describing the density values of ESFCs and the DESFC lengths. While extreme maximum value distribution is the reasonable model for the growth rates of DESFC. Fatigue damage is subject to an evolutionary random process from an initially chaotic (nonordered) state, to an independent state, and finally to a loading history-dependent state.