Citation

Harlow DG. Int. J. Reliab. Qual. Safety Eng. 2005; 12(6): 535-550.

Affiliation

Mechanical Engineering and Mechanics, Lehigh University, 19 Memorial Drive West, Bethlehem, PA 18015, USA

Copyright

(Copyright © 2005, World Scientific Publishing)

DOI

10.1142/S0218539305002002

PMID

unavailable

Abstract

Probability analyses are increasingly being used for reliability and durability assessments for life prediction of engineered components and systems. Nevertheless, many of the current analyses are predominately statistical rather than probabilistic. Fatigue life prediction has historically been based on the safe-life or the crack growth approaches, both of which are empirically based. Consequently, they do not adequately reflect long-term operating conditions, or identify the sources and extent of their contributions to variability. A comparison between probability and statistical approaches for fatigue life prediction is developed herein. Using simple crack growth models, the variability inherent in S-N response can be related to key random variables that are readily identified in the models. The identification and quantification of these variables are paramount for predicting fatigue lives. The effectiveness of probability modeling compared to statistical methodologies is shown through the analysis of an extensive set of S-N data for 2024-T4 aluminum alloy. Subsequently, the probability approach is demonstrated with S-N data for SUJ2 steel, in which two distinct failure modes are operative. Variability associated with manufacturing and material variables.