STP1106: Fatigue Data Pooling and Probabilistic Design

    Yu, MT
    Research engineer, graduate student, and professor, University of Waterloo, Waterloo, Ontario

    Lu, FJ
    Research engineer, graduate student, and professor, University of Waterloo, Waterloo, Ontario

    Topper, TH
    Research engineer, graduate student, and professor, University of Waterloo, Waterloo, Ontario

    Irving, P
    Head of Advanced Technology, GKN Technology Limited, Wolverhampton, West Midlands

    Pages: 17    Published: Jan 1991


    Abstract

    In this paper, the authors propose data pooling techniques that allow designers to extract statistical measures of fatigue resistance and its variability from several sets of fatigue data for a given material. As the number of fatigue data banks and the exchanges of fatigue data continue to increase, these techniques will enable designers to obtain fatigue statistics from pooling of data sets that could not be obtained from any one of the invariably small sets of data. When stress-life fatigue data are plotted on log-log scales, the Basquin relationship, a simple inclined straight line, is used to fit the data. When strain-life fatigue data are plotted on log-log scales, two inclined lines are used to fit the data. Elastic and plastic strain fatigue data are plotted based on Basquin and Coffin-Manson relationships, respectively. Fatigue data sets from the same population are pooled together based on the variance, slope, and intercept of the samples. Applications of the data fitting and pooling are presented using sample data sets. A probabilistic design concept is illustrated using pooled material data and arbitrary load histories.

    Keywords:

    materials databases, computers, materials data banks, fatigue (materials), databases, strain (materials), stress (materials), life, pooling, statistics, probabilistic design, reliability


    Paper ID: STP17684S

    Committee/Subcommittee: E49.01

    DOI: 10.1520/STP17684S


    CrossRef ASTM International is a member of CrossRef.