

Random Fatigue Limit on a P/C




Impetus There are at least three practical problems with the traditional, large numbers of sN tests, approach to estimating fatigue limits:
Because of the enormous scatter in lives at the stress levels of prime interest to HCF, and the expense of testing to very long cycle counts, a better method for estimating and validating fatigue limits is needed. An innovative approach to describing fatigue limit behavior was proposed recently by Pascual and Meeker. They postulated a random fatigue limit (RFL) model in which each specimen has its own fatigue limit in much the same way that each specimen has its own fatigue lifetime if tested at a sufficiently high stress. This random fatigue limit is explicitly included in the sN model. Maximum likelihood methods (described elsewhere) are then used to estimate the parameters of the sN equation as well as the parameters of the fatigue limit distribution. The percentiles of the fatigue limit distribution are easily calculated from the estimated parameters. The random fatigue limit model produces the proper shape of the median sN function and the type of scatter typically seen in fatigue tests at HCF stress levels, as is illustrated in Figure 1, for Ti6Al4V.
Reference: (Pascual and Meeker, "Estimating Fatigue Curves with the Random FatigueLimit Model," TECHNOMETRICS Vol. 41, No. 4, p.277302, November 1999) [next]

Mail to Charles.Annis@StatisticalEngineering.com 