Distribution of failures for increasing strain accelerated fatigue tests for elastomers can be approximated by a Gaussian distribution.

A step increasing strain accelerated fatigue test has been developed and validated for the evaluation of candidate elastomeric materials for the artificial heart program. Whereas standard fatigue tests can be approximated by a log-normal or Weibull distribution, the increasing strain accelerated fatigue test has the general appearance of being normally distributed (i.e., a Gaussian distribution). The hypothesis that the data is indeed normally distributed was examined using a variety of statistical tests. The mean and median were equivalent in all data sets compared, as they would be for normally distributed data. There was very little positive or negative skew found in data collected under a wide variety of conditions. The data was found to have a slightly stronger than expected central tendency (positive kurtosis), but most of this disappeared when the data were normalized. Chi-squared analysis found normally distributed data in most subset of the data except for those with small numbers of test specimens per test. Normalized test data was not found to differ significantly from a Gaussian distribution by the Kolmogorov-Smirnov test. It therefore appears that increasing strain accelerated fatigue test data can be approximated by a normal distribution. This allows for easy data interpretation and aids in the extrapolation of incomplete data sets.