MATERIALS FATIGUE LIFE DISTRIBUTION - A MAXIMUM-ENTROPY APPROACH

A rational probability distribution for materials fatigue life is prop osed using the Maximum Entropy Principle (MEP) and the sample informat ion available. It has been shown that this distribution is most natura lly a truncated normal distribution. The expression of the distributio n as well as the relationships between the distribution parameters and the maximum entropy coefficients (or Lagrangian multipliers) is given explicitly. It is further shown that the maximum entropy estimators ( MEE) are equivalent to the classical maximum likelihood estimators (ML E) and the moment estimators (ME) provided that proper sample statisti cs are chosen as the approximations of the population parameters. A pr ocedure has been proposed for estimating the maximum entropy parameter s. Numerical examples showing the effects of the standardized truncati on point, the sample mean, and the truncation point have been given to demonstrate the significance and usefulness of the work.