STATISTICAL EVALUATIONS OF TEST METHODS FOR FATIGUE LIMITS

摘要

Several simulation-based algorithms for generation of fatigue limit data using the staircase, the ray-projected, and the parallel-projected methods are developed based on a given statistical distribution of the fatigue strength at the life of 10~7 cycles to failure. The simulated test results are statistically evaluated against the given distributions with a target mean and two different coefficients of variation (COV (N_f)) of cycles to failure. It is concluded that in the regular COV (N_f) (say 30% for the cycles to failure), the parallel-projection is the best, then followed by the Dixon and Mood approach. In the extremely large COV(N_f) (say 2,200% for the cycles to failure), the Dixon and Mood method is recommended.