Planning and statistical processing of the results of tests for plotting quantile fatigue curves

Confidence limits for the quantile values of the fatigue strengths for plotting fatigue curves are established, and a procedure for planning fatigue tests is developed. A normal distribution function is adopted as the hypothetical distribution of the logarithm of the fatigue life and fatigue strength. The results show that the experimental values of the confidence coefficient depend only slightly on the total number of tests; acceptable results are obtained for 20-40 tests. The consistency of the experimental and theoretical data are most strongly influenced by the fracture probability and the degree of extrapolation of the fatigue curve. Recommended values of the relative error of determination of a fatigue characteristic are indicated.