Statistical evaluation of strain-life fatigue crack initiation predictions

Most of the existing methods for estimating εN parameters are based on a relatively limited amount of experimental data. In addition, sound statistical evaluation of the popular rules of thumb used in practice to estimate fatigue properties is scarce, if available. In this work, an extensive statistical evaluation of the existing Coffin–Manson parameter estimates is presented based on monotonic tensile and uniaxial fatigue properties of 845 different metals, including 724 steels, 81 aluminum alloys, and 15 titanium alloys. The studied Coffin–Manson estimates include the methods proposed by Muralidharan and Manson, Bäumel and Seeger, Roessle and Fatemi, Mitchell, Ong, Morrow, Raske, as well as Manson’s universal slope and four-point correlation methods. From the collected data, it is shown that all correlations between the fatigue ductility coefficient ε′f and the monotonic tensile properties are very poor, and that it is statistically sounder to estimate ε′f based on constant values for each alloy family. Based on this result, a new estimation method which uses the medians of the individual parameters of the 845 materials is proposed.