A framework is presented for a strain-based fatigue reliability analysis. The analysis-related experimental methods and test data are made out first. The random models, considering the entire material constants as dependent random variables using the Coffin-Manson law and the modified Ramberg-Osgood equation, respectively, for random cyclic strain-life and stress-strain relations are then successively proposed with considerations of survival probability and sampling size related confidence. Reliability methods starting from possible initial conditions, including nominal random cyclic stressing, random local cyclic straining, and inspected local cyclic straining, are, respectively established on a basic consideration of the random cyclic straining applied and capacity interference. Some deficiencies have been overcome from the assumption of incomplete independent random variables, the lack of consideration of the random cyclic stress-stress relations, and the empirical selections of partial statistical parameters in existent methods.
展开▼
