Current design lives for US Air Force turbine engine materials are based on a 1 in 1000 rate of nucleation of an engineering sized crack (B0.1). These lives are determined from models fitted to test coupon fatigue data at many different loading conditions. It has been shown that this methodology can sometimes lead to excess conservatism, and it often does not fully incorporate understanding of the mechanisms that drive crack initiation, growth, and fracture. A mechanism based probabilistic life forecasting methodology has been previously proposed with the objective to improve the prediction of minimum fatigue life or design life through understanding of the type and frequency of the material mechanisms that lead to early or immediate fatigue crack initiation.
An approach is proposed and demonstrated for the estimation of probabilistic mechanism-based design life prediction confidence bounds. These confidence bounds on the calculated B0.1 or minimum life predictions are dependent on the quality and quantity of the data used in the analysis. The effect of additional data from either small crack growth tests or microstructural characterization or fractography analysis on the extent of the calculated confidence bounds is shown. The analysis presented can ultimately be used to describe a relationship between the required confidence in the design life predictions to the cost of the test program required to collect the necessary data. Comparisons are made between data requirements and the level of confidence in empirical statistical predictions from fatigue test data and the new probabilistic mechanism-based design life predictions for laboratory specimens in a turbine engine material.