Abstract: Probabilistic life prediction of nickel-base superalloy requires the modelling multiple complex random phenomena. The fatigue crack growth rates at various stress ratios are probabilistically predicted using the Paris formula and the modified Paris formula with threshold value Kth for fatigue crack growth test data of GH4133B superalloy used in turbine disk of aero-engine at room temperature. The theoretic fatigue crack growth lives Nfp at various survival probabilities P predicted by integral based on the Paris formula and the modified Paris formula are calculated and compared with the test ones Nft. It is shown that Nfp obtained by integral based on the modified Paris formula is well fitted with Nft at P=50%, while both Nfp obtained by integral based on the Paris formula and that obtained by integral based on the modified Paris formula are all well fitted with Nft at P=90%, and Nfp obtained by integral based on the modified Paris formula possesses a high reliability at P=99%. A 3D Nf-Fp,m-Fp,a fatigue life surface are plotted based on the nonlinear regression analysis of fatigue life test data, and a parameter  is introduced to characterize the effect of mean load Fp,m and load amplitude Fp,a on fatigue life Nf. It is suggested that the effect of Fp,m on Nf is larger than that of Fp,a on Nf.

Key words: fatigue crack propagation, life prediction, probability, stress ratio, turbine disk