Probabilistic fatigue evaluation of notched specimens considering small sample properties under multiaxial loading

Abstract

Probabilistic fatigue evaluation is the key to ensure the service integrity of large-sized mechanical equipment, but in practical engineering, the probabilistic fatigue models coupled with multiaxial stress conditions and small sample information characteristics are still insufficient. Accordingly, the virtual subsample augmentation method is effectively coupled with the improved Bootstrap method to propose a general probabilistic fatigue evaluation process. Specifically, the random sample regeneration method is first used to expand the small sample fatigue data, and then life distribution information under different stress levels is fused based on the backwards statistical inference method, the P-S-N curve is fitted with the integrated life dispersion information. Secondly, a new fatigue damage parameter based on the inhibition function is proposed to modify the P-S-N curve and predict fatigue life. Meantime, the multiaxial fatigue test of Q355(D) alloy steel with different notch sizes is carried out, and the model is verified in conjunction with existing fatigue test data of TC4-DT alloy and TC4 alloy. The results show that the proposed model has a certain universality for notched specimens, the modified stress field intensity method can accurately characterize the impact of notch and size effects on fatigue damage of specimens.