Uncertain fatigue analysis under incomplete information

Abstract

Fatigue life prediction of structures under incomplete information besides being computationally demanding is more prone to higher model uncertainty magnitudes. Therefore, assuming aleatory-type input data (loads, material properties, and geometrical parameters), fatigue-induced sequential failures, and the finite element method were combined into a code under an uncertainty quantification framework. The proposed scheme provides multiple levels of calculation for a planar tubular truss under fracture mechanics approach. The maximum entropy principle was the approach to deal with the incompleteness of the inputs. In addition, a bi-level sensitivity analysis was performed to distinguish the most impacting factors from the least ones. The calculation indicated that, in terms of variance, the most impacting parameter related to fatigue life is the slope of “notch growth rate versus stress intensity factor range” curve. The fatigue notch growth boundaries and fatigue notch semi-width boundaries obtained in this paper bring out the most unbiased fatigue notch design mapping of the truss presented. Thenceforth, the designer can predict the notch semi-width within a required confidence level.