Application of evidence theory in fatigue reliability analysis for impellers with consideration of hybrid uncertainties

This paper carries out a fatigue reliability analysis for impellers used in hydraulic pumps, with the consideration of both aleatory uncertainty on the geometric dimensioning, material property, load borne, and epistemic uncertainty on the fatigue strength coefficient and fatigue strength exponent. First of all, the Basquin's law is introduced and utilized as the basis for the fatigue reliability modeling. Afterwards, on the foundation of aleatory uncertainty modeling, a parametric finite element simulation analysis is executed to extract the dispersive characteristics of the strain response combining with the construction of interval model to quantify the epistemic uncertainty of fatigue model parameters. Finally, the fatigue reliability evaluation of the impeller with the hybrid uncertainty of probability-interval is implemented, and the belief and plausibility curves of the impeller fatigue reliability are obtained, respectively. It is found that the reliability evaluation results obtained from the fatigue strain range obeying a lognormal distribution is more conservative, compared with those obtained from the fatigue strain range obeying a Normal distribution. Such an observation can provide guidance for promoting the further reliability design and optimization of impellers.