The validation of some methods of notch fatigue analysis

Abstract

This paper is concerned with the testing and validation of certain methods of notch analysis which the authors have developed theoretically in earlier publications. These methods were developed for use with finite element (FE) analysis in order to predict the fatigue limits of components containing stress concentrations. In the present work we tested and compared these methods using data from standard notches taken from the literature, covering a range of notch geometries, loading types, R‐ratios and materials: a total of 47 different data sets were analysed. The greatest predictive success was achieved with critical‐distance methods known as the point, line and area methods: 94% of these predictions fell within 20% of the experimental fatigue limits. This was a significant improvement on previous methods of this kind, e.g. that of Klesnil and Lucas [(1980) Fatigue of Metallic Materials, Elsevier Science]. Methods based on the Smith and Miller [(1978) Int. J. Mech. Sci. 20, 201–206] concept of crack‐like notches were successful in 42% of cases; they experienced difficulties dealing with very small notches, and could be improved by using an ElHaddad‐type correction factor, giving 87% success. An approach known as ‘crack modelling’ allowed the Smith and Miller method to be used with non‐standard stress concentrations, where notch geometry is ill defined; this modification, with the same short‐crack correction, had 68% success. It was concluded that the critical‐distance approach is more accurate and can be more easily used to analyse components of complex shape, however, the crack modelling approach is sometimes preferable because it can be used with less mesh refinement.