Abstract
In the 2009 addendum to the ASME BPV Code [1], new design fatigue curves were adopted for austenitic stainless steel and nickel-based alloys. These changes followed publication of NUREG/CR-6909 [3] which, in addition to describing a methodology to account for environmental effects on fatigue, challenged both the mean air curve underlying the design fatigue curve and factors used to define the design position from this mean behaviour.
The methodology to build the design curve starts with the mean air curve obtained by laboratory testing, a correction is made for mean stress effects and then the resulting curve is translated by coefficients, referred to within this paper as design factors, applied on stress or on cycles, with the lower of the two translated curves defining the design curve. The translation by these design factors is to account for factors present in service applications but not present in the laboratory testing and also to provide a conservative position accounting for scatter in fatigue behaviour.
Following a detailed review of the appropriate factor on cycles in NUREG/CR-6909 [3], the revised ASME design curve replaced the historic design factor of 20 on cycles with a factor of 12, although a later revision of NUREG/CR-6909 [4] further refined this to below 10, the factor of 12 remains. The historic design factor of 2 on stress was not reviewed in any detail NUREG/CR-6909 and was not changed at this time. However, a number of reviews previously and since [5],[6] and [7] have suggested that a design factor of 2 is excessive compared to the aim of describing a 95% bounding curve position.
This paper presents a series of statistical analyses of the same austenitic stainless steel data used to generate the mean curve fit underlying the current design curve, supplemented with additional, high-quality, modern data to demonstrate that a substantially-reduced factor on stress would be appropriate to the aims of describing a 95% bounding curve.
In conjunction with other recent literature, these analyses present a strong basis for revision of this historic excess-conservatism while maintaining a suitably-conservative position. However, as such blanket conservatisms are reduced, the limitations, simplifications and assumptions of the current curve form and design curve construction must be discussed and evaluated, this wider context is described to support further discussion.
