INCEFA-SCALE TEST DATA COMPARED TO ENVIRONMENTAL FATIGUE DESIGN CURVE METHODOLOGIES

INCEFA-SCALE is a five-year project within the European Commission HORIZON 2020 programme starting in October 2020. Nuclear plant operators have generally observed that the number of failure events attributable to Environmentally Assisted Fatigue (EAF) are fewer than predicted by current assessment methodologies. It is recognised that a contributor to this discrepancy is the transferability of data from laboratory-scale tests to real nuclear components. This is the main knowledge gap addressed by INCEFA-SCALE. This is being achieved through the combination of testing, analysis, and materials characterisation to provide laboratory testing data, and the required meta data to understand the resulting behaviour, generated under conditions that are more representative of those experienced by plant components. The EAF assessment methodology described in NUREG/CR-6909 Rev 1 designed for use with the ASME BPVC Section III fatigue design curves (using factors of 12 on life and 2 on stress/strain) provides an international benchmark and technical justification. This benchmark has been the subject of over a decade of research defining the levels of conservatism provided and proposing alternative approaches to clearly define appropriate margins in EAF assessment methods. One such approach, initiated by the Japanese ‘Subcommittee on Design Fatigue Curve’and further developed by Rolls-Royce, offers modifications to the design curve approach through the use of ultimate tensile strength dependent mean curves, justification of a new approach to transference factors, the use of the Smith-Watson-Topper mean stress correction approach and explicitly accounting for variable amplitude loading in the high cycle regime of fatigue. Substantial research has been conducted and improvements proposed for the low-cycle regime of the design curves (where the factor on life applies for the ASME approach). However, research data in relevant high temperature water environments relating to the transference factors influencing the factor of 2 on stress/strain is limited. Part of this value is defined by loading history and understanding how to account for this phenomenon in a fatigue assessment requires the production of research data that enables the behaviour of materials subjected to variable loading to be understood. Variable loading is known to affect hardening and mean stress behaviour. Additionally, failure modes where a crack can initiate under larger loading amplitudes and grown to failure by sub-endurance limit loading may be important. This means laboratory testing under relevant environments and variable strain amplitudes is required to advance this area and define new benchmark methods that can adequately account for such behaviours while providing an appropriate level of conservatism. This paper focuses on the presentation and analysis of test data produced by the INCEFA-SCALE project. The data investigates the effects of loading history, environment, and surface finish, through the application of single and variable amplitude waveforms to 316L stainless steel uniaxial fatigue specimens in both air and simulated Pressurised Water Reactor primary coolant environments. By altering the above variables, the aim of this paper is to provide an understanding of how the method outlined in NUREG/CR-6909 and the alternative method proposed by the Subcommittee on Design Fatigue Curve accounts for these effects. It is shown that at time of writing the data would provide support for reducing the NUREG transference factor on stress to 1.4, however a recommendation on reducing the factor on cycles requires more analysis. The reduction in the stress factor is consistent with the approach of the Japanese ‘Subcommittee on Design Fatigue Curve’. There is however inconsistency between the use of elastic pseudo stress amplitudes used in fully reversed design life predictions, and the yield stress value used to govern the boundaries of stress-based mean stress corrections which should be considered in any new design curve approaches.