This paper considers finite element analyses that have been performed to support a fatigue endurance testing programme. This programme is aimed at understanding the influence of Light Water Reactor (LWR) environment on the fatigue life of austenitic steels under thermo-mechanical loading. Testing has typically been performed on membrane loaded fatigue specimens under isothermal conditions. However, a new test facility at Amec Foster Wheeler has been developed to enable hollow specimens to be subjected to thermal and mechanical loading for a range of thermal cycles.
This work has provided a theoretical underpinning of the observed difference in lifetimes between pressurised and unpressurised specimens, and therefore provides a means of mapping data from hollow specimen testing back to solid specimen data. The parameter used to quantify these differences was the Von Mises equivalent strain. This accounts for the additional contributions from radial and hoop strains (which come about due to the internal pressure) and therefore gives enhanced strain amplitude which is fed into the S–N curve. By comparing unpressurised and pressurised lifetimes this way, a direct comparison could be made with test results. This serves two purposes; one is to provide mechanistic understanding of the difference in lifetimes. Secondly the approach will develop an assessment methodology to treat hollow specimens so a direct comparison can be made to bar specimen S–N curves. To provide further confidence in this mechanism being the dominant factor behind this difference, an independent calculation was carried out using multi-axial fatigue life models, namely Brown-Miller and Fatemi-Soci. Good agreement was observed with these models indicating that the Von Mises strain parameter was a valid parameter to characterise the multi-axial strain behaviour at the initiation site.
The increase in Von Mises strain between pressurised and unpressurised specimens was found to be a factor of 1.20 on average. This was slightly less than the experimentally derived strain differences of about 1.25. However, there was another aspect of the test results that required investigation. It has been observed from tests that the hollow specimens had a tendency to fail at a region just after the shoulder, which is where the specimen increases in thickness. Upon inspection of the plastic strains from FEA, it became clear that there was a small increase in strain on the inside surface due to this geometric feature of the specimen. The strain at the high strain region was a factor of 1.025 higher than the centre of the specimen after shakedown. Therefore, when the two effects of geometry and incremental plasticity are combined, the agreement between experiment and FEA is better. There is still a slight difference between the observed factor and the predicted factor, and reasons for this discrepancy are discussed in the paper.