Influence of Material Model on the Fatigue Life of Turbine Exhaust

Sammanfattning

This thesis investigates how different material models affect the predicted low-cycle fatigue life of a turbine exhaust component made from heat-resistant austenitic stainless steel. Three material models were evaluated: perfect plastic (PP), linear kinematic hardening (LKH) and non-linear kinematic hardening (NLKH). The models were calibrated using experimental isothermal low-cycle fatigue data fitted to Ramberg-Osgood curves. The calibration of each material model was performed as a sensitivity study to evaluate how changes in calibration parameters influence the predicted fatigue life. The fatigue life predictions from the different constitutive material models were also compared to evaluate the extent to which the material model formulation affects the results. Fatigue life was predicted using a strain-based method. To ensure reliable results, the fatigue life should be based on a stabilized cyclic response. Therefore, a simulation with multiple cycles was performed to determine how many cycles were required for the response to stabilize. In addition, the influence of creep was investigated by comparing the fatigue life from simulations performed with and without creep. An additional study was also performed using a Neuber-based method to evaluate how the choice of fatigue assessment method influences the predicted fatigue life.The investigation of cyclic stability showed that a sufficiently stabilized response was reached after two cycles. Therefore, two cycles were used in all subsequent simulations, as this was considered a reasonable balance between reliability and computational efficiency. The sensitivity study showed that the predicted fatigue life is influenced not only by the selected material model, but also by the calibration procedure. Although the different material models identified similar critical regions in the component, the fatigue life differed between the models.Creep had a limited influence on the average fatigue life of the component. This may be because some nodes reached a strain-controlled condition. Since the strain-based method is used, it does not fully capture the effect of stress relaxation during hold time. Therefore, a fatigue assessment method that accounts for stress relaxation effects is recommended for further evaluation. However, evaluating the critical fatigue life, creep had a more pronounced local influence, especially in the bolt-hole regions. The Neuber-based method predicted shorter fatigue life and is inherently more conservative than the strain-based method.Overall, the study shows that the choice and calibration of constitutive material model have a significant influence on fatigue life prediction. To better understand the causes of the observed differences, it is recommended that the calibrated material models be applied to a simpler and more controlled geometry. This would make it possible to better isolate the influence of the material model and evaluate how the results transfer to more complex conditions, such as those present in gas turbine components.