Fatigue Reliability Analysis of Turbine Disk Alloy Using Saddlepoint Approximation

In this paper, a new fatigue reliability analysis method based on saddlepoint approximation (SPA) was proposed for calculating the probability of failure of turbine disk alloy in a low cycle fatigue (LCF) regime. Firstly, two LCF life prediction models based on total strain energy density and Support Vector Regression (SVR) meta-model are presented for turbine disk alloy GH4133 under different loading conditions at 250 degrees C. Compared with the SWT model, modified Walker model and Response Surface (RS) model, the predicted lives by the proposed models are within a factor of +/- 2 and a factor of +/- 1.1 respectively. Secondly, based on the fatigue design criteria, the probabilities of failure are calculated using SPA for the explicit and implicit performance functions using two proposed LCF models and viscosity-based model. These three models have provided the reliability design rules for GH4133. Finally, the failure probabilities curves between SPA and the designed fatigue lives are achieved. The reliability analysis results were found to be in good agreement with the calculated results of test data. These results show that SPA is very apt for the fatigue reliability analysis of turbine disk under different loading conditions using only a small number of samples without any distribution assumptions for random variables. Moreover, it can be used to estimate the system's probability of failure with a large number of random variables or high nonlinearity of performance functions. The effectiveness and accuracy of the combination of the fatigue models and SPA for fatigue reliability analysis are verified using three examples.