Probabilistic Life Assessment of Turbine Vanes

Life of a gas turbine vane generally depends on different factors such as scatter of material properties, load variation and manufacturing tolerances. However, deterministic finite element (FE) life analysis gives just a discrete value typically based on the nominal or worst case conditions. It precludes considering sensitivity to the input parameters and obtaining the expected life range. To consider the possible variations of the input parameters from their nominal values, a probabilistic approach has been applied to compute the LCF (Low Cyclic Fatigue) and creep life distributions for the uncooled vane. The deterministic 3D FE life assessment of the gas turbine components is based on the input data such as physical and mechanical properties of the base material and coating at operating temperatures, nominal geometry of the component, thermal and mechanical loadings. Each of the above mentioned inputs has its own scatter band characterized either by average and minimum values of mechanical properties (tensile strength, LCF, creep) or by variations of manufacturing tolerances; thermal boundary conditions and gas pressure distribution. The probabilistic life analysis has been performed in order to assess individual impact of each input on vane’s life scatter. LCF and creep life distributions as well as variation of the base metal oxidation layer thickness have been obtained for each scatter factor and for their overall contribution. It is seen from results that LCF and creep lives of the analyzed vane have been influenced mainly by material properties and secondarily by OTDF (hot gas temperature variation in the circumferential direction) and uncertainties of thermal boundary conditions, which depended on the operation conditions of the engine. Manufacturing tolerances and alternation of ambient air temperature in the compressor intake have the lowest impact. The derived model is useful for the risk analysis or maintenance planning. For instance, it has been shown how probability of small fatigue crack indication in one vane can be extended onto the overall probability for the failure detection of n vanes at the stator stage during regular inspection. The probability of micro crack growth due to creep after the determined amount of operating hours for the single vane may be also redefined into the overall stage probability for the detection of n such vanes. To perform validation, normalized field data have been used for comparison with the analytical predictions. Good correlations between the field data and analytical predictions have been shown.