Robust Probabilistic Analysis of Deterioration-Induced Aeroelasticity in an Axial Turbine

The surface of a real-world turbine blade differs from the geometry of its model. Manufacturing tolerances, wear, and regeneration during the turbine’s lifetime impact the aerodynamic behaviour, stage efficiency, and parameters as previous probabilistic studies have shown. Aerodynamic changes lead to aeroelastic variances in downstream rows that can have a significant impact on high-cycle fatigue due to a possible increase of a blade’s vibration amplitude. The present study expands an existing tool chain to uni-directional aeroelastic full three dimensional simulations. The test object is the last stage of a low-pressure five-stage axial turbine; the geometric variances are applied to characteristic airfoil parameters of the vane by Latin hypercube sampling. A required sample size for an accurate sensitivity analysis is estimated a-priori and verified by computational simulations. The aerodynamic analysis identified the stagger angle, trailing-edge thickness, maximum thickness, and maximum camber as the most important parameters among the characteristic blade parameters. For large deviation in the stagger angle a non-monotonous influence on isentropic efficiency was found, which shows limitations of a sensitivity analysis based on the Spearman rank correlation coefficients. A variance-based sensitivity analysis was used for a more detailed analysis, which is capable to detect such non-monotonous relationships. In the aeroelastic simulations, the chord length, maximum camber, and the trailing-edge angle have the highest influence on aerodynamic forcing and damping on a downstream rotor blade of the changed stator blade. Furthermore, direct correlations between aerodynamic stage parameters and aerodynamic damping are shown, which allows reduced order modelling. The correlation between work coefficient and vibration amplitude was overlooked by the Spearman’s coefficient but identified by the variance-based approach. A maximum vibration amplitude is identified, indicating a potential of using probabilistic studies for adjusting the safety factors for high-cycle fatigue during the design process of blades.