This paper presents the application of probabilistic methodology to a mechanical component. The probabilistic analysis approach assigns Probability Density Functions to sources of uncertainty and variation and then propagates the PDFs through a physics-based model to produce PDFs of model responses. This paper discusses how appropriate PDFs are selected for boundary condition uncertainty, model uncertainty, and manufacturing variation. A Latin Hypercube experimental design provides a series of

more »

... esign points that fill the entire design space. The Latin Hypercube is run through a physics-based model to relate model inputs with analysis outputs. With this data, the model is emulated with a Gaussian Process. The Gaussian Process serves as a fast running approximation of the physics-based model. The emulator is coupled with lifing equations for a Monte-Carlo analysis that yields probability distributions for model outputs. Furthermore, sensitivity analysis quantifies the relative effect of uncertainty and variation on part life. A jet engine turbine component is used as an example of the application of the general methodology. Nomenclature a, b ,c-thermal mechanical fatigue life equation coefficients í µí±˜ 1...5-b-spline coefficients pf-profile factor í µí¼Ž-standard deviation R 2-coefficient of determination í µí±‡ í µí±™í µí±œí µí±í µí±Ží µí±™-local temperature value at given radius í µí±‡ í µí±Ží µí±£í µí±"-the average temperature of the temperature profile í µí±‡ í µí±-the cooling air temperature í µí±‡ í µí±"-the average hot gas path temperature í µí¼‡-mean\nominal AMO-Advanced Modeling and Optimization.