Reliability prediction of monolithic structural ceramics involves stress analysis through Finite Element Analysis (FEA) and a probability of failure prediction code such as CARES that considers the strength of ceramics as a statistical (Weibull) distribution. Though the strength is considered as a statistical distribution, stress in a given volume or surface element is discrete in current practice. However, uncertainties in FEA input such as material properties or boundary conditions make the predicted stress uncertain, thereby resulting in predicted reliability being uncertain as well. A probabilistic framework has been developed for reliability prediction to estimate component reliability by treating the uncertainties in loading, boundary conditions, and material properties as random variables. The framework consists of an automated, closed loop process integrating algorithms for assessing failure probability due to general randomness. Efficient and accurate computational methods for reliability analysis have been implemented in the framework for significant savings in computation time. The methodology is demonstrated on a gas turbine component. The analysis shows that reliability is compromised significantly by a design based on mean values of the random parameters, while an overly conservative design based on a worst case scenario will result in rejection of too many components at unnecessarily high proof test loads. This method ultimately leads to the determination of an optimum proof test level to assure component reliability amidst several sources of uncertainties in reliability prediction.

This content is only available via PDF.
You do not currently have access to this content.