Investigation of Sources of Uncertainty for Stress Prediction in Dynamic Structures

the validity of the inferred results. Quantification of potential uncertainties is aimed at improving the estimation of a system’s performance, as well as identifying critical operating conditions. Previous work by the authors demonstrated the sensitivity of the stress prediction process to uncertainty in sensor location and geometric/material properties in the case of a turbine engine bladed disk, where simulated experimental data were employed. This study represents an extension to the previous work, where the model-based inference technique is applied to real test data, so that the inherent uncertainties related to sensor placement, accuracy and orientation, as well as the presence of noise in the data can be explicitly accounted for. The process is demonstrated on a reasonably large set of experimental data obtained from tests on simple plate structures, and processed according to procedures used during engine blade testing. The data are studied to statistically assess uncertainty during the stress evaluation process, and will be used to validate the uncertainty estimation process. In addition to evaluating uncertainty, this study provides necessary first steps for developing a procedure for statistically rigorous incorporation of experimental information into the systems model, which is aimed at improving the accuracy of its estimated vibratory response.