Uncertainty Quantification in Modeling and Measuring Components with Resonant Ultrasound Spectroscopy

ABSTRACT: Resonant Ultrasound Spectroscopy (RUS) is a nondestructive evaluation (NDE) method which can be used to quantify the elastic properties of a wide range of materials by measuring their resonance spectra. Modeling of RUS for critical engine components, such as Ni-base superalloy turbine blades, can predict the effect of various material states and damage on RUS frequencies. Accurate modeling of RUS frequency response to material states requires an understanding of propagation of uncertainty due to model, material and measurement error. Vibrant Corporation and Computational Tools have demonstrated capabilities for modeling RUS in previous and current work. In the Phase I project for quantifying uncertainty, the team will identify sources of error for modeling, measurement and materials. It will then perform parametric studies of RUS frequency sensitivity to those sources, and calculate confidence limits on the model outputs. The work will establish a foundation for a Phase II project. Phase II will introduce physical components for model/confidence limit validation, quantify uncertainty for more complex geometries that are representative of turbine blades, and evaluate ways to improve the confidence in model results. BENEFIT: Development of modeling capabilities for RUS/PCRT will have significant commercial benefits for Vibrant, the DoD, and Vibrant"s commercial customers. Vibrant will make predictive modeling an integral step in its process for creating commercial sorting solutions. Feasibility studies of new applications can use modeling to evaluate PCRT"s sensitivity to various defects of interest before time and money are expended on collecting parts and configuring test fixturing. Using predictive modeling will enable Vibrant to assess the effect of defects that are hard to generate or verify in physical parts. In previous projects, a large number of blades had to be collected and destructively analyzed and characterized to produce a training set of parts for a sorting solution. Modeling tools can populate a training database of resonance frequencies while requiring far fewer physical parts. Modeling will also be a crucial tool for more comprehensive material state awareness of a component, which is a major long-term goal for AFRL NDE research efforts.