An Error Propagation Method for Uncertainty Quantification of Resonant Ultrasound Spectroscopy

ABSTRACT: Resonant Ultrasound Spectroscopy (RUS) has been widely used to measure material elastic modulus. However, the accuracy of RUS is affected by many uncertainty factors in measurement, material and models. Determining the reliability and confidence level of RUS measurement requires comprehensive uncertainty quantification and error propagation analysis. We propose to develop an innovative uncertainty quantification approach for RUS (UQ-RUS) based on Bayes"Theorem and efficient high dimensional random space propagation and sampling algorithms to address the problem of isolating and quantifying uncertainty sources to the material moduli estimation process, resulting in a more reliable and confident characterization of material status. Phase I program is to demonstrate the feasibility of UQ-RUS with several representative random factors. In Phase II program, we will continue the development of the UQ-RUS architecture with the focus on optimizing the model and algorithms, performing comprehensive analysis of random parameters and validation tests. BENEFIT: UQ-RUS has significant commercial potential for applications in material design, manufacturing and maintenance. Without accurate knowledge of material properties, large safety factors need to be applied on engineering structures, which significantly increases operation and maintenance cost, especially for aerospace components. The proposed uncertainty quantification tool can potentially provide distribution, expectation, and standard deviation of critical material properties; therefore greatly reduce costs and labors. Both government and commercial organizations will benefit from the development of UQ-RUS. In addition, reliable and efficient simulation tools for material characterization are in high demand from both military and commercial industry. A significant return on investment (ROI) can be realized from the use of modeling and simulation software tools in partnership with traditional experimentation (on the order of $3 to $9 for every $1 invested in these tools and associated expertise). We expect the market for high fidelity stochastic material simulation tool will be huge.