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 for material characterization, defect detection, process control and life monitoring for critical components in gas turbine engines, aircraft and other systems. Accurate forward and inverse modeling for RUS requires a proper accounting of the propagation of uncertainty due to the model and measurement sources. A process for quantifying the propagation of uncertainty to RUS frequency results for models and measurements was developed. Epistemic and aleatory sources of uncertainty were identified for forward model parameters, forward model material property and geometry inputs, inverse model parameters, and physical RUS measurements. RUS model parametric studies were then conducted for simple geometric samples to determine the sensitivity of RUS frequencies and model inversion results to the various sources of uncertainty. The results of these parametric studies were used to calculate uncertainty bounds associated with each source. Uncertainty bounds were then compared to assess the relative impact of the various sources of uncertainty, and mitigations were identified. The elastic material property inputs for forward models, such as Young’s Modulus, were found to be the most significant source of uncertainty in these studies. The end result of this work was the development of an uncertainty quantification process that can be adapted to a broad range of components and materials.