UC Santa Barbara

The objective of this dissertation is the development of both forward and inverse modeling frameworks that extend the nondestructive evaluation (NDE) capability of resonant ultrasound spectroscopy (RUS) for materials science applications through utilization of modern computational resources and a rigorous application of the fundamental physics of mechanical resonance. First, a forward modeling framework (FMF) based on 3D linear-elastic finite element models is developed to explain the complex manner in which resonance frequencies of a specimen change as damage is accumulated. Numerous examples of damage common to Ni-based superalloys for structural applications in extreme thermo-mechanical environments, like those found in turbine engines, are specifically addressed including: creep, fatigue and recrystallization. Then, an inverse modeling framework (IMF) devised with the goal of quantitative NDE of materials based on RUS measured resonance frequencies is introduced. Built upon a full-Bayesian approach to the inverse problem, the novel IMF devised here affords substantially simplified experimental methods by estimating elastic constants and crystal orientation parameters simultaneously. The potential of the IMF is demonstrated by evaluating a series of novel Co and CoNi-based single crystal superalloys, with robust convergence behavior as compared to optimizer-based approaches described in the literature. Together, the finite element-based FMF and the Bayesian IMF demonstrate the immense value and insight gained by coupling experiments with physics-based models for state-of-the-art materials characterization and NDE.