Process Compensated Resonance Testing (PCRT) is a nondestructive evaluation (NDE) method based on the fundamentals of Resonant Ultrasound Spectroscopy (RUS). PCRT is used for material characterization, defect detection, process control and life monitoring of critical gas turbine engine and aircraft components. Forward modeling and model
inversion for PCRT have the potential to greatly increase the method’s material characterization capability while reducing its dependence on compiling a large population of physical resonance measurements. This paper presents progress on forward modeling studies for damage mechanisms and defects in common to structural materials for gas turbine engines. Finite element method (FEM) models of single crystal (SX) Ni-based superalloy Mar-M247 dog bones and Ti-6Al-4V cylindrical bars were created, and FEM modal analyses calculated the resonance frequencies for the samples in their
baseline condition. Then the frequency effects of superalloy creep (high-temperature plastic deformation) and macroscopic texture (preferred crystallographic orientation of grains detrimental to fatigue properties) were evaluated. A PCRT sorting module for creep damage in Mar-M247 was trained with a virtual database made entirely of modeled
design points. The sorting module demonstrated successful discrimination of design points with as little as 1% creep strain in the gauge section from a population of acceptable design points with a range of material and geometric variation. The resonance frequency effects of macro-scale texture in Ti-6Al-4V were quantified with forward models of cylinder samples. FEM-based model inversion was demonstrated for Mar-M247 bulk material properties and variations in crystallographic orientation. PCRT uncertainty quantification (UQ) was performed using Monte Carlo studies for Mar-
M247 that quantified the overall uncertainty in resonance frequencies resulting from coupled variation in geometry, material properties, crystallographic orientation and creep damage. A model calibration process was also developed that evaluates inversion fitting to differences from a designated reference sample rather than absolute property values, yielding a reduction in fit error.