Validation of process compensated resonance testing (PCRT) sorting modules trained with modeled data

Process Compensated Resonance Testing (PCRT) combines the collection of broadband resonance data with advanced pattern recognition to produce a fast, accurate, and automated non-destructive inspection for aerospace, automotive, and power generation components. To create a PCRT targeted defect inspection (Sorting Module) the resonance spectra of statistically significant populations of characterized acceptable and unacceptable parts are needed to train PCRT algorithms to recognize the frequency patterns that indicate defects in the midst of normal, acceptable material and geometry variations. In cases where a sufficient number of parts are not available, spectra from physical part populations can be supplemented with ’virtual’ spectra generated with PCRT forward models. Previous work investigated the creation of model-trained sorting modules for the detection of creep deformation and crystal orientation for coupon and turbine blade geometries made from single crystal Ni-based superalloy. In this work, model-trained Sorting Modules for crystal orientation and creep deformation were created and then validated with statistically significant populations of physical coupons with and without creep deformation, and with a range of crystal orientations.