Uncertainty Quantification in Modeling and Measuring Components with Resonant Ultrasound Spectroscopy

ABSTRACT:The objective for SBIR solicitation AF141-152 is the definition, development and execution of an uncertainty propagation analysis for modeling RUS frequency variation for Ni-base superalloys. Ensuring the accuracy of RUS modeling techniques requires a proper understanding of propagation of uncertainty in model results from various sources of error. In Phase I of the project, the Team identified sources of error and developed an uncertainty quantification (UQ) process that they applied to models and physical measurements of Ni-base superalloy coupons. For Phase II, the team proposes to apply that UQ process to additional geometries through a stepwise progression to more complex shapes. The team will compare modeled UQ results to RUS measurements for physical samples. These comparisons will enable verification and validation of the UQ results. Examinations of alternate RUS hardware configurations will be explored in a DOE format. Case studies to validate the uncertainty quantification process will be conducted. The team will also lay out plans for a transition to Phase III, which will feature a prototype RUS system in a production environment.BENEFIT:Development of modeling/UQ capabilities for RUS/PCRT will have significant technical and economic benefits for the DoD and commercial customers. Vibrant will make forward and inverse modeling with UQ an integral step in its process for creating inspection solutions. RUS/PCRT inspections will have better resolution for detecting the conditions of interest while reducing scrap rate. The new capabilities may also preserve serviceable life for some life-limited parts. RUS/PCRT enhanced with modeling and UQ techniques may reduce the fleet-wide cost of failure by reducing the frequency of field failures through improved detection of defects in critical components. While developing RUS/PCRT applications, forward modeling can be used to evaluate sensitivity to various defects of interest before time and money are expended on collecting parts and configuring test fixturing. Forward modeling can also enable the virtual population of a training database of resonance frequencies including the variations/defects of interest, requiring far fewer physical parts for a training database. Inverse modeling enables the determination of material state for physical components using measured resonance frequencies. Forward and inverse modeling will also be a crucial tool for more comprehensive material state awareness of a component, a major long-term goal for AFRL NDE research efforts.