To achieve the objectives of condition-based maintenance plus prognosis (CBM+), and realize its potential, the location and size of damage at any length scale, e.g., either a crack or a microstructural perturbation, needs to be determined with statistical metrics to feed prognostic reasoners and risk assessments. Previous work by Victor Technologies has focused on developing estimation-theoretic metrics for model-based inversion algorithms in eddy-current NDE. During Phase II, we will build on the Phase I feasibility study to demonstrate and validate a general statistical theory of uncertainty propagation with appropriate metrics, and apply the results to more challenging three-dimensional problems, including those in which sizing and location of flaws are required, as well as materials characterization. Further, we will continue the development of high-dimensional model representation (HDMR) algorithms for application to statistical problems in NDE, which we first studied during Phase I.
Benefit: The technology that we develop in this proposal will be applicable to the aerospace, nuclear power, materials characterization, civil infrastructure, and many other industries, so our research will have commercial benefits that extend far beyond military applications.
Keywords: model-based inversion, voxel-based inversion, estimation-theoretic metrics, stochastic reliability metrics, high-dimensional model representation, nondestructive evaluation
