Crack sizing from vibrothermography data using neural networks

The US Air Force is interested in integrating vibrothermography nondestructive inspection, or sonic IR, as a complementary tool to Florescent Penetrant Inspection (FPI) and Eddy Current Inspection (ECI) for detection of fatigue-like cracks in metal turbine engine components [1]. Vibrothermography produces qualitative information about the presence/absence of defects that generate hit/miss data. However, measurements methods for precise defect size determination are lacking. Determining the size of the defect would be critical to expand the usefulness to fracture critical life limited turbine engine components. Therefore, an accurate and reliable method for determining the size of a crack is necessary. We propose a Crack-Size-Finder method based on combining experimental data with synthetically generated one. The synthetics data is used as a ruler to measure the crack dimensions from experimental data.