Automated processing of thermographic derivatives for quality assurance

The use of thermography as a nondestructive testing method has increased significantly in recent years. However, implementation is usually based on visual interpretation of image results by a trained inspector. Although this is accepted practice in the aerospace industry, where thermography is widely used, it is not appropriate for automotive manufacturing, where higher production rates demand higher inspection throughput that can only be accomplished with automated detection. The logarithmic derivatives of flash thermography time histories provide an excellent basis for automated detection. A unique model, based on the logarithmic derivatives of a series solution for the surface temperature of a flash heated plate with adiabatic boundary conditions, provides a template for identifying pixels that deviate from uninterrupted diffusion that is characteristic of a defect-free sample. We demonstrate automated defect detection using this model on a composite sample.