Stochastic inverse thermographic characterization of sub-pixel sized through cracks

The present paper describes an approach for inferring the presence and nature of tiny flaws in thin metallic panel components. The flaws are selected to be reminiscent of nascent-stage through cracks that might appear in the thin aluminum skins of aircraft, for instance.A laser heating source is used in conjunction with a low cost microbolometer-based digital IR imaging system in order that image processing might be applied to uncover promising locations to examine further, for the possible occurrence of tiny flaws. These local regions are subsequently considered during the solution of a stochastic inverse problem; aimed at inferring the existence and character of these “unseen” flaws that fit within an individual pixel associated the imaging field of view. The study is computational; employing surrogate experimental data.

► This paper reports on an algorithmic means, for characterizing flaws in thin metallic plates.
► Optically unresolved flaws in a given pixel may be inferred from behavior of surrounding pixel.
► MCMC approaches highlight underlying uncertainty accompanying flaw parameter estimation.
► Currently available, inexpensive IR cameras seem to be practical for such applications.