Automatic defect detection from thermographic non destructive testing

Abstract. This paper illustrates an innovative algorithm for the detection and localization of partial detachment defects affecting thermal barrier coatings (TBCs), using the Long Pulsed Thermography (LPT). The scientific literature offers many applications, proving the LPT effectiveness in providing clear and intelligible thermal image contrasts. The proposed post processing technique directly operates on the raw thermograms acquired from the specimen surface. The algorithm aims at improving the polynomial fit of the logarithmic time history of the surface temperature during the cooling stage. The enhancements consist of the introduction of additional parameters, such as the fit zero-intercept and its standard deviation whose efficiencies are compared with the already exploited fit slope and determination coefficient ones. The considered parameters, called damage classifiers, are set in image maps from which the possible defects are deducted.