Information condensation in defect detection using TSR coefficients images

The Thermographic Signal Reconstruction (TSR) technique, based on the logarithmic polynomial fitting of the temperature time-evolution, improves the efficiency of pulse-stimulated thermography. However, the non-destructive evaluation of the inspected structure requires the analysis of the full temperature-time sequence. This drawback impacts all transient thermal NDE method whatsoever. A new post-processing development of the TSR data, based on the use of the very limited number of polynomial coefficients, avoids this time-consuming task by producing a unique composite image of the defects, containing information on their depths. This image results from an RGB-basis projection of three of the most pertinent coefficient images. The performance of the method is compared with that of the classic TSR method, in the case of a carbon/epoxy coupon containing artificial defects simulating delaminations. Comparisons are also made with three other well-known data processing techniques: Pulse Phase Thermography (PPT), Principal Component Thermography (PCT) and HighOrder Statistics (HOS). In each case, the relevance of the new approach is proven: not only is it simple and rapid, but it also guarantees a defect detectivity at least as good as the one provided by the other techniques. Moreover, the RGB projection, successfully applied to all four methods (TSR, PPT, PCT and HOS), is shown to be an efficient operation to compress and synthesize the thermal data which was, so far, scattered in several images. It may be considered as a general post-processing technique useful for any such method currently used in NDE.