Analysis of the accuracy of a neural algorithm for defect depth estimation using PCA processing from active thermography data

In the paper a neural algorithm, which uses an active thermography for defect depth estimation, is presented. Simulations of the algorithm, for three datasets representing different phases of the heat transfer process developing in the test sample were performed. The influence of the emissivity error of the test sample surface on the accuracy of defect depth estimation is discussed. The investigations were performed for test sample made of the material with low thermal diffusivity.

► The method for characterization of defects with active thermography is presented.
► The regressive artificial neural network was used for defect depth estimation.
► The impact of the emissivity error on the depth estimation was determined.
► The results are valid for materials with low thermal diffusivity.