E. Grinzato et al., APPLICATION OF NEURAL NETWORKS TO THERMOGRAPHIC DATA REDUCTION FOR NONDESTRUCTIVE EVALUATION, Revue générale de thermique, 34(397), 1995, pp. 17-27
A methodological study on the use of artificial neural networks for th
ermal defects characterization is presented. An implementation of an a
lgorithm called dynamic thermal tomography, by means of a tree like ar
chitecture is described. Experimental results obtained on plastic spec
imens, containing artificial defects, confirm the potential of the pro
posed data reduction procedure. The evaluation of depth, and particula
rly of the thickness of internal voids, proved extremely satisfactory.