On the use of the artificial neural network to determine heating parameters for active thermography in composites, a feasibility study

Abstract

Abstract : "This study aims to design an Artificial Neural Network (ANN) able to provide its user, in a numerical and experimental context, with a combination of heating parameters in an active thermography process. This combination allowing the user to reliably detect a known in size and depth delamination, which is one of the most important sources of damages in composite structures. This work features the realization of a thermographic database using a numerical model under the finite element software COMSOLVR. The ANN itself was developed using a parametric asset and then parametrized in the frame of this study. In order to train and validate the ANN some chosen heating parameters were used: the angle of the lamps in heat process, the distance between the test structure and the lamps, the power of the filament inside each lamp and the time for which the lamps will be turned on (heating time). The results highlight the feasibility of a combination of an ANN with thermographic finite element simulation to boost the accuracy of a thermogram. Moreover, it also points at the best ways to realize this tool, opening the field for further research."