Defect detection in pulsed thermography: a comparison of Kohonen and Perceptron neural networks

In this paper, two neural network approaches are compared for defect detection using thermal evolution, phase and amplitude data acquired in the pulsed thermography approach with pulsed phase thermography processing. The tested approaches are based on Perceptron and Kohonen neural networks. Examples of results are presented for each technique with the three types of available data, in the case of flat-bottom holes in aluminum. Results show that the Perceptron using phase data gives better results being less influenced by disturbances.