Effect of lock-in thermography test parameters on classifying defects in CFRP by means of a convolutive neural network

Lock-in thermography is a well-established non-destructive technique for detecting defects in composite materials. The qualitative analysis of defects is a challenging task and usually is assessed by an expert operator after the application of suitable algorithms. In this regard, deep learning algorithms are very attractive since they allow to speed up and automatize the identification and characterization of defects. In light of this consideration, the aim of this work is to investigate the influence of lock-in thermography set-up parameters on the capability of a temporal convolutional neural network to characterize defects in a carbon fiber-reinforced polymer specimen. Moreover, to make the lock-in technique suitable for industrial applications, a comprehensive study of reducing both the experimental test time and the processing time has been carried out. The performance of the CNN has been evaluated as a function of some lock-in test parameters such as the number of acquired frames per cycles and the number of excitation cycles. The obtained results have been critically discussed through qualitative and quantitative analyses.