Experimentally validated defect depth estimation using artificial neural network in pulsed thermography

Infrared thermography (IRT) proved to be a valuable technique in non-destructive testing due to its contactless nature, real-time application, and wide area coverage. However, the major hurdle to widespread application of IRT is quantification; more specifically quantifying the depth of defects from the acquired thermograms in low conductive host materials. In this paper, an artificial neural network (NN) is employed to detect defects depth in composite samples, coupled with a Pulsed Thermography PT setup, to complement prior work when NN algorithms were coupled to a line-scan thermography. Carbon Fiber Reinforced Polymer (CFRP) coupons with embedded and flat bottom holes defects were designed via 3D printing; to precisely control the defects morphology and depth. Firstly; the current study presents a proof of concept using a Multiphysics FEM simulation model of the inspection process, to generate the training sets of data, so that the developed NN is assessed in a deterministic (noise free) environment. Then, the proposed NN was further tested experimentally to validate its accuracy and performance. The accuracy of the developed NN for the synthetic data was more than 97% and for the experimental data was around 90%.