Deep Autoencoder for Non-destructive Testing of Defects in Polymer Composites

Abstract

Infrared thermography (IRT) is an efficient non-destructive testing technique, which is widely applied in defect detection of polymer composites. However, the nonlinear nature of the thermographic data and the adverse effects of noise and inhomogeneous background prevent IRT from delivering satisfactory results. A novel deep autoencoder thermography (DAT) method is developed to enhance the contrast between defects and background. The multi-layer structure of the deep autoencoder is used to extract the features. Then, the results of the middle-hidden layer are visualized to show the effects of removing noise and uneven background. As a result, the defect is highlighted in the visualized images. The feasibility of the DAT method is verified using the experiment of carbon fiber reinforced polymer specimen.