Automated detection and characterization of defects in composite-metal structures by using active infrared thermography

Several composite-metal samples with artificial defects of varying size and depth were experimentally investigated to demonstrate effectiveness of using a line scan thermographic nondestructive testing in combination with a neural network in the automated procedure of defect detection and characterization. The proposed data processing algorithm allowed defect thermal characterization with a practically accepted accuracy up to 16% and 51% by defect depth and thickness respectively. Characterization results were presented as distributions of defect depth and thickness correspondingly called depthgram and thicknessgram. For training a neural network, it was suggested to prepare input data in the form of non-stationary temperature profiles processed by using the thermographic signal reconstruction method.