Application of Evolutionary Neural Network in Infrared Nondestructive Test

Evolutionary neural network (ENN) was used to quantify the defects in lock-in thermography nondestructive test. The embedded specimens containing the multi-defect information were tested, and time-dependent temperature information measured from pixels of an NIR camera´s focal-plane array detector imaging the surface specimens provided characteristic parameters for network training. The results show that the error from the network after training was less than 2% and can be referred to in engineering application.