DCT and DWT feature extraction and ANN classification based technique for non-destructive testing of materials

Abstract

Eddy current (EC) nondestructive testing (NDT) based on probe impedance and magnetic flux density in the defect regions is considered in this work. For this a numerical model is introduced for the development of the EC-NDT system using 3D finite element analysis. This model is used to simulate the material defects and prepares the data required for computer simulation of the EC-NDT system. The image data for three types of cracks in two types of materials are processed by discrete cosine (DCT) and discrete wavelet (DWT) transforms for feature extraction. Depending on the extracted features, the defects are classified using artificial neural network. A series of computer simulation experiments are carried out to assess the performance. 100% correct crack identification is achieved.