Abstract
This paper presents a methodology to detect defects in materials using simulation. Wavelet transform and neural networks are used as feature extraction and classification tools, respectively. We first use the raw signal of the defect as an input to the neural networks. Then, the wavelet transform of the input defect signature is applied to the neural networks. The results of both methods are analyzed and their performance are compared and discussed. It is found that using Wavelet transform as a pre-clustering scheme before applying data to the neural networks can provide better classification results as compared to the case that does not use it. Our scheme is efficient and accurate.