Feature Extraction and Classification of Welding Defects using Artificial Neural Network

Abstract

The detection and classification of the defects in the welded components are very important in order to ensure the structural integrity of the fabricated components of the test blanket module (TBM). RAFM steel is used as structural material for the TBM, therefore ultrasonic based technique are the most suitable for high sensitive defect detection. In this work ultrasonic pulse echo technique is used to perform the experiment and the ANNs (artificial neural networks) technique is used to detection and classification of the defects in the welded region. For this study, artificial defect (Side drilled hole, notch and flat bottom hole) are fabricated in the welded region. In this paper this data acquisition from different type of defects and extraction of feature from these signal are discussed. Artificial neural network will be used for the classification of the defects.