Density-Based Dynamically Self-Parameterized Clustering for Material Inspection

Abstract

Abstract The ultrasonic based nondestructive testing (NDT) is the common technique used to perform material testing using ultrasonic signals. These signals are difficult to interpret and the examiner has to focus on every sampling signal to observe the changes in characteristics of signals. The core target points of the proposed work are used to identify the size and position of the defects as fast as possible. For that, an unsupervised machine learning approach is proposed to analyze defects such as shrinkage, porosity, crack, discontinuity, lack of fusion, lack of penetration and overlap. This would be helpful in the domain of material science and knowledge mining to study the structural integrity of metals during the manufacturing process and can be applied in automobile industries to increase the quality of manufactured parts. The proposed work incorporates a novel Density-Based Dynamically Self-Parameterized Clustering for Material Inspection (DBDSPCMI) method to effectively predict and identify the defect size and its position. The proposed method proves to be effective with an accuracy of 97.04% in measuring defect size and 95% in identifying defect position.