An autonomous technique for weld defects detection and classification using multi-class support vector machine in X-radiography image

Abstract

Non-destructive tests are a major evaluation process in the metal, oil, and gas industries. In these industries, weld defect inspection is one of the important parts of testing. Manual inspection may complicate for proper justification and it produces false identification during detection of weld defects in the human perspective less situations. X-radiography testing is utilized in the weld defect inspection and now this method is mostly outdated further it is enhanced into analysis on X-radiography digital image. Therefore, an autonomous weld defect detection and classification is required for the error free inspection. This paper proposed an autonomous technique for weld defects detection and classification using multi-class support vector machine (MSVM) in X-radiography images where weld defects such as porosity, gas pores, tungsten inclusions, longitudinal cracks, lack of penetration, and slag inclusions are considered. Three modules are involved in this proposed method. In the first module, the images were smoothened using modified anisotropic diffusion method. The segmentation process was performed using improved Otsu’s method in the second module. Finally, the features of the region of interest are extracted and given as input to the multi-class support vector machine with the kernel Gaussian radial basis function. The results of proposed scheme were compared with artificial neural network, Bayes and MSVM with other kernel functions like multi-layer perceptron and polynomial. The implementation and experimental outcomes reveal that the proposed scheme can positively detect and classify the weld defects in X-radiography images.