Weld Defect Detection using Iterative Image Reconstruction Methods

Non Destructive Evaluation (NDE) is a technique to detect the defects without affecting the intended application or the structure, which is mainly used for damage assessment and quality control. NDE is basically used for automation of defect quantification. This paper presents a method for weld defect detection using iterative method for image reconstruction. Image reconstruction is a technique of reconstructing the image of object from the radon transform of the object, where radon transform is obtained by passing an energy source through the object in various angles and collecting the projected data on a detector. Weld defects such as slag inclusion, lack of penetration, and burn through can cause serious disaster if it is undetected at the time of weld process. The system of linear equation under consideration in Computed Tomography (CT) are basically very large in size, so the direct method takes too much of computational time which makes it impractical for implementation. Cimmino's method and Simultaneous Algebraic Reconstruction Technique (SART) are iterative methodologies for image reconstruction which are best suited for solving large scale linear system where direct method fails due to very large time consumption. This paper compares the experimental results of image reconstruction of weld defect radiographic images using SART and Cimmino's method for various test images with varying number of projections.