Computer System for Detection and Classification of Welding Defects

No one can deny the importance of pipelines in our lives today because they are used to transport flammable liquids such as petroleum from various sources over long distances. Before being used, the materials of these pipes and their welding joints must undergo extensive testing to ensure that there are no flaws or discontinuities. Even a small error could result in a major catastrophe. There are many nondestructive testing techniques that can be used to achieve this, including ultrasonic testing. A phased array is one of the ultrasonic testing techniques that can explore any discontinuity with a high degree of accuracy in a short amount of time while having no negative environmental effects. We tried to find a unique characterization for each type of welding discontinuity during its inspection by ultrasonic phased array device of Omniscan X3 type, and then created a computer application using expert system on MATLAB to differentiate between welding discontinuities types to identify and characterize them. Additionally, we employed a convolutional neural network (CNN) to separate photos of some types that have comparable values, like different sorts of porosities. The program demonstrated its effectiveness and reduced the dependency on human experts.