Abstract
Abstract A weldment is the joining of two elements; therefore, they function as a single unit. However, a “defective weldment” refers to a welded structure that is deficient and weak, resulting in failure at the joint. The aim of the study is to find the initial shielded metal arc welding parameters under which two different steels, AISI 4140 and AISI 1018, were welded from fractures obtained by Charpy impact test. The three initial welding parameters are the type of filler material (AWS E8018 and AWS E9018 electrodes), the application or not of heat treatment, and the number of welding passes (2 or 3). To achieve the above, circular specimens have been designed for testing in the impact machine, allowing for the capture of fracture images from which attributes are extracted through a deep learning model. One novel aspect is the proposed methodology for designing circular specimens with a welded joint in the center with a machined notch to induce fracture under impact loading, allowing the weld penetration to be measured with the naked eye. The other innovative aspect is the deep learning model designed from scratch, consisting of 6 feature extraction layers, a flattening layer, and 4 fully connected classification layers; the model provides a $$100\%$$ 100 % identification of the weld parameters, similar to how a fingerprint identifies its owner; it is also independent of subjective expert opinion, and it is efficient, effective, and low cost.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call