Development of a computer vision image recognition tool for underwater wet welding

• Ano de defesa: 2022
• Autor(a) principal: Jorge Luis Ortiz Solano
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Universidade Federal de Minas Gerais; Brasil; ENG - DEPARTAMENTO DE ENGENHARIA MECÂNICA; Programa de Pós-Graduação em Engenharia Mecanica; UFMG
• Palavras-chave em Português: SMAW; UWW; Visual sensing in welding; Bubble phenomenon; Engenharia mecânica; Soldagem subaquática; Arco de soldagem submersa
• Link de acesso: http://hdl.handle.net/1843/49904

The focus of this work is to analyze the dynamics of the bubbles resulting from the welding process, on the effect of different formulations of coated electrodes, in both polarities and varying the welding current. The visual analyses of the bubble formation have been accepted as a fundamental tool in studying the welding process due to its correlation effect with the process stability and so, the final weld bead. Also, the shielded metal arc welding process is still the dominant technique used in marine and submarine engineering due to its reasonable cost of implementation and suitable performance. The proposed work includes the study of three different stick electrodes specifically designed for Underwater wet welding, including, two electrodes developed in the Robotics, Welding, and Simulation Laboratory, and a commercial one specific for underwater welding. The differences under direct and inverse polarity are highlighted by bubble evolution and arc stability analysis. This work is developed through the use of a quality imaging system, a power supply, a data acquisition system, and a detection and tracking algorithm. The image processing algorithm identifies the bubbles generated in the shielded metal arc underwater welding experiments. Also, the algorithm facilitates the calculation of the projected area of the bubble and its equivalent diameter. Then, it is able to track the bubbles for monitoring purposes as well as to calculate their frequency. As the imaging system and the signal are synchronized, it is possible to determine the relationship between the bubbles and the electrical characteristics of the arc. It was identified that the welding voltage presents a behavior similar to a sine wave and depending on the slope of the curve, its signal will have two specific directions, being able to be a positive slope during the growth of the bubble or a negative slope during its separation from the welding arc area. In the experiments carried out it was found that in the moments in which the voltage decreases to values lower to 5 Volts, it does not explicitly mean that there is a short circuit, it is simply an attenuation of the arc. A possible inversely proportional relationship between the mean voltage of the experiment and the frequency (in bubbles per second (b/s)) was also identified. And, a proportional relationship between this voltage and the calculated equivalent diameter. Also, there is a possible relationship is presented between the coefficient of variation of the current and that of the diameter of the bubbles, however, this could only be valid for the consumables and welding parameters presented in this study. The observation of the experiment videos suggests that, unlike the other experiments, during the formation of the bubble, the basic electrode, agglomerated with a polymer binder, adopts a flakes-like appearance and it expands and contracts in volume oscillating before breaking into several smaller bubbles. This bubble phenomenon was also observed even without the presence of short circuits or voltage drops. Also, the increase in the diameter of the bubbles of this electrode could be related to the presence of polymer particles in the additional material. With regard to the phenomenon of the bubble formation in the rutile flux coated electrode, a smoother phenomenon of formation and detachment of the bubble was perceived, which could be related to its greater stability in the electrical parameters of the process.