Desenvolvimento, otimização e automatização de novos modelos de fontes de calor para simulação numérica de processos de soldagem a arco

Abstract

Joining processes by electric arc welding involve thermal, mechanical and metallurgical phenomena of relative complexity, requiring arduous mathematical modeling. The thermal aspects govern the mechanical and microstructural behavior of the welded part. The thermal analysis of welding is a three-dimensional non-linear transient problem. Its modeling is based on the concept of heat source to contemplate the entry of heat into the geometry of the welded joint. Several heat sources are available in the literature, however, to obtain the desired accuracy, these sources have geometric parameters to determine, which tends to demand a long computational time, even with the use of optimization methodologies. For several cases of welded joints, sometimes the results are not satisfactory. In this work, numerical analysis of single pass welding are presented, for four joint configurations (butt, overlap, corner and T), having as base materials AISI 1020 carbon steel and AISI 304 and AISI 316 stainless steel, in thicknesses of 3,00 and 6,00 mm, and using the GTAW and GMAW welding processes. The numerical thermal simulations were carried out using the Finite Element Method, via ANSYS Multiphysics software, considering three classic heat sources (Gaussian, Conical and Double Ellipsoid), and two new proposals of heat sources with variable volumetric profile called FCV and FFV. Heat transfer by convection and radiation to the environment, and material properties as temperature-dependent were considered. The fused zone shapes and thermal cycles of the joints, obtained from the experiments developed at the Research Laboratory in Welding Engineering (LAPROSOLDA - UFU), located in Uberlândia, Brazil, are compared with the results of the simulations. Such data, obtained by macrographys and thermocouples, were used in a new optimization methodology to find the optimal parameters of the heat sources presented, through inverse problem techniques and Genetic Algorithms, and a new technique called Reduced Geometry. A software (OWSO) and a numerical welding library (WeldLib) were developed, which requires minimal user interference, making the process partially automated. Accuracy analysis were carried out in relation to experiments, as well as the computational times obtained. A good agreement for thermal results was obtained, and the new heat sources developed showed superior results when compared with classic ones. Besides, the new optimization methodology resulted in decreases of around 90% in computational time.