Eventos Anais de eventos
ENCIT 2022
19th Brazilian Congress of Thermal Sciences and Engineering
PARALLELIZED NUMERICAL SIMULATION RUN IN GPU FOR THE ANALYSIS OF THE THERMAL EFFECTS IN THE APPLICATION OF DIFFERENT SHIELDING GASES IN AUTOGENOUS WELDING
Submission Author:
Arthur Mendonça de Azevedo , RJ
Co-Authors:
Arthur Mendonça de Azevedo, Ernandes José Gonçalves do Nascimento, Elisan dos Santos Magalhães
Presenter: Ernandes José Gonçalves do Nascimento
doi://10.26678/ABCM.ENCIT2022.CIT22-0145
Abstract
Modern-day engineering designs are very sensitive to development and manufacturing costs efficiency. The competitive globalized market prioritizes fast and affordable technological advancements to suit commercial tendencies and avoid obsolescence. The first numerical methods revolution had greatly reduced design costs by requiring fewer experiments to build equivalent devices. However, after nearly fifty years, the application of traditional CPU (Central Processing Unit) based numerical methodology still requires expensive equipment such as processing cluster type computers, room climatization, and higher electricity capacity no-brake systems. The requirement of this high-priced set of machinery gives room for new alternative solutions with higher hardware appliance efficiency. The present work is an example of the application of recent advances in GPU (Graphics Processing Unit) parallel computing, resulting in a faster, less costly, and more reliable numerical methodology. Here, the latest developments and features of an in-house parallelized CUDA-C (Compute Unified Device Architecture) language code were exposed. The software was applied in the analysis of the thermal effects caused by the application of different shielding gases in a simulated autogenous welding process. The heat conduction partial differential equation (PDE) with an added phase-change term is transformed into an algebraic equation and discretized in a three-dimensional domain by the application of the Finite Volume Method (FVM). Additionally, the enthalpy method is applied to account for phase changes in the material by evaluating the drop in temperature due to the latent heat of fusion. Temperature-dependent thermal properties are applied to the welded specimen material. The resultant temperature and material liquid fraction fields were discussed. The simulation outputs were also exposed as a function of shielding gas type. The present research outcomes had suggested that the analyzed methodology has enough potential for replacing traditional CPU (Central Processing Unit) processing techniques, thus greatly reducing costs by minimizing time expenditure and required laboratory apparatus.
Keywords
autogenous welding, Numerical phase change, Finite volume method, GPU processing, CUDA-C language
