Eventos Anais de eventos
COBEM 2023
27th International Congress of Mechanical Engineering
Performance Verification of the Volumetric Thermal Capacitor Method Applied in Moving Heat Source Autogenous Welding Simulation Run in GPU with Enhanced Parallelized Code
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.COBEM2023.COB2023-0467
Abstract
The previous simulation of manufacturing processes allows the prediction of possible failures, the estimation of the final mechanical characteristics, and production cost reduction. The Computational Fluid Dynamics (CFD) techniques revolutionized modern product design by also requiring a lower number of experiments to achieve equal or better-quality results. The simulation outcomes provide several advantages such as easy variable probing at any point of the domain, operational risk reduction, elimination of experimental space and required apparatus, and a much faster solution. The computational time needed to produce reasonable numerical solutions has greatly dropped since the beginning of the CFD era. However, there is still plenty of room for large computing enhancements through better code and hardware coupling. To fulfill this gap, an example of the applicability of a parallelized CUDA-C language code is presented in this work. The authors verified the performance of The Volumetric Thermal Capacitor (VTC) method applied in the Laser Beam Welding (LBW) process simulation. Here, the algebraic form of the transient heat conduction Partial Differential Equation (PDE) was discretized by the Finite Volume Method (FVM) and computed in the Graphics Processing Unit (GPU) for a faster solution. The melting process was accounted for through the application of the enthalpy method employed with the VTC method. Here, a fixed grid approach allows for the prediction of a sharp phase change front through the calculation of the liquid fraction parameter. The heat loss due to radiation and convection effects as well as temperature-dependent properties were accounted for to increase the solution accuracy. The Successive Over-relaxation (SOR) solver was applied to obtain the final matrix governing equation solution. The applied hardware consisted of an RTX 3090 graphics card with 24 GB of video memory with double precision floating point operations capability. The research findings that the VTC method is an efficient model to analyze phase-change processes when it is compared with the classical model. Furthermore, it suggested that GPU processing offers reasonable reliability but also the great advantage of a much faster computational solution when compared to traditional serial computing.
Keywords
Volumetric Thermal Capacitor Method, GPU processing, CUDA-C language, Finite volume method, autogenous welding
