Comparative Evaluation of Emerging 3D-CFD Tools

The use of High Performance Computing (HPC) in three-dimensional Computational Fluid Dynamics (3D-CFD) was identified as one of the main source of the evolution of the sector, and it was identified that, in the current CFD market, both open source software and license-based commercial software are developing advanced HPC integrations, such as parallel computation and cloud computing. Open-source and license-based software both have advantages and disadvantages related to software costs, different user experience and reliability. The objective of this thesis work is to reproduce some case studies using both types of tools, in order to compare them to a benchmark obtained in a well-established commercial CFD software. Considering these factors, two software were identified to be of interest for an evaluation, due to their qualities in terms of workflow optimizations and promised performances: OpenFOAM and Simerics-MP+. OpenFOAM is the most popular open-source CFD tool in the market, while Simerics-MP+ is a growing license-based software. The first case considered was a Flow in a Pipe, an analysis of the flow of an incompressible fluid through a multi-outlet pipe in steady state, performed in both Simerics-MP+ and OpenFOAM. In Simerics-MP+, a sensitivity analysis was performed, and it was observed that it was able to replicate all expected flow behavior observed in the commercial CFD tool, as well as obtain extremely promising results in terms of time consumption, requiring 81.4% less time than the benchmark in the most basic case. On the other hand, OpenFOAM adoption encountered several difficulties with layer generation when using internal utilities. Despite providing good results in terms of computational times, the solver does not accurately replicate the benchmark case, especially underestimating flow separation, meaning additional steps in mesh generation and solver setup has to be made, unlike Simerics-MP+, which has better performance with comparable mesh quality. In the second case, Cabin Comfort, the internal air environment within the cabin of a car was analyzed by simulating the airflow patterns generated by the HVAC system. This simulation aimed to evaluate how effectively the incoming air distributes throughout the cabin, contributing to passenger comfort and thermal regulation. Several mesh configurations and physical modelizations were tested in Simerics-MP+ and compared to a benchmark case. Simerics-MP+ successfully replicated the internal flow dynamics within the volume. However, the analysis revealed that excessive mesh refinement near surfaces is not advisable, as it may lead to unnecessary computational overhead without improving accuracy. Consequently, further investigation is necessary to precisely characterize the interaction between wall functions and mesh quality, in order to identify an optimal balance between accuracy and computational efficiency. Simerics-MP+ provides exceptional results in terms of performance, with similar scalability to the benchmark software, but solving 39% less time. This result is proof of the validity of the software that can be established as a strong option for 3D-CFD simulations, especially for time-consuming applications.