Comment pouvez-vous optimiser vos simulations CFD pour une meilleure précision ?

Optimizing the accuracy of CFD simulations significantly depends on selecting the appropriate models for the specific flow phenomena being studied: Turbulence Models: Choosing the right turbulence model (like RANS for less computational demand or LES/DES for complex, transient phenomena) is key for accurately capturing flow details. Multiphase Models: For simulations involving multiple phases, selecting the correct approach (Eulerian, Lagrangian, or Volume of Fluid) is crucial to accurately model the interactions between phases. Reaction and Combustion Models: Selecting the appropriate combustion model based on the type of combustion (premixed, non-premixed, or partially premixed) to predict flame characteristics and emissions correctly.

Mesh refinement is a critical aspect of CFD simulation that directly impacts accuracy. I have always compared mesh to the resolution in photographs. Photos with large pixels are not clear and the resolution is low, it is difficult to understand what the photo taken is. As we reduce the size of pixels, we increase the resolution of the photo and get a more accurate and clear image. With mesh independency study, implementing local mesh refinement, optimizing mesh quality, choosing the appropriate mesh type, you can significantly improve the accuracy of your CFD simulations. This approach allows for a more precise representation of the physical phenomena being modeled, enhancing the reliability of the simulation results.

Mesh independency tests are a crucial step for getting a high quality mesh. Take a critical parameter such as temperature in specific regions and see how its calculated value changes with the mesh refinement. If we reach a point where further refinements will not change or affect the temperature (or any other monitored parameter), or the change is very small (matching a pre-established acceptance criteria), then we have found a mesh we can work with.

It is important to perform a comprehensive literature review before starting a CFD study. You can use the example closest to your problem as a basis. Then, depending on the similarity of the results, you can continue. The most accurate-most costly method to verify the results is experiment. A cheaper-complex method is to confirm the analysis analytically. What differentiates CFD from 'Colorful'FD is the measurability of the analyzed situation and the physical accuracy of the results. Check for the Newton's Laws (conservation of mass, conservation of energy, etc.). If the general condition looks correct both physically and 'visually', validation can be performed by experimental or analytical methods.

Sensitivity analysis is another step that should not be overlooked. Sensitivity analysis allows us to answer questions like "What would happen with my output Y if my input X changes 5% (for example)?" That kind of questions are sometimes more complicated or costly when answered with experiments. If we want to understand how the different shapes of an airfoil affect the laminar-turbulent transition region using a rig, for example, we would need to build the different airfoils, which in turn would imply an increase in time and cost (due to the manufacturing of the airfoils). That apart from the time required to set up and run the tests with the different airfoils.