Until the end of the 20th century, CFD was typically a tool used in an academic environment, or in large aerospace companies. Back then, CFD was a very expensive and time-consuming activity. But over the past 10 to 20 years, CFD methods have developed and computational power has grown exponentially. As a result, CFD becomes more reliable, cheaper and more accessible. Despite these trends, there are still some prejudices about CFD. In this article we discuss four of them.
The equations behind most CFD software are the Navier-Stokes equations. These equations are proven to be correct. There is no discussion about that. However, solving the Navier-Stokes equations analytically, is not possible. Therefore, it is necessary to simplify the equations and solve them in an iterative way. In order to simulate a gas or liquid flow in a specific environment, the user needs to determine which simplifications to apply. This depends on the situation at hand. Moreover, the user needs to choose the right turbulence model, the right boundary conditions and a suitable numerical method.
As such, the result of a CFD simulation largely depends on these choices, and here it can go wrong. Sometimes the user choses the wrong simplifications or models in order to get fast results. In other cases, the computer model might ask for unknown boundary conditions. In those cases, the user must make assumptions, resulting in extra uncertainties.
If the results of a CFD simulation do not match reality, then this is not due to bad CFD-software, but due to wrong inputs or assumptions of the user. The experience of the user and the possibility to validate CFD by comparison with experiments, determines the reliability of a CFD simulation.