I quite like Anders' approach, but there is also something else: the fluid dynamicist has a virtually continuous range of analysis options available to him, ranging from sticking a wet finger in the air and guessing right through to a fully qualified, mesh and turbulence model independent, fully converged, time accurate CFD solution, or a multi-million $ wind tunnel test series. Between these extremes are classical theoretical methods as well as "quick and dirty" CFD solutions. ALL of these methods have their appropriate applications. And yes, q&d CFD is a lot better than a WAG (wild assed guess) or even a SWAG (scientific WAG). But I still think you need to know what you are doing ....

In response to CFD monkey man's statement "What goes on inside the Black Box is really of no concern to the user". This is true if the only concern is the interpretation or validation of the CFD results. Here you need the so-called degree in aerodynamics or Mechanical Engineering or Chemical Engineering in the case of multi-phase flows and reactions!

On the other hand, to DRIVE the CFD car safely towards meaningful solutions, an understanding of the inner works of the black box comes in very handy. At one of our local University we have constructed a M.Eng course that deals with a wide spectrum of Numerical issues applied to the finite volume method. Amongst other, issues addressed are grid non-orthogonallity, higher order discretisation, pressure solutions on co-located grids etc. Short pre-defined tutorials on a commercial code outlining the typical numerical induced errors and remedies to them is part of the course.

Feedback from industry and the students revealed that the students attending this specific course were able to set up a CFD problem and reaching meaningful solutions much faster. The other outcome was that students (say from industry) who were exposed to CFD in their daily activities, managed to understand the numerical issues of a quite difficult course much easier.

You should have experience non convergence of a CFD solution due to a badly placed non-orthogonal cell to appreciate the beauty of non-orthogonal discretisation as implemented on general unstructured grids. AND/OR You should know the detail of the non-orthogonal implementation to realise that this specific non convergence is due to that badly placed non-orthogonal cell !!!!!

To be a CFD EXPERT, make sure you have as much knowledge of both the inner works of that CFD black box as well as the fundamentals of the aerodynamics that you try to solve. In reality, an Expert CFD group need employees from both groups to reach ultimate CFD solutions.

May your CFD solutions converge!

Correct. Too quick and too dirty, and you may even mislead the design process in the wrong direction.