[ConnClark]

Originally Posted by Sven7

I'll speak for myself because ConnClark surely has something to say about it too. (CC, what do you do for a living, if I may ask?)

Engineering. Mostly in electronics.

I have "calibrated" it based on ChazinMT's recommendations but noticed slight discrepancies between the data and my tuft testing. Not sure what to do about that, but once you get it locked in it should be pretty accurate for most tests, right?

The answer is no.

Here are a few quotes from the author of Flow Illustrator

"I (S.Chernyshenko) am often asked how to use Flow Illustrator for research. The short answer is that you cannot do this. "

"Flow Illustrator is optimised for speed and robustness, with accuracy being traded in. Treat the movies as artist impressions. The artist is you. This means that, however 'correctly' you select the parameters, the results will be inaccurate."

"Realistic conditions for flows past, for example, cars correspond to very high Re (like 10000 or more). However, if one ignores small-scale turbulent fluctuations always present at such high Re (which one has to do in Flow Illustrator since the characteristic length scale of those fluctuations is typically smaller than the pixel size) one should introduce so-called turbulent viscosity. In Flow Illustrator the effect of turbulent viscosity can be crudely modelled by decreasing Re. Re calculated using turbulent viscosity instead of laminar viscosity is called effective Re. Very crudely, for bluff bodies (like cars, or any body the flow past which is separated) effective Re, if based on the body largest cross-section dimension, is usually between 100 and 1000. Note that the laminar, numerical, and turbulent viscosities differ not only in magnitude but in many other respects, so that one can never be accurately modelled by the other. "


Flow Illustrator More Info page

I have no training outside Phil's lectures and other things on this site. I don't have the money for Hucho's book so it's mostly just studying what people say and studying tuft and smoke images (including independent "research" via tuft testing on the Probe). Still, when CC says "99.9%" of people don't know what to look for, I would wager that that other 0.1% is the members on this site, reading this thread, and using the software. If not us, who?

Curious to see what others say.

Let me quote the author of Flow Illustrator one more time. "So, if, for example, you want to have a realistic movie of the flow past a particular vehicle moving at 80Km/h, you should somehow imagine first what that movie should be like. For this, one probably has to take a few years university course in aerodynamics, then do a PhD in car aerodynamics, and perform a number of experiments with visualisations and so on - that is to become a specialist in the field. Once you know, what the movie should be like, you can play with Flow Illustrator's parameters until what it gives you looks realistic. "

Since it requires for a person to know what the actual output to look like to get something representative of reality this renders this software useless for experiments. So unless the members on this site, reading this thread, and using the software have a PhD in car aerodynamics and are a specialist in the field, every image produced by this software is what they want to see and not a real representation of what can be expected in the real world.