Differences in Javafoil and XFLR5 polars

[Don Moretti]

Hi,
I'm trying to design a prop with javaprop. Therefor I calculated the polars in javafoil and in xflr5.
when I imported those polars into javaprop I noticed that there were big differences between the same airfoil when it was calculated with javafoil and xflr5. I used a A18 foil at RE number 40000 and 80000.
Does someone know why there are so big differnces? I also tried a NACA 5512-43 foil at RE 300000 with the same big differences.
I dont know how to proceed now and dont know which polars to take for the prop calculation.
Any help is appreciated.
Thanks

[Don Moretti]

No one?

[quequen]

Don, I can't help you but I'd suggest you reading Hepperle's "JavaFoil Users Guide.pdf" and "XFLR5 v6.02 Guidelines.pdf" . XFLR5 is strongly based on Mark Drela's XFOIL, no one has published yet a comparison between them. Their implementations of panel methods could be slightly different.
I quote Hepperle:

All calculations are performed by a computer code of my own. JAVAFOIL is neither a rewrite of Eppler’ PROFIL nor of Drela’s XFOIL program. The boundary layer module is based on the method which is also found in the initial version of the Eppler program. Additions include new stall and transition models. The panel method was developed with the help of the survey of panel methods found in [13][13] Katz, Plotkin, “Low-Speed Aerodynamics”, McGraw-Hill, 1991..

Can you attach your .wpa file?

[Don Moretti]

Hi quequen,
thanks for your reply.
I did read both guidelines and I know that javafoil uses a different calculation method than xfoil.
But the differences are just enormous. So I wonder whom shall I believe and which data shall I take for the prop calculation??

I attached the .wpa file for the A18 airfoil I analyzed with XFLR5 v6.
regards

[Petros]

I am not current on these programs, but many years ago I worked doing aerodynamics. We always wind tunnel tested the shapes developed using CFD, and those programs were very expensive three dimensional CFD programs run on big high speed mainframe computers. I remember the words of one of our top aerodynamics engineers, he said that no one can accurately predict drag with any degree of reliability or accuracy. It was generally thought that if the CFD drag prediction was anywhere close to the actual flight testing, it was purely random luck.

Xfoil is a pretty simple 2D freeware program (it has been around for a long time) compared to what we were using than. The polar shape is dependent on accurate drag prediction. Each of these programs uses different methods and assumptions about how the drag is calculated, therefore I would be surprised if these polars are even similar, let alone match. The problem I seem to remember is how they treated the boundary layer, all of the methods use assumptions to calculate approximations, no direct method of calculating drag exists as far as I know. Each of the various methodologies only has a very narrow range where it was reasonably close to real life operations. We had always considered the CFD models only for prelinary design, and we often adjusted the CFD results when we had actual flight test data for similar shapes. Many times aircraft builders pay big time penalties for not meeting range/speed gaurantees they made based on preliminary design predictions done using CFD. the builders that met their range/speed guarantees always adjusted their preliminary drag data with actual flight test data of similar size and configuration aircraft.

That is why I suspect you get different answers. We just are not as capable at predicting what nature does as the current popular mythology about using computers to predict outcome, be it fluid mechanics, global warming, stock markets, election results or anything else.

[Don Moretti]

Hi Petros,
sure, there is no absolute precise method for calculating drag, but with some assumptions the existing methods and calculations get quiet close to the experimental results.
Take a look here:
http://www.mh-aerotools.de/airfoils/javafoil.htm
under the tab "Validation" you find a comparison of experiment, javafoil and XFLR5 for a NACA Foil.

I know I can't expect precise polars, but I'd think I'd get at least some close tendency to reality.
The problem I got is that javafoil and XFLR5 are not close together at all with their predictions, so I wonder whom to believe more...
Actually, how do you guys design props? How do you decide which foil to take?
Do all base their decisionmaking on windtunnel tests? I can't afford a windtunnel unfortunately...

[daiquiri]

Hi Don,
Did you try to check some older threads about Javafoil, where these issues were discussed? For example, this one: Javafoil
I have tried to compare the results from Javafoil and Xfoil relative to NACA 0012 section and have discovered that results from Javafoil are far from the known wind-tunnel data, while XFOIL predicts the characteristics of the said section pretty well when Ncrit=6 is used.
For me that's sufficient to not trust the output from Javafoil, and hence I'm using the Xfoil when I need to create and analyse a new airfoil. And even then a lots of caution and knowledge is necessary when interpreting the results.
Cheers

[Don Moretti]

Thanks Daiquiri,
thats some good info for me.
I have not found appropriate posts yet, so I'm happy about every good link!
I kept Ncrit=9 like suggested, but maybe I will play around a little with this...
My experience tells me the same that javafoil is not very accurate, almost every polar looks the same...

[daiquiri]

And please note that the choice of discretisation of the section is important for predicting both the Mach number effects and the point of boundary layer separation. As a general rule I tend to refine the grid in such way to have a lots of points where the airfoil curvature is high and in those where a positive (increasing) pressure gradient can be expected. And these areas are alpha-dependent (especially at high alphas), so you might need several iterations of grid refinement to get a decent result.
Cheers