[HutchMan]

Hi all,

New to the forum and to CFD but my thesis is on the simulation of flow around an airfoil so thought this would be a handy resource.

I'm performing an initial inviscid flow sim on a NACA4412 before I move on to turbulent models. Basically I'm following this tutorial which I understand is widely used:

https://confluence.cornell.edu/displ...ver+an+Airfoil

I'm trying to compare my sim data with the Abbott and Von D data and whilst I get a lift curve slope of about 2Pi as expected for a inviscid sim, I can't get the zero lift angle to agree with the Abbott data of -4 degrees. At -4 my sim gives a Cl of about 0.035.

It's as if my correct lift curve has just been shifted along the alpha axis to the left a bit so that the zero lift angles don't agree. At first, myself and my supervisor thought my coordinates were slightly off orientation wise (i.e. not at zero AoA) but I've checked them with other coordinates and confirmed that the chord line runs parallel with the x axis when imported into GAMBIT, as it should.

I've even downloaded the mesh from that Cornell tutorial and run that in Fluent only to get similar results of Cl = 0.035 at -4.

My thinking is that there must be something about the Abbott data that slightly changes the effective AoA or something along those lines.

Or perhaps this is a mesh resolution problem? Just throwing the problem out there for someone to hopefully shed light on. It'd be much appreciated.

It's irritating as I'd like to move on to turbulent models knowing my inviscid sim is as accurate as can be.

[cfd seeker]

Mesh given in the cornell university tutorial is not fine enough to capture all flow features. May be mesh refinement will solve your issue

[HutchMan]

What would you suggest as a suitable mesh size?

[cfd seeker]

well it depends on the flow conditions you are trying to model and also the flow features of your interest

[HutchMan]

I simply want to get the most accurate Cl readings for an inviscid model just to validate the method. I won't really be looking at particular flow features until the turbulence modelling.

[cfd seeker]

Just double the mesh given in the cornell university turorial. I am quite sure you will get good results. BTW starters normally find difficulty in getting good results for the drag not the lift but don't know why you are facing problem with lift, are you sure about your case setup, boundary conditions, solver etc?

[HutchMan]

Yes pretty certain that the boundary conditions and solvers are all set up fine.

It's not that I'm getting bad results for Cl it's just at the zero lift angle I get around 0.034, and I expected it closer to zero. Should I be expecting to replicate the zero lift angle in an inviscid model?

Is it better to adapt the mesh in FLUENT or remesh in Gambit?

Thanks for all the advice so far.

[cfd seeker]

It's not that I'm getting bad results for Cl it's just at the zero lift angle I get around 0.034, and I expected it closer to zero.

ooohhh ok but don't you think that 0.034 is close enough to zero? I guess you have some confusions about the results of fluent, it will never give you a 100% correct result as compared to experimental results....small amount of error will always be there e.g if experimental data is showing zero lift angle of attack to -2.3 deg then it will only be the matter of luck to reproduce exactly this answer with fluent, instead you should try with -2.3+-0.3(lets say) to have zero lift angle of attack. Small amount of error will always be there in the results due to descritization errors, numerical errors, false diffusion etc

Is it better to adapt the mesh in FLUENT or remesh in Gambit?

It always better to go back and refine the mesh in the mesher(Gambit in your case)

[HutchMan]

Originally Posted by cfd seeker

ooohhh ok but don't you think that 0.034 is close enough to zero? I guess you have some confusions about the results of fluent, it will never give you a 100% correct result as compared to experimental results....small amount of error will always be there e.g if experimental data is showing zero lift angle of attack to -2.3 deg then it will only be the matter of luck to reproduce exactly this answer with fluent, instead you should try with -2.3+-0.3(lets say) to have zero lift angle of attack. Small amount of error will always be there in the results due to descritization errors, numerical errors, false diffusion etc

It always better to go back and refine the mesh in the mesher(Gambit in your case)

Ok well I have a couple of things to try now but I think I may just try some turbulent models without worrying about the inviscid at the moment.

Most importantly is that the lift curve slope seems right. And I guess 0.034 is pretty close enough for such a simple mesh. If I refine it and get better results I post it up.

Thanks for your help so far!