Almost all work using symmetry (half model) mesh 8 (auto), 600,000-1.45M fluid cells, many runs using refinement lvl 1, periodic start 100, every 50, relax for 40 after achieving goals. (most runs, AMD 64 6400+ with 4GB 800MHz DDRII, about 1h 45m - 2hrs)
I would think the program would do a decent job on this problem depending on R'number. You need the R'numbers to be less than say 6 million. The turbulence model apparantly heads south above this number - possibly a bit sooner - you should chek against a cylinder or other.
Also you will need to have higher mesh density at the leading and trainling edges to get adequate resolution of the curvature.
The program does a decent job of settinghte external flow boundaries if you specify the free stream velocity in the wizard during set up. The lower the velocity (mach number) the large the domain needs to be. However you probably need a moving floor if it is a ground effect vehicle - so that domain boundary would be a solid with a surface velocity equal to and aligned with the free stream velocity.
Obviously at higher alpha you will need a much finer mesh at locations of flow separation. I would guess that at large alpha the drag estimate errors might tend to grow.
I would run tests on a known case with similar physics (or as close as you can get) to get the method sorted out.
Thanks to Bill for his response!
Though the 'concept' has a length and span that would generally be considered tiny, with a 5m root chord, a 6m span and a 28m2 lifting area! (Not so far from being an ugly rectangular planform! ) This gives a mean chord of 28/6=4.666m and therefore mean Re=6M@68km/h (take off speed for less than half payload!) and Re=12M@136km/h (max speed)
From several postings I'm beyond the Re capabilities of the program!
With 2D external, including moving ground (@alt=5% root chord or 0.3m direction double checked using vector plot! ) I have tried MANY runs with 2 or more refinements: (mostly 300-2000 iterations but have tried up to 17000!)
Can always get convergence (all green goals)
Have gone to MANUAL "criterion" (tightened to approximately 1% of objective MTOW lift and 1% objective drag (MTOW/75)
(using Mesh 8, auto, manual minimum gap set to 0.3m)
Using 'Solution Adaptive' refinement, lvl=3, (periodic, table 2, commencing after 2 travels then every travel)
Screen shot with modest angles of attack (4.5-7 degrees) at 80km/h attached!
Leading/Trailing edge mesh density down to about 2mm using "lvl=3"
EVEN with all this, , the difference between a 4 refinement and a 5 can be MORE than 20%!!
Is it just me? Did anyone see those pictures?
I checked them after I posted and again just now, the images are returning into China from where they came!
Can send more with up to +-17.3% for same foil, same alpha, same speed/Re same mesh settings (8/lvl=3 min gap 0.3) just a difference in the number of iterations, refinements or portion/width of the extrusion within the boundaries!
Pictures seem to be there now. So Karl what's the issue again? In my experience the lift estimates should be decent once you have eliminated mesh convergence as an issue. The issues of there being significant differences between mesh level settings does not seem in any way troubling to me. Some problems you need a ton a cells at it to resolve descretization effects. The drag estimates are going to be affected by the divergence of the turbulence model so L/D estimates will be similarly affected. L/D estimates are going to be a bit tough regardless even in the R'number range and will require very high mesh densities. I have run 2D simulations of some modified (laminarized) NACA 0012 sections and at moderate alpha the program did match what was in Abbot & Doenhoff. However, the "convergence" trace did not exactly level off to some nice asymptote - it was all over the map and I just flet lucky that it happen to stop right at an L/D of 32 as I recall, which was pretty close. The other thing you need to realize, probably do, is that the program is running the time average NS approach so even a steady state analysis is a really a transient till the flow settles down after the inital transient from the implusive sart or whatever the initial condition is. It will never really settle down to no variations - the wake reagion will always have fluctuations no matter how long the solution runs for - same in a wind tunnel test. And since the tangential forces (drag) is a manifiestation of the wake/B'layer well you are going to have some noise present in the results. Pretty sure none of that added much....am I missing something on what you are trying to sort out?
I was the Project Cordinator for the runner up in the 1987 World Solar Challenge in 1987. (Ford)
You may be aware of the strong following that continuues to exist for the 'Nurflugel' (wing only) designs that go back to, at least, 1930 (Horten Bros) and starting perhaps a little later, Northrop (Grumman)!
I have a concept for a surface effect craft, I hope to productionize, (China? ;( ) ALREADY a 'contradiction' there as (at least) the Russians were adament that a large rear wing is necessary to stabilize SE craft!
A Nurflugel has no rear wing!
1) Controlling (minimizing) C of M variation, during differing load cases SOLVED
2) During a lengthy (CFD) development (due to funding/land/building/language) Flow (3D) started telling me I had THRUST (yes NO DRAG! As mentioned in first post, perpetual motion?? )
3)First belief is Flow is in err!
4)I read what I can on 'tip induced thrust' and start to rationalize that; a)a long wing with a thin tip (winglet) GAINS some THRUST to offset the drag in the non-tip regions b) if this tip induced THRUST is being MAXIMIZED/HARNESSED by the low aspect ratio and the central 'DRAG' portion is SHORT maybe these MINIMIZE overal total drag (BUT NET THRUST??)
5)Yaw stability: believed to be NO PROBLEM due to ultra-wide landing areas! (sea/lakes/widish rivers that the wind tends to allign with)
6)What is an ACHIEVABLE l:d goal: with 37:1 claimed by the World's most efficient turbine powered aircraft ( http://en.wikipedia.org/wiki/Virgin_Atlantic_GlobalFlyer ) that has TWO and a HALF hungry fuselages??? (spread the high fuel load across the wing!) 72:1 claimed by (http://en.wikipedia.org/wiki/Eta_Aircraft_eta) powered glider?? The Nurflugel 'Brigade' believe/swear by the idea that 'integration/unitization' reduces wetted area, skin friction and therfore losses. The SE/WIG guys make some interesting claims! I decided with ALL surfaces 'clean' and lifting, IN SE/GE maybe (DESPITE very low 1.28:1 aspect ratio!) >75:1 is achievable! (chalk the rest/THRUST up to error!)
Thats what I went back to 2D to look at! Crawling before I can run again!
But what I see in 2D is +-17.3% as mentioned earlier!
Can I email you or skype you? innov8ive at excite com.
Feel free to email me firstname.lastname@example.org.
Well, net thrust is not on. Either the code is outside its bounds or is inadequate for the job. Net thrust is indicative of a low prediction on the drag forces. The thrust component is there due to the low pressure suctions that exist on the leading edges. However the pressures and tangential force computation should be larger than the the low pressure suctions at the front.
In 2D I run at level 8 or use the manual meshing to get the density in th right places. I would not use the refinement option on a highly dense mesh in 2D. I would guess that you would need in the order of multiple million cells on maybe a symmetry model, depending on the size of flow separated region in a 3D model. Given the slow speed the domain boundaries need to be quite a distance away from the object. High density betweenthe ground and the body is probably also required. The prgram works pretty well but you need to sort out your method. If you are a US citizen you could get a copy of PMARC and compare the results. PMARC stands for Panel Methods Ames Research Center. It is essentially a public version of a once market dominant low speed aero code called VSAERO by Analytical Methods, Inc. (Frank Devorak & Brian Maskew). I think they got paid to write it under an SBIR back in say 1990, might have been earlier and it probably isn't the whole story either. I am sure if you google it something should come up. Might be a better choice and it should be pretty cheap if you are a US citizen.
Oh and by the way, Ihave had net thrust predictions using VSAERO as well but it is inherently a potential flow code with a transpiring wall b'layer model and I can't remember what we did about it. It was a long time a go.
Bill thanks for the email address, did spend half an hour at your site last night!
Will email you directly when things get touchy on 'Public Disclosure' wrt IP!
Will try to keep it here if our communications may help others!
Have started a new (but related) thread!