Learning Forum

[charles delepelaire]

hi all, ti try to use fluent, so i've tried to simulate flow fluid around a 2mm ball inside a 1 liter fluid box (100x100x100 mm)

ball is the ridiculous sized point at the center of the french flag

 

so... from here we can saw : 

Cd of a smooth sphere around 0.47 at 5m/s and around 0.08 at 30m/s (see graph)

i've made a coefficient drag report file with it's plot with Y force Vector set to 1.

and the computation zone setted to my ball mesh

 

anyway i'm not able to find these values, and furthermore i got plethore amout of value dependant of intialisation caractéristic like hybrid/standard or pressure/density based... etc...

 

so what can i do to get repeatability, accuracy and "good" value with my simulation ?

[Nikhil Narale]

First thing, any reason for this? "plot with Y force Vector set to 1." 

Is the fluid flowing along the Y axis? 

[charles delepelaire]

yes fluid flow along Y axis

[Nikhil Narale]

How good is your mesh near the object (ball)? Can you post the screen grab of the cut section of the volume mesh?  

[charles delepelaire]

 

 

here the mesh near the ball :

here at the middle of the ball :

 

 

[Nikhil Narale]

 

This doesn’t look like a cut section. Did you add inflation layers? 

 

Edit: Just saw the second image. 

[charles delepelaire]

 

no inflation no (adding too much node inside student license)

 

[Nikhil Narale]

For getting accurate results, it is very critical to have a mesh with good resolution and quality, at least near to the object of your interest. I would suggest you to look for aero meshing tutorials (airfoil for that matter) on the internet. 

Also, keep a keen eye on the y+ value and make sure to maintain it as per the best practice. To know more, check this: 4.18.1. Overview (ansys.com)

If you are not able to access the link, please refer to this forum discussion: Using Help with links (ansys.com)

 

Nikhil

[charles delepelaire]

ok so do you suggest to make it in 2D and analysis one vector at a time ?

is needed to get a symetric mesh ?

[Rob]

2d has it's own problems as the wake separation may be asymmetrical and 2d-axisymmetric will mask that. Inflation shouldn't add too much to the cell count, and some careful decomposition should mean you can tet only the near sphere region and hex the rest. 

Also, check sphere projected area and then read up on wind tunnel blockage factors. The domain looks fairly big relative to the sphere so you may be able to save some cells that way. 

Finally, Student is intended to help you learn how to use the software. That includes realising you need a bigger mesh to get the right answer, and is why we also have Research licences. 

[charles delepelaire]

hello all,

is this new meshing seems better to you ?

 

 

 

i didn't used inflation but i did a sphere of influence sizing more finely to my ball, and applyed a refinementat the same point at all.

and did the best i can (in my opinion) to get better mesh quality graph

[Rob]

Use inflation and resolve the wake region. The graphs will show you the quality of the cells, what you're lacking is resolution of the boundary layer, separation and wake regions. Read the Body of Influence part here https://ansyshelp.ansys.com/account/Secured?returnurl=/Views/Secured/corp/v222/en/flu_ug/tgd_user_sf_types.html%23tgd_user_sf_boi  to show how I'd mesh this. In Ansys Meshing the functions are there, but the images are better in the Fluent Meshing documentation. 

The data you get from your model will not be accurate, but the general principle is shown. 

[charles delepelaire]

i did another one mesh :

i let my body of influence around my ball (radius ball * 1.2), kicked out the refinement method, and add a a pre First-layer-thickness inflation method with growth-rate set to 1.5 and try to keep all cells quality, 

what do you think about it ?

[Rob]

Better. You may find you need more mesh via a BOI in the wake and more inflation layers but give it a go. 

[charles delepelaire]

 

"You may find you need more mesh via a BOI in the wake"

what do you mean exactly ? 

make a better refinement/cell quality using BOI in the drag region like in this picture

[edit] how can i add quote to my post here ?

[Rob]

Yes, like the image. You may not need to go as far downstream though. 

No idea about the second bit of a quote. Two posts are fine. 

[charles delepelaire]

ok but the probleme is how can i make a sphere of influence shaped like an ellipsoïd ?

[Rob]

Create a block & chamfer or curve & rotate: you do this in SpaceClaim or DesignModeler.  As it's just a shape to guide the mesh refinement it's not required to be a perfect shape - a cuboid wedge will be fine. 

[charles delepelaire]

i maked this new one :

 

does it seems better for you ?

[charles delepelaire]

up

[charles delepelaire]

 

re-up

[EDIT] is there a reason when i up my message it's not upped ?

 

[Rob]

The system sort is new threads, or recent activity, you may have the wrong sorting option selected. As I also receive emails to tell me something has been added there's no need until you've not heard for a few days. Bumping after a day isn't appreciated, and will result in a miffed support engineer. 

Given how coarse your new mesh is, how big a jump in cell size over the interface and how close it is to the region of interest, then no, it's not better. Have a look at the airfoil/aerofoil tutorials (might be NACA) and see how they've been meshed. Then read up on body of influence as you've split the zone out. 

[charles delepelaire]

 

and this one ?

are mesh order feature important ?

is there a way to compare 2 meshing quality/skewness/aspect-ratio graph ?

[EDIT] thanks for information about updating post ^^

 

[Rob]

Much better. The order shouldn't matter as the tool ought to take the smallest cell from each: it's why it takes some time before building the mesh. 

[charles delepelaire]

i have this warning :

"Hard edges are not supported and may be ignored for pre-inflation, please check your model carefully."

is it important ?

[Rob]

Depends on where the hard edges are! A hard edge is an edge that doesn't form part of a surface perimeter: https://ansyshelp.ansys.com/account/Secured?returnurl=/Views/Secured/corp/v222/en/wb_dm/dm3dRepairHardEdges.html has an image showing what they are. 

[charles delepelaire]

when i go to warning-> show problematic geometrie it's seems it's edge of my ball and edge outside my enclosure.

 

so will i try to make my fluid flow simulation anyway ?

[Rob]

If the mesh checks OK, yes.  Warnings are something to check out and understand, they may not be important on any one occasion but MUST be checked. 

[charles delepelaire]

 

this is the mesh quality :

Mesh Quality: Minimum Orthogonal Quality = 2.81099e-01 cell 61822 on zone 2 (ID: 61823 on partition: 0) at location (-2.28725e-02, -9.77421e-02, 6.04073e-03) Maximum Aspect Ratio = 8.38496e+00 cell 85645 on zone 2 (ID: 85646 on partition: 0) at location ( 1.20744e-03, -8.94578e-04, 3.44995e-04)

 

i maked it polyhedra, that’s good ?

 

[Rob]

Cell quality is good, you'll know if the resolution is suitable once it's run. 

[charles delepelaire]

so the probleme persist, why can't i get planned result ?

[Rob]

What are you seeing? Have you checked the reference values for the coefficients? 

[charles delepelaire]

 

oops i forgot to edit my precedent post :

 

surface area : 3.14E-6 m² (section surface of my ball)
pressure : 101325 Pa
length :0.002 (diameter of my ball)
velocity : 5m/s
fluid velocity at inlet : 5m/s

 

density based and steady solver

all reamaining seems to be default

computation from my ball, 200 iterations (need i to add more until i get convergence ?) 

i tried laminar and k-omega analysis with different option...

 

 

[Rob]

If the flow is over about Mach 3 then density based may be needed, but mostly pressure based is fine. 

Convergence is recommended, otherwise the results tend to be a little wrong. That's why we have convergence residuals, and plot monitors. 

[charles delepelaire]

 

 

 

 

hahaha

hi everyone, after few day to take fluent in hand i think i have finally understood why my first results went wrong.

i made enclosure, ball and BOI volume from Fusion 360, and export them to step files, so maybe there is a problem beetween fusion and fluent.

here are the fusions files : grosfaignan/3d-files (github.com)

if anyone can test them and help me.

 

but now i have another problem, i have a good drag coefficient, but in a wrong direction :

since i used spaceclaim 3d model, inlet and outlet are aligned to X and not to Y, so why getting good results along the wrong axis ? (in screenshot above, P drag coefficient in Y is around 0.49 for a 5m/s velocity at inlet, good result, wrong axis)

[EDIT]hahaha no, after a second test at 30m/s when i need to get a 0.1drag coef it seems it’s a false positive

[EDIT2] another misunderstood :

if i have flow velocity setted from inlet to outlet, so in this case parrallel to X axis, why Y and Z coefficient drag calculation are different ? see beelow :

===================
dc-x
Cd p = -16.622681
===================
dc-y
Cd p = 3.9403545
===================
dc-z
Cd p = -1.9091926

[Rob]

Looking at the flow field the wake isn't aligned with the axis, that'll influence the coefficients, and mess with the convergence. 

Step tends to transfer well, so unless the scaling went awry that shouldn't cause a problem. 

[charles delepelaire]

how can you see that ?

and what do you mean exactly ?

how can i correct alignement ?

[Rob]

By looking: remember I've been doing this longer than you! Wakes, especially on spheres and cylinders tend not to be symmetric as you can get eddy shedding: look up von Karman vorticies.