[Danial Q]

What is this all about (somebody translate it plz) ? So the bottoom line is " ICEM CFD". Am I right?
Anyways, could you mention the thread name where this happened (i mean the discussion) so that I could know, how this all started.
Well right now, I have tried sweep method with edge sizing which produced Hexa elements (cells what ever...) and I am politically positive that it is gonna be good this time.

Thanks

[Far]

Well, it was already traslated (not using Google translate).

Well the discussion was about the convergence problems. I just read that you are using the "hex dominant" method, so I recalled that Simon once said it is not good for CFD. So I put key words in search and found these(these are from first few search results).

Besides the choosing the appropriate numerical parameters, mesh is also major parameter in CFD (not always the main culprit). As far as the numerical parameters are concerned Glenn, already made the comprehensive comments.

Can you please show the latest mesh and some details i.e. quality and mesh size.

As far as ICEM is concerned, I believe you can get the very good quality grids in very less time

[Danial Q]

Well, new mesh is on stove yet so I would post the recipe as soon as i get it done (My system is liek a dead man right now) .Though I am not using it right now as I mentioned it in previous post. If i am right, PSYMN is simon and if he is, he suggested sweep method.

Thanks

[Danial Q]

I have attached an image regrading my mesh parameters. I hope it would be helpful for an advice for me. And for edge sizing , right now, i am using default settings except Beahvior >>> Hard, And Bias is ON, with factor 1. Type >> element with default size of element.

Thanks

[Danial Q]

Now, what is this (ERROR) about???Is it just memory thing because I have a lot of memory in TB's in my system then why this showed up?

+--------------------------------------------------------------------+
| ERROR #001100279 has occurred in subroutine ErrAction. |
| Message: |
| The total number of words needed for the Real stack exceeds the m- |
| aximum supported by the flow solver: 2,147,483,647. The solver w- |
| ill continue with the stack size clipped to that value. |
| |
| |
| |
+--------------------------------------------------------------------+

+--------------------------------------------------------------------+
| ERROR #001100279 has occurred in subroutine ErrAction. |
| Message: |
| *** Run-time memory allocation error *** |
| Not enough free memory is currently available on the system. |
| Could not allocate requested memory - exiting! |
| |
| |
| |
+--------------------------------------------------------------------+

[ghorrocks]

The error message says exactly what the problem is - you have tried to set up a simulation bigger than CFX can handle in a single process. The solution is to obviously go to multi-processor. Also make sure you are using 64 bit CFX, not a 32 bit version.

[Danial Q]

Do you think reducing meshing quality would help because at the moment I got like 8million elements. I don't have multiprocessor option available.
Thanks

[ghorrocks]

If you do not have parallel then your only option is to make a coarser mesh. Other options may exist (eg use single precision, not double) but from what I know of your case they do not sound like a good idea.

[p.galimutti]

TB really???!! Memory issues are related to RAM not hard drive.

where do you get your meshing parms from?

for your case my advice is don't make the mesh too fine (assuming you have low computational resources) or too coarse.
for complicated problems like these you need fine mesh at areas of steep gradients and coarser outside. try adaptive mesh

[Danial Q]

Yeah , you are right regarding RAM memory, it might be low but if it is about Hard drive then ofcourse not. I set them (parms) by myself as i wanted it a really fine mesh.
So, I have a mesh parameter like " (Hexahedral) mesh size >> 20 cells per 10 microns" is this good enough for this case and how to employ this parameter?? Infact i did not understand this cells per micron thing.
Any suggestions please.

Thanks

[ghorrocks]

As I said previously, the error message is not referring to RAM. It is saying CFX cannot handle the size simulation you gave it, regardless of RAM. You need to go either 64 bit or multi-processor.

There are no general rules for mesh size. You need to determine the mesh size required for your simulation with a mesh size sensitivity analysis. That will tell you exactly what mesh is required, and you can design your simulation around that.

[Danial Q]

That parameter with number of cells is what i have as a data, i was just asking that how to epmloy that as i did not see anything like cells' number in mesh module. And how to do this mesh sensitivity analysis??is it available in documents.

[p.galimutti]

Originally Posted by Danial Q

That parameter with number of cells is what i have as a data, i was just asking that how to epmloy that as i did not see anything like cells' number in mesh module. And how to do this mesh sensitivity analysis??is it available in documents.

Start from coarser mesh (use default meshing settings ) and monitor some variable you think is important, e.g. mass flow, velocity, pressure , temp.
Now increase the mesh size and see how the variables change from previous run. The mesh that shows less difference in variables that you monitor is your final mesh, after that however you refine your mesh you're likely gonna get same results.

BTW let me tell you something from what I understand, your computer see's the problem as DNS. I don't understand why you begin with superfine mesh and that too on a slow computer. You're just trying to jump to the last (maybe not last if your mesh sensitivity shows some what coarser mesh than you have now) step. Just try step by step!!

[ghorrocks]

Precisely. From what I know of your simulation you are going to have to do quite a bit of work to get it working, so why do that on a fine mesh which means everything takes hours? Do the debugging on a coarse mesh which runs in seconds, and once that is working refine the mesh to get accuracy.

[Danial Q]

Got it. I have completed a run and i am pretty sure that the mesh quality is fine as I have got values for all of three monitor points first time. May be after post processing i would have a better idea. But could you please have a look if it is OK so far.
Mesh Statistics |
+--------------------------------------------------------------------+
| Domain Name | Orthog. Angle | Exp. Factor | Aspect Ratio |
+----------------------+---------------+--------------+--------------+
| | Minimum [deg] | Maximum | Maximum |
+----------------------+---------------+--------------+--------------+
| splat | 90.0 OK | 2 OK | 53 OK |
| substrate | 38.7 ok | 6 ok | 37 OK |
| Global | 38.7 ok | 6 ok | 53 OK |
+----------------------+---------------+--------------+--------------+
| | %! %ok %OK | %! %ok %OK | %! %ok %OK |
+----------------------+---------------+--------------+--------------+
| splat | 0 0 100 | 0 0 100 | 0 0 100 |
| substrate | 0 3 97 | 0 <1 100 | 0 0 100 |
| Global | 0 <1 100 | 0 <1 100 | 0 0 100 |
+----------------------+---------------+--------------+--------------+

[ghorrocks]

Your mesh looks pretty terrible. Why don't you use an regular 1:1 aspect ratio hexas, and with only a few thousand nodes so it runs fast for debugging?

[Danial Q]

I think for solid domain (Bottom cube in pic above) very tight mesh would be unnecessary as it is not playing any significant role in system . But you think that mesh in fluid domain is also really terrible while mesh stats looks OK like;
Orthogonality : 90 (Recommended greater than 20 in documents)
Expansion Fcator: 2 (Recommended less tahn 20)
Aspect Ratio:53 (recommended less than 100)

Which Aspect Ratio ?? : First aspect ratio OR Last aspect ratio. The default meshing does not even give all monitor points' values.

[Danial Q]

I have some strange values in my output report as "Avg scale information", could you please mention the reasons for them. I cant understand that why volume fraction for liquidNi is zero while it should have value from (0 -> 1) and mass fraction is also weired.
Global Length = 1.1696E-04
Minimum Extent = 4.0000E-05
Maximum Extent = 2.0000E-04
liquidNi.Mass (Conservative) = 1.2954E-23
liquidNi.Mass (Normalised) = 1.2958E-23
liquidNi.Volume = 1.6197E-27
liquidNi.Volume Fraction = 1.0123E-15

Air at 25C.Mass (Conservative) = 1.8960E-12
Air at 25C.Mass (Normalised) = 1.8960E-12
Air at 25C.Volume = 1.6000E-12
Air at 25C.Volume Fraction = 1.0000E+00

Thanks

[ghorrocks]

You average scale information simply means that almost no Ni has entered the domain. It is almost entirely air.

If I was you I would start with a 1:1 aspect ratio mesh. Do not use refinement until you know exactly what refinement is suitable. And in your case I am pretty sure the answer is not much (if any) due to the numerical instability of your approach.

[Danial Q]

yeah i know that no Nickle is entering in domain from information above, but how is that possible? i mean if it is true then why? while I get values for temp in post processing, it shows that domain has temperature change from opening temp to some room temp. It makes me think that Ni should be present in domain to make this temp change happen.

Thanks