Learning Forum

[Christian Neuer]

Hi everybody, 

I have a strange problem. When I start up my simulation with one core it runs just fine (but takes forever), when I start it with more then one core then it throws an error. But if I start the same simulation an uncheck the "Dynamic Mesh"  it also starts running (I manually terminated that simulation though bc. it would make no sense since I need the dynamic mesh).
So it seems the dynamic mesh has problems with multicore. Is there a solution?

Thank you,
Christian

Console output:

Expansion in mm = 0
Velocity in m/s = 0
Expansion in mm = 0
Velocity in m/s = 0
Expansion in mm = 0
Velocity in m/s = 0

Global Courant Number [Variable Time Step Criteria] : 0.01

Updating solution at time level N...

Tmax in K = 306.15
Expansion in mm = 0
Velocity in m/s = 0
Tmax in K = 306.15
Expansion in mm = 0
Velocity in m/s = 0
Tmax in K = 306.15
Expansion in mm = 0
Velocity in m/s = 0 done.
physical-dt 5.0000e+00

Updating mesh at time level N...

Expansion in mm = 0
Velocity in m/s = 0
Expansion in mm = 0
Velocity in m/s = 0
Expansion in mm = 0
Velocity in m/s = 0
Expansion in mm = 0
Velocity in m/s = 0
Expansion in mm = 0
Velocity in m/s = 0
Expansion in mm = 0
Velocity in m/s = 0
Expansion in mm = 0
Velocity in m/s = 0
Expansion in mm = 0
Velocity in m/s = 0
Expansion in mm = 0
Velocity in m/s = 0
Expansion in mm = 0
Velocity in m/s = 0
Expansion in mm = 0
Velocity in m/s = 0
Expansion in mm = 0
Velocity in m/s = 0
Expansion in mm = 0
Velocity in m/s = 0
Expansion in mm = 0
Velocity in m/s = 0
Expansion in mm = 0
Velocity in m/s = 0
Expansion in mm = 0
Velocity in m/s = 0
Expansion in mm = 0
Velocity in m/s = 0
Expansion in mm = 0
Velocity in m/s = 0
Expansion in mm = 0
Velocity in m/s = 0
Expansion in mm = 0
Velocity in m/s = 0
Expansion in mm = 0
Velocity in m/s = 0
Expansion in mm = 0
Velocity in m/s = 0
Expansion in mm = 0
Velocity in m/s = 0
Expansion in mm = 0
Velocity in m/s = 0
Expansion in mm = 0
Velocity in m/s = 0
Expansion in mm = 0
Velocity in m/s = 0
Expansion in mm = 0
Velocity in m/s = 0

Error at Node 3: Interface node number mismatch! (13072, 13032)
Error at Node 5: Interface node number mismatch! (13648, 13664)
Error at Node 2: Interface node number mismatch! (13032, 13072)
Error at Node 1: Interface node number mismatch! (15040, 15136)
Warning: multiple subdivision required on zone 62576. Subdivide using Constant Ratio.
Warning: multiple subdivision required on zone 62558. Subdivide using Constant Ratio.
Error at Node 0: Interface node number mismatch! (14496, 14504)
Error at Node 6: Interface node number mismatch! (14504, 14496)
Error at Node 9: Interface node number mismatch! (11880, 11944)
Error at Node 8: Interface node number mismatch! (9472, 9384)
Error at Node 4: Interface node number mismatch! (17296, 17336)
Error at Node 7: Interface node number mismatch! (8384, 8336)

Error: error: wta[1](string)

Error Object: ()

[Christian Neuer]

 

Update/additional information:

via try and error I learned that I can restart a saved simulation with more than one core during timesteps where the cell count is constant (=dynamic mesh isn’t creating new cells) but as soon as the cell count starts changing again the simulation terminates with the same error even if it has run for a few timesteps with more than one core!

Thank you,
Christian

 

[Rob]

Can you check both the dynamic mesh set up and starting mesh boundary conditions. Some images may help point you towards possible problems. 

[Christian Neuer]

 

 

 

Sure, since I didn’t know how to attach files I made a pastebin for the UDF https://pastebin.com/4gbPyeRZ (in this version of the UDF I reduced the velocity again because I got the “Update-Dynamic-Mesh failed. Negative cell volume detected” error too often and I read the reason for this is when the mesh moves more than 50% of the size of the adjacent cell per time step)

Now a few screenshots, I hope everything you need is there:





The first 3 pics are from the settings, then 2 pics from the mesh before and after the expansion and behind the link are 2 pics from the real world experiment. What is not clearly visible is that there are 4 bodies (or I think zones is more correct here), not 3. The real world fire-resistant glass consists of a 2.6mm glass pane 1.4mm intumescent interlayer and another 2.6mm glass pane (total 6.6mm). Afterwards the total height is a little more than 20mm so the intumescent expanded about 14mm.

Chemical composure and material properties are unkown (to us). At the moment we split the 1.4mm layer in 2 parts, one is 1.3mm thick with the material properties we experimentaly determined at room temperature and the second is 0.1mm thick and this is the one that actually expands with the dynamic mesh. (I don’t want to go too much into detail here but there are ofc. several mass source UDFs for the mass that emerges/disappears in one or the other region and for the H2O-vapor that gets created and disappears through the outlets.)

I hope I provided enough information, thank you,
Christian

 

 

 

[Rob]

I'm not permitted to download any files. Looking at the above, check the layering part to ensure there are no zones to the sides, and if there are that they're not connected. I think the non-deforming parts should be moving, probably with rigid body. Won't you just need a mass sink on the reacting side, and then mesh motion is linked to that? Ie there is no mass source into the solid. 

[Christian Neuer]

 

So you are probably right that it would work with only one mass sink (and do the rest with temperature dependant material properties), but I have to admit I didn’t build the whole model from the ground up. In fact a colleague of mine set it up in the way that there are 2 zones, one that is filled (at the start) with the intumescent in gel form (its mass is updated minus the water-vapor that leaves throgh the outlets and the mass that is generated in the other zone) and the other zone that gets filled with the foam. Even then I have highly temperature dependent values in both zones but a few things are constant, like the gel is always transperant and the foam never. However, no matter how this is realized it shouldn’t have an influence whether I can use multicore or not.

Regarding the other files, they should appear when you click on the links ….. if you aren’t permitted to click on links I wonder why we are allowed to post them. Anyway, here is the UDF:

1. #include "udf.h"
2. #include "dynamesh_tools.h"
3.  
4. static real velz = 0.0;
5. real tmax = 0.;
6. real value =0.;
7.  
8.  
9. DEFINE_CG_MOTION(wall_move_temp_z,dt,vel,omega,time,dtime)
10. {
11.     Domain *d;
12.     real temp;
13.     int ID = 62532;
14.     Thread *t;
15.     cell_t c;
16.  
17.     d = Get_Domain(2); /* Get the domain using ANSYS Fluent utility */
18.     t = Lookup_Thread(d, ID);
19.     /* Loop over all cell threads in the domain */
20.  
21.     begin_c_loop(c,t)
22.     {
23.         temp = C_T(c,t); /* get cell temperature */
24.         if (temp > tmax || tmax == 0.)
25.         {
26.         tmax = temp;
27.         }
28.     }
29.     end_c_loop(c,t)
30.     # if RP_NODE /* Perform node synchronized actions here. Does nothing in Serial */
31.     tmax = PRF_GRHIGH1(tmax);
32.     # endif /* RP_NODE */
33.     Message0("\n  Tmax in K = %g",tmax);
34.  
35.    
36.    
37.    
38.    
39.    
40.     /* reset velocities */
41.     NV_S(vel, =, 0.0);
42.     NV_S(omega, =, 0.0);
43.    
44.  
45.     if(tmax >= 393.15 && value <= 14.0) //starts at 120°C, stops at 14mm
46.     {
47.         velz= 0.00002;  //,meter per second -> 1mm in 50s
48.         value=value+velz*dtime*1000.0;
49.     }
50.     else
51.     {
52.         velz = 0.0;
53.     }
54.     Message(" \n  Expansion in mm = %g", value);
55.     Message(" \n  Velocity in m/s = %g", velz);
56.     vel[2]=velz;
57. }

[Rob]

The UDF looks OK at a glance, check against the example as I'm not overly familiar with them and we don't debug. I suspect it's a problem in how the mesh motion is set: check the motion/remeshing and compare with what your colleague did. 

Staff are not permitted to download files, but other users are: we're not the only people who help on here. 

[Christian Neuer]

OK, I will continue to try and play with the settings. Ah and in case it wasn't clear, my colleague had exactly the same problem and gave the case to me hoping I'd figure it out :-) 

As soon as I find a solution I will post again!

[Rob]

Right, in which case work through the whole set up very carefully! See if a fixed motion (via the UDF) works too. For troubleshooting it's generally a case of switching everything off and adding stuff until the model breaks: then figure out why!