README.md 4.17 KB
Newer Older
ylan's avatar
ylan committed
1 2 3 4 5 6 7 8 9 10 11
## 352,625 (non-touching) pebbles in an annulus cylinder (Last updated: Feb. 15, 2021)
ann352625 N, Version 3.3, E98,782,067

- New features (after Jan, 2021):
  - mesh smoother+optimizer in Nek5000
  - tol adjustment in edge collapse
  - new scale (dsep = 2.0 = k(2.5), Rs0=0.8388, Rs1=1.0)
  - new boundary configureation (extrusion instead of projection)
  - Fix edge insert, edge collapse, Eadj, and Qadj

- Notes
12 13 14 15 16
  - DEM data from Yiqi (StarCCM++).
  - smaller edge collapse tol depends few local locations
  - re2 is NOT using a regular Nek format. See userdat2 in usr for proper boundary BC setups.
  - usr doesn't need dependencies (`my*.f`) if userchk is empty.
  - Geometries, pebble locations are stored in `ann350k.dat`
ylan's avatar
ylan committed
17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32
  - Sphere Centers saved into `*txt` files    
    `pebble_centers_re2.txt`, sphere centers matched in re2   
    `pebble_centers_rad_1.txt`, after rescaled by `1.0/0.91945`   
  - To check what scale the txt files has, use the `chk_txt.py`    
    For example, `python3 chk_txt.py pebble_centers_re2.txt` will produce
    ```
      Input file:  pebble_centers_re2.txt
      Nsph 352625
      Xmin   -82.2655
      Xmax    82.2703
      Ymin   -82.2662
      Ymax    82.2711
      Zmin   -60.1758
      Zmax    60.3377
      Rmin    24.1010
      Rmax    82.3009
33 34
      dtouch 1.997404   << min distance between any two spheres' centers
      dsep 1.999376     << "average" length to seperate spheres, = kissing number at 2.5 
ylan's avatar
ylan committed
35
    ```
36 37
    (`modulde load python` on Summit)    
    (It takes some time to compute dtouch)    
ylan's avatar
ylan committed
38 39
    dsep is about to 2, which makes sense for the radius=0.91945 
    
ylan's avatar
ylan committed
40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65

- Geometry and BC in re2 file:
  - linear mesh, no curved sides
  - Sphere distribution: free-free (selected from some range in Z, free = non-flat)
  - Sphere radius = 0.91945 (Wall)
  - Cylinder (inner) radius = 23.101 (W01, convert to Wall in production run)
  - Cylinder (outer) radius = 83.300 (W02, convert to Wall in production run)
  - Zbot of the cylinder = -64.275 (W03, convert to Inflow in production run)
  - Ztop of the cylinder =  66.437 (W04, convert to Outflow in production run)
      ```
        xyz min    -83.301      -83.301      -64.276
        xyz max     83.301       83.301       66.438

        mesh metrics: (linear mesh)
        GLL grid spacing min/max    : 1.18E-03 3.22E-01
        scaled Jacobian  min/max/avg: 4.33E-02 1.00E+00 4.09E-01
        aspect ratio     min/max/avg: 1.05E+00 4.60E+01 6.79E+00
       
        mesh metrics: (curved mesh)
        GLL grid spacing min/max    : 4.41E-04 3.18E-01
        scaled Jacobian  min/max/avg: 2.66E-02 9.99E-01 3.84E-01
        aspect ratio     min/max/avg: 1.05E+00 7.72E+01 1.07E+01
      ```

- Mesh file (`ann350k_t1r11.fld+X+time=0.0`)
  - lx1=3 GLL points projected onto curved sides
66 67
  - no issue for interpolation onto lx1=8 and mid-multigrid levels at lx1=4
  - no issue for lx1=2 corase grid
ylan's avatar
ylan committed
68 69 70

- Production run
  - Use re2 for boundary conditions, but use restart file for the curved mesh
71
  - We use dimless scale to run Re5000 with vz=1.0 (Re10k seems ok, but no time/allocationto test)
ylan's avatar
ylan committed
72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97
  - NekRS version (after udf changes)
    ```
      commit 82a68b6bbf550c84e3757ffcf06b783578a133d3
      Author: Stefan <stgeke@gmail.com>
      Date:   Wed Jan 20 21:02:30 2021 +0100
      
      Fix zero usr file generation
    ```

  - Runs (check `dat_logfiles`) (1600 nodes on Summit):
    ```
     - run1, Re50
       step= 500   t= 1.00000000e-01  dt=2.0e-04  C= 0.30  UVW: 3  P: 16  eTime= 8.79e-01, 7.41745e+02 s
     - run2, Re5k
       step= 5000  t= 3.50000000e-01  dt=5.0e-05  C= 0.10  UVW: 1  P: 7   eTime= 4.02e-01, 1.75611e+03 s
     - run3, Re10k
       step= 3000  t= 9.50000000e-01  dt=2.0e-04  C= 1.19  UVW: 1  P: 25  eTime= 1.09e+00, 2.44549e+03 s
     - run4, Re5k
       step= 2490  t= 1.24800000e+00  dt=2.0e-04  C= 0.95  UVW: 2  P: 35  eTime= 1.49e+00, 2.37206e+03 s
     - run5, Re5k
       step= 5000  t= 2.15000000e+00  dt=2.0e-04  C= 0.77  UVW: 2  P: 42  eTime= 1.82e+00, 5.22005e+03 s
     - run6, Re5k
       step= 10370 t= 4.02400000e+00  dt=2.0e-04  C= 0.83  UVW: 2  P: 23  eTime= 1.11e+00, 1.04075e+04 s
     - run7, Re5
       step= 11000 t= 6.15000000e+00  dt=2.0e-04  C= 0.96  UVW: 2  P: 28  eTime= 1.26e+00, 1.04812e+04 s
    ```