Restart options

new developments, future functionalities, hints for further applications of MICRESS
deepumaj1
Posts: 104
Joined: Fri Feb 13, 2015 11:00 am
anti_bot: 333
Contact:

Re: Restart options

Post by deepumaj1 » Tue Apr 17, 2018 3:26 am

Hi Bernd,
This trick is not working for me. The grain structure is not read in. The simulation starts with a uniform (empty) phase field map, which is then filled by the voronoi in grain input.

Deepu

Bernd
Posts: 1504
Joined: Mon Jun 23, 2008 9:29 pm

Re: Restart options

Post by Bernd » Tue Apr 17, 2018 5:22 pm

Hi Deepu,

There seems to be a misunderstanding: My proposal was not to set any grain by grain input (apart from the default grain 0).

Maybe, it is not clear for you that reading the restart file comes after initial grain input! Then, grain 0 is replaced with the restart file content which contains your grain information...

Bernd

deepumaj1
Posts: 104
Joined: Fri Feb 13, 2015 11:00 am
anti_bot: 333
Contact:

Re: Restart options

Post by deepumaj1 » Tue Apr 17, 2018 6:07 pm

Hi Bernd,
Thanks for pointing this out. It works fine now.

Deepu

deepumaj1
Posts: 104
Joined: Fri Feb 13, 2015 11:00 am
anti_bot: 333
Contact:

Re: Restart options

Post by deepumaj1 » Wed Apr 18, 2018 6:57 am

Hi Bernd,
With the restart structure only option, can I decrease the grid size, without changing the overall size of the domain? For example, I want to restart a simulation with 400x400 cells of 0.25 grid size by changing these to 1000x1000 cells with grid size of 0.1, so that the overall size of the simulation domain is maintained. Is this operation possible at present?

I gave it a try and saw that the domain is sort of mirrored up to fill up the 1000x1000 space and sometimes it ends up in warnings saying:
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
Warning in spfAddPhInkr, t=0.6887666E-03
Wrong fractions at nTupelp= 6419
sum = 0.000000
x,y,z = 132 1 800
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
Deepu

Bernd
Posts: 1504
Joined: Mon Jun 23, 2008 9:29 pm

Re: Restart options

Post by Bernd » Wed Apr 18, 2018 4:28 pm

Hi Deepu,

Essentially, the new "zoom" functionality of MICRESS allows only zooming by factors of 2 or 3 in each direction. The zoom factor needs to be an integer number which is 1 by default. If it gets too big, the result probably will be just noise which "materializes" into random structures...

The implementation in a first step creates a structure with multiple identical grid points without interpolation. Therefore, the factor must be integer. In a second step, the interfaces which now are much too broad have to be rearranged using "initialisation" steps like for sharp interfaces. However this "initialisation" process is now more complicated because a special diffusion algorithm must also rearrange composition such that the phase compositions and their gradients are reconstructed in the best possible manner. For this reason, you also need about 10 times more steps to adjust profiles compared to initially sharp interfaces.

Thus, the zoom method is restricted to small factors like 2 or 3 and always may create some artefacts. Therefore it is recommendable to play a bit with the number of initialisation time steps. If you need a higher factor, you could try do do it in several steps with some intermediate simulation time to reestablish a smooth microstructure.

The possibility to have different factors in different directions and to freely rescale grid spacing also would allow to mimic simple deformation steps.

Bernd

deepumaj1
Posts: 104
Joined: Fri Feb 13, 2015 11:00 am
anti_bot: 333
Contact:

Re: Restart options

Post by deepumaj1 » Thu Apr 19, 2018 4:45 am

Hi Bernd, - Question..
The zoom factor needs to be an integer number which is 1 by default.
Can I manually control this zoom factor? (Lets say I don't want zoom, but still want to change the number of cells in the domain)
I have made some trials. As I observe, the zoom feature gets activated when I change the number of cells in the domain. Right?

Following are couple of my trials:

Domain: 400x400, grid: 0.25
I
restart_2.png
restart_2.png (9.48 KiB) Viewed 7897 times
Domain: 800x800, grid: 0.1
restart_1.png
restart_1.png (19.58 KiB) Viewed 7897 times
If you calculate the real domain size here, both of them are 100x100 microns. What I would like to have in the second case is the same microstructure as the first case (instead of zooming out), on a finer grid size (normally what we expect when you refine the grid size). Is this possible?

Deepu

Bernd
Posts: 1504
Joined: Mon Jun 23, 2008 9:29 pm

Re: Restart options

Post by Bernd » Thu Apr 19, 2018 12:37 pm

Dear Deepu,

The size of the domain has nothing to do with the zoom factor! If you use "restart structure_only" you get the prompted e.g. the following input sequence:

# Restart options
# ===============
# Restart using old results?
# Options: new restart [reset_time | structure_only]
restart structure_only
# How many restart files shall be read?
1
# For each restart file a grain number and (optionally) a
# shift (in grid cells) and zoom factor for all 3 dimensions
# as well as a character for rotation options must be specified:
# grain number [shift X (int) shift Y (int) shift Z (int)
# zoom X (int) zoom Y (int) zoom Z (int) rot(string)] ?
# Rotation options: "xz+90" "xz-90" "xz180" "xy+90" ... "yz180"
0 0 0 0 1 1 1 xz+90
# Name of restart file?
M247_DS_Sol_TTNI7_HR_Selection_r
#


I have highlighted as bold the place where you input the zoom factor for each direction. In this line of input, only the first number is required (namely the number of the grain which shall be used as mask for the content of the restart file. The optional parameters must be given in the correct order, but may be skipped at any place if you do not need those which come afterwards. Default values are no shift (0), zoom factors of 1, and no rotation. If you just want to have a zoom with a factor of 2 without shifting (i.e. the position of read microstructure relative to the coordinate origin (0,0,0) stays the same for the domain stored in the .rest file and the new domain), you would input:

0 0 0 0 2 2 2

The size of the new domain has no direct influence on the way the initial microstructure is read from the restart file. However, if the new domain size is bigger or if you apply a strong shift, it may happen that microstructure from the restart file cannot fill the grain mask defined within the new domain. Then, if the original simulation which created the .rest file had periodic or symmetric boundary conditions, the restart microstructure is continued correspondingly. Otherwise, the mask will not be filled, which should be avoided because, especially in case of partial filling, numerical trouble is to be expected...

Bernd

deepumaj1
Posts: 104
Joined: Fri Feb 13, 2015 11:00 am
anti_bot: 333
Contact:

Re: Restart options

Post by deepumaj1 » Thu Apr 19, 2018 1:30 pm

Hi Bernd,
Thanks for pointing out these options. Didn't notice in screen output since I used old driving file in MICpad. It works great. :)

Deepu

Post Reply