MICRESS Forum

I would interest in running a simulation starting from a microstructure, which is the results of a previous simulation.
Although this is straightforward for 2D simulation, by reading the initial grain structure from a KORN ascii file, I do not know if a similar procedure may be also used for reading a 3D microstructure. In other words could it be possible to read a “3D” ascii file with a geometry as AnzX, AnzY and AnzZ?

Best Regards

Pina Mecozzi

Dear Pina,

welcome to the MICRESS forum!

Yes, it is possible to read initial structures from prior simulations in 3D! The difference is just that in 3D you have to use the vtk format. If the results of the prior simulation have been written in the normal MICRESS format, you can convert them by using DP_MICRESS.

Another way would be to use the restart file. The option "reset_time" removes the time history, so that you can start at t=0 and with a new initial temperature. The problem of using the restart file is that the geometry must be the same as in the prior simulation, and that both simulation must have been made using the same MICRESS version and format (windows/Linux, 32bit/64bit). Furthermore, not all parameters can be changed freely. The main advantage of using the restart file is that the diffuse interfaces are directly imported without going back to sharp interfaces.
Currently I am working on a more general way of reading in initial structures from a restart file which furthermore will allow combining initial microstructures (i.e. from file, from restart or definition inside MICRESS) in one simulation!

Best wishes

Bernd

Hi Bernd,

I have been trying to start a new simulation using 3D microstructure from another one. Following your suggestion, I converted the *.*korn file using DP micress to both vtk ASCII and binary format and then tried to read this file from a new simulation of micress. Unfortunately, everytime I ended up with an error massage saying that the point data is provided not the cell data. It will be kind of you if you can give me some suggestions.

Regards,
Razzak

* edited * * edited*
In DP_MICRESS 7, we decided to convert to point data because it is more comfortable to display with Paraview, i.e. using the contour filter.
However, DP7 does no interpolation to get point data from MICRESS' cell data if you do not change the grid geometry. The cell data is just moved from the centers to the corners of the grid. You can easily convert the data back by changing the keyword 'POINT_DATA' to 'CELL_DATA' with a normal text editor for ASCII data or with a hex editor if you have binary data. Additionally, the dimensions have to be raised by 1 to match the amount of data.

With the next update of MICRESS and tools, we will make the data format more consistent.

Ralph

Hi Bernd,

Thank you for the reply.

After converting the ASCII point to cell data, the input grain structure file is readable using MICRESS 6.003 but the same file can not be read using MICRESS 6.100. The microstructure is in 3D and I would like to take the advantage of MICRESS 6.100's parallel option to accelerate the simulation. Please let me know a possible solution for this problem.

Regards,
Razzak

I have tested the following workflow successfully:

1. Generate an initial grain structure with MIC 6.003 in MIC format
2. Converted it to VTK ascii point data with DP_MICRESS 7
3. Converted it manually by changing POINT_DATA to CELL_DATA and raising the dimensions by 1
4, Read in the grain structure from this file with MIC 6.100 parallel The field specifier 'korn' has to be added in the driving file, resp. is chosen when no specifier is given.

Is the field specifier 'korn' perhaps missing in your VTK file? Should be like this 'SCALARS korn float' in the header of the field. If not, please send us (support@micress.de) the result file in MIC format. I will have a look at it.

Best regards,
Ralph

Sorry, I made a mistake in my first post above. I edited it once more.
Definitely, while converting back from point to cell data, the dimensions have to be raised by 1. The product of the dimensions always represents the number of nodes of the structured grid.

Ralph