Queries regarding Stats Generator

[DANISH KHAN]

Hi,

I am interested in using DREAM 3D to generate microstructures with various statistics and use it to predict the effect of statistics on mechanical properties for my research. For this I am using Stats Generator. In this regard, I have certain queries regarding the terms and terminology used in Stats Generator and their meanings:

1) Is there a more detailed manual for Stats Generator? Or is there any literature/paper/book reading which I can understand most of Stats Generator terminology?

2) What is meant by Omega 3 distribution?

3) What is it meant by Radial distribution function for precipitate? Is it for inter-precipitate particles distance distribution? In that same context , what is meant by box sizes?

4) How are the ODF statistics for the orientations being generated? . What is meant by Weight for particular Euler angle set in Weight and spread tab (specifically the notion of multiples of random distribution) ?  In the same tab , what is meant when you explain sigma as "Spread to use in blurring out the orientation chosen" ? What is the use of Sample points?

Pardon me for asking lots of queries. :) Being a new user of DREAM 3D I am finding most of the terminology very new to me. A small suggestion would be to please make the manual a bit more detailed so that any new user can understand the meaning of the terms and what is meant by them.

Looking forward to your response.

Thanks!  

[Sean Donegan]

Hi Danish,

Please see below for some answers to your questions:

On Thursday, January 21, 2016 at 6:18:15 AM UTC-5, DANISH KHAN wrote:

Hi,

I am interested in using DREAM 3D to generate microstructures with various statistics and use it to predict the effect of statistics on mechanical properties for my research. For this I am using Stats Generator. In this regard, I have certain queries regarding the terms and terminology used in Stats Generator and their meanings:

1) Is there a more detailed manual for Stats Generator? Or is there any literature/paper/book reading which I can understand most of Stats Generator terminology?

The documentation for StatsGenerator is pretty light at this point, but it is a known issue that we hope to address moving forward.  Most of the terminology is adopted either directly from statistics or from materials science literature.  If a particular term is confusing and not properly defined in the documentation, feel free to post here and I'll clarify.

 

2) What is meant by Omega 3 distribution?

Omega 3 is a description of the shape of a 3D object.  You can think of it as relating to the exponent in the equation of a super-ellipsoid; varying the parameter changes the 3D shape from a sphere through a cube-octohedron.  The following references have detailed explanations of omega3:

[1] Representation and Reconstruction of Three-dimensional Microstructures in Ni-based Superalloys, AFOSR FA9550-07-1-0179 Final Report, 20 Dec 2010.

[2] On the use of moment invariants for the automated classifcation of 3-D particle shapes, J. MacSleyne, J.P. Simmons and M. De Graef, Modeling and Simulations in Materials Science and Engineering, 16, 045008 (2008).

[3] n-Dimensional Moment Invariants and Conceptual Mathematical Theory of Recognition n-Dimensional Solids, Alexander G. Mamistvalov, IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 20, NO. 8, AUGUST 1998, p. 819-831.

 

3) What is it meant by Radial distribution function for precipitate? Is it for inter-precipitate particles distance distribution? In that same context , what is meant by box sizes?

The radial distribution function describes the distribution of distances between particles.  The box size in this context is the total sampling volume over which the radial distribution function is approximated; if you plan on matching the RDF in your synthetic microstructure, then the box dimensions here should be (x, y, z) dimensions of your the structure you plan on creating, in voxels, to ensure proper sampling.

 

4) How are the ODF statistics for the orientations being generated? . What is meant by Weight for particular Euler angle set in Weight and spread tab (specifically the notion of multiples of random distribution) ?  In the same tab , what is meant when you explain sigma as "Spread to use in blurring out the orientation chosen" ? What is the use of Sample points?

The ODF statistics are generated by binning orientation space (specifically, Rodrigues orientation space) with a bin size of 5 degrees.  This results in a specific number of bins per point group (for example, for cubic high symmetry, which has the point group m3-m, the number of bins ends up being 5832).  Then, a number of orientations are chosen from this binned space equal to the number of sampling points entered.  How the orientations are chosen is based on the entered weight and spread values.  The weight in this sense is in units of multiples of a random distribution, or MRD.  MRD represents the likelihood of a given orientation appearing relative to its frequency of appearance if the orientation distribution were completely random.  For example, and MRD of 5 for an orientation A means that for the given ODF, orientation A is 5 times more likely to appear than if the ODF were completely random.  Sigma specifies whether the orientations can be chosen "smeared" about the entered orientation.  Specifically, it determines the number of bins about the orientation that may be chosen during the sampling.

 

Pardon me for asking lots of queries. :) Being a new user of DREAM 3D I am finding most of the terminology very new to me. A small suggestion would be to please make the manual a bit more detailed so that any new user can understand the meaning of the terms and what is meant by them.

No problem!  Please ask as many questions as you need, it helps both you and the community to have these things answered.

[Sean Donegan]

Hi Ankit,

If you want to ensure that your features/particles are physically aligned preferential to some direction, make sure that you are modifying the axis ODF in StatsGenerator, not the normal ODF.  The axis ODF will effect the morphological orientation of the particles, while as the plain ODF will only effect the crystallographic orientation.

Also, if you want a large majority of your particles to be aligned along a given orientation, you should increase the weight value; this number will determine how strong the alignment ends up being.  Try using a weight value closer to 50,000.  The number does have a numerical meaning; its value interacts with the number of bins in orientation space for the given symmetry being used.  For the case of an axis ODF, orthorhombic symmetry is used, and since DREAM.3D uses 5 degree bins for all our ODFs, the total number of bins for orthorhombic is 46,656.  This means that if you use a weight value greater than 46,656 for a particular orientation, you will saturate the ODF with just that orientation.  This way, you can enforce that all your particles have nearly the same orientation.  If you want some deviation from the specific orientation, then you can reduce the weight value.

hope that helps,

-Sean

On Thursday, January 21, 2016 at 6:18:15 AM UTC-5, DANISH KHAN wrote:

[Stanislav Buklovskyi]

Hello everyone,

Just adding a little bit to the discussion. I was wondering if there is a reliable way to determine the "weight" and "sigma" parameters for the axis ODF angles?

Let's say I have a histogram of angles (see picture attached) with a preferred orientation (in this case 45 deg.). What would be the generally accepted way to approach such a problem? Is that enough input information? Any suggestions are welcome. 

Best,

Stas

среда, 23 марта 2016 г. в 12:45:49 UTC-4, Sean Donegan:

[Michael Jackson]

As Sean stated in the email reply, if you want *all* of the grains/features to have a specific morphological orientation then you would want to use a weight of 50000. I would start with an angle of 45, 0, 0 with a weigth of 50000 and sigma of 1 and see if that results in a synthetic microstructure that suits your needs. If you need a bit more spread in the tight distribution then adjust sigma upwards (2 or 3 or 4). If you want other angles to be included then reduce the weight from 50000 to 40000 to 25000 and see if that gives you what you are looking for.

 

--

Mike Jackson

--
You received this message because you are subscribed to the Google Groups "dream3d-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to dream3d-user...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/dream3d-users/d151d439-f24f-4f2d-a320-8a43b10e1bccn%40googlegroups.com.

[Buklovskyi, Stanislav]

Thank you, Mike.

 

I got the idea. Simple iterations for different coefficients to get something similar to the histogram.

 

I just recently started working with Dream3D and I am not very familiar with a generally accepted ways to do things, so I was wondering how this approach of picking coefficients would be justified? Is it just driven from experience?

 

Stas

[Michael Jackson]

Yep. Experience. Just a bit of trial and error iterations.

 

--

Mike Jackson