absolute area fraction by variant

[trudator]

Hi Frank,

Is there any way to plot the absolute fraction of each variant in the whole grain graph? To be more specific, is that possible to plot it using EBSD data instead of grain data? Thanks Frank!

Linh

[frankNiessen]

Hi,

this might not be the most elegant way of doing it, but you can compute the variant ID of each EBSD pixel this way:

    pGrains = job.grains(job.mergeId(job.ebsdPrior(job.csChild).grainId));
    isParent = pGrains.phaseId == job.parentPhaseId;
    pGrains = pGrains(isParent);
    cEBSD = job.ebsdPrior(job.csChild);
    cEBSD = cEBSD(isParent);
    varIds = calcVariantId(pGrains.meanOrientation,cEBSD.orientations,job.p2c,'variantMap',job.variantMap);

Then you can check how often each variant ID is present within varIds which should give you the relative area fraction based on EBSD data.

Please let me know if it works.

Best wishes
Frank

[trudator]

Hi Frank,

I think it's the same for the dataset used in example 1 too. As I said, even though I have used a couple of cleaning methods to smoothen/denoise my ebsd data (left image), the variant map (right image) still returns with many white pixels. My question is how to achieve plot the variant map based on ebsd data after smoothening/denoising. I have attached my code here for better use.

I wanted to point it out since it would affect my area fraction vs. variant/packet graph in some ways (not too much in example 1, but could be a huge impact on another dataset, like mine). Here below is the two graphs from ebsd and grain data from example 1.

Thanks Frank!

[frankNiessen]

I think you misunderstand the way EBSD data and grains work in MTEX.

This is the EBSD dataset from Example 1, it contains several unindexed points.

Then we compute the grains, this operation consumes all the unindexed points into the grains to reduce the effect of fragmentation

The map you show on the left handside is a reconstructed EBSD map in which the unindexed points have been filled with a half-quadratic filter, it is not the original EBSD dataset.

This operation does not extrapolate the variant IDs as well, so you cannot get a variant map from such postprocessed maps.

The variant map function works either on the ebsd dataset

plotMap_variants(job,'linewidth',3);

or on the grains dataset

plotMap_variants(job,'grains','linewidth',3);

but not on further postprocessed EBSD data.

[trudator]

I see. Thanks for your interpretation, Frank. So way back to my previous question: in case I have some huge deviation between EBSD- and grain-based area fraction vs. variant chart (either remove noise or local unindexed pixels or not), which one you would recommend using (for the sake of data transparency)? Thanks, Frank.

Here is your recommendation above.
I would either go for the EBSD data (Assuming you have cleaned it properly before reconstruction) or grain data weighted by the grain area.

[frankNiessen]

It really depends on the case and there is no wrong and right as long as you clearly state what you have done in your methods. I would probably trust the cleaned (and not smoothed) EBSD data the most, so removal of CI<0.1 and of small grains. If you don't expect a different tendency of the variants to be succesfully indexed this should give you the right picture.

Best wishes
Frank

[trudator]

Sorry, I forgot to say thanks to you Frank!
Thanks Frank so much for helping me out.