Automated vs manual steel microcleanliness

[steelmtllrgst]

Hi Friends.

I am looking at purchasing a new optical microscope that has automated inclusion analysis (image analysis) per ASTM E45. When the unit was demonstrated, it counted roughly 4X more D-thin inclusions in E45 Method D rating than our metallographer found. Another metallographer from a separate company verified our manual results. This is very typical of what we see with automated versus manual ratings.

Can someone help me to determine:
- Do the automated systems overestimate the ratings?
- Do manual ratings underestimate the inclusion amount?

Thanks in advance!

 

[winstonsk]

I'd be curious as to how the quality of the polish influences the quality of the results from the automated system. My human eyes can distinguish between polishing artefacts and stringers pretty well, but I don't know how well software might handle that. Additionally, you might also consider the lighting conditions used by the automated camera. If the camera is using oblique lighting or similar illumination, I could imagine that would impact the quality of the automated data.

These are just some thoughts--I don't have experience with these types of automated softwares.

 

[metengr]

My only tip would be to consider contacting Mr. George Vandervoort through his web site below. He is probably one of the most knowledgeable in the art of metallography and would be an opinion to consider for inclusion analysis.

http://www.georgevandervoort.com/me...oundaries-in-heat-treated-steels-article.html

 

[mrfailure]

There are 2 factors that you will need to address. First, as Winston noted, you need to review the polishing protocol to make sure no artifacts are left behind that will be counted. At a minimum you should make sure to use a set of clean polishing and grinding cloths to reduce transfer. The other factor is to adjust the threshold sensitivity in your software so it correctly differentiates actual inclusions. You should end up with automated results that are close to your manual examination results after you make these adjustments (unless the manual method is not being performed correctly). Good luck!

 

[steelmtllrgst]

Thanks Friends.

 

[Maui]

Contact LECO about their image analysis software (if they still sell it). I helped them develop the software for their image analysis system back in 1995. One of the features that was absolutely necessary to have is the ability to exclude certain artifacts from the analysis that the software would pick up during the rating process. No matter how carefully you polish your specimens and threshold the system, you will still pick up false readings. And you need to have the ability to review the fields that were scanned and exclude those extraneous features. Otherwise, you will always question the results.

Maui

 

[steelmtllrgst]

Hi Maui. The Leica software was able to do this and was quite user-friendly for removing artifacts. The problem is more on the high number of D-thin for E45 Method D, it would take longer to go through each field for these tiny inclusions than it would to just rate it manually.