Windchill AI Parts Rationalization

[avillanueva]

PTC Launches New Windchill AI Parts Rationalization Capabilities

I remember seeing this at last year's conference. Very impressed at how quickly this was put together. I am even more impressed that its available with release. Not impressed at how little information has been published about it beyond a short demo video. If anyone has more information, please share in replies. I do have the slides from the last conference which go into great detail. A document explaining security and data protection would be very helpful since it would be a no go at my company if that was not addressed. 

Just looking at what I know about this tool, I had two concerns which might lessen the ROI. Perhaps this would be an issue not in my company but a different manufacturer would see a greater need.

1. If we are using shape information (geometry) to identify duplicate parts, we have already missed the cost avoidance portion. Say Bob fails in their search for an existing part, CADs up what they need and do all the design work to release it. Its only then would this AI tool at the next run determine that we have a duplicate. It would be great if Bob was stopped earlier in their efforts.  Where else would we see savings once a duplicate was already created?
2. How often would this need to be run? Say Becky kicks off this tool to analyze the system, follows the steps to identify and disposition duplicates. All findings are addressed or at least we knocked out the big portions. Are we going to see deminishing returns?  Is the pricing model one where we can run this as a service, clean things up, and then move on?

Curious to know your thoughts.

[Fadel]

@avillanueva maybe this HC link can give some details : https://support.ptc.com/help/windchill/ai_plugin/en/#page/Windchill_Help_Center/parts_rationalization/AIPluginIntroduction.html# 

[avillanueva · Accepted solution]

Ok, that answers some of my questions. Looks like there is AI credit being billed. Reference pricing would be nice even if it said contact PTC sales for accurate and current numbers. We also see that this is a continual indexing process. Much like SOLR, we can kick off indexing for a select set of contexts. Docs say 2 credits per part and that if there are changes, it can reindex. So in theory, if we index a set of libraries and disable indexing, we can search the resulting data for duplicates with no use of credits? 

As for the use cases, interesting that scale is not considered when classifying as a duplicate. This would be useful if I was to find duplicates for the purpose of classification. Hey, all of these look like bolts. These all look like washers. And how many times did I say not to download fasteners with modeled threads! Assisting in classification makes sense but dimishing returns.

Not sure what was the use case for change notice integration. I guess it was if you found a dup, you can add to a change notice to obsolete it? What else would you want to do about it? This would make more sense if it considered scaling as part of the data set but it says it does not.