Consultancies

Here’s a story we’ve seen a few times in the last 12 months at ShapeSpace:

A small/medium sized engineering company has been happily using CAD for 10 or more years. Things have gone well,  perhaps the design team has doubled in size,  and tens of thousands of parts and assemblies have been created. But now it’s becoming clear that the organisation of the data is getting a bit out of hand.  Lots of time is being spent looking for data.  Duplication is everywhere as files have been copied between folders.  Something needs to be done.  So the decision is made – we need a PDM or PLM system.

Yes, we can use batch tools to upload files and product structure, but we know our data is badly organised and so we’ll need to clean things up. We can bring in consultants and set up an in-house team to tackle the data. But now we’re faced with the scale of the problem. Perhaps we  have 50,000 files and 10% need some attention prior to migration into the PLM system. That’s 5000 files. Say, as a wild underestimate, each needs 10 minutes of someone’s time to open,  do a where-used,  and make some decision on what to do. Thats’s 50,000 minutes, or 833 hours, or 104 working days – solid. So we throw our hands in the air, decide to park the data somewhere and migrate it into our PLM system as it’s needed.

Yes,  PLM will help to keep future data in order, but it is actually no help to the problem we started with – organising our existing data.

Oleg Shilovitsky has written about legacy data here.  However,  ’legacy’ comes with an implication that the data is old – in fact, all the data at this point is legacy data.

We should look for help in the data itself.

Most of the information needed to help with the migration is already present in the data, but we just lack the tools to utilise it. François Guillaumin wrote an interesting blog post on how he used a semantic analysis to classify parts prior to a migration.  We (ShapeSpace, along with our partners AESSiS) have used our shape similarity algorithms to determine groups of geometrically identical parts from the CAD data, and then analysed attributes to determine duplication.

I think tools to data-mine and data-cleanse within product data is an area where much more should be done. After all your PLM system can only be as good as the data in it.

I’d be interested to hear your thoughts….