What is the best way to reverse engineer a die in Creo?

[pimm]

We are partnering with another shop that has sparse data for converting their dies for machining purposes.  Sometimes we get part models, sometimes dirty geometry die models and sometimes absolutely nothing.

In the most recent situation all we have is a scan of the top and bottom dies.  The die impressions themselves don't have any sharp edges, and the radii sometimes wipe out the draft walls.  There would be a combination of revolves, extrudes, lofts and there are locked transitions.  We are tasked with reverse engineering these dies to make parts that size accurately and within a timely matter.

I wish that I could share a screenshot of the actual dies but that would be proprietary in nature.

I have been told that machining the actual STL is very slow and not very clean.

Can you suggest the best way to convert the STL scans into at least sealed, accurate to machine surfaces?  It would be nice to have these parametric, but that would require these surfaces to be analytical which wouldn't be easy to do.

[KenFarley]

In the past I've had to do something like this, trying to build a part with parametric features based on STL data and such. As with all such endeavors the method that works best "depends". Depends on what kind of assumptions you can make about the types of surfaces that comprise the part, what kind of tolerance you need, what you can surmise from the parts that are made with the dies you're trying to reproduce, etc.

What I've done as a general method is to take the STL data as is, in its own part file. Make an assembly with the STL derived part and my new "properly" modeled part, and as I build the nice model, check that I'm seeing agreement with the STL points and such. It's tremendously tedious and if you've got lofted surfaces it will be difficult.

STL data for machining is probably horrible, because you are machining little triangular planar surfaces (facets). And you'll probably always be able to tell that it's what you did.

[pimm]

Ken:

When we have a scan of the part itself, we do take advantage of stacking the STL in assembly with the model being built.  For us this works fairly decently even though an IGES or STEP works much better.

When trying to replicate die geometry from a STL it certainly complicates things.

[tbraxton]

I have had to deal with this many times for tooling and parts. Using the STL file as you have seen is really tedious and not very useful for anything but trivial geometries. If you have scan data, then you should have the raw point cloud data available. There are software packages that can generate analytical geometries from the scan data. Geomagic Design X and Zeiss reverse engineering software are two options. For a one-off job, it will not make sense to buy this type of software ($$$$), but you can outsource the work to a vendor that can get you a 3D model in Creo (typically a STEP file) based on alignment and fitting to the point cloud from the scan.

Geomagic X has LiveTransfer™ Function. The LiveTransfer™  command allows real-time transfer of whole or partial data created in Design X to target CAD systems. The LiveTransfer™  command transfers the geometries, topologies, modeling histories, and parameters of a modeling job to a CAD system that can then be edited.

PTC also has some reverse engineering capability but IME it is not best in class and requires purchasing a license for it.

[pimm]

I certainly would like to farm the point cloud out instead of modeling a bunch of one and done models; especially if our company doesn't purchase the Geomagic Pro software.  We do have the Polyworks software but from my past experience the modeling package is quite tedious.

[pimm]

Just noticed what you said in terms of the Geomagic LiveTransfer Funchtion:  I am not sure that this would work with geometry that has very little trace of analytical features but this is very impressive.