3D Scans and Reverse Engineering in Fusion 360

[Anonymous]

I am currently investigating purchasing a 3D scanner to help me in creating CAD models of existing parts. The parts I want to model are mostly injection-molded plastic shells for things like consumer electronics, but I want to end up with a parametric model rather than just a mesh. The parts I am reverse engineering often have complex spline-style curves, which I can't just readily model using my trusty old caliper and bevel protractor. So a 3D scanner seems like a good way to capture those curves without having to physically cut the original parts, trace the curves in 2D, and then scan and trace into a Fusion sketch.

I'm looking at the Einscan Pro scanner. It's about $4k. I'm just a hobbyist, so I'd rather not drop that kind of money unless I'm pretty confident I can use it to achieve my purposes with Fusion. Does anybody have experience doing this kind of thing with Fusion 360?

My concerns are:

• I remember reading there is a limit on the number of polygons a mesh can have for import into Fusion. Perhaps this has increased since I read the number, but I remember thinking it was much smaller than most 3D scan poly counts. Anybody know what the limit is now?
• I saw in a YouTube tutorial someone offhandedly reference that there was a way to divide a mesh (in the new Mesh workspace) up into planes/sections of that mesh and use it to create splines that give you the contours of the mesh. Is this true? That's exactly what I want, as if I can get the basic curves of a scanned geometry, some basic extrusions and lofts will give me a parametric version pretty easily. But I can't find any info on this new functionality.
• How is size calibration handled? Do most scanners work out the real-world physical size of the object? Or is there some way to calibrate a mesh to its real-world size based on two known points (like we do with canvasses)?

Any insights or thoughts would be greatly appreciated!

[PhilProcarioJr]

@Anonymous

To be completely honest...if your planning to use "Only" Fusion and a 3d scanner your headed for a lot of long tedious work. On the other hand if you use other 3d software with Fusion you will get work done and be happy with the results.

1) As to the limits...that's computer dependent, but I would say a safe zone for a good computer is around 20,000-50,000 polygons and still be able to work.

2) Yes you can do this but I'm not sure your going to get the desired results on complex objects.

3) Most scanners output scale correctly and you shouldn't have to do any calibration between the model and Fusion.

[Anonymous]

@PhilProcarioJr: Interesting. What other 3D software, for example, would you suggest? I believe most 3D scanners come with software that generates STL/OBJ meshes, so I wasn't planning on actually scanning directly into Fusion (didn't even know that was possible, if it is). My planned workflow was something like: scan with the Einscan Pro, clean up and simplify the mesh in the scanner software, and then import into Fusion to trace the contours of the scanned geometry as a jumping off point for a parametric model.

This is the kind of thing I'm often trying to model:

I had an original part for the base there but in order to model it, I had to cut it with a bandsaw, trace the contours, scan them, insert as a canvas, and trace.

This feels kind of retrograde to me. But maybe a 3D scanner (at least the kind available in this price range rather than the $15k plus industrial ones with reverse engineering software) just isn't the right solution to the problem?

[PhilProcarioJr · Accepted solution]

@Anonymous

The problem is you want to buy an optical scanner so your scans will not be 100% accurate. Next is scan data is almost always dense, triangulated data which is always a nightmare to work with. So to really work with those scans you need to retopologize your scan meshes to really get good usage out of them. I do this work everyday so I know how much of a nightmare it can be to work with scan data.

To retopologize meshes I use topogun, and Zbrush although Mudbox is great for this also.

When you use the cross section tool in the mesh tools the results can be less then desirable so I would always plan on other software to save the day.

My typical workflow for scan data is retopologize the mesh into quads then bring that mesh in to either create surface to solid data or T-Spline to solid data.

My suggestion to you would be to download some free scan data and try to convert it into a Fusion model to see if this is the route you want to take. A lot of people I know give up that idea once they see how much work is involved to get true accurate solid models from scans...

[Anonymous]

We are evaluating where Fusion 360 might fit into our reverse engineering workflows also.

We have been using Geomagic Design X for the last few years which is a great product but not at a price point that's suitable for occasional use.  We've also used Spaceclaim with reverse engineering plug-ins in the past.  The debate is always whether one models in one of the specialised tools (like Geomagic Design X) or uses a mesh-based program to do data capture/ mesh clean-up/ mesh alignment/ mesh reduction etc. and then use a CAD tool like Fusion 360 to do modelling , drawings/CAE/CAM etc.

From an initial peek, Fusion 360 can handle lightweight meshes, can create polyline sections which you can fit sketches to and would allow you to create prismatic CAD models from mesh data. Is it is as sophisticated as Geomagic Design X? Not by an order of magnitude, but at an order of magnitude lower price. So for occasional use, you could use the software that comes with your scanner to generate a mesh and align it to the global origin. If you needed more work on the mesh, there are free programs out there (including Autodesk Meshmixer) that can help.

The last post was spot on with his comments about accuracy and precision from a hand held scanner. Mold-makers (and ourselves) generally use scanning arms which combine contact and non-contact methods to generate accurate data for reverse engineering machined parts like molds. But our Faro Edge HD scanner is at the $50,000 price point with software like Design X. Don't expect a $4k handheld to produce the same results.

[Anonymous]

What about Autodesk Recap and Autodesk Remake (discontinued?)?

Is Recap a solid alternative to DesignX?

[SC_Precision]

I can't speak high enough for Fusion in this case.  I am a novice to be sure, but I am having great luck bringing in scans with 2 million facets into Fusion, even working with them.  I know there are ways to get MUCH more efficient that some of these experts have chimed in with, but I have been developing my process for a couple years part time and am really please with how it is working for me.  I am using scans as a design "guide" and I make parametric models only by using a scan as a guide. 

The scans I have been using have been very accurate, and I have used the top of the line $100k+ Creaform metrology grade scanner, a David, and recently doing a lot with the new Einscan top end model.  I am getting results that match physical dimensions within a few thousandths of an inch.  The Creaform is a metrology grade unit and is more capable, but the DAvid and Einscan are performing excellent for what I need. 

You can see a few pics of my process below.  I am happy to share any info on the actual design side inside of Fusion as well, but here is the flow from raw scan, to 3D solid model, to finished part.

Loving Fusion for sure.  Gave up on Solidworks, Pro-E, and a few others and Fusion360 is now my one and only CAD/CAM............

raw scan. over 2 million facets! (yes, overkill)3D model "carved" around the scan data (actually a different model than the scan shown but you get the idea)3D model, machined, anodized, and in real life use.

[Anonymous]

Case, that is awesome - thanks for sharing! We do mainly castings such as cylinder heads, blocks, sumps, pumps, superchargers etc. for pre- and post-war autos and bikes etc. so do appreciate that what you've achieved with Fusion 360 deserves a very large Kudos point! Thanks for blazing the trail!

It would be interesting to get a handle on the Fusion 360 roadmap as to whether there is any development planned on the reverse engineering front. For sure, Geomagic Design X has great functionality that helps speed-up reverse engineering - for example it will auto-estimate a fillet size from scan data and can do scan-to-CAD deviation etc. etc. but at a massive cost - the annual maintenance charge here in the U.K. is nearly $4,000 per year including taxes!

The experience you've had with large poly count being usable is really interesting. We will keep plugging away on trialing F360, but so far, so good!

I need to get my head around how to align the scan data accurately in Fusion. In Design X you would typically align the scan roughly to the global origin using best fit to regions (they might be planar or cylindrical features for example), then take a slice through some geometry that had machined features, fit construction geometry to the polylines and use the resultant sketch to accurately align the scan via 3-2-1 alignment or similar. Time spent on alignment is time well-spent. How do you handle that process in Fusion?

Well done again!

[cekuhnen]

@Anonymous

I actually do a lot of reverse engineering in my research work (bio medical products) I design at work.

If you can use Fusion for this type of work is more influenced by what type of shapes you have to scan and rebuild.

If we are talking about objects that were constructed using basic geometry tools (box sphere cylinder fillets) then the tools Fusion offers are good.

You can make mesh cuts via the Sketch > Create mesh section tool.

If however your objects are more curved then you will quickly run into two issues.

1. Surfacing tools in Fusion lack abilities such as fine tune controls for surface blending. to help here you need to add sketches with blend curve rails

however the sketch engine has since a long time few serious bugs which break coincident and thus G1 G2 constraints making this not usable as a parametric design.

Direct Modeling will be fine.

2. T-Splines after the initial increase of tools was never really further refined and seems to be on hold for a long time. particularly for reverse engineering surfaces when

projected new surfaces over them T-S has the starting tools but lacks depth making it more complicated to use which is why I do not model at all in TS but use it only

as a converter. Like Phill mentioned you can and should also use addition programs to do the retopology.

TS unlike in Rhino also lacks some tolerance displays for reverse engineering surfaces.

Do you have some more images of what you want to reverse engineer?

To me it seems the AD team put down the foundation of tools yet did not refine them as much as needed to make them also very meaningful for me as a designer.

However considering that they work on multiple parts of the software this might come later.

Currently the sketch engine is getting a lot of attention to fix some of the foundational problems which later will enable other problems to be fixed afterwards.

[Anonymous]

Hey guys. I'm not sure why this thread picked up interest again after I initially posted it some time ago, but I'm glad to see that it has and wanted to provide an update. After this discussion above and some more investigation, I ended up getting an EinScan Pro and have now successfully used it on a few reverse-engineering projects in Fusion. I'm astonished by the scanner's accuracy (entirely sufficient for the type of work I do) and generally pretty happy with the Fusion workflow.

Fusion tends to be a little crashy (at least on the Mac) when using scan data—even after significant mesh reduction in MeshMixer—but other than that it's pretty smooth sailing, especially with the Mesh workspace tools that allow tracing a mesh's intersection with a plane to a smoothed sketch curve.

An example would be the below carry case (an original sci-fi TV prop that we're replicating), where we chose to idealize the geometry slightly to make it symmetrical (etc) but wanted to follow the original geometry as much as feasible.

This will be used to create a vacuum-forming mold and allowed us to work completely non-destructively with the (somewhat valuable) original rather than having to do silicone casting against the interior.

I've also used it for tighter-tolerance applications where I built a CNC-machined housing in Fusion around a 3D scan of a computer keyboard plate and PCB, and it fit the actual part perfectly.

I haven't yet gotten into really complex compound curve modeling with reverse-engineered objects, but per @cekuhnen's comments, I think the limitations there are more in the realm of inherent Fusion limitations not the scanning workflow per se.

[cekuhnen]

@Anonymous

sweet project

yeah as stated when your geometry is more linear or being build from basic geometry block fusion will do very fine.

a car body is a different story or a toy.

I just added the opinion based on all the projects I did and some people who are new to scanning and reverse engineering and fusion lack

to understanding of what will work we’ll and what not (yet).

Crashes so far I experienced less - rather slowdowns.

i did Work with models that have up to 4 million points but since the mesh section tool slows down fusion significantly when more sections are added

i often to some simplification of the models externally and then send that data to fusion to work further with.

[Anonymous]

Thanks folks for sharing your experiences. I think we've come to the conclusion that we'll reserve F360 for simpler prismatic shapes and stick with Design X for the more complex ones.

I enclose screenshots of a machined casting with core features which combines more "freestyle" autosurfaced external shapes (to save modelling time) with various sketch-based features like holes, cuts, cast internal galleries etc. Both the pattern-making and machining are driven from the same model, but stopping at different stages of the history/feature tree. This is a project that would be tough in F360 from what I have found so far, but is the stuff that Design X swallows with ease.

Glad to hear the sketching tools are getting a make-over and perhaps we'll see some more reverse engineering stuff come up on the roadmap.

Machined duplicateHybrid model combines autosurfaced exterior and parametric-based recaptured design intent.Patterns and core boxes milled from model

[cekuhnen]

@Anonymous

Design X Is also heavily specialized on reverse engineering which makes comparing Fusion to it (which is based on Inventor tools) a little tricky.

I would say Fusion where it is today is a good all-rounder. Not super at anything but offering a lot of tools for many tasks.

One also has to consider the extremely generous price tag for Fusion Autodesk provides which is number one reason why I switched my students

from SW to Fusion. And often Fusion is also enough for them.

But tasks we cannot do in Fusion (lack of modeling tool depth - sketching bugs) we then simply do with other apps and supplement Fusion this way.

More apps to learn but also a more diverse toolset at their hand I think.

[SC_Precision]

I looked at the website for Design X again, sure would like a demo on that.  Looks great by the description.  You can see my post above is nothing at all prismatic!  Its all curves and shapes.....  But I make do in Fusion and am happy with my flow now, mainly because I have a strong feeling that Design X costs a bit more than I want to pay!  I know , if it is really that fast I am paying with time.......but really still just happy with Fusion.

[chris.itterly]

Hi All,

I'm on this same path, after receiving a demo of the EinScan product this week, it should meet my needs for initial mesh capture.

The reverse engineering workflow is of particular interest to me, due to lack of experience.

• For geometric-like objects, are there any videos showing how to do this with F360?
• Geomagic Dexign X looks to be an Artec product with a $20k price tag, is this correct?
• I've watched many demo videos of Ansys SpaceClaim. It looks like a powerful tool set for reverse engineering, although it's $3.5K.

F360 has been great so far, enabling me to design and 3d print many things.

Also being a hobbist, budget is a real factor.

What other resources should I be exploring to connect the mesh to solid model gap?

Thanks

Chris

[Anonymous]

Hello Chris,

I've been working quite some time with RE services, scanners etc...

Currently and over 5 years, i have used ATOS Scanners( GOM.COM) at work , along Ansys Spaceclaim and totally replaced oud workflows with SoliwWorks and Design X, also for my personal projects use F360 and i'm loving it.

The one thing i can recommend to everyone is GOM Inspect to complement what everyone is doing. Its FREE, and you can repair, reduce, align any STL . Also compare with the designed CAD. and iterate changes. And finally do a inspection report to serve as warranty for the work done.

If F360 had this functionality, it would definitely step up

[chris.itterly]

Thanks for the guidance, I'll check out GOM Inspect too.

In preparation for getting the scanner, I'll be trying out several RE free trials this month:

SpaceClaim

Rhino

Mesh2Surface

I found some sites that allow download of scan data, with a few interesting components that should enable me to get my feet wet.

https://www.laserdesign.com/sample-files/

https://www.artec3d.com/3d-models

[Anonymous]

Hello, Chris! 

How it's going? Could you please tell something about using the software you were searching for?

Me and my fellow are making brief tutorials about 3d scan data processing in the most popular softwares and it'll be nice if you share your experience.

Thanks!

[masterskywalkerog]

can you rebuild or work from that scan ? 

[masterskywalkerog]

so pretty much  get i to the point where you can trace it n rebuid it correct ?