General CAD question: 3d point clouds from hard probe.

[epichouse]

Hi. This is a very general question not necessarily specific to Fusion 360, although I have briefly been looking at the software. (I currently use Inventor 2009). This appears to be a busy forum with great general knowledge, so I thought I'd ask here.

I have a Renishaw cyclone probe scanning machine. It was mainly used to copy-machine broken 3d sculpted die/mould form tooling (as it could copy-machine the forms direct from the point cloud data without needing to surface or mesh anything!).

However, a major problem is not being able to give people models of the scan data. One of the main stumbling blocks seems to be that it is a probe scanner, from which the software does not output point cloud data that is accurate to the 3d form.

For example, if you imagine a simple square block that has been scanned (on the top face and down the sides), the bottom of the probe ball is "Zero" when it touches on the top face of the block, but, the X and Y 'zero' is the side of the probe ball.

Therefore on the scan point data, you get a waterfall effect as it goes over the corner and then down the side, as the bottom of the probe slowly and gradually turns into the side of the probe.

On a complex 3d form, it will be the same, gradually deviating as the form changes severity of shape. 2d shapes are fine, tracing around the edges of things. They are always true to size.

My question generally are:

What is this effect called?

Does Fusion 360 or any other Autodesk product have functions to translate the points to cater for the probe diameter (and thus re-calculate/ generate a true representation of the form)?

If Fusion or the latest Inventor cannot do this, does anybody know what would be needed to do it?

(The software can export an STL mesh, but again, it is not accurate to shape for the same reasons).

Many thanks,

[Fueler]

I also would like to know this.

[I_Forge_KC]

If you're looking for an automated solution in the Autodesk family, look to PowerInspect and PowerShape.

PowerInspect is a metrology software designed to take the place of the software that came with your inspection machine. It can provide a lot of mesh and point cloud tools (in addition to the metrology and quality tools).

PowerShape is a high-end modeler that is built on a different kernel than Inventor/AutoCAD/Fusion. It is capable of reverse engineering meshes and point clouds into BRep surfaces. It also has the ability to work with hybrid models that are mesh, surface, and solid. While Fusion (and Inventor) can have meshes in the files for display purposes (and some conversion), the tools in PowerShape for this are FAR superior.

https://www.autodesk.com/products/powershape/features

[Anonymous]

(Somehow managed to use an old log-in, will repost below to avoid confusion).

[epichouse]

Thank you for your reply.

I have taken a quick look at your link to the Powershape products (which I do vaguely remember as once being Delcam). It seems that Autodesk are in the process of making it look a bit more like Inventor, with the ribbons and the view cube. Looking at some of the demonstration video's, I wish I'd have had access to some of those tools in recent years when fixing up some complicated CAD models ready for designing fixtures around! It could have saved me days, if not weeks of work.

In terms of the mesh/point cloud editing, the video entitled "Reverse Engineering" is particularly impressive. I cannot quite see how our old probe-scanner can present the same kind of raw data, but if it could be presented in the same way (and the data re-shaped as described in my original post) it could be very useful indeed.

The problem with this solution though may come with the infrequency of this kind of work. When we used to be a main manufacturer of stamping dies and a third-party supplier of spare parts for other people's die forms, we did a lot of replication work on the probe scanning machine - which would be CNC re-cut right off the back of the point-cloud data. It was always "top down" scans of 3d die forms, never including the sides or underneaths of those objects, which were always plain.

Now, most of that work has disappeared - the larger companies bought their own, more sophisticated, laser and light scanners. We only get complicated things to scan every now and again, and these tend to be products which need to be scanned all over, not just 'top down' with the probe scraping over a surface to a boundary. These new kinds of customer don't tend to want machining to be done, they increasingly want a reference copy of their product. I am familiar with Rhino, and have thus far had to try and recreate these objects as best I can using that. It is hard work and very time consuming - and not all that accurate.

Given the infrequency of this kind of work, I'd not be sure the management would warrant a whole new system in order to cope with it. However, I will look into what the rental arrangements are. If they are like other Autodesk products, it might be possible to just use it when needed. It may also open another can of worms, in that, if we do go ahead and look into getting Powershape as a permanent addition, they may be better off investing also in some kind of hand-held laser scanner - and they would then need to make sure we can pull in the work to cover the outlays....

Anyway, thanks for your reply, it has been useful to see the tools in Powershape and being quite a fan and long term user of Autodesk products, I'd rather stick with things in the Autodesk umbrella if possible.

[I_Forge_KC]

While the Delcam portfolio is obviously the powerhouse of metrology and CAM for Autodesk, it isn't the only path to reverse engineering. I've been working a bit with @Fueler offline on an alternative method. It is substantially more hands-on (with a steep learning curve), but in the end will get you to quite similar results.

Those steps include:

• Using MeshLab (a free, open-source tool) to create a mesh surface out of the point cloud data
• Taking that mesh surface into InstantMeshes to retopologize it into a quad-heavy (if not pure quad) mesh
• Taking that quad mesh into Fusion for conversion to TSplines - Here it can be massaged and manipulated to jive with known details (e.g. hole sizes)
• Converting the TSpline into BRep data (still in Fusion) for downstream parametric modeling

One key difference between this workflow and the one shown in those PowerShape feature videos is that the PowerShape tools end up creating parametric features (for some tools). The Fusion-centric workflow ends up creating a dumb solid. The thing is though (as the PowerShape video with the lion head seal alluded to), that more complicated parts generally end up as dumb solids anyway.

I should have included the same video set for PowerInspect. PI is where things like the probe ruby offset can be managed.

https://www.autodesk.com/products/powerinspect/features

In a typical reverse engineering scenario at a mold shop, you'd see both softwares side by side. PI to gather and process the inspection data and then PS to rebuild the solid model.

[I_Forge_KC]

I would be remiss if I left off the other toolset for reverse engineering...

Autodesk Alias is used extensively in the automotive industry (I'd wager 99% of the cars manufactured today are built around Alias data). One of its key features is taking scan data (from the hand-sculpted clay models) and fitting surfaces onto those scans. It's interesting that in such a modern age, the concept of hand-carved car bodies is still front and center. Physically seeing (and feeling) your form lends itself to much better quality than digital only methods. Since all of that art has to be captured and digitized for tool development (as well as downstream engineering), that workflow is exceptionally well documented. Alias is kind of like Rhino or Moi but with the power and precision of a high-end solid modeler. It certainly isn't for the faint of heart or the casual user.

This is a training site, but just glancing through the lectures can give you insight for sure.

http://autodeskautomotivetraining.com/product/alias-class-a/

Hopefully, there is some good info in my ramblings somewhere.

[epichouse]

Thanks a lot, I_Forge. You have provided some useful information - which I will certainly need to further investigate. For example, with the ruby offset, it may be one thing to predetermine the probe information and then gather the points manually....but it may be something very different to already have points from the Renishaw Cylcone 2 and then have some software re-adjust those points back to a truly representative model shape. Here is a Cyclone 2 in action (https://youtu.be/BGhv7zsvTAo?t=294) to show you how it works. I will certainly take a look into the non-parametric workflow too, because it is rare we'd need a parametric part and we'd likely need to export such a parametric part back out to the customer as a dumb Iges, Step file anyway.  Its all food for thought. Thanks again.

[Fueler]

My Centroid works much the same way.

https://www.youtube.com/watch?v=IcVv1So1n_8

[epichouse]

That's an interesting piece of kit. I've never seen one of those. It does work in very much the same way, only the Renishaw seems to scrape over the surfaces whilst the Centroid seems to repeat-contact to collect the points. I suppose that you'd be in the same boat as us, in that, for example on the horse in the video there, you'd be able to copy it on a milling machine but not be able to provide a customer with an accurate CAD model of that horse...

If the company I work for ever decides to upgrade the scanning facilities, or would like to have more of a connection between the design department and the toolmakers on the shopfloor, I'd tend to mention this Centroid equipment to them so that die forms can still be copied and machined directly from the point data as the are now, but on more 'product design and capture' jobs, such as joysticks, mice, things with a true multi-sided 3d form, a laser scanner and reverse engineering surface modelling would be used. (We only have 3 axis milling anyway).

[Fueler]

That's pretty much it in a nutshell. The reason Centroid used that small horse for a demo is that you can turn around and cut the new piece after digitizing AS LONG AS you use the same size end mill as the ruby on the end of the stylus.

IE a 5mm stylus will require a 5mm end mill to pull this feat off.

They leave us with no hints on how to do anything larger and compute the offsets versus the 5mm stylus.

I just can't see a future in carving a 1 square foot piece down with a 5mm end mill. 

They only thing they offer is that the converted file which is a .txt can be read by many CAD programs. 

and that is where we stand trying to convert points into Fusion so new pieces and drawings can be generated. 

I probably not talking to the right person but Centroid does not seem anxious to work with me on this but they might talk to a big dog like Autodesk. 

Looking at the Renishaw slide move made me think that there is some super stuff going on there reading force versus position. Too much force moves it up a bit. 

[epichouse]

Oh dear, that doesn't sound too great about having to use the same probe size. I assumed it would take the points then know what to do with it with any kind of cutter and any kind of machining cycle.

The renishaw machine is old technology now, but I think it was a very clever piece of kit for its time. For the work we needed to do, it was good enough. Machining straight off the point-cloud was, well, just 'wow'. Spending hours, days, trying to resurface scan data, fix it up, make it watertight, export to a CAM program....all not necessary. I don't know if any other machine or CAM program can do that, even a decade and a half later. I have never heard of one and I would be interested to hear of any that can, and I don't include STL meshes in that, I just mean the raw point cloud data.

The Tracecut software, once the object has been scanned, lets you use any kind of cutter at any kind of size, with roughing cycles, different (but limited I suppose) finishing cycles. You can set areas to miss, or mirror the data, etc, so it did quite a lot and although a bit quirky sometimes, it did it well. I never had any trouble with the machining code, other than for stupid things I did when setting up the program cut information. The only thing you had to be aware of was that you couldn't mill any radius or areas smaller than the probe it was scanned with. It doesn't have high speed machining or anything fancy like REST machining, but there's different patterns of finish, different directions, pitches, cut depths, etc. 

The software can export STL mesh, TXT files of points, Iges point data......but the kicker comes in that the data is not true to the shape of the scanned object. The waterfall effect difference of the bottom of the probe being datum to the side of the probe being datum (when moving over a 3d object) gives an object that looks like the article, but isn't right. You can ask the software to 'create a true model' and it tries to shrink up/reshape this difference, but the results are not great, particularly around the edges of the area that has been scanned. When trying to replicate CAD data from this, I do ask it to do this, then create an STL mesh and DXF slices through the part in X and Y directions, then import all of this data into Rhino3d to try and make some rational sense of it.

For truly 3d objects which are multi-sided or need scanning three or four times in different orientations with the probe, we have no means to "match up" that data to create a fully populated point cloud that depicts the real world part. In that regard, it is kind of limited to the horses in the video.

I am pretty sure modern hand held laser scanners would be vastly superior for data capture and further manipulation in 3d software.....but we'd lose the ability to machine parts right off the Renishaw point data and thus start having to fully surface up that data to get models ready to process for successful CNC programming in CAM software. 

[Fueler]

I used the 5mm stylus since it is what I use on a daily basis and was already dialed / zeroed in.

I think I will go with a smaller one on the next scan to see if that waterfall effect is reduced. Hopefully this makes the edges more sharp. 

As I tinker with software perhaps the stylus size will be a moot point if the software knows what diameter produced that curved edge.