CAD Forum

tldr; I want to mimic (and hopefully improve on) the process of reverse engineering with CMM, but am not sure what the options and scales of affordability are regarding software workflows.

I work at a machine shop that my coworkers would classify as a "Job" shop, and we were visited today by a CMM saleswoman from Keyence.

We had previously discussed CMM because we have a client who generates more rejections because of theirs, but this was the first time I had seen one in action.

While my coworkers were primarily concerned with the limited precision of the model (+/- .0003 in) compared to the tolerances of jobs they would be validating (they claimed half a tenth would be their need to justify the investment), my primary interest was in it's capacity to quickly reverse engineer worn, walk in, and low precision parts, turning what I would call the "mystery meat".

Part of this was based on my understanding of our business needs, but also from my previous work and hobby experience with parts specification, 3D printing, and my experiences with the pros and cons of different CAD systems and styles.

In my previous experience as a parts manager for a manufacturing and saw mill operation, I've had to deal with a lot of mystery meat, and I know from experience that getting it documented and reproducible (especially affordably) is an uphill battle, even with very liberal tolerance requirements. It's a lengthy process for myself, and while smaller medium shops can bring more experience skill, and speed to the process, that ought to make it a viable service to offer, it often isn't where they make their money and they seem to have limited interest in doing such work even when asked, or even pressed. In my experience at this shop, the specification issue often creates hidden costs and losses (i.e. unbilled time for CAD and customer handling).

I've recently begun learning parametric and sketch based modelling, starting with Fusion 360 and Freecad, and immediately recognized how similar the process can be to machining operations (i.e. establishing parallel surfaces for reproducibility, defining paths to target geometry and reference points, mapping measurements to coordinate systems).

As a 3D printer, I'd already had an eye on the improving market for consumer scanners, and the sales demo really confirmed to me how closely related these paradigms are.

What I saw was a point cloud scanner that semi-automatically inferred parametric models, only to be limited to outputting them into less well defined formats. The whole thing was bundled as a proprietary package that combines the shortcomings of each stage.

So my thought process is, if the precision doesn't justify the cost for output validation, and you're going to have to combine it with a software workflow anyway, can't one just combine the more affordable consumer 3D scanner options with an affordable software workflow to get the same benefits.

My brain tells me this is an obvious option, but after some basic research, I'm not really sure what roads to pursue, or if the market really cares enough to offer the kinds of tools that would be needed affordably.

At the end of the day, to justify a significant software investment, I'd need something smarter than merely give me a surface model that I then need to redraw as parametric from scratch (like using a reference image as a starting point in Fusion). Being able to (even semi-) automatically detect, holes, centers, planes, steps, chamfer/fillets, tangents, perpendiculars, and spliny profiles doesn't seem like too tall an order, and there are clearly companies making big money selling moderately impressive tools to do it. But I'm not finding clear options on how to quickly and coherently reduce the complexity from a point cloud scan to a constructive geometry or parametric definition.

My searches seem to be hindered by the disparity between the different styles of 3d modelling. Many software offerings seem to offer part of the process, but the ones I've found marketed towards reverse engineering are either locked in a high precision scanner bundle, or are sold as an expensive plug in to an even more expensive CAD package.

CAD/CAM products are almost always egregiously overpriced, in my opinion. Most of renown, and many imitators, are completely inaccessible to me on my personal budget, and incredibly hard to justify to most of the businesses I've worked with or for. Even at our job shop, it is incredibly hard to justify further investment in a CAD product versus another machine that can actually put out parts. I'd go further to add that my experience with the name brands and the off brands tends to be the same, it's overpriced for what you actually get, and often half broken and awkward in mysterious ways.

So I'm left to wonder, is this simple dream just out of the little guys league, because market forces keep it in the domain of big players, or are there viable options to leverage the increasing financial viability of 3D scanners to rapidly reverse engineer tangible items and murky 3D models.

In either case, what do the required software, workflows and cost structures look like?

So not to say that this can't be solved but there's quite a bit here that is all over the map here. Let's take a quick step back to start to define what the box looks like...One thing about adapting/acquiring a new process and knowing when it will pay off is to know what your current costs are. If that isn't known then no amount of high or low costing solution can ever be compared properly. Also the company has to ask itself am I ok still faxing paper vs using emails? Sure faxing works but it's a ton less efficient.

Second, what are the typical sizes of the models you want to scan? Not all scanners can do everything, especially when looking to get different levels of accuracy and precision.

Yes as a whole these ecosystems are initially closed because it allows for reproducibility of process and output. Not all scanners work the same way...some are laser based, others structured light...etc so it's not possible, nor does it make sense for them to be open source in the beginning. There is also a TON of range of functionality that is offered in different packages that help take the 3D mesh data and convert it to CAD. This isn't an easy feat and requires a ton of intelligence to be built into the software to be able to create and output the many various ranges of what's needed for reverse engineering. The CAD softwares are trying to get better with working with mesh data to help bridge this gap but F360 has a LOOOOG way to do it. So for the moment you're going to have to work with the software that comes with the scanner.

This is a harsh one but I always like to keep it honest.... it's not that something is expensive, it's just that "our" wallet can't afford it. Which is a BIG difference. You'd no sooner walk into BMW's showroom, asking for a car with the expectation that you'd get it for the price of a Honda Civic. CAD/CAM software, if amortized over the course of say 5 years should be paid for, after that it's pure $$ outside of the annual maintenance. But again this all goes back to part of my initial statement about knowing what the current cost are. This isn't to say that it has to be 100% dialed in but from what you've described it needs to be much more buttoned down.

What you might consider is a hybrid solution.... outsource the 3D scanning to a company that can do all that you need for a year. What this does is help to prove the use case for needing the technology in-house. While this is happening you can start to get the current metrics dialed in so that by the end of that year you'll have hard data to work from and inform yourself and the company one way or the other if this is a viable route to go.

When you have 20 ft +, 10000 lb parts to CMM, you'll want a laser scanner.
At this scale 0.0003" is lots of accuracy.
I don't see a "job shop" need 0.00005". 0.005" will be really good.
I've work with the about 10000lb parts to IR camera internal parts that need 0.00005".

With proper software, laser scanner is great.
You use your CAD model and the software compare scan with it and tell you the deviation.
You can "CMM" it on the mill/CNC. Without tearing down the setup and finish machining after.

Scan to model again heavily depends on software but still mostly for "organic" parts.
You can compare scan to model to see if you are close.